See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/302590255

Seismic Behaviour of Steel Storage Pallet Racking Systems (SEISRACKS2)”,

Technical Report · January 2014

DOI: 10.13140/RG.2.1.2325.5929

CITATIONS READS

7 1,293

19 authors, including:

Carlo Andrea Castiglioni Claudio Bernuzzi

Politecnico di Milano Politecnico di Milano
158 PUBLICATIONS   857 CITATIONS    57 PUBLICATIONS   441 CITATIONS   

SEE PROFILE SEE PROFILE

Benno Hoffmeister
RWTH Aachen University
73 PUBLICATIONS   175 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PROINDUSTRY - SEISMIC PROTECTION OF INDUSTRIAL PLANTS BY ENHANCED STEEL BASED SYSTEMS View project

FUSEIS View project

All content following this page was uploaded by Carlo Andrea Castiglioni on 18 September 2016.

The user has requested enhancement of the downloaded file.

 ISSN 1831-9424 (PDF)
ISSN 1018-5593 (Printed)

Seismic behaviour of steel

storage pallet racking
systems
(SEISRACKS2)

Research and
Innovation EUR 27583 EN
Interested in European research?
RTD info is our quarterly magazine keeping you in touch with main developments (results,
programmes, events, etc.). It is available in English, French and German. A free sample copy or
free subscription can be obtained from:

Directorate-General for Research and Innovation

Information and Communication Unit
European Commission
1049 Bruxelles/Brussel
BELGIQUE/BELGIË
Fax +32 229-58220
E-mail: research@ec.europa.eu
Internet: http://ec.europa.eu/research/rtdinfo.html

EUROPEAN COMMISSION
Directorate-General for Research and Innovation
Directorate D — Key Enabling Technologies
Unit D.4 — Coal and Steel

E-mail: rtd-steel-coal@ec.europa.eu
RTD-PUBLICATIONS@ec.europa.eu

Contact: RFCS Publications

European Commission
B-1049 Brussels
 European Commission

Research Fund for Coal and Steel

Seismic behaviour of steel storage
pallet racking systems
(SEISRACKS2)
Carlo Andrea Castiglioni, Alper Kanyilmaz, Claudio Bernuzzi, Alberto Drei
Politecnico di Milano
Piazza Leonardo da Vinci, 32, Milano-Italy

Hervè Degee, Catherine Braham

University of Liège
Place du XX Août, 7, B-4000 Liège, Belgium

Benno Hoffmeister, Cristoph Heinemeyer

Aachen University - Institute of Steel Construction
Templegraben 55, 52062 Aachen, Germany

Ioannis Vayas, Adamakos Konstantinos

National Technical University of Athens
Heroon Polytechniou Str. 9, 15780 Athens, Greece

Stefano Sesana, Barbara Orsatti

SCL Ingegneria Strutturale
Via Volta 3, 22100 Como, Italy

Jan Hermanek, Jan Willem Frederiks

NEDCON
Nijverheidsweg 26, 7000 AA Doetinchem, Netherlands

Tito Cudini, Leo Rovere

MODULBLOK S.p.A
Vanelis 6, 33010 Pagnacco, Italy

Denis Jehin
STOW INTERNATIONAL N.V
Industriepark 6b, 8587 Spiere-Helkijn, Belgium

Oliver Kraus
FRITZ SCHÄFER GmbH
Fritz-Schäfer-Straße 20, 57290 Neunkirchen, Germany

Dimitris Bakalbasis
COMPUTER CONTROL SYSTEMS S.A
Kifissias 94-96, 15125 Athens, Greece

Grant Agreement RFSR-CT-2011-00031

1 July 2011 to 30 June 2014

Final report
2014 Directorate-General for Research and Innovation EUR 27583 EN
LEGAL NOTICE
Neither the European Commission nor any person acting on behalf of the Commission is
responsible for the use which might be made of the following information.
The views expressed in this publication are the sole responsibility of the authors and do not
necessarily reflect the views of the European Commission.

Europe Direct is a service to help you find answers

to your questions about the European Union

Freephone number (*):

00 800 6 7 8 9 10 11
(*)  Certain mobile telephone operators do not allow access to 00 800 numbers or these calls may be billed.

More information on the European Union is available on the Internet (http://europa.eu).

Cataloguing data can be found at the end of this publication.

Luxembourg: Publications Office of the European Union, 2014

Print ISBN 978-92-79-53896-4 ISSN 1018-5593 doi:10.2777/931597 KI-NA-27-583-EN-C

PDF ISBN 978-92-79-53897-1 ISSN 1831-9424 doi:10.2777/686466 KI-NA-27-583-EN-N

© European Union, 2014

Reproduction is authorised provided the source is acknowledged.

Printed in Luxembourg

Printed on white chlorine-free paper

 TABLE OF CONTENTS

FINAL SUMMARY ................................................................................................................. 7

SCIENTIFIC AND TECHNICAL DESCRIPTION OF RESULTS ........................................................ 13
1. INTRODUCTION ....................................................................................................... 13
1.1 Weaknesses in FEM 10.2.08 [4] ................................................................................. 15
1.2 Comparison between RMI-2008 [9] and FEM 10.2.08 [4] .............................................. 19
1.2.1 Definition of the seismic action ............................................................................... 19
1.2.2 Importance factor ................................................................................................. 20
1.2.3 Definition of the seismic mass ................................................................................ 21
1.2.4 Definition of the behavior factor ............................................................................. 21
1.2.5 Figures for the behavior factor ............................................................................... 22
1.2.6 Methods of analysis............................................................................................... 23
1.2.7 Second order effects ............................................................................................. 25
1.2.8 Combination of the seismic actions ......................................................................... 26
1.2.9 Overturning ......................................................................................................... 26
1.2.10 Control of the rotational capacity of the connections ................................................. 26
1.2.11 Test methods ....................................................................................................... 27
1.2.12 Design of pallet beams .......................................................................................... 28
1.2.13 Influence of the type of bracings............................................................................. 28
1.2.14 Strength of bolted connections ............................................................................... 28
1.2.15 Beam-end connector stiffness to consider in the analysis ........................................... 28
1.3 Structural typologies studied in SEISRACKS2 ............................................................... 29
1.4 Case studies ............................................................................................................ 30
2. COMPONENT TESTS.................................................................................................. 31
2.1 Joint Tests on Upright-To-Beam Connections ............................................................... 31
2.1.1 Loading in down aisle direction ............................................................................... 31
2.1.2 Loading in cross aisle direction ............................................................................... 35
2.2 Column-base tests ................................................................................................... 36
2.2.1 Down aisle tests ................................................................................................... 36
2.2.2 Cross aisle tests ................................................................................................... 37
2.3 Substructure tests .................................................................................................... 38
2.3.1 Braced longitudinal frames ..................................................................................... 38
2.3.2 Cross frames tests ................................................................................................ 40
2.3.2.1 Partner A (frame for high seismic zone) type 4F ....................................................... 40
2.3.2.2 Partner B (frame for moderate-to-high seismic zone) type 1C .................................... 40
2.3.2.3 Partner C (frame for moderate seismic zone) type 1A................................................ 41
2.3.2.4 Partner C (frame for high seismic zone) 2a-B ........................................................... 41
2.3.2.5 Partner D (frame for low seismic zone) 1B ............................................................... 41
2.3.2.6 Partner D (frame for moderate seismic zone) 2b-B ................................................... 41
2.3.2.7 Partner D (frame for high seismic zone) 2b-B ........................................................... 41
2.3.3 Ductility............................................................................................................... 42
2.3.4 Upright stiffness ................................................................................................... 42
2.4 Component method .................................................................................................. 43
2.4.1 Active components................................................................................................ 44
2.4.2 Application of the resistant model ........................................................................... 44
3. WAREHOUSE TESTING .............................................................................................. 45
3
3.1 Monitoring of a Pallet Rack ........................................................................................ 45
3.1.1 Data processing .................................................................................................... 45
3.1.2 Assessment of the events ...................................................................................... 47
3.1.3 Earthquake recordings........................................................................................... 48
3.1.4 Damping.............................................................................................................. 48
3.1.5 Conclusions .......................................................................................................... 49
3.2 Identification of the linear dynamic properties of racks on the base of ambient vibration or
hammer tests .......................................................................................................... 49
3.2.1 Partner A-Lab tests ............................................................................................... 50
3.2.2 Partner B – Test realised by ULg ............................................................................. 51
3.2.3 Partner C – test realized by RWTH .......................................................................... 53
3.2.4 Partner D – test realized by ULg ............................................................................. 53
3.3 Identification of dynamic properties of pallets/merchandize ........................................... 54
4. FULL SCALE PUSHOVER TESTS .................................................................................. 57
4.1 Design of full scale test specimens ............................................................................. 57
4.2 Test procedure......................................................................................................... 57
4.3 Test frames ............................................................................................................. 58
4.4 Test results ............................................................................................................. 59
4.4.1 Unbraced specimens ............................................................................................. 60
4.4.2 Braced specimens ................................................................................................. 61
4.5 Calculation of q-factor .............................................................................................. 62
4.6 Conclusions ............................................................................................................. 64
4.6.1 Unbraced racks .................................................................................................... 64
4.6.2 Braced racks ........................................................................................................ 65
5. NUMERICAL STUDIES ............................................................................................... 67
5.1 Preliminary numerical analysis ................................................................................... 67
5.2 Numerical static pushover analysis ............................................................................. 68
5.3 Pushover analysis of 3D racks .................................................................................... 72
5.4 Post-testing calibration of numerical models ................................................................ 72
5.4.1 Calibration of Beam-end connectors in down aisle direction ........................................ 73
5.4.2 Calibration of Base-Plate behavior in down aisle direction .......................................... 73
5.4.3 Calibration of upright’s behavior ............................................................................. 74
5.5 Calibration of cross-aisle frames ................................................................................ 75
5.6 Calibration of Full-Scale racks in down-aisle direction ................................................... 76
5.7 Effect of pallet loads on global behaviour of racks ........................................................ 77
5.8 Incremental Dynamic Analyses (IDA).......................................................................... 77
5.9 Pallet-merchandize interaction ................................................................................... 79
5.9.1 Maximum applied forces over a pallet beam during the pallet’s sliding ......................... 80
5.10 Application example ................................................................................................. 81
5.11 Correction coefficients for horizontal bending ............................................................... 82
5.12 Results ................................................................................................................... 83
5.13 Buckling length modification factor ............................................................................. 84
6. SEISMIC DESIGN METHODOLOGY .............................................................................. 87
6.1 Assessment of the behavior factors ............................................................................ 88
6.1.1 Unbraced racks (Down-aisle).................................................................................. 88
6.1.2 Braced racks (Down-aisle) ..................................................................................... 89
6.1.3 Upright frames ..................................................................................................... 90

4
6.2 Evaluation of Unit load weight for seismic design (Q p,rated) and rack filling grade reduction
factor (RF) ......................................................................................................................... 91
6.2.1 Specification of Qp,rated and RF ................................................................................. 91
6.2.2 Reliability of loading assumptions based on rack filling factor ..................................... 92
6.3 Assessment of the response spectrum modification factors ............................................ 93
6.4 Assessment of the methods of analysis ....................................................................... 95
6.5 Qualification of testing for components, joints and substructures ................................... 96
6.5.1 Beam-end connector tests (BEC) ............................................................................ 96
6.5.2 Floor connection tests ........................................................................................... 98
6.5.3 Upright frame shear test ........................................................................................ 99
6.5.4 Full scale tests in down-aisle direction ................................................................... 100
7. SOFTWARE TOOL DEVELOPEMENT............................................................................ 103
7.1 Theoretical analysis of the design method ................................................................. 103
7.2 Software components development .......................................................................... 104
7.3 Software features ................................................................................................... 105
7.3.1 Analysis ............................................................................................................. 105
7.3.2 Results .............................................................................................................. 105
7.4 Parametric studies.................................................................................................. 105
8. CONCLUSIONS ...................................................................................................... 109
9. EXPLOITATION AND IMPACT OF THE RESEARCH RESULTS .......................................... 111
LIST OF FIGURES ............................................................................................................. 113
LIST OF TABLES ............................................................................................................... 117
REFERENCES ................................................................................................................... 119
ANNEX FOR CHAPTER 2..................................................................................................... 123
ANNEX FOR CHAPTER 4..................................................................................................... 142
ANNEX FOR CHAPTER 5..................................................................................................... 147

5
6
FINAL SUMMARY
Static steel pallet storage racks are particular structures composed of specially designed cold-formed
steel elements allowing easy installation and reconfiguration, and they are widely used in many
warehouses to store various kinds of goods. Despite their lightness, racking systems carry very high
live load, by far higher than the dead load, opposite to what happens in usual civil engineering
structures. Prediction of the structural behaviour of these structures is not an easy task because of
their particular structural components, i.e. members made of high slenderness thin-walled profiles
hence prone to global, local and, for the uprights, distortional buckling problems. Moreover beam-
to-upright and base-plate joints exhibit a strongly nonlinear behaviour.
Furthermore structural elements, members and joints, are not standard since developed by each
industrial partner with proprietary systems. Due to their peculiarities, specific rules are necessary for
these steel structures that cannot be considered as classical buildings. Reference can then hardly be
made to usual Structural Design Recommendations and Standards. The most recent Design
Standards for steel storage racks in normal use conditions provide thus a combined numerical-
experimental approach in which the design structural analysis is supported by specific tests to
evaluate the performance of the key components, i.e. members and joints. The design of storage
racks is getting even more difficult for storage racks installed in seismic zones, where they must be
able to withstand horizontal dynamic forces. Moreover, in seismic conditions an additional limit state
must be considered beside the usual global and local collapse mechanisms, i.e. a limit state
represented by the fall of pallets due to rocking or sliding, with subsequent damage to people, goods
and to the structure itself.
Within SEISRACKS2 project, various issues have been investigated by means of component testing,
full scale (push-over) testing and in-situ testing of racking systems in operating warehouses, as well
as by numerical simulations using the ad-hoc numerical tool set up during the previous SEISRACKS
project. Additionally a software tool for the design of racks under seismic loads have been be
developed implementing an innovative method that combines nonlinear analysis with multimodal
spectral analysis. These improved design rules will allow building safe structures with an optimized
seismic behavior and with a significantly reduced risk of economic losses. Since the European rack
industry operates on a highly competitive worldwide market, the development of duly validated and
optimized seismic design rules providing a mastered safety level for a limited increase of the cost is
likely to improve the competitiveness of European companies in countries exposed to a moderate to
high seismic risk.
The project is organized in 7 operative work packages which will be presented hereafter ( Figure 1).

Figure 1 Flowchart of adopted Methodology during SEISRACKS2 project

7
WP 1 Definition of Case Studies
A reanalysis of FEM 10.2.08 [4] and a comparison of FEM 10.2.08 [4] and RMI-2008 [9] identified
the relevant weaknesses of the FEM 10.2.08 [4], requiring the improvement in the knowledge of
important aspects of the seismic behaviour of racks, such as the assessment of design spectrum
modification factors, damping of the rack structures, behaviour factors indicated for low ductility
design concept, effects of eccentricities of the schemes due to construction, proper rules for the
ductile design concept.
Eight models, covering a wide range of constructive typologies and different static schemes, were
identified by the IPs, focusing on technical solutions for which current FEM10.2.08 [4]
recommendations are felt less efficient. Each partner designed, according to daily practice and
following FEM recommendations, two configurations: one unbraced for low/medium seismicity, and
one braced for high seismicity.
The solutions designed for the upright frames resemble either the normal practice or new proposal
adapted just for seismic conditions. The aim is to compare the different way of connection of the
diagonals, and to confirm which is the most effective one in seismic conditions; to confirm the q
factors actually permitted by the Norms; to identify the weaknesses or skills of commonly used
constructive typologies.
The solutions designed for vertical bracing resemble common practice (although the solution with
cables is less used); by the comparison of different design possibility and different stiffness of the
frame - braced and unbraced – the most effective solution shall be found with respect to grade of
seismicity. The design is based on low ductility rules; the checks of components is performed
according to EN 15512:2009 [3] and EN1993 [2].
WP2 Component Testing
In this work package, assessment of component (beam-to-upright, upright-to-base, upright-to
bracing connections) behaviour, local ductility of cross frames, verification of applicability of
monotonic test results to the assessment of cyclic performance, and application of the components
method for characterizing rack connections have been accomplished.
The setup of the down aisle tests is different from the one proposed in EN 15512:2009 [3] for beam-
end connector tests, and was developed within the scope of this research project to provide
information about the plastic deformation capacity under realistic support and loading conditions
(Figure 2).

Figure 2 New test setup developed within SEISRACKS2

Influence of loading on the connector behaviour is investigated by testing the racks with different
pay load (0%, 50% and 100% of service load). For each load case and for each industrial partner (4
industrial partners) one monotonic and one cyclic test were performed.
Tests results provided information related to failure modes, influence of payload, differences between
cyclic and monotonic behaviour, moment rotation characteristic, compatibility of these test results
with those derived from EN 15512:2009 [3] standard tests and effectiveness of safety bolts.
Cross-aisle tests have been performed at RWTH for the assessment of the moment rotation
characteristic of the beam to upright connectors and of the influence of pallet loads on the cross aisle
deformation behaviour of the beams.
In addition friction tests have been performed to allow determining the influence of pallets on the
cross aisle deflection resistance, in accordance with FEM 10.2.08 [4] (Figure 3).

8
 Figure 3 Setup of frictions tests (left: FEM tests; right: performed tests)

In total, 27 friction tests were performed on the pallets representing the 50% of the maximum service
load (approx. 400 kg). For the tested products it can be stated that the moment resistance in the
connectors for cross aisle bending is negligible and pallet friction mainly controls the resistance in
the cross aisle direction.
The column base tests are carried out at RWTH for the assessment of the moment rotation
characteristic of the connection between the upright and the column base (Figure 4). The influence
of loading on the base-plate behavior is investigated by testing the specimens with different axial
load. Down aisle loading causes bending in the base-plate connections. On the contrary, bending in
the base-plate connection seems to be negligible under loading in cross aisle direction.
Constant loading
(force controlled)
Rotation of base plate Rotation of connector

Pendulum H1

Variable loading
(displacement controlled)

Upright
H2

Cj,plate Cj,connector
Column base
1 1 1 Rigid steel plate
──C──── = ──C─j,plate
─── + ─C───── Figure 4 Column base tests
j,total j,connector

Substructure testing is carried out at ULIEGE on cross-frames and braced longitudinal frames under
horizontal loading, aimed at the definition of standardized procedures for the assessment of the local
ductility of cross-frames as well as of the longitudinal frame bracing properties.
4 types of cross frames have been identified based on the geometrical pattern of the diagonals as
well as 4 types of diagonals’ position and connection (Figure 5). 8 case-studies have been prepared
(2 by each IP plus one extra for one Partner) with the objective of getting a wide range of situations
(design for low/moderate/high seismicity, D/Z/X type of cross bracing).
All cross-frames have been modelled using a finite element analysis tool developed by the University
of Liège in partnership with a design office.
The following failure modes have been observed during the tests:
• Local and global buckling of diagonal in compression
• Shearing of bolt of diagonal in tension
• Local shear at diagonal intersection
• bearing of diagonal in tension
• buckling of diagonal in compression
• buckling of diagonal in compression with local crushing of diagonal’s extremities
• Torsional buckling of uprights
• bending of bolts

9
 Figure 5 Global buckling of 2nd and 3rd diagonals accompanied with shearing of bolt

WP 3 Warehouse Testing
This work package includes monitoring of a real warehouse on the long period during operation
activities, identification of the linear dynamic properties of racks, and identification of dynamic
properties of pallets/merchandize.
During the previous Seisracks1 project [6], an installation near Athens had been continuously
monitored for a few months to record rate of occupancy, operations and accelerations. This activity
was extended during the whole Seisracks2 research. Data are recorded continuously on site, and
transmitted to the remote server of NTUA. A large number of data related to everyday activities is
being recorded at a 200 Hz sampling rate, and re-analysed by POLIMI.
The objective of warehouse testing consists in defining the range of periods and damping of a real
structure and to calibrate numerical models. These tests are performed by ULIEGE and RWTH on
site, on existing structures in active warehouses chosen by the IPs (one warehouse for each IP). The
geometrical and material properties of the structures to be monitored are chosen very similar to the
case studies of this research.
The dynamic properties (frequency, damping) of a large range of stored merchandizes were identified
on the basis of a push-by-hand excitation on top of the stored good with a quick release or an impact
given by the human waist. For each of the chosen pallet a tri-axial sensor has been fixed on top of
the goods.
WP 4 Full Scale Testing
This work package aims at the assessment of the global behaviour of full scale racks in down aisle.
Specimens for longitudinal tests are 4 levels (8.0m) and 2 bays (5,6m). Eight specimens provided
and designed by four industrial partners (conventionally named IPA, IPB, IPC and IPD for privacy
reasons, according to a non-disclosure agreement) were tested (4 unbraced, 4 braced racks).
All the tests were performed outdoor, in a testing facility owned by Marcegaglia Buildtech and
managed by Politecnico di Milano where full scale racks could be easily mounted and loaded (Figure
6).

Figure 6 Deformed shapes at the last loading increment of IP A (left) and IP C (right) structures

WP5 Numerical Analyses

First, preliminary linear and nonlinear static analyses have been performed according to the prFEM
10.2.08 recommendations [25], then these models have been calibrated with the results of
experimental studies on component and full scale tests. With the calibrated models, q-factors have

10
been calculated and parameter studies have been carried out to investigate the influence of
occupancy rate and merchandise properties.
Inelastic structural properties of beam to upright connections and base-plate to the upright
connections are treated by using nonlinear link elements in the model. The input data of the link
elements are obtained from experimental data which are provided by each IP.
WP6 Assessment of Seismic Design Methodology
Current seismic design methodology including assessment of q-factors, Qp (specified unit load
weight), Rf (rack filling rate), ED1, ED2, and ED3 spectrum modification factors, impact of second
order effects have been assessed. Crucial conclusions have been provided for seismic design of racks,
which are based on the extensive numerical and experimental studies of SEISRACKS2. Designer’s
manual for seismic design of racks including innovative and non-conventional methods has been
prepared.
WP 7 Software tool development
The current analysis method for racks under seismic loading (the multimodal spectral analysis) does
not allow for the consideration of geometrical nonlinear effects due to the linear superposition of the
modal responses.
A new software tool and analysis procedure has been developed to allow a multimodal spectral
analysis with due consideration of 2nd order effects. This method is a multimodal spectral stepwise
non-linear analysis. It offers the following advantages against other approaches that were examined:
 It uses seismic data that is well defined in every national design code (parameterized spectra,
importance categories, regional acceleration limits etc.).
 It uses fast, reliable and reasonably accurate algorithms and procedures (modified Newton-
Raphson, CQC).
 It can be combined with existing software.
 It is easy to use and understand even for users without specialized knowledge (simple
presentation of variables and data, user is not required to know the implementation details).

CONCLUSIONS
SEISRACKS2 has been a ground-breaking project within the RFCS program. It combined
experimental and numerical tools to improve, validate or invalidate the current seismic design rules
for racking systems, thanks to a sound collaboration between rack industry, design offices, and
universities. The project provided the most extensive experimental and numerical work ever
conducted in Europe, to investigate the component and global behaviour of racking systems, and
their behaviour under seismic actions. Moreover, SEISRACKS2 project resulted in the development
of design guidelines and recommendations for seismic design of racking systems. Novel results of
SEISRACKS2 project can be summarized as:
1) A complete comparison has been performed between European (FEM 10.2.08 [3]) and
American (RMI:2008 [9]) seismic design provisions together with an assessment of
weaknesses in FEM 10.2.08 [3]
2) New testing procedures have been developed for:
o Qualification of beam-to-upright and column base-plate joints in cross-aisle
and down aisle directions
o Estimation of the global ductility of full scale rack systems in down-aisle and
cross-aisle directions
3) A component method has been developed to characterize the resistance of beam-to-upright
connections.
4) By means of continuous monitoring (1 year), ambient vibration and hammer tests, structural
damping values have been calculated for the racks in an existing warehouse, and typical rack
configurations provided by 4 industrial partners of the project
5) For the first time, results of push over tests have been published, which took into account
real-scale geometry and loading on different pallet racking systems, provided by 4 world-
wide renowned industrial partners of the project. Thanks to these tests, ductility
performances of several types of racks (braced and unbraced) have been investigated.
Crucial improvements have been suggested to consider in the seismic design of pallet racks:
o The behavior factors q=1.5 or q=2 for unbraced racks are reliable figures provided
that the soft storey collapse mechanism is prevented; this can be achieved improving
the bending strength and the rotation capacity at the base, either by installing floor
beams or base-plates adequately designed.

11
 o The properties and the design of the floor connections derived from non-seismic
design appears to not be adequate for seismic conditions, also considering of the
variability of the axial force in the upright during a seismic event.
o Full scale testing of unbraced racks appear necessary to calibrate the numerical
models in particular with reference to the bending stiffness of the uprights.
o Stricter requirements have been introduced in prEN16681 [22] for the design of floor
connections of unbraced frames.
o A new rule has been added in prEN16681 [22] on the design of the bolted connections
of the diagonals, requiring a minimum distance between the bolt and the edge of the
diagonal equal to 2.5 times the bolt’s diameter, that shall be considered in
conjunction with the general rule for bolted connections requiring the bolt’s strength
at least 20% larger than the bearing strength. This rule is a constraint to design
stronger connections, in addition to improve the development of the plasticity at the
bolt’s hole
o The assessment of the behavior factor should be required when a rack non regular
in plan (braced rack) is designed using a behavior factor q>1.5; this assessment
should be done through a 3D numerical pushover analysis including torsion.
o For braced racks, the experimental calibration or validation of the numerical pushover
through a full scale test should be required in ductile design or when the behavior
factor in cross aisle direction is smaller than the one considered in down aisle
direction.
6) Calibrated numerical models have been developed based on the experimental studies, and
parameter studies have been performed, and significant observations have been made on the
filling rate, pallets location, merchandize-structure interaction.
7) A designer’s manual for seismic design of racks has been prepared
8) A new numerical analysis procedure has been developed (multimodal spectral stepwise non-
linear analysis), together with a new software, which takes into account second order effects.

12
SCIENTIFIC AND TECHNICAL DESCRIPTION OF RESULTS
1. INTRODUCTION
Due to increasing mass production of goods and increasing consumption levels, demand for storage
space increases. As the land becomes more valuable everyday both in terms of economy and
environment, highly optimized, reliable and safe storage systems are needed.
Storage racks are widely adopted in warehouses where they are loaded with tons of (more or less)
valuable goods. The loss of these goods during an earthquake may represent, for the owner, a very
large economic loss, much larger than the cost of the whole rack on which the goods are stored or
of the cost for its seismic upgrade (Figure 7). Racks are also more and more frequently adopted in
supermarkets and shopping centres, in areas open to the public. The falling of the pallets, in this
case, may endanger the life of the clients as well as that of the workmen and employees, involving
not only Civil and Penal Code considerations about the liability of the owners, but also economic
considerations related to the insurance coverage. Hence, solution of the problems connected with
safe and reliable design of steel storage racks in seismic areas has a very large economic impact.

Figure 7 – Rack collapses after Emilia-Romagna Earthquake 2012, Italy

Racks are industrial products, with a very high level of optimization of the components, in terms of
amount of material used, production and installation procedures. Historically in Europe, the racks are
generally considered as working equipment, and only in few cases their design must comply with
building codes. Only recently, the market has started to request racks capable to withstand seismic
actions, both for safety reasons and to protect the goods. For this reason, the development and
industrialization of the product “industrial rack” diffused on the market has been mainly based on its
load-bearing capacity under the vertical load of the stored product, taking into account only
horizontal actions of low or moderate intensity produced by the interaction with handling equipment
and by the effects of the imperfections; the consequence is that their basic design and optimization
is relatively far from usual seismic resistance concept. The compliance with the seismic requirements
often leads to the provisions for non-standard construction for racking manufacturers.
Storage racks are composed of specially designed cold-formed steel elements that permit easy
installation and reconfiguration, consistent with the merchandising needs of a warehouse retail store.
The shapes have sections in class 3 and 4. Proprietary moment connection frames are typically used
as structural system in the down-aisle (longitudinal) direction, while braced frames are typical for
the cross-aisle (transverse) direction (Figure 8). Braced frames are also used in down-aisle direction
if required for overall stability reasons

Figure 8 Typical pallet rack configuration [1]

Typical stored pallets have plan areas of about one square meter and can weight up to approximately
10 to 15 kN. Storage rack bays are typically 1.0 to 1.1 meter deep and 1.8 to 2.7 meters wide and

13
can accommodate two or three pallets. The overall height of pallet rack structural frames such as
usually found in retail warehouse stores varies between 5 and 6 meters, while in industrial warehouse
facilities, racking system can reach much more considerable heights up to 12-15 meters or more.
Prediction of the structural behaviour of these structures is not an easy task because of their
particular structural components, i.e. members made of high slenderness thin-walled profiles hence
prone to global, local and, for the uprights, distortional buckling problems. Moreover beam-to-upright
and base-plate joints exhibit a strongly nonlinear behaviour. Furthermore structural elements,
members and joints, are not standard since developed by each industrial partner with proprietary
systems. Due to their peculiarities, specific rules are necessary for these steel structures that cannot
be considered as classical buildings. Reference can then hardly be made to usual Structural Design
Recommendations and Standards. The most recent Design Standards for steel storage racks in
normal use conditions provide thus a combined numerical-experimental approach in which the design
structural analysis is supported by specific tests to evaluate the performance of the key components,
i.e. members and joints. The design of storage racks is getting even more difficult for storage racks
installed in seismic zones, where they must be able to withstand horizontal dynamic forces. Moreover,
in seismic conditions an additional limit state must be considered beside the usual global and local
collapse mechanisms, i.e. a limit state represented by the fall of pallets due to rocking or sliding,
with subsequent damage to people, goods and to the structure itself.
There is a well-established design methodology of thin gauge profiles, in the field of static loads,
codified in EN1993 [2] and specialized for the application of industrial racking structures in the EN
15512:2009 [3], which remains the reference methodology for the design of the pallet racks and the
definition of their experimental design parameters.
In some cases, and especially for the construction of large structures with high static engagement
(specifically silos or clad racks warehouses), the components used are made of hot-rolled profiles,
for the design of which reference is made directly to the applicable parts of the EN1993 [2].
In the dynamic field, as specified above, the FEM 10.2.08 [4] is based on the design criteria of EN
1998 Part 1-1:2005 [5], adapting certain rules and introducing a number of modifications to allow
its application to this particular type of structure; for racking structures or their parts in traditional
steelworks construction, reference is directly made to the relevant parts of the EN 1998 Part 1-
1:2005 [5]. In particular, the following principles have been derived:
 The description of the seismic action by means of the response spectrum and the definition
of the design spectrum
 The reference design methods
 The classification of the seismic design approach in “low dissipative” and “dissipative”; in
particular, the one of largest practical interest is the “low dissipative”, which is applicable to
the standard industrial racks made with cold formed thin gauged profiles.
 The definition of the “behavior factor” q
 The criteria for the hierarchy of resistance
FEM 10.2.08 [4] has been developed by ERF – European racking Federation – Work Group Seismic
design, under the umbrella of the FEM – Federation Europeen de la Manutention - Section X since
the year 2000, and it was officially published in 2010. The actual edition (Rev. 1.04 dated May 2011)
includes some editorial amendments.
The previous Research Seisracks [6] had the FEM 10.2.08 [4] – draft 2008 as reference; the edition
published by the FEM has included some relevant changes.
Recently, based on the results SEISRACKS2 project, a new European Standard (prEN16681 [22])
has been submitted for publication, which overcame some of the problems of FEM 10.02.08 that are
analysed in the following paragraphs.
Corrections and amendments are introduced to take into account specific properties and behavior of
the industrial racking structures. These are based on experiences and researches, derived from
studies developed in Europe, independently or specifically during the development of the FEM Norm,
and from research and experience performed especially in the USA in the past years and other Norms
for racks design, specifically the RMI-ANSI.
Among the aspects that have been developed and established in the FEM 10.2.05 [4] requiring
theoretical and experimental investigation, are the following:
 The modifications of the design spectrum produced by the dynamic interactions between
bearing structure and supported load (the “unit loads”)
 The behavior and modelling of the connections under dynamic actions and low cyclic fatigue
(beam-end connectors, floor connections, upright frame diagonals)
 Methods to quantify the effects of second order and geometrical nonlinearities
 The assessment of the effect of friction between unit loads and support beams
 The methods to evaluate the experimental design parameters in the dynamic field
 The use of structural schemes not considered by EN1998 [5], using special cautions and rules
14
The application of the FEM 10.2.08 [4] is limited to the “adjustable pallet racking system” (APR),
which is the most diffused on the market, while there are several other typologies of industrial racks
that can only in few cases and partially be assimilated to APR.

1.1 Weaknesses in FEM 10.2.08 [4]

In this chapter, main characteristics of the seismic behavior of the industrial pallet racks, and their
representation in the FEM 10.2.08 [4] are described and analysed.
Design spectrum modification factors: The effects of the response of industrial pallet racks to seismic
actions is described by the response spectrum defined in the Eurocode 8, which is modified in the
FEM 10.2.08 [4] to take account of the features of the behaviour of racks. The design spectrum
modification factors introduced in the FEM 10.2.08 [4] are three; their application modifies the
ordinate of the spectrum and the period of vibration.
i) ED1 modification factor: This coefficient takes into account the effect of the sliding of the
load on the rack’s beams. Its analytical formulation has been derived through numerical
simulations of unbraced racks in down aisle longitudinal direction. Its extension and
validation to the cross aisle direction and to the down aisle for braced racks is required. An
aspect requiring investigation is the influence of the flexibility of the load on the supports
(Figure 9). Besides its influence on the design spectrum, the coefficient of friction determines
locally the stability of the load on the supports and hence the risk of its falling, therefore it
requires a precise and reliable assessment.

Figure 9 Unit load on supports

ii) ED2 modification factor: This coefficient affects the mass of the product and consequently
it modifies the period of vibration of the structure and the entity of the inertial actions, related
to the mass of the product. The numerical values of ED2 reported in the FEM 10.2.08 [4] are
currently based on engineering judgment applied with safety criteria, on the basis of the
approach and figures used by other norms (RMI). Their figures need to be extensively
assessed using reliable scientific approaches for different families of stored products.
iii) ED3 modification factor: This coefficient affecting the ordinate of the spectrum takes into
account over-strengths and global effects of the interaction between racks and stored loads
that are not explicitly considered in the mathematical formulation of the spectrum and in the
other modification factors. The value 0.67 of ED3 has been estimated with engineering
judgment, comparing the effects of global seismic action obtained from other referenced
Norms, and its value needs to be properly investigated and assessed.

Low ductile and ductile design concepts: The design of industrial racks is extensively carried out
under the low ductility concept, which is recommended by the EN1998 [5] only in low seismic zones.
The industrial racking structures can hardly be designed using ductility, and for this reason the
extension of the low ductile design to all seismic conditions has to be verified and validated.
Damping of the rack’s steel structure: The 3% damping is considered in FEM 10.2.08 [4] for the
elastic response of the racking steel structure; this assumption, derived from Seisracks [6] research,
needs to be confirmed and validated. The damping of the rack loaded with pallets has to be
determined as well, to validate the design approach with the “modified design spectrum”.
Values of θ permitted by FEM 10.2.08 [4]: EN1998 [5] prescribes the limitation of the interstorey
drift sensitivity coefficient θ to be less than or equal 0.3, while FEM 10.2.08 [4] allows any value of
θ prescribing specific method of analysis and rules. This deviation from the EN1998 [5] rules should
be assessed for ductile design.
Formulas for calculation of θ: The difference in the results with the application of the two formulas
reported in chapter 2.4 of the FEM 10.2.08 [4] should be assessed and considered.
MRSA method of analysis: The modal response spectrum analysis (MRSA) with second order effect
should be referenced and validated for the application to racking structures. The methods to include
directly the second order effects in the modal response spectrum of analysis are currently based on
a normal MRSA performed considering the stiffness matrix of the racking structure [K] equal to the

15
sum of the initial elastic stiffness matrix [KI] and the geometric matrix [KG] calculated for the
distribution of the axial forces:
[K] = [KI] + [KG] [Eq. 1]
Bracing configurations in cross aisle and down aisle direction: Typical bracing schemes used for the
upright frames (Figure 10) including diagonals in compression (such as K, D and Z configurations)
are not allowed by EN1998 [5], but they are instead permitted by FEM 10.2.08 [4] in low dissipative
design concept. Their design must be performed limiting the behaviour factor q to the value of 1.5,
provided that diagonals are designed with a safety coefficient 1.5 applied to member and their
connections, without consequences on other elements (uprights and floor connections). This rule is
a precaution versus the buckling collapse of the diagonals or the brittle failure of the connections,
when the static scheme is such that the stability is fully provided by compression elements. The
reliability of this approach requires being validated. Effect of spine bracings in ductility of down-aisle
frames should be investigated (Figure 10).

Figure 10 Typical bracing schemes used for the upright frames

Figure 11 Down-aisle bracings

Assessment of the behaviour factors: Behaviour factors q reported in chapter 3.4 of FEM 10.2.08 [4]
are based on the ones reported in EN1998 [5], but they need to be assessed and validated, taking
into account the different typologies of constructions. In particular there are schemes of bracings,
like K or D or Z schemes, that are not allowed by EN1998 [5], but are used in the low ductile design
concept introducing additional safety rules.
Floor fixing: Industrial racks are in general supported on concrete floor slabs, to which are connected
by means of post-installed anchor bolts. At the time of the publication of FEM 10.2.08 [4] criteria for

16
the design and certification of the anchor bolts in seismic zone were not available. The requirement
in FEM 10.2.08 [4] of using anchor bolts in cracked concrete conditions was derived from ACI 318-
05 Annex D [7], but criteria to determine the capacity of post-installed anchor bolts in seismic
conditions were not available in Europe so it was not possible to provide more detailed requirements.
Actually the ETAG-001 Annex C [8] design rules assume “concrete without tensile components of
stress” around the anchor bolt, regardless cracked or uncracked conditions, and this requirement
(not very clear when anchor bolts are working in tension) reduces the applicability of the method.
Bolt in the beam-end connection: The behaviour factor for the unbraced rack in down aisle direction
is 1.5 in general, and it can be increased to 2 when the Beam-end connector is secured to the upright
with a bolt (with washers under both bolt’s head and nut) installed above the centre of gravity of the
section of the beam. This assumption is based on engineering judgment, and it is supported by some
experimental evidence in tests performed under Seisracks [6] research; nevertheless the behaviour
factor of the frame in general and the value of the behaviour factor q=2 requires a rational
background.
Pallet beam design for horizontal bending: The friction between unit load and pallet beams provides
a diaphragm effect positively affecting their bucking length and bending moment in the horizontal
plane. The quantification of this reduction is derived from engineering judgment, but there is no
background on the values of the reduction coefficients. The reduction coefficients require rational
validation.
Shear stiffness and strength of the upright frame: The upright frame generally exhibits a shear
stiffness lower than the one that can be evaluated theoretically using the properties of the members
and the schemes of the axes. This effect is mainly due to the local behaviour of the connections
between upright and diagonals, and it is strongly affected by the type of connection and the way the
shear action is introduced in the upright. The linear dynamic behaviour (period of vibration) of the
upright frame is strongly influenced by this effect, resulting in a sensible increase of the period of
vibration and a reduction of the seismic action challenging the rack’s components. Upright frames in
which bracings introduce torsion in the upright, like in most of the typical construction arrangements,
usually exhibit an apparently high shear flexibility, but this is a consequence of the torsion of the
upright, that can only be equilibrated by the bending of the diagonals.

Figure 12 Examples of bracing member connections with large restraint

The FEM 10.2.08 [4] allows taking into account the beneficial effect of the shear flexibility of the
upright frame reducing the seismic action, using the same design parameters determined according
to the EN 15512 [3], that are based on the assumption of low horizontal actions arising from
imperfections or operation loads. At the moment there are no design methods or protocols to
evaluate the shear strength of the upright frame at the ultimate limit state; “secondary” effects on
strength of uprights (torsion) and diagonals (bending) are not considered in the design. It is
necessary to investigate this aspect, which is critical for the structural stability;
In particular:
 Special attention should be paid on the interaction between shear and bending on the upright
frame, which currently is not taken into account.
 For practical applications, a correlation with shear stiffness in static conditions (EN15512:2009
[3] tests) and for dynamic analysis should be very useful
 The need of specific tests required for ULS design of the upright frame should be investigated
and, in case, a testing protocol established. Design requirements for diagonals-uprights
arrangements introducing torsion in the upright are required as well.
Behaviour of base-plates in cross aisle direction: The floor connections of the uprights are usually
modelled with hinges in cross aisle direction, and the eccentricity between the base of the upright
and the scheme point of the first element of the bracings has to be modelled as well; the consequence
in the numerical results is a bending moment in the upright around the weak axis, due to the
relatively small shear action at the upright’s base, varying linearly along the axis of the upright from
17
zero to a maximum value at the intersection with the frame bracings element. Because the base of
the upright usually is not a physical hinge, despite being considered as such in the numerical model,
a percentage of bending moment can be considered absorbed by the base-plate.
This can be considered as a source of overstrength of the structural system (a structural resource
not considered in the design), or, alternatively, it can be accounted for the design, by modelling the
connection also in cross aisle direction with its own response law (elastic or elasto-plastic rotational
hinge), or considered using a method of limit analysis, applying a bending moment at the base of
the upright compatible with the strength of the base-plate

The base-plates are in general not symmetrical with respect to the longitudinal plane perpendicular
to the upright frame (Figure 13), so a suitable modelling may be difficult and in any case there are
no test protocols in the standards; therefore, the first option above should be investigated to improve
the evaluation of the behaviour or spectrum modification factors.

Figure 13 Loads acting on base-plates

Eccentricities of the schemes: Chapters 8.6 and 8.7 of the EN 15512 [3] allow the rack’s designer to
ignore the effects of the eccentricities of the schemes of the racking structure typical of this type of
construction. This can be considered acceptable under the assumption of EN 15512 [3] of small
horizontal actions, due to imperfections and operational loads, but it is also considered in the analysis
of the racks in seismic conditions, when horizontal actions are much larger. The safety of this
assumption and the effects of these eccentricities should be properly investigated and assessed
(Figure 14)

a. Eccentricities in spine bracing b. Eccentricities in frame c. Eccentricities in

bracing cross aisle
direction
Figure 14 Eccentricities of the schemes

Parameters derived from tests: Several properties affecting the dynamic response of the rack are
derived from tests performed according to the “EN 15512 [3] Appendix A protocols for non-seismic
design”:
• The beam-end connector stiffness and strength
• The floor connector stiffness and strength (base-plate)
• The shear stiffness of the upright frame

The use of these values determined experimentally for operating conditions should be validated also
for seismic design. Alternatively, practical testing protocols to derive these parameters for seismic
18
applications, or correlation rules with values obtained through standard protocols for design in non-
seismic conditions, need to be established.
Distribution of the masses: In chapter 3.3.2, FEM 10.2.08 [4] specifies the distribution of the masses
to consider in the seismic analysis. In principle all the possible configurations of the unit loads should
be considered to find the most severe distribution, but this is impossible in the design practice and
for this reason the Norm has specified the following as relevant design conditions, to be considered
with unit loads at 100% of their weight. In the down aisle direction, the situation of rack fully loaded
is assumed as the dominant one. In cross aisle direction two loading situations are considered: rack
fully loaded and only top level loaded; the first one for the strength and stability of components and
connections, the second for the purpose of the check for the floor anchors. The intermediate
distribution of the unit loads are considered unlikely to occur. For the cross aisle direction and for
the down aisle direction of braced racks, the design for uplift action of the floor connections may be
underestimated when the fundamental period of vibration is in the lower range of the constant
velocity branch of the design spectrum. In these cases the reduction of the mass may produce a
relatively small reduction of the fundamental period and a relevant increase of the ordinate of the
spectrum, which is not balanced in terms of seismic shear by the reduction of the mass. The
consequence can be an increase of the seismic action at the base level, resulting an uplift less
counterbalanced by the reduced weight of the unit loads. This situation is likely to occur in practice,
because the probability of consecutive bays fully loaded in a rack is relevant and the weight of the
unit loads equal to the design one is quite rare. For this reason rules to define the minimum number
of configurations to study for uplift action covering the critical situations should be better investigated
and defined.
Design of racks on suspended floors: The design of racks on suspended floors or platforms is
mentioned in chapter 2.4.4 of FEM 10.2.08 [4], but no rules or guidelines are provided. This is a
quite widespread situation in retail areas.
Ductile design concept: Ductile design concept appears very difficult to apply for racking structures,
in particular because of the requirement of “uniform dissipation” along the height of the bracing
systems (to be applied for example for upright frames). CEN TC344/WG5 has decided to provide
information only for low ductile concept, and make direct reference to EN1998 [5] for the ductile
design concept. An investigation on this subject may be useful for the development of effective
structural solutions.
1.2 Comparison between RMI-2008 [9] and FEM 10.2.08 [4]
In this chapter the seismic design procedures of FEM 10.2.08 [4] are compared with the ones of the
reference Normative in the USA for racks design:
 The Rack Manufacturer Institute RMI-2008 [9], MH16.1:2008 ANSI-RMI “Specification for
the Design, design and Utilization of Industrial Steel Storage Racks” and Commentary
 FEMA – Federal Emergency Management Agency - and NEHRP - National Earthquake Hazards
Reduction Program, FEMA 460 [10]– September 2005 “Seismic considerations for steel
storage racks located in areas accessible to the public”
The comparison is focused for FEM 10.2.08 [4] only on the low ductile design concept, because it is
the most relevant for the storage rack industry.
1.2.1 Definition of the seismic action
The seismic action is conventionally defined with the same bases by both the Norms as in most of
the seismic design Codes by means of an elastic response spectrum 5% damped with 10% probability
to be exceeded in 50 years, corresponding to a return period of 475 years; the shape of the spectra
are slightly different because of the numerical formulas used.
FEM 10.2.08 [4]: Chapter 2: The horizontal and vertical spectra are the ones defined in EN1998 [5];
the single parameter defining its intensity is the design Peak Ground Acceleration (PGA) for the
reference return period of 475 years agR. The parameters defining the spectrum take into account
the attenuation of the soil. The design acceleration is corrected by the importance factor I, modifying
the reference return period of the earthquake. Unlike the EN1998 [5], FEM 10.2.08 [4] doesn’t
consider the lower bound limitation of the ordinate of the spectrum, relevant at long periods. The
vertical component of the seismic action needs to be considered only when cantilever beams support
loads and when columns are supported by beams.
RMI-2008 [9]: Chapter 2.6: The design spectral acceleration parameters are defined according to
the International Building Code, as follows:
 SS and S1 are the mapped spectral accelerations for short periods and at 1 sec period
respectively used to define the site seismic intensity. They are determined for the USA as
per USGS [11] Open-File Report 01-437 “Earthquake Spectral Response Acceleration Maps”.
They are also published in the RMI-2008 [9] and they can be found from the USGS website;
estimations of SMS and SM1 in all the world can be found on the same website.
19
  Fa and Fv are two site coefficients depending on site class, classified on the basis of the soil
characteristic. The above are the basic parameters to define the seismic intensity at the site.
 SMS and SM1 are the maximum considered earthquake site spectral response accelerations
for short periods and at 1 sec period respectively; they are obtained as follows:
SMS = FaSS [Eq. 2]
SM1 = FvS1 [Eq. 3]
 SDS and SD1 are the 5% damped design spectral response acceleration at short periods and
at 1 sec period respectively, and are determined multiplying by 2/3 SMS and SDS respectively
The “seismic response coefficient” (design spectrum) is defined as:
CS = SD1/(T*R) [Eq. 4]
CS ≤ SDS/R [Eq. 5]
The lower bound cut-off is defined as:
CS  0.044 SDS [Eq. 6]
If S1 > 0.6 then CS  0.5*S1/R if S1>0.6 [Eq. 7]
Where
T = period of vibration
R = response modification factor
The vertical component of the seismic action is not considered.
Apart from differences in the formulas and in the shape of the response spectra, the definition of the
seismic action is very similar for FEM 10.2.08 [4] and RMI-2008 [9].

1.2.2 Importance factor

In FEM 10.2.08 [4] Chapter 2.2.3, the Importance factor I is correlated to the return period and it
modifies the reference peak ground acceleration agR (defined for I = 1.0), so that the acceleration
parameter of the spectrum is defined as ag = I agR.
The Importance factor is also correlated to the design life of the rack, to be specified for each
application, which can be 30 or 50 years.
To define I, four “Importance classes” are defined as reported in Table 1.

Table 1 Importance factors according to FEM 10.2.08 [4]

In terms of probability of exceedance, the various importance factors have the following
consequences:
Importance class I: 20% probability of exceedance of the seismic action in 30 and 50 years.
Importance class II: 10% probability of exceedance of the seismic action in 30 and 50 years.
Importance class III: 5.8% probability of exceedance of the seismic action in 50 years.
Importance class IV: 3.65% probability of exceedance of the seismic action in 50 years.
RMI-2008 [9]: Chapter 2.6.2 states that, unless the rack is used to store hazardous material, storage
racks are to be deemed Occupancy Category II structures.
The system Importance factor Ip is defined in the RMI-2008 [9] and applied in the formula to calculate
the base shear. The following are the situations and the figures:
Ip = 1.5 if the system is an essential facility;
Ip = 1.5 if the system contains material that would be significantly hazardous if released;
Ip = 1.5 for storage rack in areas open to the public, (e.g., in warehouse retail stores)
20
Ip = 1.0 for all other structures;
If a displacement based evaluation of the rack structure is performed in either of the two principle
directions of the rack, Ip may be taken as 1.0 in that direction (this approach is referenced in FEMA
460 [10])
The FEM 10.2.08 [4] allows considerably greater seismic risk in some situations.
1.2.3 Definition of the seismic mass
In FEM 10.2.08 [4]: Chapter 2.3.3, the specified weight of the unit load QP,rated for the seismic design
is defined as the “specified value of the weight of the unit loads in the compartment, upright frame
or global down aisle design (see EN 15512 [3]) as given in the Contract document”
This value is not necessarily the maximum weight of the unit loads for the local design of the pallet
beams, it can be smaller and it must be specified considering, on the safe side, the turnover of the
unit loads on the rack.
The weight of the mass to be considered in the design is corrected by means of the two modification
factors:
 Rack filling grade reduction factor RF, taking into account the average filling of the rack, with
the following limitations:
RP  0.8 in D.A. direction [Eq. 8]
RP = 1.0 in C.A. direction [Eq. 9]
In fact, in the cross-aisle, the verification of the upright frame is a local verification of lateral resisting
systems without redundancy
 ED2 to consider the influence of dynamic behavior of goods on the rack structure
Values of ED2 are defined in Table 2.8 for different type of stored product and packaging.
Thus, the weight of the seismic mass is:
Ws = ED2*RP*QP,rated [Eq. 10]
In RMI-2008 [9]: Chapter 2.6.2:, the maximum weight of the product that will be placed at any
position of the storage rack PLMaximum is considered as basis for the design of the stored rack.
The weight of the mass is determined multiplying PLMaximum by means of two coefficients:
 Product Load reduction factor PLRF = PLMaximum / PLAverage rated [Eq. 11] where PLAverage is defined
as “the maximum total weight of product expected on all the beam levels in any row divided
by the number of beam levels in that row“.
In warehouse retail areas open to the public, PLAverage must be considered equal to PLMaximum.
In cross aisle direction PLRF = 1
 0.67 reduction factor for all typologies of stored product
Thus, the weight of the seismic mass is:
Ws = 0.67*PLRF*PL [Eq. 12]
RP is the equivalent of PLRF.
ED2 is the equivalent of the 0.67 reduction factor.
1.2.4 Definition of the behavior factor
In FEM 10.2.08 [4]: behavior factor q is defined according to the EN1998 [5]; with reference to EN
1998-1:2004 chapter 3.2.2.5 clause (3)P, the definition of the behavior factor q is:
The behavior factor q is an approximation of the ratio of the seismic forces that the structure would
experience if its response was completely elastic with 5% viscous damping, to the seismic forces that
may be used in the design, with a conventional elastic analysis model, still ensuring a satisfactory
response of the structure. The values of the behavior factor q, which also account for the influence
of the viscous damping being different from 5%, are given for various materials and structural
systems according to the relevant ductility classes in the various Parts of EN 1998. The value of the
behavior factor q may be different in different horizontal directions of the structure, although the
ductility classification shall be the same in all directions.
In RMI-2008 [9], the reference for the definition of the response modification factor R is the NEHRP
- 2000 edition “Recommended provisions for seismic regulations for mew building and other
structures” Part 2 [12] – Commentary chapter 5.2.

21
It includes the effects of plastic behavior of the materials, structure overstrength, and damping other
than 5% of the critical damping.
The Commentary states also in the text that “The R values, contained in the current provisions, are
largely based on engineering judgment of the performance of the various materials and systems in
past earthquakes”.
1.2.5 Figures for the behavior factor
Both FEM 10.2.08 [4] and RMI-2008 [9] consider the schemes of unbraced frames or braced frames
with concentric bracing.
FEM 10.2.08 [4] Chapter 3.4 states that EN1998 [5] defines low ductility concept for behavior factors
not greater than 2. In any case, the value of q need not be less than 1.5.
According to EN1998 [5], for low ductile design concept there are not prescriptions regarding the
control of the collapse mechanism and overstrength.
In the definition of the behavior factor FEM 10.2.08 [4] takes into account mainly the effects of the
behavior of the steel structure of the rack, while other effects are considered through other
coefficients.
The values of the behavior factors, in low ductility design concept, range between 1.5 and 2.0.
i) Upright frames
The behavior factor q is correlated to the scheme of the bracings; there are three main categories of
bracing patterns for the upright frames:
-Bracings where resistance is provided by diagonals working in compression, without the possibility
of formation of other mechanism when a diagonal buckles and consequently the upright is subject to
primary bending; this is the case of K, D, Z bracings, and X bracings without horizontal members.
For these configurations, lower behavior factors is applied (q=1.5) and, for the upright frame
bracings, extra safety factor 1.5 is applied to the design of the diagonals and their connections
-X bracings with horizontal members, with diagonals working in tension and compression; in this
case q=1.5 is used but no extra safety coefficient for the design of the diagonals and their connections
is required
-X bracings with horizontal members, with diagonals working only in tension; in this case q=2 is
used.
ii) Vertical bracings
For the vertical bracings the same criteria used for the bracing scheme of the upright frames apply.
iii) Moment resisting frames (down aisle direction)
In general q=1.5 is applied; q=2 can be considered if the beam-end connectors are secured with
bolts (with washers below both bolt’s head and nuts), this assuming improved performance under
cyclic loads.
RMI-2008 [9] Chapter 2.6.3 defines R=4 in the braced direction, regardless the scheme of the
bracing and the type of connection, and R=6 in the unbraced direction.
The braced direction is in general the direction of the upright frame and the down aisle direction
when the rack is braced; the unbraced direction is in general the down aisle direction when the rack
is not braced.
The comparison between the figures q and R can’t be done directly; a quantitative comparison can
be done between q/(ED1*ED3) and R.
When q=1.5: q/(ED1*ED3) ranges between 2.25 (ED1*ED3=0.67) and 3.75 (ED1*ED3=0.4) [Eq. 13]
When q=2.0: q/(ED1*ED3) ranges between 3.0 (ED1*ED3=0.67) and 5.0 (ED1*ED3=0.4) [Eq. 14]
For the braced direction the figures defined in FEM10.2.08 are largely more conservative than R=4
defined by RMI-2008 [9] in almost all the practical situations, apart from some extreme ones with
very strong earthquakes when there are relevant effects of sliding of the pallets on the beams.
For the unbraced direction the figures defined in FEM 10.2.08 [4] are always more conservative than
R=6 defined by RMI-2008 [9]; it must also be emphasized that the condition in which ED1*ED3 <
0.67 (ED1<1) in down aisle direction is quite unlikely to occur because this would imply very strong
accelerations on the unit loads and consequent strong bending moments in the vertical members.
It must be emphasized that RMI-2008 [9] allows the use of R=6 with spectral reduction of the action
only when the rotational capacity of the beam-upright connection is demonstrated by test to be
greater than the rotational demand (this is compulsory for racks designed in seismic design category

22
D); alternatively, the maximum spectral ordinate must be considered to calculate the seismic action,
equal to:
Cs = SDS/R (Table 2, Table 3) [Eq. 15]

Table 2 Seismic design category as per ASCE-7 [16] 2010 chapter 11.6

Table 3 Risk category category as per ASCE-7 [16] 2010

1.2.6 Methods of analysis

In FEM 10.2.08 [4] Chapter 2.4, the reference method of seismic analysis is the modal response
spectrum analysis (MRSA); simplifications of the method can be applied depending on the structural
regularity in plan (Table 4).

23
 Table 4 Methods of analysis in FEM 10.2.08 [4]

The structural regularity affects the allowed simplifications of the analysis and the value of the
behavior factor (Table 5).

Table 5 Consequences of structural regularity on seismic analysis and design in FEM 10.2.08 [4]

The criteria defined to establish when the lateral force method of analysis is allowed are fixed to
ensure that the higher modes of vibration have limited influence on the response.
In order to use the MRSA for racks non-regular in plan, it is necessary to consider in the proper way
the torsion and the possible related resonance.
In RMI-2008 [9] Chapter 2.6.2, the reference method is the lateral force method of analysis based
on a seismic base shear V determined as follows:
V = CsIpWs [Eq. 16]
Where
Cs is the seismic response coefficient (design spectrum)
Ip is the Importance factor
Ws is the weight of the seismic mass
The usual method of distribution of the base shear varying linearly along the height is considered.
The effects of non-regularity in plan are mentioned in chapter 2.6.7 “horizontal shear distribution”
where is generically mentioned that the seismic design shear at each level shall be distributed to the
various vertical elements of the seismic force resisting system at the level(s) under consideration
based on the relative lateral stiffness of those elements.
24
The displacement-based method referenced in FEMA-460 chapter 6.5.1 can also be applied; the
method is based on simplified kinematic models for racks unbraced in the down aisle direction and
with a regular geometry and set of properties.
The modal response spectrum analysis is not referenced in the RMI-2008 [9] Code but it appears be
used in the practice.
1.2.7 Second order effects
In FEM 10.2.08 [4] Chapter 2.4, the method to consider second order effects depends on the
sensitivity of the structure to the lateral displacements, measured by the parameter , defined as
Inter-storey drift sensitivity coefficient for the fundamental mode.
 is defined as in the EN1998 [5], but FEM 10.2.08 [4] provides and alternative definition which is
simpler to be applied and that is consistent with the definition of the Eurocode formula:
 = q Ptot/Pcr.E [Eq. 17]
where
q is the behavior factor
Ptot is the total vertical load
Pcr,E is the Euler critical load
Second order effects need to be considered either directly in the analysis, performing a geometrically
nonlinear analysis, or indirectly by amplifying the effects of the horizontal action by:
1/(1-θ) [Eq. 18]
Amplification of 2nd order effects by factor shown in Eq. 18 is not recommended if θ>0.3 as the
results tend to be unduly conservative.
In RMI-2008 [9] Chapter 2.6.4, the amplification of the displacements due to second order effects is
considered in the displacement-based method referenced in FEMA-460 Appendix A, and in the design
procedure described in the Code only for the evaluation of the rotational demand of the beam-end
and floor connections.
The 2nd order amplification provided by FEMA-460 Appendix A for the down aisle direction only, is
based on the simplified analytical model represented in Figure 15; in the kinematic model adopted
all the flexibility is concentrated at the connections.

Figure 15 Kinematic model in RMI 2008

The second order amplification factor is defined by the formula:

[Eq. 19]
Where
kc = secant rotational stiffness of the beam-end connector
kbe = flexural rotational stiffness of the beam-end assumed to be given by
kbe = 6EJb/L [Eq. 20]

25
kb = secant rotational stiffness of the base-plate
kce = flexural rotational stiffness of the base upright-end, assumed to be given by
kbe = 4EJc/H [Eq. 21]
hpi = height of the shelf i
Wpi = weight of the loads at level I (not scaled by 0.67)
Nc = number of beam-to-upright connections
Nb = number of base-plate connections
The numerical comparison with results of analyses performed using large displacement analysis
shows a relatively good approximation, but generally with underestimated 2 nd order amplification.
1.2.8 Combination of the seismic actions
FEM 10.2.08 [4]: chapter 2.4.5 requires combining the seismic action in the two horizontal directions
(and eventually in the vertical direction):
a) 0.30×EEdx “+” 0.30×EEdy “+” EEdz [Eq. 22]
b) EEdx “+” 0.30×EEdy “+” 0.30×EEdz [Eq. 23]
c) 0.30×EEdx “+” EEdy ”+” 0.30×EEdz [Eq. 24]
where:
“+” implies “to be combined with”
EEdx, EEdy and EEdz are the action effects due to the application of the components of the seismic
action respectively in the x and y (horizontal) and z (vertical) directions.
RMI-2008 [9] do not to require the combination of the components of the seismic action.
1.2.9 Overturning
For the design of the base fixing to the floor, FEM 10.2.08 [4] chapter 2.3.7 allows to consider,
besides the situation of rack fully loaded, only the top level loaded.
In cross aisle direction the masses of the unit loads need explicitly to be placed at the level of their
centre of gravity.

Figure 16 Forces causing overturning in FEM 10.2.08 [4]

In addition to the prescription of FEM 10.2.08 [4], RMI-2008 [9]: Chapter 2.6.8 requires to consider
in this verification the rack fully loaded with 67% of its rated load capacity.
1.2.10 Control of the rotational capacity of the connections
In the low dissipative design concept the control of the rotational capacity of the connections is not
required by FEM 10.2.08 [4]. The rotational capacity must be demonstrated in the dissipative design
concept (q>2) based on the criteria of EN1998 [5].

26
RMI-2008 [9] chapter 2.6.4 requires that the rotational capacity Max of the beam to column
connection be demonstrated by test to be greater than the rotational demand, D
If the connection rotational capacity is not controlled, or the rotational capacity is less than the
rotational demand, the design with R=6 is permitted (except in Seismic design category D)
considering the horizontal action defined with the maximum ordinate of the spectrum (C S=SDS/R).
The rotational demand is determined as:
Cd (1+αs )Δs
ΘD = [Eq. 25]
htotal

Where
Cd = 5.5 is the deflection amplification factor (from FEMA 460 [10] chapter 6.2.6)
htotal is the height of the top shelf level
s is the second order amplification factor from FEMA 460 [10] calculated using the same Ws as the
vertical load
s is the seismic displacement determined using the same stiffness as used to determine the period
for the base shear and using the calculated base shear.
The high values of the seismic response modification factor R in down aisle direction need therefore
to be considered in conjunction with the control of the rotational capacity of the connections.
RMI doesn’t make distinctions between low ductile and ductile design.
For FEM 10.2.08 [4] this is not relevant in low ductile design concept.
1.2.11 Test methods
For low ductility design concept, FEM 10.2.08 [4] makes reference to the test protocols of EN 15512
[3] Appendix A, so in a range of relatively small horizontal actions and displacements.
Only for ductile design, tests are requested to assess the rotational capacity with procedures derived
from EN1998 [5] for the beam-end connector (chapter 5.3) and floor connection (chapter 5.5).
For the design of the unbraced frame in the down aisle direction, plasticity is assumed to be
concentrated in the connections.
A protocol for the loading history is prescribed in Annex G, derived from Seisracks [6] research, quite
difficult to apply in practice.
A testing protocol is described in Chapter 9.6 of RMI-2008 [9] and Commentary; the drift angle limits
 for various performance levels are indicated in Figure 17.

Figure 17 Testing protocol for beam-upright connection in RMI 2008

27
The connection must satisfy both the Strength and the Drift Angle requirements; the Test Specimen
must sustain the required Drift Angle for at least one complete loading cycle.
The test results also includes the beam-to-column moment-rotation characteristics and “dynamic
spring relationship”, that means a reduced value of the stiffness of the connection (with respect to
the one determined for non-seismic conditions) to be used in the analysis.
The rotational demand for the floor connection shall be not less than the one for the Beam-end
connector.
1.2.12 Design of pallet beams
FEM 10.2.08 [4] Chapter 4.2.4 provides a detailed procedure for the design of pallet beams
challenged by the horizontal inertial forces, in addition to the weight of the pallet and the actions in
down aisle direction.
The procedure takes into account the influence of friction between unit load and support beam, both
in terms of limitation of the action and restraint provided by the pallets resulting in a positive
diaphragm effect.

Figure 18 Forces acting on pallet beams in FEM 10.2.08 [4]

RMI-2008 [9] doesn’t prescribe any additional requirement for the design of pallet beams in seismic
conditions.
1.2.13 Influence of the type of bracings
In FEM 10.2.08 [4], the behavior factor q is correlated to the scheme of the bracings; in the case of
schemes where resistance is provided by diagonals working also in compression, without the
possibility of formation of other mechanism when a diagonal buckles, extra safety factor is applied
to the design of the diagonals and their connections
In RMI-2008 [9], the typology of bracings both for upright frames and spine bracing is usually the
concentric bracing, without any distinction between typologies
1.2.14 Strength of bolted connections
In FEM 10.2.08 [4], when the pattern of the bracing is such that the resistance is provided by single
diagonals working in compression, an extra safety factor 1.5 is required to design bolted connections
(Chapter 3.4).
In addition, in general the bolt’s strength must be 1.2 times larger than the bearing strength (this
requirement needs not to be applied when the bearing strength of the bolted connection is greater
than q times the calculated bolt shear due to seismic action.)
In RMI-2008 [9], despite the high values of the response modification factor R, RMI-2008 [9] doesn’t
mention any rule for the design of connections.
1.2.15 Beam-end connector stiffness to consider in the analysis
In FEM 10.2.08 [4] Chapter 3.3.3, the stiffness of the beam-to-column connection and of the floor
connections (base-plates) are derived from static loads tests performed according to EN 15512 [3],
specified for non-seismic conditions.
The reference value of nominal bending strength of the beam-end connections for which the stiffness
is calculated can be chosen equal to η1.0 times the bending strength obtained by the test.
A value of η<1.0 will normally give a higher stiffness and hence higher seismic action but
accompanied by a higher plastic reserve moment capacity. Conversely η=1.0 will give lower stiffness,
lower seismic forces but a lower plastic reserve.
This justifies that q=2.0 may be considered in any case.
28
In RMI-2008 [9] – Chapter C2.6.5, the beam-end connection stiffness to be considered for the
analysis and the design of the connection components is the secant stiffness derived from the
moment-rotation curve obtained from static test, consistent with the base shear applied loads and
resulting displacements. The connection stiffness is in general in the lower moment range. This rule
implies an iterative procedure to calibrate the numerical model.
1.3 Structural typologies studied in SEISRACKS2
Experimental studies of SEISRACKS2 are based on 4 types of upright frames with different
geometrical pattern of the diagonals (Figure 19), and 4 types of longitudinal frames with different
diagonal configuration and connection details (Figure 20).

Figure 19 Upright frames to be used in the tests and numerical analyses.

The system of the longitudinal bracings are composed of vertical bracings and horizontal bracings,
placed at each load level.
Vertical bracing systems are composed of either X bracings with extra uprights and extra horizontal
elements; diagonals working only in tension made by flats or rounded bars with turnbuckle; diagonals
working in tension and compression made either by tubes or channels, or bracings made by cables
with turnbuckle and post elements.
3 types of horizontal schemes are identified; for the details of components and type of connection
see Figure 20.

Figure 20 Horizontal bracing system configuration

29
1.4 Case studies
The design of the case studies has been carried out by the 4 industrial partners, according to their
daily practice and following FEM recommendations. Each partner designed almost 2 configurations,
one unbraced for low seismicity and one braced for high seismicity. Rack industrial partners are
referenced in this report as IPA, IPB, IPC and IPD due to a non-disclosure agreement. Partner B has
designed the racks braced and unbraced in both situations of low and high seismicity. Partner D has
designed 3 cases: unbraced in low and moderate seismicity, braced in high seismicity, to consider
three different types of upright frames.
Thus, in total the case studies are 11:
 4 cases unbraced for low seismicity
 1 case braced for low seismicity
 1 case unbraced for medium seismicity
 4 cases braced for high seismicity
 1 case unbraced for high seismicity
The design parameters are summarized in Table 6. The choice of the parameters is related to:
 Capability of the testing facilities that influences the dimensions of the racks
 Correspondence to the real practice that influences the number and position of beam levels and
the weight and dimensions of the unit load;
 Compliance to the reference Norm for the seismic design of adjustable pallet racking (the FEM
10.2.08 [4]), that indicates the values of parameters Ed2 and Ed3, the friction coefficient, the
importance class of the structure, the type of design spectrum to be used in relation to seismicity
level.

Height of the frame 8m

Number and height of the beam levels 4 levels @ 2 m nearly
Length of the rack 6 bays; 3 pallets per beam
Max Acceleration (Low Seismicity) 0.12g-0.15g (Soil C, Type 2)
Max Acceleration (High Seismicity) 0.25g-0.30g (Soil C, Type 1)
Pallet mass 800 kg
Ed2 1.0 (concrete blocks placed on wooden pallets)
 (spectrum cut-off) 0.2
Friction coefficient 0.375 (wooden pallet on steel beams – normal warehouse
conditions)
CL 0.67
CH 1.5
Importance Class 2
Design life 30 years  γ1=0.84 (normal use of the racks)
Ed3 0.67
Table 6 Geometrical configuration of the racks and design parameters

A standard EURO pallet is taken as reference for the unit load, with the following dimensions:
800 mm width x 1200 mm depth x 1500 mm height
Each IP designed its rack according to low dissipative design concept, applying a q factor between
1.5 and 2, depending on specific structural features of their products.

30
2. COMPONENT TESTS
2.1 Joint Tests on Upright-To-Beam Connections
Test on beam-to-upright connections were performed in both down aisle and cross aisle direction, to
assess the moment-rotation response of the joints under cyclic loading. The response parameters
derived by re-analysis of the experimental results are used for the numerical simulation of full scale
structures.
2.1.1 Loading in down aisle direction
In total 48 connections have been tested under regular conditions (2 connections per test setup)
plus 4 additional connections in 2 tests focussing on the influence earthquake bolts. As described
above in each test the moment resistance of two connections was determined so that the testing
program covers all test cases as described in the proposal. In general, a very high ductility is achieved
by all products.
The test setup in Figure 21 represents one shelf of a rack system with a bay width of 2.70 m; a
common bay width which allows to load 3 pallets. The height of the frame (h top + hbottom, is
approximately 1m and the beams are installed at approximately half height). To load the rack with
pallets two frames, a front and a back frame are needed. The front frame is composed of the rack
parts (uprights and beams) to be tested. The uprights are perfectly hinged at the supports and at
the top so that the sway of the frame is constrained by the beam to upright connectors only. The
back frame is a kinematic frame made of hollow sections with the beam perfectly hinged to the
columns.

Figure 21 Down-aisle test setup

The payload was applied by loaded pallets. This application of the payload had the advantage of a
realistic loading situation: correct hogging moment, correct and constant shear force at connectors,
redistribution of moments (left to right or reverse) allowed.
This test setup gives result for positive and negative Beam-end moments in one test and it also gives
results for the redistribution of the moments from one end to the other one. This information could
not be derived in component tests performed according to EN15512:2009 [3]. The beam-end-
moments can then be clearly derived with the measured forces by the application of mechanical laws.
The rotations in the connector are determined by measured displacements top and bottom on the left
and right side divided by the vertical distances d as indicated in Figure 21.
All results are given as relative values, where the load and the moment are related to their maximum
value achieved in monotonic tests without payload. Deformations and rotations are given with their
absolute values.
For each load case (0% payload, 50% and 100% payload) and for each industrial partner (4 industrial
partners) one monotonic and one cyclic test were performed. The monotonic pushover tests were
performed under controlled conditions, to derive the yield deformation ey, to be used as reference
parameter in cyclic tests. These were performed according to ECCS [17] cyclic testing procedure,
but with a modified amplitude factor: one cycle with the amplitude factor 0.25, 0.50, 0.75 and 1.0
and 3 cycles with the amplitude factor 2, 3, 4, …(compared to ECCS [17] where the amplitude factor
is 2, 4, 6, …).
31
The test specimen were delivered by industrial partners A to D. The selection of profile sections was
decided by the industrial partners according to their experience. The products of industrial partner B
were assembled with an earthquake bolt, while all other were assembled with the safety pin.
In the re-analysis two components of the bending moment were considered:
 Mtotal: the sum of the moment due to payload and due to pushing.
 Mpush: the resistance to pushing of the frame.
Compared to the cantilever tests in accordance to EN15512:2009 [3], differences may be expected
in this test setup due to the possibility of redistribution of moments from one Beam-end to the other
as it may also be expected in real rack systems. The determined moment rotation curves are in
good agreement with the design curves derived from EN 15512 [3] tests. It was also observed that
the moment resistance for positive and negative moments of the connection is comparable (Figure
22).
Mtotal/Md Mtotal/Md
3 3

2 2

1 1

0 0
-0.20 -0.10 0.00 0.10 0.20 -0.20 -0.10 0.00 0.10 0.20
-1 -1

-2 -2

-3 -3
Rotation in rad Rotation in rad
A: 0% mono - ML A: 0% mono - MR B: 0% mono - ML B: 0% mono - MR
A: 50% mono - ML A: 50% mono - MR B: 50% mono - ML B: 50% mono - MR
A: 100% mono - ML A: 100% mono - MR B: 100% mono - ML B: 100% mono - MR

Mtotal/Md Mtotal/Md
3 3

2 2

1 1

0 0
-0.20 -0.10 0.00 0.10 0.20 -0.20 -0.10 0.00 0.10 0.20
-1 -1

-2 -2

-3 -3
Rotation in rad Rotation in rad
C: 0% mono - ML C: 0% mono - MR D: 0% mono - ML D: 0% mono - MR
C: 50% mono - ML C: 50% mono - MR D: 50% mono - ML D: 50% mono - MR
C: 100% mono - ML C: 100% mono - MR D: 100% mono - ML D: 100% mono - MR

Figure 22: Mtotal rotation characteristics from monotonic tests

Figure 23 shows the change of moment and rotation of the rack when it is laterally pushed aside,
without considering the moments caused by vertical pallet loading. It can be seen from the diagrams
that the resistance of the rack against lateral deflection is practically independent from the loading
(2nd order effects are corrected as described above). This is a very important observation as it allows
for significant simplifications for design on the basis of moment rotation curves. Further it could be
observed from the tests that the positive (push-) moment resistance of all products at the right
connector of the frame (when pushing to the left) also reaches the design resistance – a new result
that could only be achieved with the presented test setup. This also allows for significant
simplifications for design on the basis of moment rotation curves.

32
 Mpush/Md Mpush/Md
2 2

1.5 1.5

1 1

0.5 0.5

0 0
-0.20 -0.10 0.00 0.10 0.20 -0.20 -0.10 0.00 0.10 0.20
-0.5 -0.5

-1 -1

-1.5 -1.5

-2 -2
Rotation in rad Rotation in rad
A: 0% mono - ML A: 0% mono - MR B: 0% mono - ML B: 0% mono - MR
A: 50% mono - ML A: 50% mono - MR B: 50% mono - ML B: 50% mono - MR
A: 100% mono - ML A: 100% mono - MR B: 100% mono - ML B: 100% mono - MR

Mpush/Md Mpush/Md
2 2

1.5 1.5

1 1

0.5 0.5

0 0
-0.20 -0.10 0.00 0.10 0.20 -0.20 -0.10 0.00 0.10 0.20
-0.5 -0.5

-1 -1

-1.5 -1.5

-2 -2
Rotation in rad Rotation in rad
C: 0% mono - ML C: 0% mono - MR D: 0% mono - ML D: 0% mono - MR
C: 50% mono - ML C: 50% mono - MR D: 50% mono - ML D: 50% mono - MR
C: 100% mono - ML C: 100% mono - MR D: 100% mono - ML D: 100% mono - MR

Figure 23: Mpush rotation characteristics from monotonic tests

Two failure types have been observed: beam failure and connector failure (Figure 24 from monotonic
tests). One failure is called a “beam failure” when cracks and/or buckles were observed at Beam-
ends, while the hooks or holes of the connector have minor deformation. In a “connector failure”
the loss of moment resistance is due to large deformation of the hooks (or their holes) while the
beam has no or minor damages. The tests have also shown that the failure modes may be different
at the ends of one beam, e.g beam failure on the left end and connector failure at the right end. It
depends on the direction of the first deformation, which failure type occurs. The failure mode has an
important influence on the degradation of the resistance of the system. The absolute values for the
two types of failure may differ significantly.

Figure 24 Failure modes from monotonic tests

The payload has a positive influence on the resistance of all investigated rack products while the
stiffness in the elastic range seems to be independent from loading (Figure 25). The increase of
resistance due to loading is supposed to originate from the enhanced forces transferred by the loaded
hook compared to the unloaded one.

33
 Force 1.5
Force 1.5

A: 0% mono 1
B: 0% mono
A: 50% mono B: 50% mono 1

A: 100% mono 0.5

B: 100% mono
0.5

-300 -200 -100 0

-0.5
100 200 300 0

-300 -200 -100 0 100 200 300

-0.5

-1

-1.5 -1

displacement in mm displacement in mm
Force Force
C: 0% mono 1.5

D: 0% mono 1.5

C: 50% mono 1
D: 50% mono 1

C: 100% mono D: 100% mono

0.5 0.5

0 0

-300 -200 -100 0

-0.5
100 200 300 -300 -200 -100 0
-0.5
100 200 300

-1 -1

-1.5 -1.5

displacement in mm displacement in mm

Figure 25 Force displacement curves of all products (A to D) for all load cases from monotonic tests

The comparison of cyclic and monotonic tests gives an overview of the degradation of the resistance
of the system. Observing the Figure 26 it can be stated that the test results are well reproducible
which confirms the adequate and sufficient quality of the test setup, testing procedure and application
and quality of the measuring devices.
A further result of the tests is the identification of the influence of the failure mode on the degradation
of the resistance, which can be identified by the comparison of the curve from a monotonic test with
the envelope of the corresponding cyclic test: the bigger the distance of the curves the bigger the
degradation of the resistance (Figure 26). From this point of view the systems with connector failure
seem to have an advantage compared to those with beam failure. However, the load level of systems
with beam failure is significantly higher.
Force Force
1.5 1.5

1 1

0.5 0.5

0 0

-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
-0.5 -0.5

-1 -1

-1.5 -1.5

Displacement in mm Displacement in mm
A: 100% mono A: 100% cyclic B: 100% mono B: 100% cyclic

Force Force
1.5 1.5

1 1

0.5 0.5

0 0

-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
-0.5 -0.5

-1 -1

-1.5 -1.5

Displacement in mm Displacement in mm
C: 100% mono C: 100% cyclic D: 100% mono D: 100% cyclic

Figure 26: Load deformation curves from monotonic and cyclic tests on systems with 100% payload

The influence of the earthquake bolt on the resistance of a frame was investigated in additional
monotonic tests on the products B and C for the unloaded situation. Differences occurred in product
B when the loading direction was changed after release. It is supposed that the difference is not
caused by deformations of the connector but by an earlier beam failure due to higher connector
stiffness with bolt. So it may not be expected that the application of an earthquake bolt increases
the performance of the connector in terms of ductility or resistance (Figure 27). The bolt may only
give additional redundancy when connector failures are expected. Here the bolt prevents the beam
to drop off the upright.

34
 Figure 27 Results with and without earthquake bolt and bolt application
2.1.2 Loading in cross aisle direction
The test setup represents a typical rack detail with a bay width of 2.70 m: two pairs of uprights allow
the installation of beams that can be loaded by standard pallets. The end frames are detailed such
that one frame is movable while the other end frame is fixed and the reaction forces R are measured.
In the tests a horizontal displacement is applied to the movable end frame and the corresponding
force F is measured, see Figure 28 in which these measures are designated as external measures.

Figure 28 Cross aisle test setup

Additional measures are the transverse rotation of the beam in the connectors, force measurements
that allow determining the shear forces H transferred from the uprights to the beams and - as an
option – the lateral displacements d of the beams between the pallets.
The measurement of the rotation angles and the global displacements allow determining the
moments in the connectors by application of mechanical rules. The result may be cross checked by
comparison with the measured shear forces in the case of unloaded tests.
In case of loaded pallets, friction of pallets may influence the deflection of the beams such that
mechanical rules for determining the moments do not apply. The change of the deflection line of the
beams is determined by the displacement measurement di in the gaps between the pallets. Together
with the shear forces H, the moments and rotation in the connections, it should be possible to
determine friction effects of the pallets.
The project proposal foresees one monotonic test and one cyclic test per industrial partner (4
industrial partners) and load case. The proposed load cases are no payload, 50% payload and 100%
payload, where 100% payload is 3 pallets of 800kg weight each. The push over and cyclic tests are
deformation controlled tests. The monotonic tests give the reference deformation e y for the
deformation controlled cyclic tests that were carried out according to the ECCS-recommendation
[17], but with a modified amplitude factor: one cycle with the amplitude factor 0.25, 0.50, 0.75 and
1.0 and 3 cycles with the amplitude factor 2, 3, 4, … (compared to ECCS [17] where the amplitude
factor is 2, 4, 6, …).
Figure 115 - Figure 118 in Annex for chapter 2 report the resulting load displacement curves. The
measured applied force in the test had to be “corrected” from second order effects that occur due to
supporting the moveable frame on columns. For a specific range of deformation the hysteresis of the
corrected force can be described by two parallel lines. The distance of the lines is defined by the
friction of the pallets on the beams and the slope gives information about the elastic deformation of
beams, beam connectors and pallets. In the tests bending of the pallet beams in cross aisle direction
was observed that is accumulated due to friction.

35
The tests on all products showed that there is no significant moment resistance in the beam to
upright connection for moments due to cross aisle deflections. For this direction of loading, friction
of pallets on the pallet beams is governing the resistance.
In addition to the test program in the proposal friction tests have been performed for the assessment
of the influence of pallets on the cross aisle deflection resistance of the racks. The friction tests were
not performed in accordance with FEM-rules (FEM 10.2.08, 2010 [4]) where the pallet beams are
inclined until sliding of the pallet. In the tests performed here the beams remain in horizontal position
while the pallet is moved by a measured external force. The results are shown in Figure 119 in the
Annex for chapter 2.
2.2 Column-base tests
These tests were carried out to determine the plastic behavior of the column bases of rack systems
under seismic loading, and to assess the moment rotation characteristics of the connection between
the upright and the column base.
The connection of the base plate to the ground (concrete slab) is not within the scope of these tests,
so that upright to base plate connections are tested and not the dowels. The possible variations of
concrete slabs (thickness, reinforcement and strength) on the compression side of the connection
and variation of applicable dowels (type and producer) on the tension side of the connection is too
large to obtain results that can be transferred to real projects.
The characteristics of this part of the connection can be obtained for a specific case application on
the basis of the dowel characteristics (provided by dowel producer) and concrete characteristics and
easily added to the moment-rotation characteristic from the tests presented here. Figure 29
illustrates the stiffness and resistance component of the complete base-plate connection.
Rotation of base plate Rotation of connector

Cj,plate Cj,connector

1 1 1
──C──── = ──C─j,plate
─── + ─C─────
j,total j,connector [Eq. 26]
Figure 29: Components of stiffness and resistance of the fixing upright to slab

The test set up according to EN15512:2009 [3] is not appropriate for the determination of the moment
rotation characteristics under earthquake exposure as two column bases are tested at once and only
small rotations are investigated.
Within the research project, the moment rotation characteristics especially under large rotation are
investigated; further, tests on both down aisle or cross aisle direction have been performed.
Respect to each direction, the upright undergoes to different type of loading. While the structural
behavior in down aisle direction causes bending at the base, in cross aisle direction the main effect
is represented by normal forces on the uprights. Bending in the baseplate connection due to cross
aisle deformation of the structure seems to be negligible.
2.2.1 Down aisle tests
For down aisle tests the test setup (Figure 30) represents a column base that is rigidly connected to
the ground. On top of the upright there is a top steel plate where the horizontal load is applied. To
allow for horizontal deflection of the top steel plate a pendulum is installed between the top plate
and hydraulic jack that applied the vertical force.
The vertical force is controlled to be constant over the duration of the test while horizontal
displacements are superimposed by a horizontal hydraulic jack measuring the applied force.
Horizontal forces from second order effects are also measured by the horizontal devices. The rotation
of the upright at base plate has been measured by two displacement transducers applied to the
upright.

36
 Figure 30 Sketch of principle test setup for down aisle tests

As in all other component tests three load levels (0%, 50% and 100% payload) have been tested in
down aisle direction. Monotonic tests give the reference deformation e y for the cyclic tests carried
out according to the ECCS-recommendation [17], but with a modified factor for ey compared to the
ECCS recommendation [17]: one cycle with the amplitude factor 0.25, 0.50, 0.75 and 1.0 and 3
cycles with the amplitude factor 2, 3, 4, …
In down aisle direction, the stiffness of the connection increases with an increasing axial load level
for all type of products; instead the variation of bending strength due to the axial loading differs
significantly between the products. The moment bending strength may be limited by local or global
buckling effects of the upright (especially when fully loaded) or by failure in the connection (failure
in welds or bolts – or large deformation of the base plate) especially under moderate loading. The
bending strength in cyclic and monotonic tests is in most cases comparable, while the rotation
capacity is significantly reduced in the cyclic tests.
2.2.2 Cross aisle tests
Special considerations have to be made for testing in the cross aisle direction as the combination of
vertical and horizontal loading on the column is very different to that in down aisle direction.
A preliminary investigation has been performed to find a test set up realistically representing the
actions on the connection during earthquake and that could be reproducible in the lab. The solution
was testing the inclined connector under varying axial vertical loads, to consider the effects of both
bending and axial load. Furthermore, in cross aisle direction detailing has a big influence on the
bending strength of the connection. A conclusion from the test is a proposal for a tests setup taking
into account the effects that could not be investigated otherwise.
For testing in cross aisle, as illustrated in Figure 30, the axial load in the upright is the relevant
oscillation measure. In addition, shear forces in cross aisle direction induce deformations in the
column base connectors. From the design case studies performed in the project it is known that the
axial force at base varies from 140 kN in compression to 40 kN in tension, while the shear force plays
in the range of ± 12 kN (under high seismic exposure). In order to test the combination of shear and
axial load the test specimen is mounted with an inclination of about 5°. Since the test facility can
apply forces in the range of ± 100 kN, the maximum compression (140 kN) cannot be applied. On
the other hand the load resistance under compression is of minor interest because the compression
force is transferred to the base plate by contact with the edge of the upright.
Vertical compression forces are transferred to the baseplates by contact forces between the upright
and the connected base plate. When the compression force decreases and changes to a tension force,
sliding occurs in the connections bolted to the upright. Some products use the notches for the beam
end connectors for the bolts which are much larger than the bolts diameter. For these products
sliding has a big influence on the stiffness of the connection. Thus the stiffness from the tests may
not be applicable for design in practice.
As the column bases are in practice always installed and do act as couple (two bases per cross frame)
it is suggested for further investigation to test the couple. For such tests a ridged cross frame should
be used so that all deformation and possible failure modes occur at the baseplate connectors only.
All the results are reported in the Annex for chapter 2.

37
2.3 Substructure tests
2.3.1 Braced longitudinal frames
The aim is to assess the efficiency of the longitudinal bracing under cyclic loading and to characterize
bracing-to-frame connections.
The system of the longitudinal bracings is composed of:
 Vertical bracing system: i) X bracings with extra uprights and extra horizontal element; ii)
diagonals working only in tension made by plate or rounded bars with turnbuckle; iii) diagonals
working in tension and compression made by tubes or channels, iv) cable stayed bracings made
by cables with turnbuckle and post element.
 Horizontal bracing system: 3 types of horizontal schemes are identified (Figure 31).

Figure 31 Horizontal bracing system configuration

Two tests for each partner (1 pushover and 1 cyclic) were carried out on longitudinal frame "one-
span-one-level" specimens with bracing.
The test setup is shown in Annex for chapter 2 (Figure 132 and Figure 133).
All graphical results are shown in Annex for chapter 2.
Horizontal bracing configuration Type 1c, diagonals working only in tension:
The hooks of the beam-end connectors close to the vertical bracing and to the load acting point,
slowly opened. At the end, connector got out, which led to the failure of the rack specimen by
disconnection of the pallet-beams and at the same time of the horizontal bracing panel (Figure 32).
Displacements corresponding to the uprights on the opposite side of spine bracings are much larger
than the other ones, as bracing panel is very stiff.

Figure 32 Deformation of the connecting hooks and disconnection of the pallet-beams.

Horizontal bracing configuration Type 3b, diagonals working only in tension:

The diagonals in compression buckle and the diagonals in tension get stiffer until the horizontal
bracing system diagonal connecting the cross-frames to the eccentric vertical bracing system failed
by rupture of the net section at the bolt hole.
Thanks to the flat area in the bracing connection, strain gauges could be applied in this case to
measure the forces on the bracing. Vertical bracing received 75% of the applied load.

38
 Figure 33 Tested system and horizontal bracing failure

Horizontal bracing configuration Type 2b, with cable diagonals:

Failure occurred by tearing of the strap connecting the bracing cable to the eccentric column. It is a
failure in the root of the welding. The structure behaved substantially elastic until the failure.

Figure 34 Structure before and after failure during the push-over.

Horizontal bracing configuration Type 1b, diagonals working in tension/compression:

The failure occurs because of excessive deformation. The vertical bracing is very stiff and the outside
panel opposite to the bracing panel moves causing the rack to deform as a parallelogram (Figure
35).

Figure 35 Maximum displacement of uprights

During the cycle test the hooks of the pallet-beam’s connector jumped out of the notches of the
upright. The reason the hooks jumped out of the notches might be due to several reasons:
 no vertical loads are applied on the pallet beams,
 no safety pin is foreseen,
 the horizontal load is applied by means of jack acting under the bottom surface of the pallet-
beams which does not guarantee a horizontal force but might be slightly inclined causing a
lift of the beams.
All beam-connectors were then equipped with bolts connecting them to the uprights through both
sides; a nut is fixed to the bolts just to have a contact of bolt-head and nut against support. The
cyclic test is continued under force-controlled conditions. When pushing the structure to the left, the
left beam-connectors push on the interior flange of the uprights causing the section to close whereas
39
the right beam-connectors pull the interior flange causing the section to open. When pulling to the
right the opposite phenomenon occurs. During the test different elements softened and the failure
of the weakest part caused failures of other elements such as:
 buckling of the horizontal bracing
 failure in shear of the upright to base-plate connecting bolt
 failure of the horizontal bracing connections
 bending of bolt of vertical bracing and bending of plate
2.3.2 Cross frames tests
4 types have been identified based on the geometrical pattern of the diagonals, and 4 types of
diagonals’ position and connection (Figure 36).

Figure 36 Types of uprights to be tested

7 case-studies have been prepared by 4 Industrial Partners with the objective of getting a wide range
of situations (design for low/moderate/high seismicity, D/Z/X type of cross bracing). The
configurations to be tested, made by the combination of frame’s type and diagonal’s type, are
summarized in the previous chapter. The total number of tests on the subcomponent is:
 for unsymmetrical frames: 3 tests = 2 pushover + 1 cyclic
 for symmetrical frames: 2 tests = 1 pushover + 1 cyclic
8m high frames have been tested, in a horizontal position for compatibility with the existing lab
facilities. All suggested cross-frames have been modelled numerically.
Some pre-test with different test setups were performed to identify the test setup that best
represents the real situation. The test setup identified is shown Annex for chapter 2.
During the pushover tests the load is increased until collapse.
Cyclic testing procedure is composed by single fully reversed cycles at displacement ductility of ¼,
½, ¾ and 1, followed by a sequence of groups of two (or three) cycles at multiples (2, 3, 4, etc.) of
the yield displacement (obtained from the pushover test) until a conventional failure criterion is met.
All the results are shown in the Annex for chapter 2.
2.3.2.1 Partner A (frame for high seismic zone) type 4F
This cross-frame is symmetric and only one push-over test is required.
The failure modes of this cross-frame which are:
 local and global buckling of diagonal in compression
 shearing of bolt of diagonal in tension

The failure load found in the cyclic test corresponds perfectly to the one of the push-over test.
The predicted failure mode is compatible with the observed failure mode. The compression design
strengths for buckling for the diagonal and for the flat part of the diagonal are very close; the failure
is then depending on the effective weakness of the installed frame, but in accordance with the
predicted failure mode (see the Figure 139 in the Annex for chapter 2).
2.3.2.2 Partner B (frame for moderate-to-high seismic zone) type 1C
As this cross-frame shows D-shaped diagonal patterns, is asymmetric, and 2 push-over tests are
required.
The diagonals of these cross-frames are neither convergent to one connecting point nor to the centre
of the upright cross-section. During the push-test one diagonal of each intersection is in tension and
the other one in compression. The bending moment appearing in this zone increases until failure by
creation of a local shear in the lower upright. This phenomenon is accompanied with an out-of-plane
movement of the upright.

40
Even though this cross-frame is unsymmetrical both push-over tests can easily be compared in terms
of maximum applied load and corresponding displacement before failure. Failure under cyclic
condition was achieved for lower values than those for the pushover test although the failure mode
for both type of tests is the same. The failure mode predicted with the finite element model is not
accurate. Although the estimation of the base shear at which failure occurs is very good, the real
failure mode could not be foreseen with the model as it assumes that the diagonals converge in one
single point in the centre of gravity of the upright which is not the case in reality (see the Figure 140
in the Annex for chapter 2).
2.3.2.3 Partner C (frame for moderate seismic zone) type 1A
This cross-frame is asymmetric and two pushover tests are required.
The uprights were subjected to torsional buckling during the test and failure occurred by buckling of
the diagonal in compression with a local crushing of the diagonal’s end zones and bearing of diagonal
in tension.
The push-over curve in one direction matches perfectly the cyclic curve whereas the push-over curve
in the other direction reaches higher loads with lower displacements. The reason is probably due to
the fact that both push-over tests were performed downwards while the cyclic down- and upwards
leading to not symmetrical response in terms of looseness of the connections.
The bearing and buckling failure modes of the diagonals could be predicted with the finite element
model for a base shear lower than the experimental one (see the Figure 141 in the Annex for chapter
2).
2.3.2.4 Partner C (frame for high seismic zone) 2a-B
This cross-frame is symmetric and only one pushover test is required. The failure load is much lower
than the one for the previous tests.
The envelope of the monotonic test matches quite well with the cyclic response. Failure in the cyclic
tests occurs before the maximum load of the monotonic tests could be reached. The observed failure
modes are:
 buckling of diagonal in compression,
 bending of bolts,
 torsional buckling of uprights.
FEM model predicts a lower failure load and is unable to identify correctly the failure mode (see the
Figure 142 in the Annex for chapter 2).
2.3.2.5 Partner D (frame for low seismic zone) 1B
This cross-frame is asymmetric and two pushover tests are required. The first monotonic test was
performed by pushing the frame whereas the second by pulling it. The cyclic test has started by
pushing the frame. The failure modes for this cross-frame:
 buckling of diagonals in compression,
 bearing of diagonals in tension.
The vertical shift of the cyclic test in the positive values with regard to the monotonic one is due to
the push down and upwards for the cyclic test whereas both monotonic are performed by pushing
downwards with a different behaviour in terms of looseness of the diagonal to upright connections
(see the Figure 143 in the Annex for chapter 2).
FEM allows a correct prediction of the failure mode but underestimates the failure loads
2.3.2.6 Partner D (frame for moderate seismic zone) 2b-B
This cross-frame is symmetric. A monotonic and a cyclic tests are performed for two configurations
each: with and without the bolt connecting the diagonals mid-span in all panels.
The first monotonic test consisted in increasing the load as much as possible without breaking the
setup itself. As the frame remained in the elastic domain loading was stopped. After removing all
bolts connecting the diagonals at their mid-length the test was restarted. This cross-frame is very
rigid. Figure 144 Figure 145 in the Annex for chapter 2 show the evolution of the load vs.
displacement at the upper beam-level with and without the bolts.
The tests without the connecting bolts could be carried out until failure. The failure mode is
characterized by buckling of diagonals of lower panels (see the Figure 145 in the Annex for chapter
2). Also in this case, FEM allow a correct prediction of the failure mode but under estimated the
failure load.
2.3.2.7 Partner D (frame for high seismic zone) 2b-B
This cross-frame is symmetric. A monotonic and a cyclic test are performed for two configurations
each: with and without the bolt connecting each other the diagonals at their mid-spam.
41
The tests without the connecting bolts could be carried out until failure. This cross-frame is very
rigid. Figure 146 in the Annex for chapter 2 show the evolution of the load vs. displacement at the
upper beam-level with and without the bolts.
The failure mode is characterized by buckling of diagonals of lower panels and bearing of diagonal in
tension (see the Figure 147 in the Annex for chapter 2).
Also in this case, FEM allow a correct prediction of the failure mode but under estimated the failure
load.
2.3.3 Ductility
Ductility of each frame is estimated with the following procedure:
On the force-displacement curve of one push-over test, horizontal line at 75% of the maximum
reached load is drawn, crossing the force-displacement curve in a point. This point is connected to
the origin and this new line extended upwards until crossing the horizontal line at the maximum
reached load. This intersection identifies, on the X axis d y. A horizontal line at 80% of the maximum
reached load which crosses the force-displacement curve in a point identifying, identifies on the X
axis dmax. The ductility of the cross-frame is given by dmax/dy.
The following table summarises the different failure modes of the tested cross-frames and the
corresponding ductility:

Failure mode Estimated ductility

Type 4F Diagonal: buckling + shearing of bolts 1.01

Type 1C Upright: local shear 1.75

Diagonal: buckling + local crushing

Type 1A 1.40 / 1.63
Upright: torsional buckling
Diagonal: buckling + bending of bolts
Type 2a-B 1.79
Upright: torsional buckling

Type 1B Diagonal: buckling + bearing 1.33 / 1.18

Type 2b-B Diagonal: buckling 1.27

Type 2b-B Diagonal: buckling + bearing 2.06

Table 7 Estimated ductility

2.3.4 Upright stiffness

Actual shear stiffness of the upright frame is affected by several factors (e.g. connection
eccentricities, connections looseness, structural imperfection, bolt bending, local distortion of
uprights), which are not normally considered in the numerical analysis of racking structures. In order
to take into account these factors, that reduce considerably the overall stiffness compared to the one
that can be obtained with simple structural models, a reduced value of area of bracing members is
considered. This value is obtained comparing the deformation of a typical upright frame tested in a
laboratory test, with the numerical model of the same frame, reducing the area of the diagonals
cross section in order to match results of real test.
This reduced cross-section is chosen in a way that it can change from one beam-level to the other.
This means that we consider one reduced cross-section for the lower diagonals between ground and
the first beam-level, another reduced cross-section for the diagonals between the first and the second
beam-level, again another reduced cross-section for the diagonals between the second and the third
beam-level and finally a last reduced cross-section for the diagonals between the third and the last
beam-level.
These reduced cross-sections Ared,i are chosen independently such as the displacements observed at
each beam-level match the displacements measured during the test.
For that we chose for each configuration a base-force Fb from the measured load-displacement curve
𝑖
which is still in the elastic domain and applied this force as follows: 𝐹𝑖 = 𝐹𝑏 10, where Fi is the
horizontal force acting on storey i.

42
Standard shear tests according to the EN15512:2009 [3] seem to provide higher values of the
diagonal’s cross-sections to be used. This is the case for most of the tested configurations except 1
as can be seen in Table 8.
% of Ag according to
% of Ag according to SR2 tests
Standard Tests EN15512:2009 Typology
Bottom  Top
[3]
A 10 4.8  2.9 Symmetric
B 6.3 2.2  4 Asymmetric
C1 2.1 3.1  5.4 Asymmetric
C2 5.3 3.4  1.5 Symmetric
D1 12.4 12.4 Asymmetric
D2 12.4 12.4 Symmetric
D3 12.4 12.4 Symmetric
Table 8 Comparison of values for diagonal’s cross-sections to be used as a result of standard tests and tests
realised in the lab during SEISRACKS2.

Specimens D1, D2 and D3, all coming from the same Partner, seem to be indifferent from the test
procedure. A percentage of 12.4 can be used for the diagonal’s cross-section, which is much more
than all other frames. For cross-frames A to C, reduced cross-section for diagonals increase over the
height of the frame for asymmetric frames whereas it decreases for symmetric frames.
2.4 Component method
One of the peculiarities of storage racks is the assembly possibility which consists in connecting
beams to uprights by means of hook-end connectors. The response of this type of assembly is usually
characterized by bending tests.
The design of racks and thus of their “assembly nodes” is based on the recommendations formulated
in the EN 15512:2009 [3] in accordance with EN1993 [2] for thin walled cold formed elements. The
peculiarity of the FEM code lies in the fact that all fundaments refer to test results (design assisted
by testing). Indeed it advocates determining the geometrical and mechanical characteristics of the
different elements with tests in laboratories. In its analysis, the FEM code considers the resistant
moment as well as the design stiffness of the assembly, but it does not introduce any notion of
rotation capacity which excludes the possibility of any ductile behaviour.
In classical metallic construction there is an analytical method used to characterize the beam-to-
upright connections called “component method”, which gives the possibility to take into account this
rotation capacity.
Within the scope of this project, the “component method” is applied to develop resistant model for
the assemblies of the storage structures. This method induces a methodological approach.
The implementation of the component method involves a 3 steps procedure during which:
 all individual components of the considered joint are identified,
 stiffness and resistance of all components are evaluated,
 assembly of these characteristics in an appropriate manner in order to determine the stiffness
and resistance characteristics of the joint.
The stage of assembling the components involves, whether it is for the evaluation of the stiffness of
the joint or of its resistance, a distribution of the internal stresses in the joint. Indeed, as a joint is
subjected to external loads, which are distributed among the different individual components
according to their instantaneous stiffness and resistance. The analytical formulas are given in EN1993
[2] and its Annex J.
A comparison between the standard method described in the FEM and the component method leads
to the conclusion that this last method is favourable. Indeed, the component method:
 is an analytical method more economic for the industrial partner because, even if there are
changes in the specifications of the product, he does not have to redo lots of tests which are
often expensive;
 allows predicting the failure mode of the assembly;
 gives indications on the rotation capacity of the assembly and thus indications for an eventual
plastic analysis.
The component method is used here to characterize one component of a joint for one Partner. Even
if the component method includes the stiffness model as well as the resistance one, only this last
model will be developed here.

43
2.4.1 Active components
Before applying the method, the bending tests performed at RWTH have been identified and classified
according to their failure modes. The main components that contribute to the failure of the beam-
to-upright connections are listed in Table 9.
Partner A  bending of the hooks. First the upper hooks open and come out of the
notches. Then, when increasing the load this weakness propagates to
the lower hooks;
 Z-shape bending of the beam-end-connectors along the uprights with
the folds along the flanges of the beam;
 bearing of the column notches with bending of the earthquake bolt.

Partner B  cracking of the beam;

 local buckling of the beam;
 deformation of the end plates;
 reduced bearing of the upright notches.
Partner C  cutting of the uprights by the hooks;
 cracking of the welds between beam and end-plate;
 Z-shape bending of the beam-end-plates along the uprights with the
folds along the flanges of the beam;
 local buckling of the beam.
Partner D  yielding of the hooks with bearing of the respective upright holes;
 cracking close to the welds between beam and end-plate;
 Z-shape bending of the beam-end-plates along the uprights with the
folds along the flanges of the beam.
Table 9 components that contribute to the failure

2.4.2 Application of the resistant model

The method was applied only to the connection of partner B. The resistant moments are superposed
to the experimental moment-rotation curve for comparison; they are represented by horizontal lines
(Figure 37).

Figure 37 Moment – rotation measured during testing in RWTH.

In theory, for the model to correspond to reality, the horizontal lines M u (ultimate moment) need to
be situated above the top of the test curve. In this case, the first component subjected to failure
during the test is “cracking of the beam”, which is also the case during testing.
Bearing of the holes should occur almost at the same time according to this model. But the
characterization of this component was performed assuming all hooks are in tension whereas only
half the hooks are in tension, the compression sustained by other half is overtaken by the contact of
the end-plate against the upright. Another reason of the low resistant moment for bearing is the fact
that was the contribution of safety bolt not taken into account.
The analytical formulas derived here take into account the nominal failure strength guaranteed by
the industrial partner and not the real value. The real value is in general higher than the nominal
values so that all lines should be higher.

44
3. WAREHOUSE TESTING
3.1 Monitoring of a Pallet Rack
During the previous Seisracks [6] project, an installation near Athens had been continuously
monitored to record the rate of occupancy, operations and accelerations. A part of the data obtained
from this monitoring has already been used (essentially those dealing with identification of linear
dynamic properties), and published. Since the measurement system was still on the site, despite a
number of accidents (more or less unexpected) occurred in the meantime, more data have been
recorded during Seisracks2 research (Figure 37).

Figure 38 Pallet rack that is monitored (left) and accelerometer at point (right)

Three bays of a pallet rack system have been monitored. In total, 32 accelerometers have been
installed; their geometrical location is shown at Figure 39. These accelerometers record either down
aisle or cross aisle accelerations of the rack. Additionally a three-dimensional accelerometer has been
placed on the ground to record base accelerations.

Figure 39 3D setup of the monitor system’s accelerometers

The 16bit data acquisition comprises a computer unit for saving the data, with a sampling rate of
200 Hz. The system works continuously and starts recording data as soon as a minimum value of
acceleration is captured by the sensors. Accelerometers act as triggers for the recordings. The trigger
value of the acceleration is chosen 0,015g for the base accelerometers and 0,025g for the sensors
positioned at the top of the structure. The duration of the monitoring was one year.
3.1.1 Data processing
To process the data of high number of records, a program has been developed. Taking the minimum
relevant acceleration value as 0,049g, the program plots every value greater than this fixed limit
and the relative sensor. Each record, that contains at least one plotted value, are classified as
“significant”. Figure 40 and Figure 41 shows the response of each sensor in terms of the number of
“significant” accelerations, and the maximum peak value reached.

45
 Figure 40 Number of “significant” accelerations for sensor

Figure 41 Maximum peak acceleration for sensor

Observing these results, significant sensors have been identified as the ones which have:
 at least 3000 recordings of accelerations above 0.049g
 at least 60 recordings of accelerations above 0.3g
 recorded at least one acceleration greater than 0.9g
Figure 42 shows a summary of respected criteria of running sensors.
It can be observed that all sensors, positioned on the external upright in the longitudinal direction,
verify the first criterion. This was due to the presence of a manually operated wrapping machine
below the rack, which caused a continuous minor vibration during its operation. The presence of this
machinery certainly affected the number of low intensity accelerations recorded but considering its
wide use in warehouses; it is believed that the collected results were significant.

Figure 42 Summary of three criteria

All criteria are verified contemporarily only by the sensors 17 and 21, placed in the lower level of the
external upright in down-aisle direction. Figure 43 shows the distribution of the acceleration intensity
in terms of relative frequency can be analysed for sensor 21. The curve follows a hyperbolic trend:
more than 70% of the values are concentrated in the first interval, between 0.05 g and 0.1 g, while
for greater values of acceleration frequency decreases progressively. Nevertheless it should be noted
that the sensor 21 has registered as many as 15 values of acceleration greater than 1g.

46
 Figure 43 Sensor 21; relative frequency of accelerations

3.1.2 Assessment of the events

As the time of registration is quite long, 60 seconds after the triggering, it was decided to define
single significant shocks contained in a file as shaking-events, in the following way: an event is a
group of data registered after an acceleration value ≥0.05g, whose last significant value (≥0.05g) is
separated from the next significant one in the whole file, considering all sensors, at least 4 seconds
(Figure 44). That is, two values ≥0.05g belong to two different shocks if they are separated for at
least 4 seconds by only small tremors in the whole rack. To identify an event, this fixed condition
must be satisfied by at least one sensor.

Figure 44 Definition of event

Each event is characterized by maximum values of accelerations recorded between all sensors.
Frequencies of the maximum accelerations associated to each event are shown in Figure 45.

Figure 45 Histogram of the accelerations

Table 10 shows the results of this preliminary analysis of the data. We have a significant file when
at least one sensor got the 0.05g value. With respect to the 2028 significant files, the shocks
registered are 2169; so only in a small percentage of files there was more than one shock, and in
only a few of them three or four shocks. Moreover among the events considered, 448 can be set as
small events, as they have only one 0.05g value of the acceleration, therefore disregarding them the
number of significant shakes could be reduced to 1721. On the other hand, there are 676 events
with accelerations larger than 0.10g. Among these 186 shakes have a peak acceleration larger than
0.30g and in 55 of them the acceleration is larger than 0.50g. It can be seen that in the months with
the largest amount of shocks, the average trial is from 10 to 15 events every day.

47
 Table 10 Event classification

3.1.3 Earthquake recordings

During the monitoring period two low-moderate intensity earthquakes have occurred. The first one
was recorded with a magnitude of 4.4 on the Richter scale, on June 4 th 2013 at 4:56 in Attica, 22
km to the east of Athens. Maximum acceleration of 0.024 g was recorded on sensor 6 (Figure 46.a).

a. On sensor 6; June 4th 2013 b. On sensor 4; August 7th 2013

Figure 46 Earthquake registrations

On August 7th
2013 another earthquake was registered in Athens with magnitude 5.1 on the Richter
scale and a hypocenter of 10km. It occurred at 120km north / west of Athens, with its epicenter near
Elatia (Central Greece). Maximum acceleration of 0.024g was recorded on sensor 4 (Figure 46.b).
3.1.4 Damping
Damping in a rack depends on many factors, some constant, as the type of connections in the racks,
and others changing in time, as the loads on the racks and their distribution, which affect the friction
forces in the connections. In order to obtain an indicative damping behavior of the rack in a real
working situation, a group of events with large peak accelerations have been analysed. The damping
was evaluated considering the decrease of the acceleration parting from the peak value, both in
positive and negative direction, taking into account the first three or four following peak values, and
determining the average of these results. Many times it’s difficult to evaluate the damping, as only
one or two significant peaks of the acceleration are available.
An example for file 11_12_201213_28_15 (date and time of the event) can be seen in Figure 24 and
Table 3 for the registration of sensor 15.

Figure 24: Sensor 15 – Peak acceleration on December, 11 2012 h 13:28:15

Table 3: Sensor 15 – Damping for record on December, 11 2012 h 13:28:15

48
In the following Table 11 are reported the results for some files with large values of peak acceleration,
registered on sensor 21 down-aisle. The reported damping values refer to sensor 21 and to sensor
17, both down-aisle. Sensor 17 is positioned immediately over sensor 21 on the corner-upright.

Table 11: Max acc. in sensor 21 – Average Damping (%) in sensors 17 and 21

In Table 12 are reported the damping values in sensor 4, cross-aisle, on the top of the corner upright.
These values were calculated for 4 files, with registered maximum peak acceleration in sensor 4.
Considering the presented values representative for damping at the top for the whole rack in cross-
aisle direction, the average value yield by these data is 7.4 %. Therefore, as a rough indication on
the expected damping, a value of 6 % seems enough conservative, considering also that the position
of the corner upright is the less restrained one.

Table 12 Max acc. in sensor 4 - Damping in sensor 4

Applying similar considerations to the results in Table 11, we have for sensor 21 an average damping
of 9.5%, with lower values of 5.4%. For sensor 17 the results are different, with an average damping
value 6.5%, and a lower value 3.5%. Nevertheless a global 6% damping factor in down aisle direction
seems acceptable to describe the dynamic behavior of the rack.
3.1.5 Conclusions
The monitoring performed permitted a description of the dynamic loading undergone by a rack in a
real industrial situation. Considering significant shocks in the different parts of the rack only those
with acceleration ≥0.05g, 10-15 shocks per day are undergone by the rack during the periods of full
activity. Apart from the corner up-right, that got many shocks and was subjected to higher
accelerations for its position in the way crossing, from one to three important shakes per week, as a
global average, with acceleration >0.30g happened. At any rate the monitoring, showed that also
large shocks are possible, with accelerations >0.5g in all parts of the racks, and in some cases the
acceleration can be near to or larger than 1.0g. The shakes had generally an impulsive shape, with
a rapid decrease of the acceleration, and lasted generally less than one seconds. The typical
frequencies of the shakes, were from 50 to 60 Hz. An analysis of some important shocks, undergone
by the structure, suggested, from a conservative point of view, an indicative global damping factor
of 6 % for the structure. This study also underlines the relevance of accidental bumping upon pallet
racks due to pallets handling and human activities, leaving to subsequent studies the possibility to
convert it to a specific design load case to include in design codes.
3.2 Identification of the linear dynamic properties of racks on the base of
ambient vibration or hammer tests
Warehouse testing here presented aims to identify the linear dynamic properties of racks, on the
basis of ambient vibrations, free response, and hammer tests or appropriate excitation of the racks.
The objective is to define the range of periods and damping of a real rack structure. These results
are then meant to be used to calibrate numerical models.

49
Four warehouses have been considered (one from each producer). Tests have been performed on
site, on existing structures in active warehouses.
The position of the accelerometers is of high importance. The signals are decoupled in longitudinal
and transversal vibrations. The longitudinal bracing is covered by the pallet beams, and the beam-
to-upright connections are considered semi-rigid.
It can be assumed that the rack will deform mainly in shear, as indicated in Figure 47.

Figure 47 Vibration modes.

To distinguish and measure both vibration modes, it was necessary to place sensors at 2 levels; one
at mid-height (for an even number of beam-levels) and one on top of the structure.
Tests on rack systems in warehouses performed by ULG and RWTH showed that a relevant exciting
of the racks was not possible. This had two reasons:
 The absolute stiffness of a rack (in down aisle direction) grows with each bay. Thus even an
empty rack of e.g. 20 bays cannot be excited with usual means.
 The mass grows with every loaded bay. Thus not only the stiffness but also the mass to excite
is too high.
In case of an earthquake, high top displacements may be expected. In all measurements in
warehouses performed yet within this project the operators of the warehouse did not allow to excite
the rack with large amplitudes as they feared the rack may be damaged.
These experiences led to the decision to perform the last dynamic testing not in a warehouse but in
the laboratory of RWTH on a full-scale rack. The measuring in the lab has the advantages of no limits
on the dynamic deflection, more time to perform the measurements (for a warehouse testing you do
only have one day, in order to reduce as much as possible interfering with warehouse operations)
and the chance to apply all available means of dynamic measuring. From these advantages the time
factor is very important as more time allows evaluate the measurements and to repeat them if
necessary or to optimise the measurement procedure. Only one rack structure of Partner A has been
tested in the laboratory.
3.2.1 Partner A-Lab tests
A rack system from Partner A has been tested in the laboratory of RWTH Aachen. The measured rack
system has 2 bays (width of 2400 mm) and 3 storing levels. The rack was mounted as in a warehouse
with usual fasteners on the ground. Two measurement configurations have been investigated (Figure
48):
 Configuration 1 has beams at the heights of 1 m, 2 m and 3 m.
 Configuration 2 has beams at the heights of 2 m, 4 m and 5.2 m (the maximum height was
limited by the lifting height of the forklift).

a. Configuration 1 b. Configuration 2
Figure 48 Investigated rack configurations
50
Measurements have been performed in 3 steps. In the first step, the influence of single rack
components has been determined. In the second step the (static) stiffness was measured. In the
third step the rack system was excited by a shaker and dynamic properties have been determined.
During assembly of the racks, several free oscillation tests have been performed to get an idea of
the influence of the rack components on the dynamic behavior of the rack system. In these tests two
kinds of excitation have been applied: impact with a punch and rocking of the rack by hand. The
loading was changed in weight and position to achieve different natural frequencies and different
displacement amplitudes.
The decay part of measured acceleration-time history curve was evaluated to determine the natural
frequency of the system. Damping was determined by half band width method. The overview of the
test results is summarized in Table 13.

Test No Description Natural frequency Damping

Frame without any beams
01 2.46 0.002
Excitation by impact
Frame without any beams
02 2.45 0.007
Excitation by rocking
Beams at 1 m in one bay
03 2.61 0.02
Excitation by impact
Beams at 1 m in one bay
04 2.49 0.02
Excitation by rocking
Beams at 1 m in two bays
05 2.55 0.03
Excitation by rocking
Beams at 1 m and at 2m in one bay
06 2.76 0.06
Excitation by rocking
Table 13 Results of free oscillation tests at different erection stages

From these results first conclusions for small amplitudes can be drawn:
 The stiffness of this rack is governed by the base-plate connection. In view of the stiffness,
beams have a minor influence as there is only a small change of natural frequency after
installation of beams.
 The damping of this rack is governed by the beam or Beam-end connectors.
As these conclusions can only be drawn for small displacements which are possibly not relevant for
earthquakes, additional shaker tests with larger amplitudes have also been performed.
On each test set up the shaker tests started with an excitation frequency of 1 Hz that was increased
by 1 Hz step by step until 5 Hz were reached. Then further tests with smaller frequency step size
were performed to find the resonance frequency.
With the stiffness of the rack K, which was determined in the static tests, the static deformation d stat
can be calculated from the frequency at the resonance point. The dynamic deformation d dyn at
resonance is the amplitude measured at the resonance point.
It was possible to determine the corresponding damping values for the upper and lower bounds:
Configuration 1:
 Magnification factor 4.5:  = 0,11
 Magnification factor 6.5:  = 0,077
Configuration 2:
 Magnification factor 5.0:  = 0,10
 Magnification factor 8.0:  = 0,063
Location and mass of loading has no significant influence on the damping of the rack structure.
The dynamic measurements showed that damping behavior of this rack was dominated by friction in
the Beam-end connectors. Friction has, especially at high top displacements, a growing influence and
reduces the stiffness of the system. From the measurements, an upper and a lower bound for
(viscous) damping values could have been identified: at low displacements amplitudes (up to 10mm)
 = 0.10, at high amplitudes  = 0.06.

3.2.2 Partner B – Test realised by ULg

Warehouse is situated in Germany. This rack is a double-entry rack where the cross-frames are
connected at 1.5 m and on top Figure 49.

51
 Figure 49 Connection of the cross frames

The transverse and longitudinal vibrations have been measured at the second beam-level (4460 mm)
and at the fourth beam-level (8560 mm), in 3 different sections on the length of the structure for a
total of 6 signals (3 longitudinal positions on both levels, one arrow per measured level).
We decided to test the rack at the lower-left part of the general layout (Position 1). A second set of
tests has been performed moving all sensors along the structure (Position 2) (Figure 50).

Figure 50 Position of the accelerometers

In order to study the dynamic properties of the rack in longitudinal and transverse direction, 4
configurations were identified:
 Configuration 1 (Transversal measurements in Position 1): All sensors have been fixed on
the pallet beams. A rotating motor with eccentric mass is fixed on the pallet-beam 1st level
4th span with the wheel turning in parallel with the down-aisle direction. The measurements
are started; after 5 seconds, the motor was activated, increasing its rotation speed, and after
85 sec it was switched off.
 Configuration 2 (Longitudinal measurements in Position 1): The sensors are turned in order
to measure the longitudinal accelerations.
 Configuration 3 (Transverse measurements in Position 2): The sensors are placed on the
pallet-beams in order to measure the transverse accelerations. At the beginning the motor
turns slowly. After 100 seconds it has been stopped for verification, some hammer shocks
have been given to the structure, then after 600 seconds the motor has been turned on again
but much faster.
 Configuration 4 (Longitudinal measurements in Position 2): For this configuration the motor
was fixed on the same pallet-beam but very close to the upright. The sensors are turned in
order to measure the longitudinal accelerations.
The following table summarises the natural frequencies and corresponding vibration modes for each
configuration of a double rack:
Natural frequencies – Vibration modes
Configuration 1 – cross-aisle 10.5 Hz and 26.3 Hz
Configuration 2 – down-aisle 10.9 Hz
Configuration 3 – cross-aisle 26 Hz and 35 Hz – vibration mode 2
Configuration 4 – down-aisle 5.5 Hz and 6.5 Hz – vibration mode 1
Table 14 Natural frequencies (B)

The natural frequencies given in the design calculation notes provided by Partner B are much lower
than the frequencies measured on site. This is because measured frequencies are coming from almost
empty racks (by chance, when the measurements took place, the warehouse was almost empty due
52
to sales season). In design calculation notes, taking into account a fully loaded rack, the natural
frequencies were: 0.19 Hz, 0.80 Hz, 1.11 Hz and 1.6 Hz, respectively from configuration 1 to 4.
3.2.3 Partner C – test realized by RWTH
The measured rack system has 20 bays with a bay width of 3.6 m and 4 storing levels including the
ground level. The top loading level is 7.6 m above ground. The bays 20 and 9 provide passes through
the racks.
This test consists in determining the plastic behaviour of rack systems under seismic exposure in
cross aisle direction. The tests aim at the determination of the moment rotation characteristic of the
beam to upright connectors and of the influence of loaded pallets on the cross aisle deformation
behaviour. It is expected that this influence of pallets is mainly governed by the friction between
pallet and beam and the stiffness of the pallet. The influence of shear forces transferred from the
beams on the behaviour of the beam to upright connection is expected to be low or negligible.
The measurements were performed on rack row 23 which is coupled with row 22 by braces on each
level. The rack was partially loaded with plastic boxes on pallets fixed by stretch film. The pallet
weight was in the range of 100 kg to 400 kg.
Measurement and excitation were performed in bays 14 and 15 by applying accelerometers (± 3g)
on the upright at each level: accelerometers B, C and D were adjusted to measure in down aisle
direction; accelerometers F, G and H to measure in cross aisle direction; two additional
accelerometers were placed at the top level of the neighbouring uprights measuring also in cross
aisle direction (J and K) and one accelerometer (M) was applied at the location of excitation (Figure
51).

F B K Level 4
J

G C
Level 3

H Level 2

B ay 14
B ay 1
5
Level 1
Figure 51 Position of accelerometers.

For excitation in cross aisle direction a 10 kg impact hammer was used. For excitation in longitudinal
direction a rope was applied at top level and strengthened by a controlled force. The impact was
applied by cutting the rope.
Results are too large to be presented here, they are available inside the annex of deliverable 3.2.
3.2.4 Partner D – test realized by ULg
The warehouse is situated in Belgium. This warehouse is divided into 2 parts; one with a lot of traffic
of forklifts for loading and unloading the racks and another part with goods that are stored for a
longer period and where the traffic is reduced. A single-entry rack in the second part of the warehouse
was tested, in order not to disturb the activity of the store.
Two types of measurements have been performed independently: transverse measurements and
longitudinal measurements. For these measurements, 6 mono-axial sensors and one additional bi-
axial sensor were installed.
Transversal measurements: The first tests consisted in giving some impacts with a impact
hammer on the uprights close to the cross frames equipped with the sensors. The transversal
stiffness of the rack did not allow to measure significant movements. So the 12 th upright from the
4th level (8m) was pulled and pushed several times transversally before releasing the impulse and
let the structure find its position after absorption of the movement. A second test has been performed
by repeated hand-and-shoulder shaking the same 12th upright from the 2nd level (4 m).
Longitudinal measurements: As any significant transversal vibrations could not be noticed only
by using the impact hammer or during the loading or unloading the rack with pallets, it was decided
to proceed with the rack shaking with the forklift.

53
Here again 2 tests were performed, one by shaking the rack longitudinally from the 2 nd level and
then from the 4th level.
The following table summarises the natural frequencies for both transversal and longitudinal
measurements, these values are an average of the different measurements:
Natural frequencies
Transversal measurements 1.61 Hz, 2.02 Hz, 2.26 Hz and 3.82 Hz
Longitudinal measurements 0.61 Hz and 1.84 Hz
Table 15 natural frequencies (D)

The natural frequencies given in the design calculation notes provided by Partner C are the same
range as the frequencies measured on site. Taking into account a fully loaded rack the natural
frequencies are: 0.75 Hz, 2.28 Hz, 3.83 Hz and 4.90 Hz for the cross-aisle direction and 0.32 Hz,
1.06 Hz, 2.09 Hz and 3.26 Hz for the down-aisle direction.
3.3 Identification of dynamic properties of pallets/merchandize
The dynamic properties (frequency, damping) of a large range of stored merchandizes were identified
on the basis of a push-by-hand excitation on top of the stored good with a quick release or an impact
given by the human waist. Several such shocks have been given in both directions of the pallets and
the vibration and damping have been recorded. The description and the results of an individual test
are detailed in deliverable reports.
For each of the chosen pallet a tri-axial sensor has been fixed on top of the goods in order to measure
the vibrations along both axes parallel to the pallet edges (Figure 52). Each recording gives two
signals. From each signal, natural period and damping ratio are identified; they are then averaged
to give the property of the pallet.
For each test several impulses have been given to the pallet goods. Only the neat decreasing
exponentials have been kept for the free-response analyses.

Figure 52 IP D, Pallet n°1, 6 boxes of projectors packed with cellophane with a total mass of 106.8 kg

Each event results in a period/frequency and a damping coefficient ξ. Figure 53 and Figure 54
summarize the mean values of the natural frequency and the damping ratio of each palletized
merchandize with their distribution around the average, for partner B and D.

54
 Pallet No

Natural frequency [Hz] Damping ratio [-]

4.8
0.11
4.6
0.1
4.4 0.09
0.08
4.2
0.07
4
0.06
3.8 0.05

1 3 4 1 3 4
Pallet No Pallet No

Figure 53 Partner B, Box-plot representation of the natural frequency and damping ratio of the palletized goods.

For Partner B, natural frequency of the stored goods on pallets varies from 3.8 Hz to 4.5 Hz; the
damping ratio variesDamping
from 0.048 to 0.093.
ratio [-]

Natural frequency [Hz]

0.11
6
Values

0.1
4
0.09
1 2 3 4 5 6 7 8
0.08 Pallet No
Damping ratio [-]
0.07 0.1
Values

0.06
0.05
0.05
1 2 3 4 5 6 7 8
Pallet No
1 3 4
Pallet No
Figure 54 Partner D, Box-plot representation of the natural frequency and damping ratio of the palletized goods.

For partner D, natural frequency of the stored goods on pallets varies from 3.25 Hz to 6.21 Hz; the
damping ratio varies from 0.0316 to 0.072.

55
56
4. FULL SCALE PUSHOVER TESTS
For each industrial partner 1 braced and 1 unbraced frame have been tested in down-aisle direction
(for a total of 8 tests in down-aisle direction). This represents a modification of the original proposal,
where it was suggested to perform 4 tests in down aisle and 4 tests in cross aisle direction. However,
as the behaviour in cross aisle direction is governed by the response of the uprights, that is
investigated during the substructure tests carried out by ULIEGE within task 2.3, during the 2nd
technical meeting it was decided to concentrate attention on the down aisle behaviour, comparing
the response of braced and unbraced structure.
By elaborations of experimental results, capacity curves have been plotted in order to estimate the
global ductility resources. Furthermore the failure mechanisms of the different rack typologies have
been observed and discussed.
4.1 Design of full scale test specimens
Eight rack specimens have been designed, and provided by the four rack industrial partners
conventionally named IPA, IPB, IPC, and IPD for confidentiality reasons, according to a non-
disclosure agreement.
The height of the full scale specimens is approximately 8m, representative of most common
configurations on the market. To maximize the masses with respect to the stiffness of the rack, and
the local effects on the pallet beams, the configuration with 3 pallets per bay has been chosen, with
a corresponding bay length of 2.8 m.
Rack specimens for down-aisle tests are composed of 4 levels and 2 bays. Upright frames of 1.1m
depth, are composed of two columns made of thin gauge cold formed profiles linked together by a
system of diagonal and/or horizontal members. One of the two bays of braced rack configurations,
is braced both vertically and horizontally.
The mass of the unit loads is 800 kg, which represents a medium-high value for the merchandise in
practice. During the experiments, unit loads are represented by concrete blocks that according to
FEM 10.2.08 [4] correspond to the stored good class “compact and constrained”. Thus the parameter
ED2 has been set equal to 1.0 in the calculations. Design parameters are summarized in Table 4.
4.2 Test procedure
Tests have been performed outdoor, where full scale racks could be easily mounted and loaded, on
a testing facility managed by Politecnico di Milano, in a compound owned by Marcegaglia Buildtech.
The racks have been assembled over a specially designed support structure. This particular support,
shown in Figure 149 of Annex for chapter 4, is made with steel beams, and has been designed in
order to avoid deformation during the tests and allow easy installation of several racks’ base-plates.
Pushover tests have been performed by pulling the racks by a hydraulic jack. The applied load has
been designed to have a triangular pattern (Figure 55.a), with loads acting on each level, increasing
from the bottom to the top. Since the pulling force is generated by one hydraulic jack only, a specially
designed load balancer was required to distribute the force with such a pattern. This load balancer
(Figure 55.b) has been designed as a series of three double supported beams, placed in such a way
that reaction forces at their ends are proportionally increasing from the first support to the fourth.

a. Triangular load pattern b. Load balancer static scheme

Figure 55 Loading scheme and load balancer

Some pictures from the site showing the connection of the load balancer to the test frame, can be
seen in Figure 56.

57
 Figure 56 Load balancer, and test frame and assembly

The load is transferred from the jack to the structure through a steel strand. To monitor the effective
load in the strand, a load cell is installed between the end of the jack and the strand itself. This is
useful for a real-time monitoring of the applied load that, together with the monitoring of the
displacements, can offer a good indication of plasticization and collapse limits, so that the test can
be easily stopped in case of danger or imminent instability.
The displacements of relevant nodes are monitored and recorded, so that the load-deformation
curve, useful to investigate racks’ deformed shape and ductility properties, can be plotted. For this
purpose twelve potentiometric displacement transducers were used. A total of 8 transducers was
placed, two for each level of the rack, in order to monitor the longitudinal backside and front side
displacements (down-aisle direction), while 4 transducers were placed backside at the second and
fourth floor on the left and right side of the rack, to control the displacements in cross aisle direction.
This configuration was adopted to register the out-of-plane displacements due to torsional
deformations expected in braced racks. The general test layout is shown in the Figure 57.

Figure 57 General test layout

The jack is controlled by a remote console where the maximum value of pulling force at the end of
each loading step and the time requested to reach step duration are set. At the same time the start
button is pressed, control units of the transducers, load cell and strain gauge start to record data
and the photos start to be taken. When the rack’s global behavior becomes completely plastic, i.e.
the rack cannot be loaded any further, the test ends, and the structure is unloaded.
4.3 Test frames
Four different unbraced racks have been tested. All specimens are composed of the following
components:
 Upright frames: Composed of two uprights in perforated thin gauge cold formed profiles
linked together by a system of bracing elements.

 Pallet Beams: Members linking adjacent upright frames, lying in horizontal direction parallel
to the operating aisle (down aisle direction). They are made of closed built up sections
composed of two cold formed “C” elements, welded to end plate connectors hooking in the
holes of the uprights.
58
  Beam to upright connections: Beam to upright joint is a proprietary system consisting in a
hook-in connector welded to the Beam-ends, fitted into dedicated regularly spaced slotted
holes in the uprights.

 Base-plate connections: Base-plates are either L-shaped or welded U-profile components

that connect the upright bases to the foundation. The uprights are connected to the base-
plates, through slotted holes in the vertical part of the base-plate, using bolts. The horizontal
part of the base-plate is connected to the foundation, using bolts. Number and configuration
of bolts depend on designer's choice.

In addition to these components, braced racks also have:

 Vertical bracings: Bracings in the vertical plane in down aisle direction, linking adjacent
upright frames. Spine bracing is located on the rear side of the rack, for single entry racks,
and it is made of vertical uprights, transverse members (mainly compressed) and diagonal
members working in tension only.

 Horizontal bracings: Diagonal bracing elements placed between beams, which transfer the
horizontal forces from the unbraced vertical plane of upright to the braced vertical plane.

As technology for rack fabrication is proprietary, due to a nondisclosure agreement among the project
partners for confidentiality reasons, only the main characteristics of each test frame are explained in
Table 35 and Table 36 in Annex for chapter 4.
Figure 150 in Annex for chapter 4, layouts of all eight configurations can be seen.
4.4 Test results
In this section, test results are presented in terms of (base shear) vs (top displacement) diagrams,
and deformed shapes at significant loading levels. The graphs are obtained plotting readings of the
top displacement transducers vs the base shear measured by the load cell.
Due to the abovementioned nondisclosure agreement among the partners, the graphs have been
normalized as follows:
 Base shear force (F) has been divided by the global yield shear force (Fy) of the frame:

𝑭
= 𝑭∗ [Eq. 27]
𝑭𝒚

 Displacements (d) have been normalized on the yield displacement of the frame (d y):

𝒅
= 𝒅∗ [Eq. 28]
𝒅𝒚

Fy and dy are calculated from the global force-displacement graphs obtained from tests as explained
in detail in section 8.

59
4.4.1 Unbraced specimens
Adimensionalized force-displacement curves for all four unbraced tests are shown in Figure 58.

a. IPA1 b. IPB1

c. IPC1 d. IPD1
Figure 58 Force-displacement curves

 Global behaviour of specimens IPA1 and IPB1 was moderately ductile (Figure 58.a.b). These
racks did not have a sudden local failure, thanks to progressive lateral bending of their
uprights, and accumulation of plastic deformation in their upright bases and Beam-end
connectors (Figure 59).
 Both specimens IPA1 and IPB1 sustained base shear forces more than their yielding values;
nearly 30% more in case of IPA1, and 10% more in case of IPB1.
 Ultimate deformation of both specimens IPA1 and IPB1 was greater than their yield
displacement values; 5 times greater in case of IPA1, 2.5 times greater in case of IPB1 ( Figure
58.a.b).
 When specimens IPA1 and IPB1 were unloaded after tests, their residual permanent plastic
deformations were greater than the elastic one; 3.5 times greater in case of IPA1, 1.5 times
greater in case of IPB1.

a b c d
Figure 59 Residual permanent plastic deformations

 Specimens IPC1 and IPD1 both had global collapse due to soft storey mechanism formed at
their first levels, since their base-plate connections lost their initial stiffness and behaved like
hinges after reaching a certain level of horizontal load. This loss of stiffness caused the
formation of plastic hinges on the uprights and beam end connectors just below the first
level, and after this point, the specimens could not sustain any more horizontal load.
 Specimen IPD1 had a soft story mechanism because of a plastic hinge formed in its central
upright, and distortional buckling occurred in the lateral ones. Distortional buckling, also
known as “stiffener buckling” or “local-torsional buckling”, was caused by the rotation of the
flange at the flange/web junction, in members with edge stiffened elements. At the end of
60
 the test, this phenomena arose along the first and second level upright’s flange; in fact the
distortion in the uprights was favoured by the absence of rigid constraints that could have
been given by the bracing system joints in cross aisle direction (Figure 60.a.b).
 Specimen IPC1 had a soft story mechanism, due to the distortional buckling centralized at
its first level, while its upper stories remained undamaged with little drifts (Figure 60.c).
During the test, this phenomena arose locally on the upright’s flanges in direct contact with
beam-end connectors surface; in fact the progressive increase of deflection in the upright,
favoured by non-rigid constraints (base-plate behaving as a hinge), was not accompanied by
a consequent deformation of the Beam-end connectors. The resulting transmission of strong
localized pressures led to a distortion of upright section in correspondence with the joint, and
the subsequent formation of a soft story mechanism. Not all uprights were subject to this
phenomenon. Those, where the upright bracings converge in the beam-to-column
connection, did not present substantial deformation. Indeed the presence of the upright
bracing in the node prevented inwards rotation of the upright flange.

a b c
Figure 60 Deformation of unbraced specimens

4.4.2 Braced specimens

The force-displacement curves for all four braced tests are shown in Figure 61.

a. IPA2 b. IPB2

c. IPC2 d. IPD2
Figure 61 Force-displacement curves

 Among all four specimens, only IPA2 showed a ductile performance, without a sudden local
failure in its elements. IPB2 and IPD2 both have had local failures, after which they are
unloaded. IPC2 was a very resistant rack, its testing had to stop because the hydraulic jack
reached its maximum limit.

61
  IPA2 showed a ductile performance thanks to the exploitation of plastic resources of its
vertical bracings that are efficiently connected to the rack joints (Figure 61.a). Figure 61.a
shows that IPA2 was able to sustain a base shear nearly 15% greater than its yielding value,
and its ultimate deformation is almost 3,5 times greater than its yielding value. When this
specimen is unloaded, its residual permanent plastic deformation is equal to 2 times the
elastic one. From Figure 62.c, it can be also observed how the middle upright is stressed and
buckled locally, due to tension bracing forces.
 Specimen IPB2 suddenly failed due to the shear failure of one of its bolts that connected the
vertical tension bracing to the bracing base-plate at ground level. When this bolt failed, all
the rack members were still in their elastic phase. Figure 61.b also shows that global stiffness
of the rack had a sudden decrease, when the compression diagonals were buckled. After this
point, no plastic behavior occurred until the shear failure of the bolt, which did not permit
the rack to take any advantage of plastic resources of the diagonals (Figure 62.a).
 For specimen IPC2, test was stopped due to the capacity of hydraulic jack. Nevertheless,
design elastic limit of the structure was reached at the end of the test. Pushover curve has
almost an elastic branch until the end of test, which means the deformations of rack members
mainly remained elastic. Slight plastic global deformation can be observed from the
descendent branch of the curve which shows a small residual deformation (Figure 61c). It
should be noted that in the substructure tests performed within WP2, this bracing element
had a brittle rupture at its connection at higher loads.
 Specimen IPD2 had a sudden local collapse caused by a connection failure of its tension
bracing at the first level. This failure was caused by the bending of the bolt that was
connecting the first level bracing to the bracing joint at the first floor. Progressive plastic
deformation until failure, concentrated in the bolt and the connection plates caused a slight
global plastic performance, which can be seen from in Figure 61.d. This shows that, if the
connection were more efficient, the specimen could have a more ductile performance.

a. IPB – bracing connection failure b. IPA – bracing yield c. IPA – upright yield
Figure 62 Deformation of braced specimens

Thanks to data recorded by transducers and the photographic surveys edited by FOTOGRAM software
[18], it was possible to reproduce the rack’s deformed shapes at significant load steps. Deformed
shape of two monitored uprights at the last load step are compared in order to observe global
torsional behavior of the braced racks due to their asymmetric configuration. Figure 151 of Annex
for chapter 4 shows that in all the cases, a moderate global torsion was observed due to asymmetry
in plan. Though it was limited thanks to the horizontal bracings used at each level.
4.5 Calculation of q-factor
The resistance and energy-dissipation capacity of a structure are related to the extent to which its
nonlinear response is to be exploited. In operational terms, such balance between resistance and
energy-dissipation capacity is characterized by the values of the behavior factor q. As a limiting case,
for a non-dissipative structure, where no (or extremely limited) hysteretic energy dissipation is taken
into account, the expected behavior factor, in general, is not greater than value of 1,5. For dissipative
structures, the behavior factor is in general larger than this value, accounting for the hysteretic
energy dissipation that mainly occurs in particular zones, called dissipative zones or critical regions.
The definition of q-factor is strictly related to the assessment of the yield strength F y and the yield
displacement dy. There are several criteria to define the conventional values Fy and dy. According to
ECCS45 [17] three different definitions are recommended:

Definition A
From the recorded base shear – top displacement curve (F-d curve), the conventional limit of elastic
range Fy and the corresponding displacement dy may be deduced as follow (Figure 153.a of Annex
for chapter 4):
62
- evaluate the tangent at the origin of the F-d curve; it gives a tangent modulus:
E = tgαy [Eq. 29]
- locate the tangent that has a slope of:
E/10 [Eq. 30]
- the intersection of the two tangents defines level of Fy;
- dy is the displacement corresponding to that intersection.
Definition B
The values correspond to first yielding identified on the experimental response of the specimen
(Figure 153.b of Annex for chapter 4).
Definition C
The values correspond to deformation n-times the deformation which would have been obtained in
a purely elastic behavior. Figure 153.c of Annex for chapter 4 gives an example with 2-times the
elastic deformation.
In order to calculate the q-factor, the pushover curves have to be linearized. The q-factor is defined
as the product of the overstrength Ω, and the ductility ratio μ = q0.
Definition B is only suitable when the shape of the pushover curve has a long elastic branch
immediately followed by plasticization and collapse. Therefore, for the racks of this study, two
different methods have been suggested taking into account definition A and another one taking into
account definition C.

Method 1 (see Figure 154 of Annex for chapter 4):

𝑭𝒖 𝒅𝒖
𝛀= [Eq. 31] ; 𝒒𝟎 = [Eq. 32] ; 𝒒 = 𝛀 ∙ 𝒒𝟎 [Eq. 33]
𝑭𝒚 𝒅′

where:
Fy is obtained from the above mentioned definition A
Fu is the maximum reached load
du is the displacement associated to the maximum reached load
d’ is the displacement associated to the intersection between the tangent at the origin and the Fu
limit
Method 2 (see Figure 154 of Annex for chapter 4):

𝑭𝒖 𝒅𝒖
𝛀= [Eq. 34] ; 𝒒𝟎 = [Eq. 35] ; 𝒒 = 𝛀 ∙ 𝒒𝟎 [Eq. 36]
𝑭𝒚 𝒅𝒚

where:
Fy is obtained from the above mentioned definition A
Fu is the maximum reached load
du is the displacement associated to the maximum reached load
dy is the displacement obtained from the above mentioned definition A
Method 3 (Figure 154 of Annex for chapter 4):

𝑭𝒖 𝒅𝒖
𝛀= [Eq. 37] ; 𝒒𝟎 = [Eq. 38] ; 𝒒 = 𝛀 ∙ 𝒒𝟎 [Eq. 39]
𝑭𝒚 𝒅𝒚

where:
Fy is obtained from the above mentioned definition C
Fu is the maximum reached load
du is the displacement associated to the maximum reached load
dy is the displacement obtained from the above mentioned definition C
For the tested frames, the q values have been calculated according to three different methods
provided above. Figure 155 of Annex for chapter 4 shows the variables used in method 1 and 2.

63
Dashed lines represent the tangent modulus E in the origin and the tangent reduced modulus E/10,
while solid lines indicate the projections of the variables Fy, Fu, dy, d’ and du values.

Figure 156 of Annex for chapter 4 shows the variables used in method 3. Dashed lines represent the
tangent modulus E in the origin, half of the tangent modulus (E/2), while solid lines indicate the
projections of the variables Fy, Fu, dy, d’ and du values.

Table 16 summarizes the results obtained by the various method adopted, for the rack specimens of
this study.
For specimen IPA1, q varies from 3,48 to 4,92. This rack has a short initial elastic branch, then it
starts to show increasing plastic deformations before reaching the maximum force Fu. For specimen
IPB1, q varies from 1,90 to 2,38. The tested rack has a short initial elastic branch, then it starts to
show increasing plastic deformations before reaching the maximum force Fu. For specimens IPA1 and
IPB1, moderately high values of q indicate reasonable ductility of these structures, and the global
capacity of these racks to dissipate energy.
For specimen IPC1, q value has been obtained as 1,41-1,51. For specimen IPD1, it is between 1,24-
1,26. For specimens IPC1 and IPD1, low values of q indicate lack of global ductility of the structures.
In fact the pallet rack system does not develop a global diffused plasticization, but a local one. The
soft storey mechanism concentrates plasticization in the first level joints.
IPA1 IPB1 IPC1 IPD1
Ω q0 q Ω q0 q Ω q0 q Ω q0 q
METHOD 1,27 3,03 3,85 1,12 1,90 2,13 1,06 1,33 1,41 1,02 1,22 1,24
1
METHOD 1,27 3,87 4,92 1,12 2,13 2,38 1,06 1,42 1,51 1,02 1,22 1,26
2
METHOD 1,15 3,01 3,48 1,00 1,90 1,90 1,06 1,42 1,51 1,02 1,22 1,26
3
Table 16 Unbraced racks, q factors according to three different methods

Table 17 summarizes the results obtained by the various methods adopted, for the rack specimens
of this study.
For specimen IPA2, q varies from 2,25 to 2,83. Moderately high values of q indicate reasonable
ductility of this structure and the global capacity of this rack to dissipate energy. For specimens IPB2
and IPD2, sudden failure of vertical bracing connections did not allow the development of a diffused
plasticity mechanism, therefore, their q-factor can be considered equal to 1,00. For specimen IPC2,
the test was stopped because of jack’s capacity (100 kN), yet design elastic limit of the structure
was reached at the end of the test. Although a local failure was not observed, at the end of the test,
deformations of the structure remained mainly in the elastic phase; therefore q value of this specimen
could not have been determined. However it should also be noted that, in the substructure tests, the
cable-braced specimen behaved substantially elastic until the failure, which occurred by tearing of
the strap connecting the bracing cable to the eccentric column.
IPA2 IPB2 IPC2 IPD2
Ω q0 q Ω q0 q Ω q0 q Ω q0 q
METHOD 1,13 2,19 2,48 1,00 1,00 1,00 1,00 1,00 1,00
n.a. n.a. n.a.
1
METHOD 1,13 2,50 2,83 1,00 1,00 1,00 1,00 1,00 1,00
n.a. n.a. n.a.
2
METHOD 1,03 2,19 2,25 1,00 1,00 1,00 1,00 1,00 1,00
n.a. n.a. n.a.
3
Table 17 Braced racks, q factors according to three different methods – IPC2 is the cable braced rack

Table 16 indicates that the third method is the most conservative among three methods, because,
as shown in Figure 156 Annex for chapter 4, the yield strength Fy of third method results greater
than those of the other methods.
4.6 Conclusions
In this chapter, seismic behavior of steel storage pallet racks has been investigated, by means of full
scale pushover tests in down-aisle direction. Unbraced and braced rack specimens, provided by four
industrial rack industrial partners have been tested. For each full scale test, deformed shapes and
experimental capacity curves have been presented.
4.6.1 Unbraced racks
Global capacity curves indicate two major performance types: ductile or brittle. Global behaviour of
the specimens IPA1 and IPB1 was moderately ductile, thanks to the diffuse exploitation of plastic
resources in their connections, and their resistant base-plate connections. On the other side, global
behavior of the specimens IPC1 and IPD1 were brittle, because of the following reasons:
64
 i. Plastic hinges formed in the uprights and beam end connectors at the first loading level,
which led to the soft storey mechanism. Insufficient base connection details caused a “hinge”
behavior at the bases, and the rotation demands were concentrated at the “more rigid” first
floor beam-to-upright connections.
ii. Second reason is the distortional buckling phenomena, spread on the first two interstories of
the structures. Strong localized pressures led to a distortion of upright sections in
correspondence with the beam-to-upright joints of the first two levels, and this also caused
a soft story mechanism. The distortional buckling of the uprights occurring in correspondence
with the beam levels, could have been avoided by connecting the upright bracing members
to the joint locations corresponding to the pallet beam connectors, as shown in Figure 157
in Annex for chapter 4. Table 35 and Table 36 in Annex for chapter 4 also highlights the fact
that upright and beam sections are weaker in case of IPC1 and IPD1.
After the investigation of the global behavior, an estimation of q-factor is provided for the rack
specimens tested in the research program. For the structures characterized by a brittle collapse, the
q-factor value range from 1,24 to 1,41 while for the structures that showed a more ductile behavior,
q-factor can reach higher values.
4.6.2 Braced racks
Only specimen IPA2 showed a ductile performance. This ductility was provided by the vertical
bracings of the rack specimen, which could have large deformations without an early failure in their
connections. Specimen IPB2 had a bolt shear failure, and specimen IPD2 had a bolt bending failure
before taking advantage of the plastic resources of their spine bracings. For specimen IPC2, test was
stopped due to the capacity limits of the hydraulic jack, although design elastic limit of the structure
was reached at the end of the test. This specimen could slightly exploit the plastic resources of its
spine bracings until the end of the test. However it should be underlined that during the substructure
tests in WP2, the bracing of this specimen had a sudden rupture at its connection at higher loads.
In particular, structural response of the specimens is highly influenced by the connection details of
their spine bracing elements. Plastic resources of spine bracings could only be exploited if their
connections were resistant enough to not to cause an early brittle local failure.
After the investigation of the global behavior, an estimation of q-factor is provided for the rack
specimens tested in the research program. For the specimens characterized by a brittle collapse, the
q-factor value is considered as 1,00 while for the specimen that showed a more ductile behavior, q-
factor could reach higher values.
Due to the asymmetry in the plan of braced racks, a moderate global torsion was observed, though
it was limited thanks to the horizontal bracings used at each level.
In general, full scale tests underlined the importance of structural detailing, on the global seismic
behaviour of unbraced and braced pallet racking systems. The observed failure modes could have
been easily avoided at low cost, with better connection details.

65
66
5. NUMERICAL STUDIES
First, preliminary linear and nonlinear static analyses have been performed according to the prFEM
10.2.08 [25] recommendations, then these models have been calibrated with the results of
experimental studies on component and full scale tests. With the calibrated models, q-factors have
been calculated and parameter studies have been carried out to investigate the influence of
occupancy rate and merchandize properties.
5.1 Preliminary numerical analysis
Racks are mainly made of thin gauge profiles and continuously perforated uprights, which ensure the
typical functionality, adaptability and flexibility needed for the huge variability of requirements in
storing goods. Their particular geometry made of highly slender elements, and the non-linear
behavior of their beam-to-column and base-plate joints, are the main parameters that mostly
influence their behavior. Some conventional numerical methods of analyses are presented here,
giving the basic steps to accomplish an appropriate design procedure.
Each rack member has basic principles of modelling in a global numerical model. These can be listed
as:
Uprights: Released for bending cross-aisle direction, and partially released for bending in the down-
aisle direction. The value of the rotational stiffness of the partial release is governed by the base-
plates, which is determined experimentally.
Pallet Beams: Released for horizontal bending and partially released for vertical bending. Rotational
stiffness of the pallet beams at their ends are determined experimentally from the beam-end
connector tests.
Diagonals: Diagonal members are connected to the uprights usually by single-bolt, which creates
severe initial looseness, and hence a modification of the initial shear stiffness of the system, in the
cross aisle direction. To predict the initial shear stiffness of the upright frames, experiments have to
be performed, from which an equivalent area of diagonal members is derived to be used in the
numerical model.
Vertical Bracing: Vertical bracings are used to withstand earthquake loads. They are usually made
of either small-diameter sections or cables, which means that they actually work only under tension.
This fact has to be simulated in the numerical model, using only the tensioned bracing members.
The presence of the bracing system, which is placed eccentrically, creates an intensive torsional
behavior on the system, a fact that obliges the engineer to use 3D simulation, which is a requirement
of EC8 and prEN16681 [22], for non-regular systems. The bracing systems can be connected to the
main rack with different methods, depending on the manufacturer. Each different system has a
unique simulation and thus, it is not possible to give a universal modelling rule for the simulation of
this assembly.
Modal and response spectrum analysis (RSA) have been performed on two benchmark models
(braced and unbraced in down-aisle direction) which are shown in Figure 63.

a b
Figure 63 Benchmark numerical models for Typical a) Braced and b) Unbraced System

In particular unbraced frames can be quite flexible in down aisle direction, which leads engineers to
take into account 2nd order effects. To consider the 2nd order phenomena in modal analyses, and in
RSA, the actual stiffness matrix of the structure has to be used; this means that the stiffness matrix
has to be updated after the application of the vertical loads. Modal analysis considering 2nd order
effects can be performed in two steps:
i. Deformed shape of the rack with modified stiffness matrix is obtained from a nonlinear
elastic analysis performed with rack’s vertical loads (self-weight of the rack and the pallet-
loads),
ii. Modal analysis is carried out using the output of the nonlinear elastic analysis.

67
Table 18 shows that consideration of the second order effects in modal analysis, increases
significantly the 1st natural period of the unbraced rack.
Case Mode Period (sec) Mass participation X (%) Mass participation Y (%)

Without 2nd 1 2,99 86 0

order 3 1,04 0 65

With 2nd 1 3,90 88 0

order 5 1,06 0 65

Table 18 Modal Results for an unbraced rack with and without the 2 nd
order effects

From the RSA, it is seen that while base shear forces are quite similar, the top displacements are
very different (Table 19). This is because the models considering second order effects use the actual
reduced stiffness matrix under the presence of vertical forces. This makes the system more flexible
and lateral displacements become higher.
Case Fx (kN) Fy (kN) Top Displacement (mm)

Cross-aisle - 31,1 -

Down-aisle 13,1 - 73

Cross-aisle with 2
nd
order - 30,5 -

Down-aisle with 2 nd
order 12,9 - 112

Table 19 Base shear and top displacements for the examined unbraced rack

5.2 Numerical static pushover analysis

Structures under investigation are specific design examples of storage racks which are provided by
4 industrial partners (IP), separated into two main categories, as braced and unbraced in down aisle
direction. The braced racks are commonly designed for high seismic zones, while the unbraced racks
are designed for medium and low seismic zones. There are also some cases, where unbraced racks
are strong enough to deal with a strong earthquake.
Examples of two main rack configurations are presented in Figure 64. The cross aisle direction is
always braced in a varied way which also depends in the seismicity of the region where the rack is
placed. First one is known as X bracing and the second one as D bracing. The major difference
between the two models is not only the number of the used diagonals but also the symmetry of the
configuration.

Braced rack b. Unbraced rack c. seismic d. Normal

configuration configuration
Figure 64- Typical bracing configurations in down aisle and cross aisle

This type of analysis requires information mainly exported by experiments. Rack members develop
inelastic behavior during extreme loading cases, and they have to be assigned plastic hinges in order
to simulate their inelastic behavior. In fact the majority of rack members are not allowed to develop
plastic behaviour, as they are classified as class 3 or 4 sections in EN1993 [2], however, the term
plastic hinge is referred here, generally, to the local material and/or geometric nonlinearity. How
each member is assigned a specific plastic hinge property is presented hereafter.
Beam-end Connectors & Base-plates: Beam-end-connectors and base-plates are special
connections, whose capacity cannot be calculated theoretically. Thus, experimental results are used
to determine their plastic hinge- properties. For base-plates, a pay load very similar to the actual
one should be applied during the experiment, to ensure that the results are representative and
realistic. The fact that external uprights have not the same axial loads as internal uprights should be
considered in the base-plate tests.
For the preliminary numerical analysis, industrial partners have provided their own values of Beam-
end connector and base-plate stiffness and strength obtained in their laboratories (Figure 65). These
values have been used as input to the multi-linear plastic links in the preliminary numerical models.
68
These nonlinear properties are introduced as a multi-linear plastic link-element with a length equal
to the half upright width. The initial stiffness is used for the linear analyses of the model, while the
exact moment-rotation curve is used for nonlinear analyses.

Beam-end connector b. base-plate

Figure 65 Moment rotation curves of the different case studies

Uprights: They are subjected to a combined action of axial forces and biaxial bending moments.
Their plastic hinges are determined using the combined capacity of the member against lateral-
flexural buckling, and/ or the section’s capacity under axial force and biaxial bending. Engineers can
use the nominal values for the capacity of each member, according to EN1993 [2]. However, uprights
are thin-walled and perforated sections and it is considered always more accurate to use
experimental and/or numerical results.
Buckling failure is introduced as an equivalent plastic hinge at the most vulnerable point of the
column, which is close to the beam-to-column connection.
The inelastic capacity of the uprights could be defined according to Eq. 40, however an accurate FEM
analysis is required to provide the exact post failure behavior of these members, due to the existing
perforations of the cross section walls, the stiffeners etc.
𝑁𝐸𝑑 𝑘𝑦 𝑀𝑦,𝐸𝑑 𝑘𝑧 𝑀𝑧,𝐸𝑑
+ + <1 [Eq. 40]
𝜒𝑚𝑖𝑛 𝐴𝑒𝑓𝑓𝑓𝑦 /𝛾𝑀 𝜒𝐿𝑇 𝑊𝑒𝑓𝑓,𝑦 𝑓𝑦 /𝛾𝑀 𝑊𝑒𝑓𝑓,𝑧2𝑓𝑦 /𝛾𝛭

Moment rotation curve of the upright model is used as plastic hinge properties in the global numerical
model developed using SAP2000 [20]. Upright members are simulated by shell elements using
elastic-plastic material with the nominal yield strength of each member. Exact section-geometry
(both radius and perforations) have been taken into account. Figure 66 presents a typical numerical
model. All upright members of case studies have been modeled in this way.

Figure 66 Typical numerical model of upright

Figure 67 shows a multi-linear curve which is used in the global model to define the actual curve
of one of the uprights.

Figure 67 Moment rotation diagram of upright and multilinear diagram, defining the plastic hinge
properties

69
Diagonals: Diagonals are modelled with truss elements and their inelastic properties are introduced
as axial plastic hinges. A diagonal member can fail either due to buckling and/or yielding in tension,
or due to connection’s failure. The connection’s failure occurs when the forces on the bolted
connections exceed the bearing capacity of the connection and/ or the shear strength of the bolts.
The properties of these plastic hinges are selected in order to describe adequately the most critical
failure between the above failure mechanisms. Sometimes more than one failure mechanisms
coexist; as a result more complicated analysis in local level are required to well-describe the way
under which a diagonal may fail. Finally, depending on the manufacturer, there are details of the
connections that are not directly norm-described and numerical analyses with finite elements
software has to be used.
In order to decide the modelling parameters to be used in global analysis, first, a diagonal channel
member is simulated by shell elements investigating initially the member’s buckling resistance. The
load is applied on the web of the channel section producing a secondary bending moment due to the
real connection’s eccentricity. Analysis with imperfections is preferred; using the shape of the
buckling mode (by a linearized buckling analysis) and an imperfection magnitude according to
EN1993 [2].
Afterwards, the bolted connection simulation is presented. A contact model is used with solid
elements in order to be more adequate even though the time consumption and the model’s numerical
instability are some serious disadvantages. The local buckling expected, leads to a great amount of
solid elements, problem that can be solved using the half model -taking advantage of the symmetry.
The appropriate boundary conditions are entered to the half model, thus it is easier to increase the
number of elements around the connection’s hole. The load on the diagonal is applied via a
displacement control of the bolt. The model is loaded in both tension and compression. The results
show that, bearing failure appears before reaching the buckling capacity of member (Figure 68).

Figure 68 Diagrams for the buckling and the bearing failure of a channel section

The conclusion is that although the bearing resistance is highly ductile, thin walled members cannot
completely develop this ductility, because of the local buckling around the holes. Figure 69 indicates
the failure modes for the two different load directions (tension and compression). In all the cases,
local buckling of the web occurred just after the initial ovalization of the hole. The plastic hinge
properties that are used in the pushover analyses are produced by the diagram of Figure 69.

a) compression load b) tension load

Figure 69 Bearing failure due to a) compression and b) tension

The same advanced analyses were performed for another provided diagonal, made by circular hollow
section, whose ends are constructed with special turnbuckles (Figure 70). This model has been
developed to determine the member’s buckling load, and the connection’s strength. Buckling load
has been found without considering turnbuckles.

70
 Geometry Buckling of CHS section
Figure 70 Circular Hollow Section with turnbuckles used as diagonal members

Then, the whole model has been simulated, with two rigid pins at the holes of the turnbuckles. One
end was fixed and displacements are applied to the other hand. Figure 71 shows the buckling of the
turnbuckles, under compression load, and the simultaneous circular-hole ovalization.

a- Buckling at flatten ends b- Ovalization of pin hole

Figure 71 Local deformations of the Beam-ends

The failure mode under compression for the selected CHS member was due to bearing, global
buckling of the member and buckling of the flattened ends, almost simultaneously. Figure 72 depicts
the load-displacement curves under compression for the member without turnbuckles and for the
whole member with endplates (turnbuckles). This second curve which is more critical, is introduced
in SAP2000 [20], as plastic hinge properties.

Figure 72 Load displacement load for simulation of only the CHS member and that of the whole assembly

Applied load during a pushover analysis is a significant parameter for the final results. There are
different theories and assumptions for the applied loads, such as a uniform load over the height, a
triangular- distributed load, and a mode-like distributed load. The first one is not always
representative, especially for unbraced systems, where the 1 st dynamic mode is the dominant. In
these cases, the triangular distribution of the load is preferred but there is always the option for a
distributed load according to the shape of a pre-defined dynamic mode. Depending on the system,
engineers can select the most appropriate load to perform a pushover analysis. On the contrary, if
higher modes could play significant role on the earthquake loading, one has to use a multimodal
pushover analysis, in which the applied load is described by more than one dynamic mode. Finally,
triangular distribution has been selected for both the braced and the unbraced racks in order to
produce results comparable to those of the full scale tests performed during WP4.
P-delta effects (2nd order analysis) have a significant influence on the overall behavior and design of
structures. 2nd order analysis is always more accurate than the 1 st order, although it can be
complicated to apply in any case. Eurocode-3 indicates when it is considered necessary to run a
second order analysis taking into consideration the P- delta effects.
First, a buckling analysis of the global structure has to be performed, to see if it is worthwhile to run
a second order analysis. If the buckling factor is greater than 10, the 2nd order effect may be

71
neglected. Figure 73 presents and compares the results derived from two different models with and
without 2nd order effects in each model. The two models present different extreme values of the
buckling factor αcr, which is defined as:
𝐹𝑐𝑟
𝑎𝑐𝑟 = [Eq. 41]
𝐹𝐸𝑑

Where: Fcr is the critical buckling load in the seismic load situation, for the 1st buckling mode and FEd
the actual vertical load of the structure.

Braced system: Buckling Factor αcr=13 Unbraced system: Buckling Factor αcr=4.33
Figure 73 Static Analyses with and without P-Delta Effect assumption

The first diagram depicts the results for a braced system, while the second one for an unbraced one,
both for the down aisle direction, where the systems are usually more flexible. The results prove the
need of taking into account the P-Delta effects in flexible systems and the fact that there are always
severe differences even if the system is stiff enough. It should be noticed that all the following
analyses take into consideration P-Delta phenomena, either it is norm-obligatory or not.
5.3 Pushover analysis of 3D racks
Nine different case studies of the various industrial partners have been simulated under triangular
horizontal loading over the height. The configurations are considered fully loaded with pallet loads,
the inelastic properties of the system are defined via plastic hinges or nonlinear link elements, as
described in previous paragraphs. The pushover curves are superimposed in order to compare
quantitatively the different configurations. There are two main groups of curves; those for the down
aisle direction and those for the cross aisle direction. Figure 74 presents the results for the nine cases
on both down and cross aisle direction.

Down-aisle Cross-aisle
Figure 74 Pushover curves for the 9 case studies for the down aisle and cross aisle direction

One can easily observe that the high seismic zone racks are quite stiffer than those designed for low
seismic zones due to the fact that the former have bracing systems on their rear plane.
5.4 Post-testing calibration of numerical models
Scope of this task is to calibrate the numerical models of previous tasks with the provided
experimental data from other work packages, in order to verify the numerical models and numerical
results. Calibration has been made for cross-aisle and down-aisle frames separately.
Behavior of down-aisle frames is mainly characterized by the behaviour of the beam-end-connectors,
base-plates, vertical bracings and the uprights. In the global model, nonlinear links have been used
to simulate the local behavior of these components. These links have been calibrated by component
test results (for Beam-end connectors and base-plates), and by refined numerical modelling (for
uprights and vertical bracings).

72
Cross aisle direction is mainly characterized by the behavior of the diagonal members, uprights and
shear stiffness of the upright frames. The calibration of the cross aisle models is completed using
either experimental or numerical data.
For the calibrations, rack configurations tested in work packages 2 and 4 have been simulated
considering the same support and loading conditions, applied constraints and the sections’ properties.
The software Opensees [19] was used for the numerical calibration.
5.4.1 Calibration of Beam-end connectors in down aisle direction
Beam-end connectors are hooked-in semi-rigid connections, which are simulated with rotational
springs. The springs are governed by a cyclic non-linear moment-rotation law that includes a
hysteretic behavior with degradation and pinching phenomena. Hooks need severe tolerance in the
corresponding holes in order to be easy-installed. This tolerance makes the stiffness gradually zero
when the assembly is almost unloaded, creating the pinching phenomenon. The buckling and/or the
bearing of the hooks and the perforations create degradation and intensify the pinching as the holes
become bigger and bigger.
From the component tests in down aisle direction, stiffness and strength of each connector have
been obtained. Although there are sufficient monotonic experimental results, only the cyclic results
have been used, as they provide more information for the dynamic analysis which will be performed.
Figure 75 presents the results of the calibration procedure for four different tested configurations,
matching the diagrams produced numerically to those produced experimentally. The calibration took
place selecting the appropriate parameters for the hysteretic model included in Opensees [19].

IP-A IP-B

IP-C IP-D
Figure 75 Moment rotation diagrams

5.4.2 Calibration of Base-Plate behavior in down aisle direction

Base-plate tests were performed at the Aachen University which provided test results for 3 different
loading conditions per assembly: unloaded, half-loaded and fully loaded (for seismic load condition).
Fully loaded case was selected for calibration as it was considered to be the most representative one
for the numerical analyses.
The numerical model consists of a single beam element which uses a rotational non-linear spring at
the base-plate point and which is pinned on the top. Figure 76 presents results of the calibration
procedure.

73
 IP-A IP-B

IP-C IP-D
Figure 76 Experimental configuration for the characterization of the base-plates

5.4.3 Calibration of upright’s behavior

Since upright tests were not foreseen during the SEISRACKS2 project, their response to cyclic loading
has been determined by numerical tools considering material and geometrical nonlinearities. Abaqus
[21] software is used to simulate a single upright member of 1m, with shell elements and with the
exact geometry provided by each IP. Behaviour along major and minor axis were investigated
separately. First, a non-linear static analysis has been performed under an axial load of 48 KN in
order to define the yield displacement of each examined upright for each direction. Afterwards, a
cyclic analysis followed applying a loading protocol, which is ±δ y, ±2δy, ±3δy, ±4δy. The resulted
moment-rotation diagrams of each member have been derived in order to be used in the global
numerical models.
For global analysis, Opensees [19] requires moment-curvature diagrams as plastic hinge properties.
Therefore, moment rotation diagrams obtained for local upright models have been converted to the
moment- curvature ones by means of Eq. 42.
𝑙
𝜃𝑝 = ∫0 𝑦[𝜑(𝜒) − 𝜑𝑦 ] 𝑑𝑥 [Eq. 42]

Where, θp is the plastic hinge rotation on each side, l y is the length of the beam over which the
bending moment is larger than the yielding moment, φ(x) is the curvature at a distance x from the
critical section and φy is the yielding curvature. Using Eq. 42 one can easily calculate the relation
between the rotation and the curvature of a critical section of a plastic hinge. Eq. 43 simplifies the
above relation.
𝜃𝑝 = (𝜑𝑢 − 𝜑𝑦 )𝑙𝑝 = 𝜑𝑝 𝑙𝑝 [Eq. 43]

Where, φu and φy are the curvatures at the ultimate load and yielding, respectively and φ p is the
plastic hinge curvature.
The plastic hinge length lp is derived from the finite element software, measuring the length of the
member, where the von Mises stress was higher than the nominal yield stress.
Figure 77 shows a typical upright member after failure at the major and minor axis respectively,
exported from Abaqus [21]. It should be mentioned that the plastic hinge is formed due to the local
buckling of the section and not due to the direct exceedance of the yield stress.
Annex for chapter 5 includes all the resulted moment-rotation diagrams from the previously
described Abaqus [21]-analyses compared to the corresponding Opensees-results. In Opensees, the
moment-curvature diagrams are introduced, with the use of Eq. 43.

74
 Bending Bending
Figure 77 Failure mode of upright for bending around the major axis

Figure 78 Boundary condition of upright model

Although the uprights suffer mainly from bending moments and axial forces, the IP-B configurations
present high values of local shear forces, due to the eccentricity between the consecutive diagonals
(see on the circle of Figure 79). The component tests showed upright’s failure due to high shear
forces, only in the cross aisle direction. After appropriate numerical simulation, the deformed member
can be seen in Figure 79b., and the corresponding resulted diagram in Figure 79c. In Opensees [19]
a linearized diagram is introduced in order to define the plastic hinge properties of the upright’s
section.

a. Under shear b. eccentricity c. Shear force-transveral deformation

Figure 79 Upright frame diagonal assembly for IP-B, with eccentricity between consecutive diagonals

5.5 Calibration of cross-aisle frames

Seven cross-aisle configurations, those included in the case-studies of task 1.3, were tested in the
University of Liege, to identify their strength and stiffness characteristics. Figure 80 shows the
numerical and the experimental results, after the calibration of typical numerical models. The results
are dimensionless in terms of the Vy and dy which are the “yield” numerical values for the base shear
and the top displacement, respectively. In the numerical models, upright and diagonals’ nonlinear
properties have been considered based on their refined numerical models.
In order to calibrate the shear stiffness, an equivalent cross-sectional area is used for the diagonal
members. The nominal characteristics of diagonals with an initial axial stiffness EA/L would
overestimate the shear stiffness of the upright frame. Thus, an equivalent section’s area is used in
order to fit the experimental results, however, the ultimate diagonals’ nominal strength remain as it
is.

75
 IPA IPB

IPC high seismicity IPC medium seismicity

IPD high seismicity IPD medium seismicity

IPD low seismicity

Figure 80 Base shear – Top displacement diagram of cross aisle frames

5.6 Calibration of Full-Scale racks in down-aisle direction

Numerical models have been developed and calibrated for the 4 unbraced configurations and two
braced configurations. The braced configuration of IP-C does not provide nonlinear information
therefore it is not examined, while the braced configuration of IP-B is not examined as this is not
included in the numerical models of task 1.3 and there is no correspondence to the numerical models
of task 5.1. The initial numerical models presented severe differences with the experimental
configurations. Although the geometry and the section’s properties have been calibrated, the results
were qualitatively and quantitatively different. Modification of the materials and section’s properties
have been done in order to fit the numerical results to the experimental results. As the full-scale
tests showed, the upright members and/or the base-plates failed prematurely, thus their properties
were those that majorly were modified for the final full scale calibration. Figure 81 presents two
examples of the push over curves of the calibrated numerical models and full scale tests related to
IPA structures, performed within WP4. IPA experimental curves have been chosen because of their
ductile behavior and little calibration needed.

76
 IPA unbraced IPA braced
Figure 81 Total horizontal force vs. top displacement for rack configurations

The general conclusion from the calibration procedure is that the nonlinear static behavior of the
racks was quite impossible to be well-predicted by numerical models. To fit the experimental results,
plastic hinge properties representing the uprights’ nonlinear behavior had to be modified. The fact
that, there were neither experimental results for isolated upright members, nor material tests, lead
us to consider the uprights’ data the most unreliable. Therefore the calibrated numerical models of
racking systems have a significantly different behavior than the preliminary models.
5.7 Effect of pallet loads on global behaviour of racks
Cross-aisle frame-tests of WP2 had no gravity loading, which is not similar to the real situation.
Therefore to investigate the effect of pallet (gravity) loads on the ductility of cross-aisle frames,
nonlinear static analysis have been performed on the calibrated numerical models. The results in
terms of base shear vs. top displacement are presented indicatively in Figure 82 for two
configurations, in order to focus on the differences that are created with and without the vertical
loads. Since the axial forces produced by the gravity loads reduce the stiffness of the matrix of
uprights, and introduce intensive nonlinear phenomena due to the P-delta effects, global stiffness of
cross aisle frames is much lower with gravity loading.

a. IP-B- High Seismicity b. IP-D-High Seismicity

Figure 82 Base Shear- Top displacement diagram

Following the procedure described in task 5.1 for the calculation of the q factor, the diagrams above,
give the values presented in Table 20. When this table is compared with the table 7 (where the
experimental values of q factors without vertical loading are shown), it can be seen that vertical
loading has a negative influence on the ductility of the frames.
q factor Bracing type

IP A Medium/High 1.1 X

IP B Medium/High 1.8 D

High 2.2 X
IP C
Low 1.1 D

High 1.2 D

IPD Medium 1.0 X

Low 1.3 X

Table 20 Estimated q factor values for the fully loaded upright bracing systems of each IP

5.8 Incremental Dynamic Analyses (IDA)

The seismic behavior of racking systems in down aisle and cross aisle direction has been investigated
by means of incremental dynamic analysis (IDA) with the statistic evaluation procedures of FEMA
P695 [24]. Two unbraced down-aisle frames, and two cross-aisle frames have been considered.
Numerical analyses (6 bays-4 levels) of racks are based on the calibrated numerical models (2 bays,
4 levels). The reason of examining only these models was because the other two unbraced racks of
the project did not present experimentally any ductility, and hence it is not useful to investigate
statistically their seismic behavior. On the other hand, the braced models present the need of 3D
77
models, which are computationally problematic, as the IDA includes the run of thousands of 3D
dynamic analyses and a huge amount of results to be edited. In the cross aisle direction the 2 main
types of bracing systems are examined; the X and the D bracing type. For IDA, 44 real ground
motions have been used (Figure 83). Behaviour factors of each system have been estimated.

Figure 83 Elastic Spectra for the 44 ground motions (coloured) and some indicative EC8 design spectra (black)

Figure 84 presents the fractile curves and fragility curve resulted from IPA rack in the down aisle
direction. A fractile curve presents the interstorey drift (Damage Measure) which leads to a specific
probability of collapse (given as legend) for a given level of spectral acceleration. The fragility curve
expresses the cumulative probability of the system’s collapse for any given level of spectral
acceleration. The symbols of the fragility curves are the numerical results and the continuous red
line is the curve of an analytical fitted lognormal distribution.

IPA Fractile curve IPA Fragility curve

IPB Fractile curve IPB Fragility curve

Figure 84 Fractile curves and Fragility curve for the unbraced rack of IP A in the down aisle

Using the procedure shown in the flow chart above, more specifically in the FEMA P695 [24], for each
system the collapse margin ratio (CMR) is defined as:
CMR=SCT/SMT [Eq. 42]
SCT is the spectral acceleration which drives to a probability of collapse 50%,
SMT=1.5Sd1/T1 [Eq. 43]
where T1 is the fundamental period, and Sd1 is the elastic acceleration for 1sec in the elastic design
spectrum. Additionally, the ACMR=SSF*CMR is defined, where SSF is a tabulated parameter
depending on the time period and the ductility of the structure (FEMA P695 [24], table 7-1). Then,
the ACMR20% is defined from table 7-3 of FEMA P695 [24], in order to be compared with the ACMR
which was defined previously. In the case that ACMR>ACMR20%, the performance criterion is fulfilled
and the q factor of the system is evaluated.
Since, racking systems have no direct limitations at the design stage for non-structural elements,
there are no direct limits to the interstorey drifts. Thus, considering a higher PGA (proportionally a
higher Sd1) and a proportionally higher q factor, the design procedure leads to the initial design
acceleration and as a result to the initial structural design of the structure, e.g. the pairs PGA=0.1g

78
& q=1.5 and PGA=0.2g & q=3 gives the same spectral acceleration. Since the design spectrum has
been modified due to the new values of PGA, the parameters of the previous paragraph (except for
the SCT) have to be redefined. The maximum values of the pair PGA/q factor, for which the
performance criterion is fulfilled, drives to the maximum available q factor values of the system;
these values are presented in Table 21 for the two unbraced models.
IP ag (g) max. q qdesign

A-Medium 0.12 3 2

B-Low 0.113 5 1.5

Table 21 maximum estimated q factor for the IP A & B for the down aisle direction

A similar procedure is followed for the cross aisle direction, for the unbraced configurations of IP A
(X bracing type) and IP D-Low (D bracing type). Estimated q factors in comparison to the initial
design values are presented to Table 22. Figure 85 presents the results for IPA and IPD.

a. IPA Fractile curves b. IPA Fragility curve

c. IPD Fractile curves d. IPD Fragility curve

Figure 85 Fractile curves and Fragility curve for cross-aisle frames

IP ag (g) max. q qdesign

A-Medium 0.12 1.5 1.5

D-Low 0.113 2 1.5

Table 22 maximum estimated q factor for the IP A & D for the cross aisle direction

The general conclusion from these numerical analyses is that pushover analyses seems to
underestimate the actual ductility of the system, while IDA presents higher q values. However the
IDA procedure needs a lot of different archetypes in order to derive an overall quantitative conclusion.
IDA lead to higher ductility values for down aisle unbraced frames, than the ones proposed in the
design codes. Though, it would not be safe to generalize this result to all systems, remembering that
other two unbraced racks tested in research had almost zero ductility. The general discrepancy of
the results show that, to estimate behaviour factor of rack structures, detailed analysis and tests
should be performed, because every industrial partner has his own particular approach to design
racks, which have a lot of particularities that can hardly be estimated in simple design methods.
5.9 Pallet-merchandize interaction
The local behavior of the racking systems is examined regarding the forces applied on the pallet
beams, the sliding phenomenon and the buckling of the pallet beams.
In the first paragraph the influence of a pallet on a beam is examined with the intention to determine
the maximum imposed forces developed on a beam before, during and after the sliding event. An
analytical method is presented, whose results are compared to numerical results, derived from
Abaqus [21] software. In the second paragraph, more sophisticated numerical analyses are
presented, in a preliminary stage, in order to investigate the influence of the pallets on the actual
bending moment diagram of a pallet beam in comparison to the theoretical one. In the third part of
this task the buckling length and buckling load of an axially loaded pallet beam are investigated
parametrically taking into account the number and the relative position of the pallets with respect to
the number of pallets and their position on the pallet beams. Finally, a capacity design is proposed
for the pallet beams, against the seismic forces/earthquake loading.

79
5.9.1 Maximum applied forces over a pallet beam during the pallet’s sliding
In order to derive a relation for the maximum horizontal forces developed on a pallet beam during
the sliding of the pallets, 1) the eccentricity of the mass, 2) the distance between the pallet beams,
3) the friction coefficient and 4) the pallet’s weight are taken into account. Norm-provisions do not
take into consideration some of these parameters.
Assuming a pallet of weight V and the fact that it is supported symmetrically by two pallet beams
(Figure 86), the vertical reaction on each pallet beam is V/2. If an extra horizontal load, indicated
with H, is applied on the centroid of the pallet, a reaction H1 is developed to the left beam (front
beam) and a reaction H2 to the right beam (rear beam).

Figure 86 Typical pallet configuration

Assuming that the horizontal load is limited by the friction force Ti, Hi should be less than Ti:
Hi<Niμ [Eq. 44]
Where, Hi is the applied horizontal load, Ni the total vertical reaction, Ti the ultimate friction force
and μ is the friction coefficient.
Assume that the distance between the center of mass and the base of the pallet is e. This eccentricity
creates an overturn moment, which leads subsequently to a pair of vertical forces on the pallet
beams; indicated with P1 and P2 for the front and the rear beam respectively.
The moment equilibrium around the front pallet beam gives:
𝐻𝑒
𝐻𝑒 = 𝑃2 𝑏 → 𝑃2 = = 𝑃1 [Eq. 45]
𝑏

Since, H=H1+H2, N1=V/2+P1, N2=V/2-P2 and the fact that before sliding H1= H2;
it must be:
𝑉 𝐻𝑒 𝑉 𝐻𝑒
𝐻1 ≤ ( + )𝜇 [Eq. 46-a], 𝐻2 ≤ ( − )𝜇 [Eq. 46-b]
2 𝑏 2 𝑏

Since N2<N1, by definition, the friction force on the rear beam is also lower and the sliding occurs
firstly in this beam, (to which the N2 refers). Thus, the equation in relation 3b gives:
𝛨 𝑉 𝐻𝑒 𝐻 𝑉𝑏𝜇 𝑉𝑏𝜇
𝐻2 = = 𝜇− 𝜇 → 𝐻𝑏 = 𝑉𝑏𝜇 − 2𝐻𝑒𝜇 𝐻(𝑏 + 2𝑒𝜇) = 𝑉𝑏𝜇 → =  𝛨2 = 2(𝑏+2𝑒𝜇) [Eq. 47]
2 2 𝑏 2 2(𝑏+2𝑒𝜇)

Using the Coulomb’s theorem about friction, the maximum friction force remains constant during
sliding. Thus, when sliding starts, the H2 reaction is given by Eq. 47. Although the object slides over
the rear beam, the front one absorbs exclusively the additional horizontal force, due to the higher
vertical reaction. Hence, assuming that H=H1+H2 and that H2 is given by Eq. 47, the equality of Eq.
46-b, gives:
𝑉 (𝐻1 +𝛨2 )𝑒
𝐻1 = 𝜇 + 𝜇 → 2𝐻1 𝑏 = 𝑉𝑏𝜇 + 2𝐻1 𝑒𝜇 + 2𝐻2 𝑒𝜇 [Eq. 48]
2 𝑏

Replacing H2 and solving for H1 it is:

𝑉𝑏𝜇 2𝑏+5𝑒𝜇
𝐻1 = [Eq. 49]
4(𝑏−𝑒𝜇) (𝑏+2𝑒𝜇)

Eq. 47 and Eq. 49 are the theoretical upper limits for the slide forces at the two beams, as a function
of the parameters examined.
Furthermore, another limit state is the overturn of the pallet. There is always the possibility of the
pallet’s overturning which in theory occurs when one of the two vertical reactions tends to zero. The
reaction of the rear beam is the most critical for overturning, which translates to:
𝑉 𝐻𝑒 𝑉 𝑉𝑏
𝑃2 < → < →𝐻< [Eq. 50]
2 𝑏 2 2𝑒

80
The corresponding horizontal design force, given by EN15512:2009 [3], is
𝐻1,2 = 𝑄𝑃,𝑚𝑎𝑥 𝐶𝜇𝐻 𝜇𝑠 [Eq. 51]

Where, QP,max is the pallet’s weight, μS the friction coefficient and Cμ,H a safety factor for the friction
coefficient.
In the design procedure (Eq. 51), the two beams are examined for the same horizontal force (H/2),
and as a result the force of the front beam is underestimated and that of the rear one is
overestimated. High values of eccentricity show that the normative document highly underestimates
the maximum imposed force of the front beam, even if the total horizontal load is well-predicted.
Finally, to avoid rocking of a pallet before its sliding, Eq. 50 gives a geometric restriction for the
eccentricity e, considering that H/V=μ:
e<b/2μ [Eq. 52]

5.10 Application example

A numerical study is also presented in order to evaluate the analytical results. Two independent
beams are simulated with shell elements in Abaqus [21], and a rigid pallet is supported on these
beams. The pallet beams are supported at Beam-ends with fixed connections in the vertical plane
and pinned connections in the horizontal plane. Figure 87 shows the model, in which the pallet’s
weight and the horizontal load are applied to the centroid. Five different centroid’s positions are
examined in order to clarify the influence of the eccentricity on the system. The reactions of each
beam are presented as numerical results in Figure 88 for 3 different cases. The horizontal reaction
of the front beam is referred as Numerical H1 and the corresponding for the rear beam, as Numerical
H2. The theoretical results for H1 and H2, given in previous paragraph, are referred as Eq. 47 and
Eq. 49, respectively.

Figure 87 Numerical model with a rigid pallet on the middle span

In this case-study, the used friction coefficient is μ=0.375 and the pallet-weight is V=800kg. The
distance between the pallet-beams is b=1.1m and the cross-section of the beams is a hollow
rectangular one with dimensions 120X40X2 mm, which is simulated with shell elements. The rigid
pallet has typical dimensions 800X1100X1500 mm and is placed on the middle of the pallet-beams.
The pallet’s center of mass is placed in a height ranging from 0 to 1.46m, in order to examine the
influence of the eccentricity e (Case 1). The maximum eccentricity is calculated equal to 1.466m
according to eq.13. In order to investigate the influence of the beam-stiffness on the problem, the
same model was examined using a very stiff pallet beam of a hollow rectangular section 120X40X10
mm (Case 2). Finally, the pallet is placed at the right side of the pallet beams (Case 3) in order to
examine the influence of the pallet’s position on the ultimate horizontal force.

On the middle of the beams 0n the middle of the rigid beams At the extremity of the beam
Figure 88 Mass eccentricity vs. Sliding force, for different cases

Summarizing, Eq. 47 and Eq. 49 give results very similar to the numerical ones. Thus, they are
considered sufficient for designing a pallet beam while taking into account several parameters of the
system. More specifically, the width b, the eccentricity of the center of mass e, the weight of the
pallet V and the friction coefficient μ are introduced. The beam stiffness and the pallet’s position are

81
not taken into account in Eq. 47 and eq. Eq. 49, however, their influence has been numerically
investigated and was not found significant.
5.11 Correction coefficients for horizontal bending
The objective of this paragraph is to investigate the interaction between pallets and pallet beams
regarding the bending moment diagrams of the latter. The motivation for this investigation is the
code-provisions that propose correction coefficients for horizontal bending in respect of the number
and the position of pallets on the pallet beams. These coefficients are used in the design procedure
of the cross aisle direction and they are presented in Table 23.

Table 23 Correction coefficients for horizontal bending, proposed in FEM10.02.8

It is considered more reliable to simplify the problem as much as possible. Parameters such as the
eccentricity of the support conditions, the eccentricity of the horizontal loads on the pallet beams
(applied on the upper flange), the flexibility of the pallet beams and/or the pallets etc. are not taken
into consideration.
The simplified problem consists of 2 independent pallet beams and 1, 2 or 3 pallets. The pallet beams
are considered rigid in the vertical plane, however they retain their actual flexibility out of plane; the
pallets are considered rigid in the transverse plane, but flexible in the longitudinal plane. Figure 89
presents clearly the actual behavior of the pallet, and graphically the entire system of the model in
Abaqus [21] software.

a. Deformed shape of a pallet, as simulated b. Simulated configuration of pallets and pallet beams
Figure 89 Model

Simulation includes beam elements for the pallet beams assigned with a general section with user
defined properties. More specifically, the moment of inertia around the major axis is 10 24 mm4,
(numerically rigid), while the moment of inertia around the minor axis is 306920 mm 4, which is the
actual value for an RHS 120X50X2. The pallets are simulated only as three rigid surfaces in contact
with the pallet beams. Each surface is tied to a given point (control point) which lies on the level of
the pallets' center of mass (CM) thus taking into consideration the mass-eccentricity. These control
points, and consequently the respective three surfaces, are connected to each other with a
“connector” element, which is rigid in the longitudinal and transverse directions but pinned in the
vertical direction. This simulation was selected in order to ensure that all surfaces of the pallet will
be always in contact with the pallet beam. Furthermore, the pallet beams are fixed in the vertical
plane and pinned in the horizontal plane.
The loads of the system are applied directly to the control points. Two different steps are used
performing the analyses; the first one to apply the actual vertical load (1/4+1/2+1/4) and the second
one to apply gradually the horizontal load until the sliding of pallets. These analyses are used to
extract the bending moment diagrams of the pallet beams when the horizontal load is applied. The
objective is to compare the numerical results to the theoretical ones. It should be noticed that each
analysis is elastic and geometrically nonlinear, using contact- friction elements for the interaction
between pallet and pallet beams.
Although, the problem has been simplified, several parameters are taken into account. More
specifically, the friction coefficient between the pallet and the pallet beam, the eccentricity of the CM,

82
the pallet’s nominal weight (Qp) and finally the position and the number of pallets. In order to
investigate the influence of the previous five parameters, the following scenarios are examined
(Figure 90) regarding the position and the number of the pallets.

Figure 90 All the possible load cases of the pallet beams (boxes represent the pallets)

The above scenarios summarize all the possible position combinations of the pallets with a given
beams' occupancy three pallets. The assumed values for the friction coefficient μ are 0.1, 0.3 and
0.5, for the pallet’s weight is 4, 8 and 12 kN and finally for the mass-eccentricity are 0.35, 0.625
and 0.9m. Each case from A to E (Figure 90) have been examined for combination of the parameters
given in Table 24.
Pallet Weight Qp Eccentricity of mass e
Case Friction coefficient μ
(kN) (m)

4 0.1 8 0.9

5 0.1 8 0.625

6 0.1 8 0.35

11 0.3 4 0.625

13 0.3 8 0.9

14 0.3 8 0.625

15 0.3 8 0.35

17 0.3 12 0.625

22 0.5 8 0.9

23 0.5 8 0.625

24 0.5 8 0.35

Table 24 Possible cases, with respect to the different values of the examined parameters

In total 55 different analyses were performed, ensuring that the influence of each parameter is
examined at least once, while the other parameters remain constant. The results from these analyses
are compared to the theoretical ones. The maximum values of each bending moment diagram
obtained numerically (labeled as M), and the respective theoretical values (labeled as M*) are given
in Table 25. Both M and M* are estimated for each step of the nonlinear analyses. L is the length of
the pallet beams (L=2700 mm) and q is the equivalent distributed load applied horizontally in the
pallet beams.
Case A B C D E
M* 25/648 qL2 5/72 qL2 1/18 qL2 8/81 qL2 1/8 qL2

Table 25 Theoretical values for the horizontal bending moments, for each case from A to E

5.12 Results
In principle, results do not verify the Correction Coefficients for Horizontal Bending (CCHB) as
proposed by the normative documents. Since, the earthquake loading has an alternate direction,
only the maximum values among front and rear beams are examined. The friction of the pallets
seems to have a positive impact on the bending moment diagrams at the first steps of the analyses
when CCHB< 1. However this is reversed after the first sliding, when the CCHB approximates or even
exceeds 1. This is due to the fact that the theoretical bending moments M* are calculated for a
horizontal force H/2, while the actual load (see paragraph 2) for the front beam is higher.
The analyses’ results are presented in Figure 91 and Figure 92 with dimensionless axes M/M* and
H/Qp, where H is the total applied horizontal load. The ratio H/Qp is the friction coefficient μ, and
the ratio M/M* is actually the Correction Coefficient for Horizontal Bending. The diagrams present
the results step by step, initiating from zero horizontal force until the end of analysis (sliding of the
pallets). All the examined cases are analytically presented in the report of the task 5.4.

83
 Figure 91 Results for case 15 Figure 92 Results for case 14

5.13 Buckling length modification factor

In this topic, the buckling length of the pallet beams is examined, on the horizontal plane. The
existence of the pallets could provide a partial diaphragm action on the system and as a result reduce
the beam’s buckling length. The same model as developed for the CCHB is also used to investigate
this problem. The pallet beams are rigid in the vertical plane and flexible in the horizontal plane. The
pallets are simulated in such a way that allows a total contact between pallet and pallet beams, as
described in paragraph 3.
The parameters that are taken into consideration in this paragraph are the friction coefficient, the
pallet’s weight, the position and the number of pallets; the eccentricity of the pallet’s mass is
considered constant at height of 625mm. The different values of the friction coefficient are 0.1, 0.3,
0.5 and 100 representing a theoretical case in which the pallets are fixed on the pallet beams. The
different values of the pallet’s weight Qp is 800kg and 1200kg. Finally, the different cases regarding
the position and the number of the pallets are shown schematically below, from A to E (Figure 93).

Figure 93 Different cases regarding the position and the number of the pallets

The analyses are performed in 3 steps. In the first step, the pallet beams are subjected to a horizontal
uniformly distributed load, in order to create an initial deflection of a bow-like shape, thus introducing
an initial imperfection on the pallet beams. The magnitude of the initial imperfection is provided from
EN1993 [2], for buckling curve C, and it is 11.5 mm. In the second step, the vertical loads are applied
to the pallets in order to create the reactions on the pallet beams and to activate the friction-contact
elements on the model. In the third step, the pallet beams are subjected to a compressive axial
loading until buckling. In order to record the behavior of the system, the applied axial load and the
horizontal displacement of the mid-span-node are monitored.
In addition to the previous models, two extra cases are presented. The first is the Euler buckling
load, and the second is a single pallet beam model, with a horizontal initial imperfection of 11.5mm.
In Figure 94 and Figure 95 the results for all the above case studies are presented, in terms of Axial
load vs. mid-span horizontal displacement. Each figure has 2 black lines, representing the Euler
buckling load and the results for the single pallet beam. The colored curves are the cases from A to
E for different friction coefficient values. The legends of the diagrams indicate both case and the
friction coefficient value. For example, the results for case C when the assumed friction coefficient is
0.3, is nominated as C-0.3.
Observing the two diagrams, one can see that the cases for friction coefficient 100 are much higher
than the Euler buckling, as the beams buckle according to the 2 nd or 3rd buckling. Thus, the existence
of the pallets could provide a partial diaphragm action on the system and as a result to reduce the
beam’s buckling length. However, case B is an exception; the pallet on the middle cannot offer any
action on the system, due to rigid body motion, without developing several reactions on the pallet
beams.
In most cases it can be said that the black lines represent an upper and lower limit. This means that
the buckling load cannot be modified in any case from A to E and for any friction coefficient value,
except from μ equals 100 which is not a conventional value. The system for any case is presented
stiffer than the single beam, but this increase in stiffness is not enough to interfere with the buckling
length (i.e with buckling load).

84
 Figure 94 Horizontal mid-span displacement- Axial Figure 95 Horizontal mid-span displacement- Axial
Load for a pallet weight 8kN Load for a pallet weight 12kN

As a general conclusion, it is considered that the buckling length has to be conservatively equal to
the length of the beam. Finally, it should be noticed, that the problem is preliminary investigated and
there are parameters that may influence the results. These could be, for example, the distribution of
the pallet’s weight; on the supporting surfaces.

85
86
6. SEISMIC DESIGN METHODOLOGY
The study of the design rules for the seismic design of pallet racks started in the year 2001 in Europe
by the working Group “seismic design” of the FEM – Fédération Européenne de la Manutention -
Section X - FEM Racking And Shelving Product Group - with the purpose of incorporating the
principles of the EN1998 [5] to the design practice of the racking industry.
At that time FEM Norms were already published and used in the market concerning design, safety
practice and maintenance of racks and shelving in non-seismic conditions, and the most relevant
reference for the seismic design of racks was the RMI Code, issued by the Racks Manufacturers
Institute in the USA. Several working documents and drafts of the Norm have been produced.
In the 2005 an experimental draft of the prFEM 10.2.08 [25] (Dec. 2005) was analysed in the scope
of the research program Seisracks1 [6] and improvements were proposed and incorporated. The
Norm FEM 10.2.08 has been published in 2010 by the ERF (European Racking Federation, joining the
European national associations of racking manufacturers), and in May 2011 the actual revision was
published (v. 1.04) [4]. The need of a reliable code of practice is well described in the chapter
“Introduction” of the document, from which a short excerpt is reported in the following:
“Racking systems are not “buildings” but a particular form of steel construction. They differ from
buildings in terms of their use, the loads that are supported, the geometrical dimensions and the
components used in their construction. These components are normally thin-gauge cold-formed
profiles and, in the case of uprights, are typically continuously perforated. This gives the required
functionality, adaptability and flexibility needed to cope with the great variation in the different types
of goods that are stored. Only rack-clad warehouses, where the racking supports not only the goods
but also floors the roof and walls, can be considered as a very special type of “building”.
The various National Building Regulations give a design approach for “ordinary steel structures”
however, the peculiarities of racking mean that the response to earthquakes is different and a revised
design approach is necessary.
Whilst the basic technical description of an earthquake is obviously the same for all structures it is
of great importance to define whether or not it is possible to apply the “general design rules”
(applicable to ordinary steel structures) to a rack. Furthermore it is necessary to consider how to
modify correctly the general principles and technical requirements, in order to take into account the
peculiarities of racking to achieve the requested safety level.
In fact, many specific physical phenomena strongly affect the structural behaviour of a racking
system during an earthquake, such as the energy dissipated in the deformation of stored goods, or
when pallets (or other unit loads) slide on their supports. Furthermore, it shall be considered that
the self-weight of a rack structure will typically constitute less than 5% of the total mass whereas in
a typical building the percentage of dead and permanent loads will be much greater. Therefore the
presence and distribution of variable loads, like pallets, on racking systems strongly affect the
response of the structure under seismic actions.”
In 2011 the TC344 of CEN “Steel static storage systems” started the works to produce a CEN Norm
for the seismic design of racks having as reference the FEM 10.2.08 [4]; the work item was opened
in 2012 and the prEN16681 [22] is now in the pre-Norm stage.
Several findings of this research program have been introduced in the Norm or have enforced its
assumptions. For this reason the actual reference as the “state of the art” of the seismic design of
racks for this work item is also the prEN16681 [22] and in the following reference is done mainly to
this Norm.
Racking systems are normally manufactured with cold formed profiles and the design is carried out
based on “low ductile design concept”, without taking into account relevant plasticity and ductility;
for the design under ductile concept direct reference to EN1998-1 is then required.
The SEISRACKS2 research has been developed considering this type of construction and the
assessment of the seismic design methodology is developed for ductility design. Several results of
the research SEISRACKS2 have been considered in the development of the new EN Norm on seismic
design of racks and in part incorporated, or used as scientific background to justify assumptions; in
this chapter these information are reported. The following items are considered:

- Assessment of behavior (q) factors for unbraced and braced racks, and upright frames
- Evaluation of Qp and Rf parameters
- Assessment of the response spectrum factors
- Assessment of the method of analysis
- Qualification of testing for components, joints and substructures

87
6.1 Assessment of the behavior factors
The resistance and energy-dissipation capacity of a structure are related to the extent to which its
nonlinear response is to be exploited. In operational terms, such balance between resistance and
energy-dissipation capacity is characterized by the values of the behavior factor q. As a limiting case,
for a non-dissipative structure, where no (or extremely limited) hysteretic energy dissipation is taken
into account, the expected behavior factor, in general, is not greater than value of 1,5. For dissipative
structures, the behavior factor is in general larger than this value, accounting for the hysteretic
energy dissipation that mainly occurs in particular zones, called dissipative zones or critical regions.
During SEISRACKS2 project, behavior factors have been estimated for several rack configurations in
down aisle and cross aisle directions.
6.1.1 Unbraced racks (Down-aisle)
The full scale tests performed in WP4 have been carried out on racks designed using substantially
different approaches:
a. Racks designed for medium seismicity, where the Beam-end connectors and the base-plates
were largely overdesigned with respect to the operating conditions, in particular the typology
and the connection of the base-plates to the upright’s bases.
b. Racks designed for low seismicity, using components very close or equal to the ones used
for similar configurations in non-seismic zone
The first typology has exhibited good performance and ductility during the pushover test, with
sufficient strength and behavior factor larger than 2 (Figure 96).
The distortion of the section of the upright at the connection with the beam was observed in a very
advanced stage of the pushover, due to the contact forces between the side of the upright and the
Beam-end connector (Figure 96b). This effect did not produce any relevant damage, but it appears
reasonable to reduce this distortion by applying a simple stiffener between the lips of the upright,
made with a bolted spacer or a small tube, on the opposite side of the upright where there is the
Beam-end connector, letting the section remain undeformed. This stiffener already exists when the
beam levels are in correspondence to the node of the upright frame bracing. Such local stiffener
should also improve the global lateral stiffness of the frame.

a. Unbraced rack exhibiting ductile behavior b. Distortion of the upright at the beam-end
during the pushover tests connector
Figure 96 Plasticity observed at the last step of test IPA1

The second typology on the contrary has exhibited a soft storey behavior after reaching a certain
level of horizontal load. The pushover curves show a brittle behavior, as consequence of the
formation of a “soft storey” mechanism produced by the collapse of the upright between the floor
and the first beam level. The photo in Figure 97 taken immediately before the stop of the tests clearly
show the formation of hinges at the base-plates and the deformation concentrated just below the
first pallet beams.

88
 Figure 97 Unbraced rack exhibiting brittle behavior due to the formation of soft storey mechanism below the
first load level during the pushover tests

These tests have produced the following improvements of the prEN16681 [22]:
 The behavior factor q =2 has been confirmed as value that can be obtained from a proper
design of the frame;
 Stricter requirements on the design of the floor connections have been introduced for
unbraced frames, which were not considered in the FEM 10.2.08 [4].
6.1.2 Braced racks (Down-aisle)
Full scale tests have been carried out on braced racks that are designed using different approaches:
a. Vertical bracing designed using appropriate detailing rules, minimizing the eccentricities and
applying basic rules of capacity design for the hierarchy of the resistances;
b. Vertical bracing designed using the ordinary arrangements and components, without special
attention to the detailing, using the design rules for the strength of the connections;
c. Racks designed with vertical bracing having an elevated strength.
The braced racks have exhibited a resistance in general considerably higher than the unbraced racks,
but, as expected, lower ductility.
The first typology was the only one exhibiting a relatively high ductility, with a behavior factor larger
than 2, thus demonstrating that q=2 can be reliably assumed provided that the design of the vertical
bracings is properly performed and in particular plasticity in diagonals can develop.
The second typology on the contrary has shown overstrength but very low ductility because of the
failure of the connections of the vertical bracing diagonals, due to the insufficient strength of the
connections. (Figure 98).

Figure 98 Failure of the vertical bracing diagonal’s connection caused by eccentricities

The third typology refers to the rack with cable bracings that has exhibited a very high strength, so
that the test could not be concluded.
In all the four tests the performance of the plan bracings was positive and no damage was observed;
also the torsion of the bracing tower was observed.
The results of these tests have confirmed the possibility to design braced racks using relatively high
behavior factors in low ductility class, provided that the design and the detailing is accurate and the
failure of the connection is prevented.
In addition, because it is assumed in the design that the low ductility in the bracings is developed at
the bolted connections because of the bearing, it is necessary to design the connection allowing
sufficient plastic deformation of the bolt’s hole.
The impact on the new EN norm was the introduction of an additional rule on the design of the bolted
connections of the diagonals, requiring a minimum distance between the bolt and the edge of the
diagonal equal to 2.5 times the bolt’s diameter, that shall be considered in conjunction with the
89
general rule for bolted connections requiring the bolt’s strength at least 20% larger than the bearing
strength. In general the failure of the bracing occurs for bearing at the bolted connection, and this
rule appears a constraint to design stronger connections, in addition to improve the development of
the plasticity at the bolt’s hole.
6.1.3 Upright frames
The assessment of the behavior factors for the upright frame configurations
(Figure 99) was based on the tests on substructures performed in the WP2.
In Low dissipative concept the basic design rules in the prEN16681 [22] require:
a. K, D, Z bracings, and X bracings without horizontal members are schemes in which the shear
is resisted by diagonal elements in compression, and they are allowed to be designed with
behavior factor q=1.5 provided that the design effect of the seismic action in all bracing
members and their connections is increased of a factor 1.5.
In bolted shear connections the shear strength of bolts F v,Rd shall be 1.20 times higher the bearing
resistance Fb,Rd of the connected profiles:
Fv,Rd/ Fb,Rd > 1.20 [Eq. 53]
This requirement needs not to be applied when the strength of the connection is q times
greater than the calculated bolt shear due to seismic action.
The “extra safety factor” 1.5 is introduced to prevent the brittle failure of the diagonals and their
connections when a behavior factor greater than 1.0 is considered. This rule should be considered
with care; in fact the tests performed have shown (for the tested configurations) behavior factors
rarely higher than 1.5, while this overstrength may affect the actions in the upright and on the base-
plates.
It shall be emphasized that uprights and base-plates are challenged by the seismic action in addition
to the vertical loads, while the diagonals work mainly for the horizontal actions.
The results of the observation and reanalysis of the tests on upright frames are summarized in the
following:
i. The behavior factors evaluated from the monotonic tests are between 1.01 and 2.06
ii. The lower values of behavior factors are achieved by configurations with D or X diagonals
without horizontals, governed by the buckling of the compresses diagonal.
iii. The higher value is achieved for X bracings designed with diagonals working in tension and
horizontals
iv. Low ductility is developed when the strength of the connections is governed by bolt’s bearing
and bolt’s bending.
v. The disposition of the diagonals “back to back” allows preventing the torsion of the upright
but not in the post-buckling of the compression diagonal; this disposition appears being in
any case favourable
vi. The bending of the bolt has to be considered in the design

Figure 99 Typologies of upright frames

The limited extension of the experimental investigation performed by Seisracks2 doesn’t allow
providing general rules based on the bracing schemes only; nevertheless some additional caution
results from the tests.
i. Extra safety factor should be considered also in the connections of the diagonals in X bracings
with horizontals and only tension diagonals active.
ii. Extra safety factor should be considered for the design of the diagonal member in
compression in X bracings with horizontals and diagonals in tension and compression; this
to prevent the consequences of the failure of the compression members on the tension
member that doubles its action. This is especially important when the failure of the diagonal
in compression is produced by some local instability that causes a sudden and not progressive
transfer of the action to the tension diagonal.
iii. The bending of the bolt should be explicitly considered in the design of the connection, with
the appropriate static scheme.
90
 iv. The “extra safety” coefficient applied to the design of the bracing members could increase
the challenge of the upright and of the floor connection. Nevertheless it shall be noticed that
the effects of the seismic action in the upright is only part of the total action, which is added
to the pallet weight
v. The eccentricity of the scheme of the bracing diagonal should be considered carefully for
local strength verifications
vi. The bearing of the hole at the bolted connection appears being the main source of the small
amount of ductility for the bracings, and it is necessary to design the connection of the
diagonal in a way that bearing can develop before failure. For this reason, in addition to the
requirement of the shear strength of the bolt 1.2 times larger than the bearing strength, also
the requirement on the minimum distance of the bolt to the edge of the diagonal profile equal
to 2.5 times the bolt’s diameter appears necessary.
vii. Due to the geometry of the upright frames and the effects of the eccentricities both in and
out of the plane of the frame the experimental assessment of the behavior factors appears
appropriate.
It has to be noticed that in low ductile design concept the development of a small amount of energy
dissipation is required; safety factors are included in the design to prevent the failure of the diagonals
unfavourable schemes (with diagonals in compression providing the required resistance), providing
extra resistance to the diagonals that are the primary elements resisting the horizontal action.
6.2 Evaluation of Unit load weight for seismic design (Q p,rated) and rack filling
grade reduction factor (RF)
QP;rated is a specified value of the weight of unit loads for the compartment, upright frame or global
down aisle design (see EN 15512 [3]). The filling grade reduction factor RF is defined to take into
account the occupancy of stored goods in the rack that can be assumed during the seismic event.
This figure is defined by the Specifier based on statistical evaluations: the prEN16681 [22] required
RF=1.0 for analysis in cross aisle direction and RF 0.8 in down aisle direction.
In the practical applications the evaluation of the design pallet weight and of the rack filling grade
coefficient are a contractual information that the User of the racks has to specify to the Supplier (see
also EN 15629:2008, 6.7.1). This information shall be based on the forecast use of the rack and it
shall be well specified in the documents that the Supplier shall consider for the design of the racks.
The Norm does not provide any information to the User how to define the values of the weight of the
unit load and of the filling grade factor to be considered, but a methodology for their evaluation can
be found in Eurocode 0; this can be necessary to optimize the value of Qp,rated and RF to be specified,
based on simulations of the flows in the warehouse, or to perform a safety assessment of an existing
warehouse.
6.2.1 Specification of Qp,rated and RF
The weight of the Unit Load should be consistent with the definition of the “quasi-permanent value”
of the action defined in EN1990:2002 4.1.3 (c)[14]:
“(c) the quasi-permanent value, represented as a product 2Qk, used for the verification of ultimate
limit states involving accidental actions and for the verification of reversible serviceability limit states.
Quasi-permanent values are also used for the calculation of long-term effects.”
NOTE For loads on building floors, the quasi-permanent value is usually chosen so that the proportion
of the time it is exceeded is 0,50 of the reference period. The quasi-permanent value can alternatively
be determined as the value averaged over a chosen period of time. In the case of wind actions or
road traffic loads, the quasi-permanent value is generally taken as zero.
The design weight of the Unit Load can be calculated, excluding the unit loads placed on the ground
floor, as the average weight on the relevant portion of the rack. For practical applications, it is
possible to calculate Qp,rated considering different categories:
 The total product weight stored on each run divided by the number of unit loads, for the
analysis in down aisle direction
 The total product weight stored on each bay divided by the number of unit loads, for the
analysis in cross aisle direction
 The total product weight stored on each compartment divided by the number of unit loads,
for the analysis of the beams
In a warehouse, or in a module of warehouse where the type of storage is homogeneous (i.e. same
typologies of unit loads stored using the same distribution rules), the maximum value of the product
weight for each category can be sampled over a time interval to be defined on the basis of the speed
of the handling of the goods.
The value of Qp,rated shall be the average of the weights sampled over a period of time long enough
to cover all the periodical fluctuations of storing in the warehouse. It may also be possible that unit

91
loads with different type and size are used and storing rules are defined in the warehouse handling
managing system to keep the allowed weight per compartment and proper clearances. In this case
it is necessary to define “equivalent unit loads” scaling each typology having one given size as
reference.
The value of RF is calculated by dividing the number of equivalent unit load present on the run /
compartment / beam pair (depending on the category) by the number of equivalent positions
allocable. The value considered for the construction of the statistical basis of R F is the one
corresponding to the value of maximum unit load weight for each sampling interval of detection; then
three values, one for each category, are obtained. The values of RF is then obtained as the mean of
the calculated value.
When the unit load weight and the rack filling grade factors are obtained with this analysis, it appears
possible to assume in all the calculations the value of RF which has been obtained from the analysis.
This derogation from the rule proposed in the prEN16681 [22], which prescribes RF=1.0 in the
transverse direction and RF0.8 in the longitudinal direction, is considered justifiable by the accuracy
of analysis and sampling, and by the statistical basis obtained from the monitoring provided over a
period significant to cover the fluctuations of filling in the warehouse.
6.2.2 Reliability of loading assumptions based on rack filling factor
The Rf rack filling factor is defined as the total actual weight of a rack, in a random loading condition,
divided by the total weight of the rack in a fully loaded condition. Instead of defining numerous load-
cases, with respect to the pallet-weight and the pallets’ position, a fully loaded rack is examined with
an equivalent pallet weight equal to the nominal one multiplied by the R f value.
Usually, the value of Rf = 1 is considered the most critical for the design; however, it is possible for
lower values to produce most unfavourable results for certain members. This can occur when the
use of Rf less than one would produce a lower time period of vibration and as a result a higher
spectral acceleration (since, the racking systems present time periods in the constant-velocity part
of EC8-spectra); although the spectral acceleration could be higher, the total mass is lower, thus the
value of the base shear is not easy to be predicted.
The effectiveness of this method has been assessed investigating numerically several load conditions
in different racking typologies and configurations. 18 different load-cases have been examined. These
are shown in Figure 100, in respect with their Rf factors.

Figure 100 the 18 different examined load-cases

Although, generally the LC1 produce the most unfavourable results, LC11 and LC14 can produce very
similar even more unfavourable results in specific members. However, the results of the
investigations have shown that there are no extreme differences changing the disposition and the
weight of the unit loads, as the maximum difference found does not exceed the 15%.
The fact that there are sometimes unfavourable results in load cases other than the fully loaded is
due to local dynamic effects; thus, the local phenomena cause local distribution of forces, which are
hard to be predicted a priori.
In conclusion, the equivalent use of the Rf factor in order to simulate more complicated loading
conditions is not always sufficient, and at local level it is necessary to consider loading distribution
which maximize the local effects on the element or part to be investigated.
This has been considered in the prEN16681 [22] considering special loading cases for the analysis of
the upright frame and of the vertical bracing; also, the analysis of the isolated upright frame based
on its local numerical model provides reliability to the design because it ensures incorporating all the
local effects.

92
In addition, in the design practice are considered the loading distributions which have relevant
probability to occur. Extra loading conditions need to be considered in special design situations, such
as racks installed on suspended floors, but all based on fully loaded rack for down aisle analysis.

6.3 Assessment of the response spectrum modification factors

Three modification factors are introduced to modify the response spectrum at the design stage, in
order to take into account the interaction between the stored goods and the racking structure.
The design spectrum modification factors ED1 and ED3 are introduced for the modification to the
ordinate of the design spectrum; the pallet weight modification factor E D2 affects the mass to be
considered in the analysis and modifies the period of vibration. The modification factors ED1 and ED3
in the FEM 10.2.08 [4] multiply the ordinate of the design spectrum to define the “reduced” spectrum:
Sd,red(T)=ED1  ED3  Sd(T) [Eq. 54]
The prEN16681 [22] defines as “modified” the spectrum affected by ED1 and ED3, and introduces a
variation to consider that only the contribution of the mass of the unit loads affects the modification
of the response:

Sd,mod = KD  Sd(T) [Eq. 55]

where:
𝑃𝐸,𝑝𝑟𝑜𝑑
𝐾𝐷 = 1 − ( (1 − 𝐸𝐷1 𝐸𝐷3 )) [Eq. 56]
𝑃𝐸

PE is the total weight of the rack in the seismic design situation (including dead and permanent
weight, live load in the seismic situation and the stored product weight)
PE,prod is the total product weight stored on the rack, in the seismic design situation
KD = ED1  ED3 [Eq. 57]
can be assumed when PE,prod is greater or equal to 90% of PE
ED1 takes into account the effects of the sliding of the unit loads on supports when the inertial forces
exceed the resistance provided by the friction; ED1 is affected by the following parameters:
i. intensity of the seismic action
ii. number of load levels, total mass and flexibility of the racking structure, expressed by the
period of vibration (dominant period in the direction considered)
iii. maximum horizontal force that can be transmitted by the pallet to the pallet beams,
expressed in terms of friction coefficient
The formula reported in the prEN16681 [22] for ED1 is
ED1 = max [0.4 ; μS/Se(T1) +0.2 ] ≤ 1.0 [Eq. 58]
where
μS is the friction coefficient of the interface between unit load and racking structure
T1 is the fundamental period of vibration of the racking structure in the considered direction
(the period with highest modal participating mass in the considered direction)
Se(T1) is the ordinate of the elastic spectrum calculated using 3% viscous damping
This formulation has been derived from Seisracks [6] research, and it is referenced in several
publications and in an Annex of the FEM 10.2.08 [4].
When pallets are restrained on the pallet beams by means of any special system (for example
materials increasing the friction between pallet and beam), ED1 is assumed equal 1.0 because sliding
is assumed not to occur. The friction coefficient to be considered for ED1 is the average value obtained
by tests or specified by the norms, because it affects the global response of the rack.

ED3 is a reduction coefficient of the seismic action which is introduced to account for the dissipative
phenomena typical of the dynamic behavior of racking structures under seismic actions, that are not
included in the mathematical formulation of the design seismic action, but that are observed on racks
that have suffered earthquakes, and from tests performed on shaking tables.
The ED3 coefficient was introduced in the final version of FEM 10.2.08 [4], and the value 0.67 was
assigned; in the prFEM 10.2.08 [25] ED3 was not considered; in the prEN16681 [22] the value of ED3
has been changed to 0.8.
A third modification coefficient ED2 is also considered, called pallet weight modification factor,
affecting the mass to be considered in the analysis. This factor includes the effects of the interaction

93
between pallet and racking structure affecting the seismic response in terms of participating mass
and modification of the period of vibration. Its definition is based on the classification of the goods
stored in the unit load, their grade of compaction and restraint. The definition of ED2 is the same in
the FEM 10.2.08 [4] and prEN16681 [22] (Table 26).
Class ED2 Stored good classes Example

Frozen goods (cold storage)

COMPACT
A 1.0 Steel sheet package
CONSTRAINED
Coils and paper rolls

Big number of pieces stored on the pallet whose size is small in

B 0.8 WEAK comparison to the pallet size, including goods stabilized by stretch
wrapping

LOOSE AND Goods that can easily move around, inside the container (e.g.
C 0.7
UNCONSTRAINED granulated materials)

D 1.0 LIQUID Unit load containing liquid that can slosh in the container

Table 26 ED2 factors

The results of Seisracks2 research showed no new significant elements that require a review of the
parameters ED1 and ED2 as they are considered in the FEM 10.2.08 [4] and prEN16681 [22].
The review of the results of some past research programs, specifically Seisracks1 [6] and Ecoleader
[15] including tests on shaking table of racks simulating strong earthquakes, have been performed;
it has allowed to assess a reliable value for ED3 spectrum modification factor.
The ED3 coefficient was not considered in the FEM 10.2.08 [4] at the draft 2005 stage, it was
introduced in the final version of FEM 10.2.08 [4], and the value 0.67 was assigned; in the actual
prEN16681 [22] the value of ED3 has been assessed to 0.8, based on the reanalysis of the results of
shaking table tests.
The FEM 10.2.08 [4] and the prEN16681 [22] assume for the unloaded rack the damping factor 3%.
This figure is derived from the results of the Seisracks1 [6] research, and it is used when the elastic
spectrum is required. The damping of the loaded rack is in general much higher, and its effect is
indirectly accounted for by the ED3 coefficient of the modified spectrum. The reanalysis of the results
of tests on shaking table performed for Seisracks [6] and Ecoleader [15] research show damping
higher than 10%.
The analysed results are the accelerograms recorded during six tests performed for the Seisracks2
and Ecoleader [15] projects, on 4 typologies of racks. A seventh accelerogram was obtained from
the monitoring of warehouse racks in Athens after a recent seismic event occurred during the
Seisracks2 research period.
In the Table 27 are reported the ranges of the frequencies of the free vibrations and the average
values of the damping calculated.

Research Test Frequency Average Damping

Seisracks1 [6] A2 0.56-0.64 17.47 DY - Cross Aisle
A3 0.42-0.46 13.23 DY - Cross Aisle
A4 0.76-1.18 17.06 DX - Down Aisle
Ecoleader [15] A 0.38-0.49 6.8
C 0.59-0.73 10.6
D 1.00-1.14 9.2
Seisracks 2
Earthquake 3.77-3.85 8.28
4/6/2013
Table 27 of the frequencies of the free vibrations and the average values of the damping from various projects

The values of damping range from 6.8% to 17.47% of the critical damping; it shall be emphasized
that in the research Ecoleader [15] the unit loads were connected to beams to prevent their falling.
Having as starting point the elastic spectrum defined by the Normative with =5% damping (in
percentage of the critical damping), the damping correction factor  is defined as:

[Eq. 59]
10
𝜂=√
5−𝜉

This figure of  is associated to the ED3. In particular ED3 is related to the damping of the loaded rack
which is greater than 5% which is the figure considered in the response spectra defined in the
Normative.
ED3= 0.67 reported in the FEM 10.2.08 [4] is related to a damping between 17% and 18%.
94
ED3=0.8 in the prEN16681 [22] corresponds to a damping between 10% and 11%, which is
considered a reliable figure.
6.4 Assessment of the methods of analysis
The reference method of analysis considered by the Norm is the modal response spectrum analysis
(MRSA); the lateral force method of analysis (LFMA) is an approximation that is applicable under
certain conditions.
The second order effects in the MRSA can be considered either directly performing a MRSA including
second order effects, called MRSA-2, or amplifying the results of the first order MRSA by the
amplification factor 1/(1-θ), where θ is the interstorey drift sensitivity coefficient; the indirect method
gives generally safe results, unduly conservative when θ>0.3.
The assessment of the method of analysis has been performed by comparing the results of analyses
performed by means of time history analyses, which are considered as benchmark. All the analyses
are linear elastic.
The nonlinear MRSA-2 is a conventional MRSA analysis in which the modified stiffness matrix of the
structure is used. The term modified is referred to the stiffness matrix which takes into account the
axial forces in the beam elements.
The time history analyses are considered the most sophisticated and reliable analyses; they require
the definition of the ground motions (accelerograms) which are introduced in the software to perform
a nonlinear, dynamic and elastic analysis. According to EN1998 [5] at least 7 accelerograms have to
be used in time history analyses, in order to have statistically reliable results for the average response
of the structure.
This method of analysis, including the geometric nonlinearity, and when appropriate, the nonlinearity
of the materials, is time consuming and complex for the preparation of the input data, and it can’t
be assumed as a reference for the daily practice, but it can be assumed as the reference benchmark
method.
The results from MRSA-2 are sufficiently accurate and close to the ones produced by time histories,
while the internal forces obtained from the MRSA are conservative only for relatively stiff structures.
Comparisons between different types of analysis have been shown in Figure 101 and Figure 102.

a. IPA down aisle b. IPB down aisle

Figure 101 Comparisons between different types of analysis for down-aisle frames

95
 a. IPA cross aisle b. IPC cross aisle (braced in downaisle)
Figure 102 Comparisons between different types of analysis for cross-aisle frames

The amplification with 1/(1-θ) coefficient is an approximate and in general conservative method to
take into account second order effects, but it is not always reliable, because it results unduly
conservative for values of θ larger than 0.3. This report proposes using the MRSA2 as the most
appropriate method of analysis which can include the second order effects; this method is commonly
implemented in commercial software. Theta method may also be used because conservative, but the
results of the analysis should be verified and considered with attention, also for values of θ lower
than 0.3.

6.5 Qualification of testing for components, joints and substructures

The Norm does not require special tests on rack’s components and substructures when the design is
performed considering low dissipative behavior; the experimental parameters required for the design
are obtained using the protocols specified in the EN15512:2009 [3] for static design. Tests are
instead required to qualify the components for dissipative behavior. For low dissipative design tests
should be in any case required to assess the behavior factors.
Testing methodologies have been developed in the SEISRACKS2 research program for beam-end
connector, floor connection, upright frame shear, and full scale tests in down aisle direction. These
methods are in general different and more complex than the standard ones normally used for the
static design. For practical applications the testing methodologies proposed, need to be completed
by specifications for the derivation of the design parameters through the reanalysis and the
application in the design.
6.5.1 Beam-end connector tests (BEC)
Monotonic BEC tests are performed to evaluate the bending stiffness and strength and the shear
strength of the joint; the reference protocols are defined in EN15512:2009 [3]. The prEN16681 [22]
doesn’t require cyclic tests for low ductile design; only for ductile design cyclic tests are required.
a. Monotonic and cyclic
Tests of typical hooked joints show that the monotonic curves envelope the cyclic curves. For low
ductile design with q=1.5 the monotonic test appears adequate because the tested joints have
sufficient capacity to address the rotation without relevant reduction of stiffness and strength. For
low ductile design with q=2 the rotation capacity can be assessed on the basis of monotonic tests as
well. In practical applications for low ductile design the cyclic tests appear to be not necessary, but
the rotation capacity of the joint has to be assessed.
b. Load history
The test setup should consider the effect of the preload produced by the pallet weight, which has
been found in general favourable. The cycles of positive-negative bending moments shall range from
the bending moment produced by the weight of the unit loads. The load history should be force-
controlled in the preload phase and displacement controlled in the cyclic phase. Cycles should be as
follows:
i. One cycle with the amplitude factor 0.25, 0.50, 0.75 and 1.0
ii. Three cycles with the amplitude factor 2, 3, 4, … (compared to ECCS [17] with 3 cycles with
the amplitude factor 2, 4, 6, …) (Figure 103).

96
 Figure 103 Load history for beam-end connector tests

c. Payloads
Because of the positive influence of the payload, the cyclic tests should be performed with different
payloads, including zero load.
d. Test setup
The test setup for the cyclic test can be either:
i. The one of EN15512:2009 [3] (cantilever test) using a reversible jack (Figure 104a)
ii. The portal test, derived from the setup used for Seisracks2, where the pair of beams are
instrumented and tested and not only one single beam; in this case the main values of the
couples of joints loaded in the same way have to be considered (Figure 104b)
iii. The RMI-2008 [9] setup (Figure 104c)

a. Cantilever test scheme in b. Portal test SEISRACKS2 c. RMI-2008 [9] setup

EN15512:2009 [3]
Figure 104 Different test set-ups for Beam-end connector test

e. Seismic bolt
The tests performed using seismic bolt have shown that the positive effect of the seismic bolt was
due to the redundancy, when the failure of the connector is relevant, in particular when the failure
is due to the bearing of the hole in the upright. However, the seismic bolt showed nearly no influence
on the ductile response of the joint (Figure 105). This result appears not in agreement with
Seisracks1 [6], where the tests have shown increased capacity with seismic bolt. It shall be noticed
that the comparison in Seisracks2 was based on monotonic test, while the connectors were tested
under cyclic loading in Seisracks [6]. Also, the beam end connector configurations were different
with respect to those tested within Seisracks [6].

a. With seismic bolt b. Without seismic bolt

Figure 105 Comparison between beam-end connector tests with and without seismic bolt

97
Summary of Seisracks2 findings
i. Different failure modes have been observed: Beam’s buckling, connector’s hook failure,
upright’s bearing. It is emphasized that the main failure mechanisms of the beam-end
connector have been observed in the research program.
ii. Moment resistance: The resistance determined is comparable to that from standard tests.
Resistance to positive and negative moments is approximately the same due to redistribution
of moments from left to right. For this reason the test-setup used in Seisracks2 or similar
ones is recommended for the assessment of seismic performance of the joints.
iii. Comparison of monotonic and cyclic test results: The first load cycle in general produces
same results as the monotonic static test. The degradation of resistance depends on the
failure mode of the joint. The pay load has a positive effect on moment resistance.
iv. The seismic bolt has negligible influence on monotonic moment-rotation behavior but
redundancy can be achieved when connector failure is relevant.
6.5.2 Floor connection tests
Monotonic tests are performed to evaluate the bending stiffness and strength of the floor connection;
the reference is defined in EN15512:2009 [3]. The prEN16681 [22] doesn’t require cyclic tests for
low ductile design; only for ductile design cyclic tests are required.
a. Monotonic and cyclic
Under monotonic loads the initial stiffness of the base connection is in general higher for larger
compression load; instead, the variation of the moment resistance as function of the compression
load differs significantly between base-plate typologies. The moment resistance may be limited by
local or global buckling of the upright (relevant for high axial loads) or by failure of the base-upright
connection (failure of welds or bolts, or large deformations of base-plates). Under cyclic loads the
moment resistance is in most cases comparable with the monotonic, while the rotation capacity is
significantly reduced.
b. Load history
First the axial preload is applied, then the rotation. During the application of the rotation, the axial
force shall be kept constant. The load history should be force-controlled in the preload phase and
displacement controlled in the cyclic phase (Figure 106). Cycles should be as follows:
i. One cycle with the amplitude factor 0.25, 0.50, 0.75 and 1.0
ii. Three cycles with the amplitude factor 2, 3, 4, …

Figure 106 Load history for floor connection tests

c. Test setup
The test setup actually included in the EN15512:2009 [3] is not easily applicable when large
displacements/rotations shall be applied. The setup used in Seisracks2 can be considered the
reference one, but it requires further adjustment (Figure 107b).
The connection to the floor should be consistent with the real application and therefore anchor bolts
of the same typology used in the application shall be installed on a concrete block, having
approximately the same characteristics specified in EN15512:2009 [3] (the performance of the
anchor bolts was not of interest for Seisracks2). The concrete block shall be well connected to the
test frame.
In order to reduce the influence of the buckling of the upright with open section during the test it is
recommended to install a spacer between the lips of the profile in the same position of the first
horizontal of the frame diagonals. In the point of the horizontal action the section of the upright shall
be restrained for the rotation around the upright’s axis.

98
 Constant loading
(force controlled)

Pendulum H1

Variable loading
(displacement controlled)

Upright
H2

Column base
Rigid steel plate
Figure 107 Floor connection test setup (H1≈3m, H2≈0.5m)

Summary of Seisracks2 findings

i. Under monotonic loads typically the initial stiffness of the base connection increases with an
increased compression load, but the moment resistance, differs significantly between base-
plates typologies.
ii. The moment resistance may be limited by the local or global buckling of the upright (relevant
for high axial loads) or by the failure of the base-upright connection (failure of welds or bolts,
or large deformations of base-plates).
iii. Under cyclic loads the moment resistance is in most cases comparable with the monotonic
one, while the rotation capacity is significantly reduced.

6.5.3 Upright frame shear test

Monotonic tests are performed to evaluate the shear stiffness of the upright frame; the reference is
defined in EN15512:2009 [3]. The actual shear stiffness of the upright frame is affected by several
factors (e.g. connection eccentricities, connections looseness, structural imperfection, bolt bending,
local distortion of uprights), which are not explicitly considered in the numerical analysis.
These influences reduce the overall shear stiffness compared to the one that can be obtained with
simple structural models; they are incorporated in the analysis by modelling the bracing members
with a reduced value of the cross section area, or adding axial springs aligned with the bracing
elements. This value of the resulting stiffness of the bracings is obtained comparing the deformation
of an upright frame measured from laboratory test, with the numerical model of the same frame,
reducing the area of the diagonals cross section in order to match the experimental results. This
principle is incorporated in the prEN16681 [22], which for low ductile design doesn’t require strength
or cyclic tests when the connection between diagonals and upright is symmetrical with respect to the
torsion of the upright (Figure 108).
The torsion in the upright is not explicitly considered in the analysis and in the verifications; when
the horizontal actions are small (like in non-seismic situations) they are assumed to be equilibrated
by the out of plane bending of the diagonals, which in static design are normally very under-stressed.

Figure 108 Actions transferred by the upright frame bracing diagonals to the upright

The problem of the torsion of the upright induced by the diagonals is considered by the EN Norm
that requires strength and stiffness tests when the shear force is unsymmetrically introduced in the
upright by the frame bracings. Tests are explicitly required to assess the stiffness and strength of
the upright frames (diagonals and uprights) when the geometrical arrangement of the frame bracings
introduce relevant torsion in the upright. prEN16681 [22] doesn’t provide any directive for the
performance of the tests. For ductile design concept direct reference to EN1998 [5] is done.
99
 a. Monotonic and cyclic
Monotonic tests are always necessary to evaluate the shear stiffness. Results of cyclic tests exhibit
similar or lower displacement capacity than the monotonic ones.
b. Failure mechanisms
The shear resistance of the upright frames is governed by the strength of the diagonals whose failure
is in general relatively brittle (buckling of the diagonals, bearing of bolt, bending of the bolt) or by
mechanisms involving the upright (torsion induced by the diagonals due to their disposition or
restraining capacity, or bending produced by the joint eccentricities although in the limits prescribed
by the EN15512:2009 [3]). The design rules of prEN16681 [22] require that the shear strength of
the bolts be greater or equal 1.2 times the bearing strength, which is considered a mechanism with
some ductility.
c. Test setup
The test setup actually referenced in the international normative are the “European” one (protocol in
the EN15512:2009 [3], Figure 109a) and the “Australian” one [64], which has been used in
Seisracks2 research (Figure 109b).

a. Test set-up in 15512 b. Test set-up in the Australian code

Figure 109 Upright frame shear test setups from different codes

The “European” protocol can be assumed to evaluate the shear stiffness. The “Australian” protocol
is on the contrary necessary to assess the strength and the failure mechanisms under both monotonic
and cyclic loading, because it allows to consider the shear-bending interaction of the upright frame
and to apply the loads according to the distributions required by the EN1998 [5].
Summary of Seisracks2 findings
i. The behavior factors evaluated from the monotonic tests are between 1.01 and 2.06
ii. The lower values of behavior factors are achieved by configurations with D or X diagonals
without horizontals, governed by the buckling of the compresses diagonal.
iii. The higher value is achieved for X bracings designed with diagonals working in tension and
horizontals
iv. When bolt’s bearing and bolt’s bending are the governing mechanisms, these provide low
ductility.
v. Mounting of the diagonals “back to back” allows preventing the torsion of the upright but not
in the post-buckling of the compression diagonal; this configuration appears being in any
case favourable
vi. The bending of the bolt has to be considered when appropriate
6.5.4 Full scale tests in down-aisle direction
Full scale tests are complex to execute and require important test setup and equipment; monotonic
pushover tests are currently the full scale test considered. This test is actually not considered and
required in prEN16681 [22].
It can be considered for the down aisle direction only because the shear test on upright frame can
be sufficient for the cross aisle direction, at least when the height to depth ratio does not exceed 10
(for higher values the experimental investigation of the large displacement effects can be
meaningful). On the contrary in the down aisle direction the full scale test allows assessing the
combined effects of the different properties of the components and connections. This test can be
used to calibrate and to validate the numerical pushover analyses performed to assess the behavior
factor.

100
 a. Purpose of the test
The purpose of the test is to evaluate:
i. The behavior factor
ii. The natural frequency, to calibrate the numerical models
iii. The damping, related to different loading distributions (measured during the loading phase
that should be executed from the top levels to the bottom).
ii. and iii. were not included in the research program, but their investigation appear feasible and
meaningful during the preparation of the pushover tests. The value of the damping may be used to
derive the ED3 coefficient that may be assumed equal to
10
𝐸𝐷3 = √ [Eq. 60]
5+𝜉

where  is the damping expressed in percentage of the critical damping. The behavior factor can be
determined in general through numerical pushover analysis. The full scale pushover test should be
required to:
- Calibrate a numerical model for the evaluation of the behavior factor
- Validate a value of behavior factor obtained numerically

b. Evaluation of the behavior factor

The behavior factor can be evaluated numerically:
i. Using the design properties of the components derived by tests.
In this case the full scale pushover test is used to validate the values obtained numerically;
the experimental value should be greater than the one obtained numerically
ii. Using the characteristic properties of the components obtained by tests.
In this case the full scale pushover test is required to calibrate the numerical models;
significant rack configurations shall be selected to perform tests and calibrate the numerical
model to be used to study other configurations.
At least the following are necessary:
- Moment- rotation curve of the beam-end connectors
- Moment-rotation curve of the floor connections, as function of the axial load N
The moment-rotation curves are derived from the set of experimental curves as polynomial
best fit.
The moment rotation points of the floor connection curve for a given value of axial load N
can be obtained by interpolation between curves for tested values of N, where the bending
moment is obtained by interpolation for a given rotation
- The other parameters (such as the shear stiffness of the upright frame) can be obtained from
standard tests.

c. Test setup
The reference test setup and testing methodology is referenced in the WP4 full scale report (Figure
110).

Figure 110 Full scale test setup

As previously mentioned, using the same setup, it is convenient to perform preliminarily dynamic
investigations finalized to determine the frequency and the damping of the system with different
loading situations recording the free vibrations by applying an impulsive action.
Summary of Seisracks2 findings for unbraced racks
i. Four unbraced racks have been tested in down aisle direction, 2 designed for medium
seismicity, and 2 designed for low seismicity with standard components.
101
 ii. The racks designed for medium seismicity with behavior factor 2.0 are characterized by
strong base-plates well connected to the upright’s base with bolts in round holes; the base-
plates are made with U profiles fitted into the upright’s section and welded to thick base-
plates. The beam-end connectors are from the standard of the manufacturers and have
seismic bolts installed.
These two racks resisted to the seismic shear and exhibited a nonlinear behavior without
initiation of collapse until the end of the test. At the end of the test the structure was
deformed and the beam-end connectors were deformed as well.
iii. The racks designed for low seismicity were designed with q=1.5 using standard base-plates
and Beam-end connectors for non-seismic applications.
These two racks exhibited brittle failure, after reaching a certain base shear. This was due
to the clear formation of a soft storey mechanism after the failure of the floor connections,
the formation of plastic hinges in the upright below the first load level and at the base, and
the torsional flexural buckling of the uprights.
In conclusion:
 The behavior factors q=1.5 or q=2.0 for unbraced racks are reliable figures provided that
the soft storey collapse mechanism is prevented; this can be achieved improving the bending
strength and the rotation capacity at the base, either by installing floor beams or base-plates
adequately designed.
 The properties and the design of the floor connections derived from non-seismic design
appears to be not adequate for seismic conditions, also in consideration of the variability of
the axial force in the upright during a seismic event.
 The full scale test of the unbraced racks appear necessary to calibrate the numerical models
in particular with reference to the bending stiffness of the uprights.
Summary of Seisracks2 findings for braced racks
i. Four braced racks have been tested, with different typologies of vertical bracings. Three were
with X scheme, one with cables
ii. The bracings of the first typology were designed taking care of the ductility of the diagonals
and their connections; the design was performed with q=2.0.
This rack has resisted the design base shear and has exhibited a regular nonlinear behavior;
the collapse occurred for distortional buckling of the upright of the vertical bracing that was
not properly designed for this effect. After the start of loading the compression diagonals
were buckled. After unloading the diagonals in tension still presented an out-of-plane
permanent deformation.
iii. The braced diagonals of the second and third typology with X scheme were designed with
standard arrangements and the collapse load was reached with the failure or excessive
deformation of the bracing’s joints exhibiting very low ductility.
iv. The rack braced with cables was tested until the capacity of the pulling equipment was
reached, without significant deformations
v. The behavior of the braced racks is governed by the response of the vertical bracings; ductile
behavior is achievable, providing a proper detailing of the bracing’s elements.
Based on the findings of Seisracks2 and considering that the extension of the results do not allow
generalizations based on the structural schemes only, the following additional design rules should be
applied.
 The assessment of the behavior factor should be required when a rack non regular in plan is
designed using a behavior factor q>1.5; this assessment should be done through a 3D
numerical pushover analysis including torsion.
 The experimental calibration or validation of the numerical pushover analysis through a full
scale test should be required in ductile design or when the behavior factor in cross aisle
direction is smaller than the one considered in down aisle direction.

Figure 111
102
7. SOFTWARE TOOL DEVELOPEMENT

The SEISRACKS2 software provides a tool to test and verify various rack structures using the
assumptions and guidelines proposed in FEM 10.2.08 and the draft EC8 documentation, as well as
the results of the SEISRACKS2 project.
The software processes the input to create an analytical model of the structure and displays a
wireframe representation of the model in a viewing window. The structure members are color-coded
by type (beam, upright, back bracing, horizontal bracing, frame bracing). The user can inspect the
model by enabling the display of node, FEM beam and member numbers or using the mouse right-
click to rotate it. Left-clicking on a member brings up a tooltip window with member section data.
New window creates a new view window which the user can manipulate to display a different view
of the structure.
The software accepts input in the form of a structured text file. The format of the file is presented in
Appendix A of the software guide “Seisracks2_documentation” where a template spreadsheet file is
provided as a guide. The guide, as well as the software description and the software itself can be
downloaded at http://eng.ccs.gr/expertise/custom-applications/seisracks2/
The software can export the model to CCS’ INSTANT steel software where it can be examined in
every detail. The INSTANT steel software is made available to the SEISRACSK2 project partners for
free with the expressed provision that they only use it in the context of the project work and not for
any commercial gain. The software requires a web-license with a unique activation code that can be
obtained from CCS with an email to info@ccs.gr.
7.1 Theoretical analysis of the design method

The current analysis method for racks under seismic loading is the multimodal spectral analysis.
However this type of analysis does not allow the consideration of geometrical nonlinear effects due
to the linear superposition of the modal responses.
To overcome the above limitation and allow for a multimodal spectral analysis with due consideration
of 2nd order effects, a new and innovative procedure has been developed. This method is a
multimodal spectral stepwise non-linear analysis. It offers the following advantages against other
approaches that were examined:
 It uses seismic data that is well defined in every national design code (parameterized spectra,
importance categories, regional acceleration limits etc.).
 It uses fast, reliable and reasonably accurate algorithms and procedures (modified Newton-
Raphson, CQC).
 It can be combined with existing software.
 It is easy to use and understand even for users without specialized knowledge (simple
presentation of variables and data, user is not required to know the implementation details).
The method can be described as follows:
1. Static non-linear analysis (step by step) for permanent (G) and live (Q) loads to compute
the RGQ response.
2. Static linear analysis of a special loading case combination with G and Q loads to simulate
the loading conditions existing on the structure at the time the earthquake happens.
3. Modification of the structure stiffness matrix to account for the geometric effect of the
deformation imposed by step 2. The 3D geometrical stiffness matrix formulation used to
calculate the effect of geometry on the structure stiffness is shown below where:

1 2
Beam bending element
L : Length
F : Axial force

The formulation shown above includes the effect of bending and shear

103
 Table 28 3D geometrical stiffness matrix formulation

4. Calculation of the critical participating modes in each seismic direction using the modified
stiffness matrix calculated in step 3.
5. For every mode i the corresponding static load Hi is calculated. This load simulates the
seismic load in the specific direction.
6. Static non-linear analysis (step by step) for G, Q and Hi loads to compute the RGQHi
response.
7. Steps 3 and 4 are repeated for every participating Eigenmode.
8. The part corresponding to the modal response is calculated using the following expression:
RHi = RGQHi - RGQ
9. The total seismic response RH for the specific direction is obtained with the application of
CQC method :
RH   RHRH
i j
ij i j

[Eq. 61]

ij 

8 i  j i   j   3/2

     
2
104 1  2  4 i  j  1   2  4  i 2   j 2  2
[Eq. 62]
Where:
ζi, ζj : critical damping for i and j modes respectively.
ρ = Tj/Ti
Ti, Tj : period of i and j modes respectively
Calculation of the final response for a seismic direction using:
R = RGQ + RH [Eq. 63]
The steps 1 and 4 require the non-linear analysis for large displacements. This is being implemented
as an extension to existing C.C.S. solvers, using the modified Newton-Raphson method. This method
was selected against the complete or arc-length variant because it was estimated to be faster to
implement and accurate enough for the specific application.

Figure 112 Newton-Raphson modified method convergence progress (2)

7.2 Software components development

The C.C.S. proprietary solver modules are being enhanced to support the execution of the new
multimodal stepwise non-linear analysis. In order to facilitate the flexibility of the analysis procedure,

104
a modular approach was implemented that divides the various analysis steps to specific modules.
The following is a simplified reference for the scope of each of the delivered modules:
ASSECS Local to global matrix assembly
DIMSCS Matrix dimensioning helper
EFFBCS Beam internal forces calculation
EQFRCS Equivalent forces calculation
ERROCS Residuals calculation
LOCBCS Local beam matrix calculation
MASDCS Mass matrix calculation
OPTCS Structure optimization
REACCS Reactions calculation
STATCS Static analysis solver
TOP2CM Combinations calculation
TOP2CS Static results expansion
TOP2EN Envelopes calculation
TOP2FS Spectral results expansion
TOPSPDEX Spectral calculations initialization
TOPSPREX Spectral calculations finalization
VVPDCS Free vibrations calculation
Table 29 Scope of each of the delivered modules

7.3 Software features

7.3.1 Analysis
Loads and masses
The structure definition contains the bay loads to be applied and these loads are also converted into
nodal masses on the bay level nodes, using the ED2 mass activation factor supplied. The masses are
applied on each structure level incrementally, starting from the top, and a separate model is created
and solved for each mass configuration.
Spectral analysis
The software takes into account the effect of the effective spectrum modification factor K D as
described in paragraph 7.5 Design parameters for seismic analysis of document TC 344 [23] WI
00344007:2013 (E). ED3 is taken to be equal to 0.67 as proposed in section 7.5.2b.
The software implements the multimodal spectral analysis procedure.
Combinations
The software produces 8 directional combinations using the seismic response results.
7.3.2 Results
After the analysis has completed successfully the software scans the results of the spectral analysis
combinations over all solved models and selects the most unfavorable results for the displacements
of each level nodes, the reactions at the supports as well as the internal forces of each beam type
(upright, beams, frame, horizontal and back braces) and can generate a report document that can
be exported to RTF or HTML formats.
7.4 Parametric studies

As part of the work done for the SEISRACKS2 project, a novel spectral analysis method has been
developed. This method has been incorporated in the software application SEISRACKS2. To
determine if and to what extent the results of this method are comparable and of practical use, a
comparison between key results from an elaborate time-history dynamic analysis performed with
Sap2000 [20] and the relevant results from the multimodal spectral stepwise non-linear analysis
performed by the SR2 software has been carried out.

105
The rack model used for the comparison has been chosen to be that of a simple system of 6 frames
with 4 bays each, without any back-braces and only a single horizontal brace at the top middle bay,
to control the extreme flexibility of the structure. Such a flexible structure was required to produce
significant second-order effects.

Figure 113 Rack model

In the tables below all the sectional and spectra data used for rack numerical modelling are reported.
Section data

Section Ax (cm2) Ay (cm2) Az (cm2) Ix (cm4) Iy (cm4) Iz (cm4)

Beam 6.64 4.42 2.21 0.09 48.38 101.97
Uprights 6.84 3.11 3.00 0.14 32.01 67.90
Frame brace 1.34 0.42 0.54 0.01 1.10 2.61
Horizontal brace 1.34 0.42 0.54 0.01 1.10 2.61
Table 30 Rack section data

Spectral data

Parameter Value Comment

Region acceleration 0.12g
Ground type C
Structure importance 1.00 Class II
Spectrum type Type 2
qx 1.50 Cross aisle
qz 1.50 Down aisle
Modification factor (ED1) 1.00 Unblocked
Stored goods type (ED2) 1.00 Weak
Table 31 Spectral data

The models were compared using their Eigen modes and found to be very close in behavior. SR2
calculates the Eigen modes using the geometric stiffness matrix and SAP2000 [20] has a similar
option and it was used to generate the following results:

106
 Table 32 Comparison between models

In order to obtain a meaningful comparison between the different methods, an attempt was made
to emulate the spectral response calculated with the Eurocode parameterized spectrum through
appropriate modifications to the time histories of the seismic events used in the dynamic analysis
(Emerville, Kalamata, Kobe, Kocaeli, Loma Prieta, Northridge and Sakarya). The goal was to produce
the same responses at the frequencies of the most important structural Eigen modes, so that the
acting forces would be similar.

Figure 114 Spectra for 8 seismic events matched to the Eurocode target spectrum

The results of the analysis using the two methods were compared in both cross and down-aisle
directions for the following metrics: base shear forces and joint displacements in global horizontal
directions. From the results, the following conclusions can be derived:
Base shears
Direction Time-History Result (KN) New method result (KN) Ratio
Cross-aisle 17.48 21.49 0.813
Down-aisle 9.93 12.05 0.825
Table 33 Base shears

Displacements
Direction Time-History Result (m) New method result (m) Ratio
Cross-aisle 0.01923 0.022284 0.863
Down-aisle 0.08154 0.070978 1.149
Table 34 Displacements

The base shear results calculated by the new method are about 18% greater than the results of the
time-history dynamic analysis. The displacements results calculated by the new method are 14%
greater in the cross-aisle direction and 15% smaller in the down-aisle direction. It should be noted
that it was expected that the spectral method would produce greater results than the time-history
analysis. In conclusion, the scale of differences found is evaluated as acceptable and the new method
results appear to correlate favourably with the time-history results for a flexible model that exhibits
significant second-order behavior.
107
108
8. CONCLUSIONS
SEISRACKS2 has been a ground-breaking project within the RFCS program. It combined
experimental and numerical tools to improve, validate or invalidate the current seismic design rules
for racking systems, thanks to a sound collaboration between rack industry, design offices, and
universities. The project provided the most extensive experimental and numerical work ever
conducted in Europe, to investigate the component and global behavior of racking systems, and their
behaviour under seismic actions. Moreover, SEISRACKS2 project resulted in the development of
design guidelines and recommendations for seismic design of racking systems. Novel results of
SEISRACKS2 project can be summarized as:
1) A complete comparison has been performed between European (FEM 10.2.08) and American
(RMI 2008) seismic design provisions in European and American standards together with an
assessment of weaknesses in FEM 10.2.08
2) New testing procedures have been developed for:
o Qualification of beam-to-upright and column base-plate joints in cross-aisle
and down aisle directions
o Estimation of the global ductility of full scale rack systems in down-aisle and
cross-aisle directions
3) A component method has been developed to characterize the resistance and stiffness of beam-
to-upright connections.
4) By means of continuous monitoring (1 year), ambient vibration and hammer tests, structural
damping values have been calculated for the racks in an existing warehouse, and typical rack
configurations provided by 4 industrial partners of the project
5) For the first time in Europe, push over tests have been performed, considering real-scale
geometry and loading on different pallet racking systems, provided by 4 world-wide renowned
industrial partners of the project. Thanks to these tests, ductility performances of several types
of racks (braced and unbraced) have been investigated. Crucial improvements have been
suggested to consider in the seismic design of pallet racks:
o The behavior factors q=1.5 or q=2.0 for unbraced racks are reliable figures provided
that the soft storey collapse mechanism is prevented; this can be achieved improving
the bending strength and the rotation capacity at the base, either by installing floor
beams or base-plates adequately designed.
o The properties and the design of the floor connections derived from non-seismic
design appears to be not adequate for seismic conditions, also in consideration of the
variability of the axial force in the upright during a seismic event.
o The full scale test of the unbraced racks appear necessary to calibrate the numerical
models in particular with reference to the bending stiffness of the uprights.
o Stricter requirements have been introduced in EN16681 for the design of floor
connections of unbraced frames.
o A new rule has been added in EN16681 on the design of the bolted connections of
the diagonals, requiring a minimum distance between the bolt and the edge of the
diagonal equal to 2.5 times the bolt’s diameter, that shall be considered in
conjunction with the general rule for bolted connections requiring the bolt’s strength
at least 20% larger than the bearing strength. This rule is a constraint to design
stronger connections, in addition to improve the development of the plasticity at the
bolt’s hole
o The assessment of the behavior factor should be required when a rack non regular
in plan (braced rack) is designed using a behavior factor q>1.5; this assessment
should be done through a 3D numerical pushover analysis including torsion.
o For braced racks, the experimental calibration or validation of the numerical
pushover through a full scale test should be required in ductile design or when the
behavior factor in cross aisle direction is smaller than the one considered in down
aisle direction.
6) Calibrated numerical models have been developed based on the experimental studies, and
parameter studies have been performed, and significant observations have been made on the
filling rate, pallets location, merchandize-structure interaction.
7) Designer’s manual for seismic design of racks has been prepared
8) To overcome the limitations due consideration of 2nd order effects, a new and innovative
numerical analysis procedure has been developed (multimodal spectral stepwise non-linear
analysis), together with a new software tool.

109
110
9. EXPLOITATION AND IMPACT OF THE RESEARCH RESULTS
Based on the results of SEISRACKS2 project, a new European Standard (prEN16681 [22]) has been
submitted for publication. Moreover, new testing protocols for full scale push over testing, and
component testing have been developed for pallet racks. A new design guide has been produced. A
new software driver, named SEISRACKS2, has been developed.
The results of SEISRACKS2 project can also be a reference for more complex shelving systems such
as automated rack supported warehouses.
Publications and any other aspects concerning the dissemination of results are the following:
 Castiglioni A., Kanyilmaz A., Drei A., Hoffmeister B., Heinemeyer C., Degee H., Braham C.,
Vayas I., Adamakos K., Sesana S., Orsatti B., The “SEISRACKS2” EU-RFCS Research Project
“Seismic Behaviour of Steel Storage Pallet Racking Systems” - Part 1: Project Overview, XXIV
Giornate Italiane della Costruzione in Acciaio, 2013, Torino, Italia.

 Castiglioni A., Kanyilmaz A., Drei A., Hoffmeister B., Heinemeyer C., Degee H., Braham C.,
Vayas I., Adamakos K., The “SEISRACKS2” EU-RFCS Research Project “Seismic Behaviour of
Steel Storage Pallet Racking Systems” - Part 2: Experimental Activities, XXIV Giornate
Italiane della Costruzione in Acciaio, 2013, Torino, Italia.

 Castiglioni A., Kanyilmaz A., Drei A., Hoffmeister B., Heinemeyer C., Degee H., Braham C.,
Vayas I., Adamakos K., Papadopoulos N., The “SEISRACKS2” EU-RFCS Research Project
“Seismic Behaviour of Steel Storage Pallet Racking Systems” - Part 3: Numerical Activities,
XXIV Giornate Italiane della Costruzione in Acciaio, 2013, Torino, Italia.

 Martens K., Experimental and numerical study of storage racking systems in earthquake
situation, Master dissertation, Faculteit Ingernieurswetenschappen en Architectuur,
Universiteit Gent, Promotor: prof. Hervé Degée, June 2013.

 Castiglioni A., Kanyilmaz A., Seismic Behaviour of Steel Storage Pallet Racking Systems, 5.
Celik yapilar sempozyumu, Istanbul, Turkey, 2013.

 Brambilla G., Chiarelli G.P., “Steel storage pallet racking systems: full scale pushover testing
and numerical analyses”, MSc Thesis, Politecnico di Milano, April 2014.

 Castiglioni A., Kanyilmaz A., Angeretti M., Brambilla G., Chiarelli G.P., Bernuzzi C., 2014,
Experimental Results of Full Scale Push Over Tests of Project SEISRACKS2 (Seismic
Behaviour Of Steel Storage Pallet Racking Systems), Second European Conference on
Earthquake Engineering and Seismology, Istanbul 25-29 August 2014.

 Angeretti M., “Steel storage pallet racking systems in seismic areas: full scale testing and
numerical simulations”, MSc Thesis, Politecnico di Milano, December 2014.

 Castiglioni A., Kanyilmaz A., Drei A., Brambilla G., Chiarelli G.P., Hoffmeister B., Heinemeyer
C., Degee H., Braham C., Vayas I., Adamakos K., Sesana S., Orsatti B., The “SEISRACKS2”
EU-RFCS Research Project “Seismic Behaviour of Steel Storage Pallet Racking Systems”,
Costruzioni metalliche, January 2015.

 Adamakos K., Vayas I., Tragverhalten von Palettenregalsystemen unter

Erdbebenbeanspruchung, Stahlbau 83, No. 1, p.36-46, Ernst & Sohn, 2014.

 Adamakos K., Vayas I., Avgerinou S., Estimation of the behavior factor of steel storage pallet
racks, 4th ECCOMAS Thematic Conference on Computational Methods in Structural
Dynamics and Earthquake Engineering, Kos Island, Greece, 12–14 June 2013

 Kanyilmaz A., Castiglioni A., Drei A., Brambilla G., Chiarelli G.P., Angeretti M., “Experimental
seismic assessment of unbraced industrial steel racks”, Journal of Constructional Steel
Research, submitted in 2015.

 Kanyilmaz A., Castiglioni A., Drei A., Brambilla G., Chiarelli G.P., Angeretti M., “Seismic
assessment of braced industrial steel racks by means of full scale push over tests ”, Journal
of Constructional Steel Research, submitted in 2015.

111
112
LIST OF FIGURES
Figure 1 Flowchart of adopted Methodology during SEISRACKS2 project .................................... 7
Figure 2 New test setup developed within SEISRACKS2 ............................................................ 8
Figure 3 Setup of frictions tests (left: FEM tests; right: performed tests) .................................... 9
Figure 4 Column base tests .................................................................................................. 9
Figure 5 Global buckling of 2nd and 3rd diagonals accompanied with shearing of bolt ................. 10
Figure 6 Deformed shapes at the last loading increment of IP A (left) and IP C (right) structures . 10
Figure 7 – Rack collapses after Emilia-Romagna Earthquake 2012, Italy .................................... 13
Figure 8 Typical pallet rack configuration [1] ......................................................................... 13
Figure 9 Unit load on supports ............................................................................................. 15
Figure 10 Typical bracing schemes used for the upright frames ................................................ 16
Figure 11 Down-aisle bracings ............................................................................................. 16
Figure 12 Examples of bracing member connections with large restraint ................................... 17
Figure 13 Loads acting on base-plates .................................................................................. 18
Figure 14 Eccentricities of the schemes ................................................................................. 18
Figure 15 Kinematic model in RMI 2008 ................................................................................ 25
Figure 16 Forces causing overturning in FEM 10.2.08 [4] ........................................................ 26
Figure 17 Testing protocol for beam-upright connection in RMI 2008 ........................................ 27
Figure 18 Forces acting on pallet beams in FEM 10.2.08 [4] .................................................... 28
Figure 19 Upright frames to be used in the tests and numerical analyses. ................................. 29
Figure 20 Horizontal bracing system configuration .......................................................................... 29
Figure 21 Down-aisle test setup ........................................................................................... 31
Figure 22: Mtotal rotation characteristics from monotonic tests .................................................. 32
Figure 23: Mpush rotation characteristics from monotonic tests .................................................. 33
Figure 24 Failure modes from monotonic tests ....................................................................... 33
Figure 25 Force displacement curves of all products (A to D) for all load cases from monotonic tests
........................................................................................................................................ 34
Figure 26: Load deformation curves from monotonic and cyclic tests on systems with 100% payload
........................................................................................................................................ 34
Figure 27 Results with and without earthquake bolt and bolt application ................................... 35
Figure 28 Cross aisle test setup............................................................................................ 35
Figure 29: Components of stiffness and resistance of the fixing upright to slab .......................... 36
Figure 30 Sketch of principle test setup for down aisle tests .................................................... 37
Figure 31 Horizontal bracing system configuration .................................................................. 38
Figure 32 Deformation of the connecting hooks and disconnection of the pallet-beams. .............. 38
Figure 33 Tested system and horizontal bracing failure ........................................................... 39
Figure 34 Structure before and after failure during the push-over. ............................................ 39
Figure 35 Maximum displacement of uprights ........................................................................ 39
Figure 36 Types of uprights to be tested ............................................................................... 40
Figure 37 Moment – rotation measured during testing in RWTH. .............................................. 44
Figure 38 Pallet rack that is monitored (left) and accelerometer at point (right) ......................... 45
Figure 39 3D setup of the monitor system’s accelerometers..................................................... 45
Figure 40 Number of “significant” accelerations for sensor ....................................................... 46
Figure 41 Maximum peak acceleration for sensor ................................................................... 46
Figure 42 Summary of three criteria ..................................................................................... 46
Figure 43 Sensor 21; relative frequency of accelerations ......................................................... 47
Figure 44 Definition of event ................................................................................................ 47
Figure 45 Histogram of the accelerations ............................................................................... 47
Figure 46 Earthquake registrations ....................................................................................... 48
Figure 47 Vibration modes. .................................................................................................. 50
Figure 48 Investigated rack configurations ............................................................................ 50
Figure 49 Connection of the cross frames .............................................................................. 52
Figure 50 Position of the accelerometers ............................................................................... 52
Figure 51 Position of accelerometers. .................................................................................... 53
Figure 52 IP D, Pallet n°1, 6 boxes of projectors packed with cellophane with a total mass of 106.8
kg .................................................................................................................................... 54
Figure 53 Partner B, Box-plot representation of the natural frequency and damping ratio of the
palletized goods. ................................................................................................................ 55
Figure 54 Partner D, Box-plot representation of the natural frequency and damping ratio of the
palletized goods. ................................................................................................................ 55
Figure 55 Loading scheme and load balancer ......................................................................... 57
Figure 56 Load balancer, and test frame and assembly ........................................................... 58
Figure 57 General test layout ............................................................................................... 58
Figure 58 Force-displacement curves .................................................................................... 60
Figure 59 Residual permanent plastic deformations ................................................................ 60
Figure 60 Deformation of unbraced specimens ....................................................................... 61
Figure 61 Force-displacement curves .................................................................................... 61

113
Figure 62 Deformation of braced specimens........................................................................... 62
Figure 63 Benchmark numerical models for Typical a) Braced and b) Unbraced System .............. 67
Figure 64- Typical bracing configurations in down aisle and cross aisle ...................................... 68
Figure 65 Moment rotation curves of the different case studies ................................................ 69
Figure 66 Typical numerical model of upright ......................................................................... 69
Figure 67 Moment rotation diagram of upright and multilinear diagram, defining the plastic hinge
properties ......................................................................................................................... 69
Figure 68 Diagrams for the buckling and the bearing failure of a channel section ........................ 70
Figure 69 Bearing failure due to a) compression and b) tension................................................ 70
Figure 70 Circular Hollow Section with turnbuckles used as diagonal members ........................... 71
Figure 71 Local deformations of the Beam-ends ..................................................................... 71
Figure 72 Load displacement load for simulation of only the CHS member and that of the whole
assembly........................................................................................................................... 71
Figure 73 Static Analyses with and without P-Delta Effect assumption ....................................... 72
Figure 74 Pushover curves for the 9 case studies for the down aisle and cross aisle direction ....... 72
Figure 75 Moment rotation diagrams .................................................................................... 73
Figure 76 Experimental configuration for the characterization of the base-plates ........................ 74
Figure 77 Failure mode of upright for bending around the major axis ........................................ 75
Figure 78 Boundary condition of upright model ...................................................................... 75
Figure 79 Upright frame diagonal assembly for IP-B, with eccentricity between consecutive diagonals
........................................................................................................................................ 75
Figure 80 Base shear – Top displacement diagram of cross aisle frames .................................... 76
Figure 81 Total horizontal force vs. top displacement for rack configurations ............................. 77
Figure 82 Base Shear- Top displacement diagram .................................................................. 77
Figure 83 Elastic Spectra for the 44 ground motions (coloured) and some indicative EC8 design
spectra (black) ................................................................................................................... 78
Figure 84 Fractile curves and Fragility curve for the unbraced rack of IP A in the down aisle ........ 78
Figure 85 Fractile curves and Fragility curve for cross-aisle frames ........................................... 79
Figure 86 Typical pallet configuration .................................................................................... 80
Figure 87 Numerical model with a rigid pallet on the middle span ............................................. 81
Figure 88 Mass eccentricity vs. Sliding force, for different cases ............................................... 81
Figure 89 Model ................................................................................................................. 82
Figure 90 All the possible load cases of the pallet beams (boxes represent the pallets) ............... 83
Figure 91 Results for case 15 ............................................................................................... 84
Figure 92 Results for case 14 ............................................................................................... 84
Figure 93 Different cases regarding the position and the number of the pallets .......................... 84
Figure 94 Horizontal mid-span displacement- Axial Load for a pallet weight 8kN ........................ 85
Figure 95 Horizontal mid-span displacement- Axial Load for a pallet weight 12kN ....................... 85
Figure 96 Plasticity observed at the last step of test IPA1 ........................................................ 88
Figure 97 Unbraced rack exhibiting brittle behavior due to the formation of soft storey mechanism
below the first load level during the pushover tests ................................................................ 89
Figure 98 Failure of the vertical bracing diagonal’s connection caused by eccentricities ............... 89
Figure 99 Typologies of upright frames ................................................................................. 90
Figure 100 the 18 different examined load-cases.................................................................... 92
Figure 101 Comparisons between different types of analysis for down-aisle frames .................... 95
Figure 102 Comparisons between different types of analysis for cross-aisle frames..................... 96
Figure 103 Load history for beam-end connector tests ............................................................ 97
Figure 104 Different test set-ups for Beam-end connector test ................................................. 97
Figure 105 Comparison between beam-end connector tests with and without seismic bolt ........... 97
Figure 106 Load history for floor connection tests ................................................................... 98
Figure 107 Floor connection test setup (H1≈3m, H2≈0.5m)...................................................... 99
Figure 108 Actions transferred by the upright frame bracing diagonals to the upright ................. 99
Figure 109 Upright frame shear test setups from different codes ............................................ 100
Figure 110 Full scale test setup .......................................................................................... 101
Figure 111 ....................................................................................................................... 102
Figure 112 Newton-Raphson modified method convergence progress (2) ................................ 104
Figure 113 Rack model ..................................................................................................... 106
Figure 114 Spectra for 8 seismic events matched to the Eurocode target spectrum .................. 107
Figure 115 Measured load displacement hysteresis and determination of friction (Industrial partnerA)
...................................................................................................................................... 123
Figure 119 Measured load displacement hysteresis and determination of friction (Industrial partnerB)
...................................................................................................................................... 124
Figure 120 Measured load displacement hysteresis and determination of friction (Industrial partnerC)
...................................................................................................................................... 125
Figure 118 Measured load displacement hysteresis and determination of friction (Industrial partnerD)
...................................................................................................................................... 126
Figure 119 Test results pallets with 50% max service load .................................................... 126
Figure 123: Test specimen after Test (Product A) ................................................................. 127
114
Figure 124: Moment Rotation curves for loading in down aisle direction (Product A) ................. 127
Figure 125: Test specimen after Test (Product B) ................................................................. 127
Figure 126: Moment Rotation curves for loading in down aisle direction (Product B) ................. 128
Figure 127: Test specimen after Test (Product C) ................................................................. 128
Figure 128: Moment Rotation curves for loading in down aisle direction (Product C) ................. 129
Figure 129: Test specimen after Test (Product D) ................................................................. 129
Figure 130: Moment Rotation curves for loading in down aisle direction (Product D) ................. 130
Figure 131: Tests on product A .......................................................................................... 131
Figure 132: Tests on product B .......................................................................................... 132
Figure 133: Tests on product C .......................................................................................... 133
Figure 134: Tests on product D .......................................................................................... 134
Figure 132 Top view of the setup with displacement senor’s positions ..................................... 135
Figure 133 Top view of the setup of the structure C with displacement senor’s positions ........... 135
Figure 137 Tests on product A ........................................................................................... 136
Figure 138 Tests on product B ........................................................................................... 136
Figure 139 Longitudinal displacements pushover test on product C ......................................... 137
Figure 140 Tests on product D ........................................................................................... 137
Figure 141 Test set-up ...................................................................................................... 138
Figure 139 ....................................................................................................................... 138
Figure 140 ....................................................................................................................... 139
Figure 141 ....................................................................................................................... 139
Figure 142 ....................................................................................................................... 140
Figure 143 ....................................................................................................................... 140
Figure 144 Total force-vs.-displacement [mm] at 8m– Configuration with bolt without failure. ... 140
Figure 145 ....................................................................................................................... 141
Figure 146 ....................................................................................................................... 141
Figure 147 Failure modes of cross-frame D/3 without connecting bolts ................................... 141
Figure 151 - General rack geometry and pallet configuration ................................................. 142
Figure 149 - Rigid base support: rendering plan, rendering top view, on site view .................... 142
Figure 150 Rack geometry ................................................................................................. 143
Figure 151 Deformed shape summary ................................................................................. 143
Figure 155 Last step deformed shapes of the specimens ....................................................... 144
Figure 153 Definitions to estimate yield force and displacements ............................................ 144
Figure 154 Methods to estimate behaviour factor – q ........................................................... 144
Figure 155 q-factor diagram for methods 1 and 2 ................................................................. 145
Figure 156 q-factor diagram for method 3 ........................................................................... 145
Figure 157 Difference between two types of beam-upright joint ............................................. 145
Figure 161 Results for case 15 ........................................................................................... 147
Figure 162 Results for case 14 ........................................................................................... 147
Figure 163 Results for case 13 ........................................................................................... 147
Figure 164 Results for case 11 ........................................................................................... 147
Figure 165 Results for case 17 ........................................................................................... 147
Figure 166 Results for case 6 ............................................................................................. 147
Figure 167 Results for case 5 ............................................................................................. 147
Figure 168 Results for case 4 ............................................................................................. 147
Figure 169 Results for case 24 ........................................................................................... 147
Figure 170 Results for case 23 ........................................................................................... 147
Figure 171 Results for case 22 ........................................................................................... 148

115
116
LIST OF TABLES
Table 1 Importance factors according to FEM 10.2.08 [4] ........................................................ 20
Table 2 Seismic design category as per ASCE-7 [16] 2010 chapter 11.6.................................... 23
Table 3 Risk category category as per ASCE-7 [16] 2010 ........................................................ 23
Table 4 Methods of analysis in FEM 10.2.08 [4] ...................................................................... 24
Table 5 Consequences of structural regularity on seismic analysis and design in FEM 10.2.08 [4] . 24
Table 6 Geometrical configuration of the racks and design parameters ............................................ 30
Table 7 Estimated ductility .................................................................................................. 42
Table 8 Comparison of values for diagonal’s cross-sections to be used as a result of standard tests
and tests realised in the lab during SEISRACKS2. ................................................................... 43
Table 9 components that contribute to the failure ................................................................... 44
Table 10 Event classification ................................................................................................ 48
Table 11: Max acc. in sensor 21 – Average Damping (%) in sensors 17 and 21 .......................... 49
Table 12 Max acc. in sensor 4 - Damping in sensor 4 .............................................................. 49
Table 13 Results of free oscillation tests at different erection stages ......................................... 51
Table 14 Natural frequencies (B) .......................................................................................... 52
Table 15 natural frequencies (D) .......................................................................................... 54
Table 16 Unbraced racks, q factors according to three different methods................................... 64
Table 17 Braced racks, q factors according to three different methods – IPC2 is the cable braced rack
........................................................................................................................................ 64
Table 18 Modal Results for an unbraced rack with and without the 2nd order effects .................... 68
Table 19 Base shear and top displacements for the examined unbraced rack ............................. 68
Table 20 Estimated q factor values for the fully loaded upright bracing systems of each IP .......... 77
Table 21 maximum estimated q factor for the IP A & B for the down aisle direction .................... 79
Table 22 maximum estimated q factor for the IP A & D for the cross aisle direction .................... 79
Table 23 Correction coefficients for horizontal bending, proposed in FEM10.02.8 ........................ 82
Table 24 Possible cases, with respect to the different values of the examined parameters ........... 83
Table 25 Theoretical values for the horizontal bending moments, for each case from A to E ......... 83
Table 26 ED2 factors ............................................................................................................ 94
Table 27 of the frequencies of the free vibrations and the average values of the damping from various
projects ............................................................................................................................ 94
Table 28 3D geometrical stiffness matrix formulation ............................................................ 104
Table 29 Scope of each of the delivered modules ................................................................. 105
Table 30 Rack section data ................................................................................................ 106
Table 31 Spectral data ...................................................................................................... 106
Table 32 Comparison between models ................................................................................ 107
Table 33 Base shears ........................................................................................................ 107
Table 34 Displacements .................................................................................................... 107
Table 35 Unbraced racks, main section properties of elements ............................................... 146
Table 36 Braced racks, main section properties of elements .................................................. 146

117
118
REFERENCES

[1] Castiglioni C.A., Seismic behaviour of steel storage pallet racking systems, Structural
Engineering Departement of Politecnico di Milano, Milano, 2008.

[2] EN 1993 Eurocode 3, Design of steel structures, 2003.

[3] EN 15512:2009 Steel Static Storage Systems - Adjustable Pallet Racking Systems - Principles
For Structural Design, 2009.

[4] FEM 10.2.08, Recommendations for the Design of Static Steel Pallet Racking in Seismic
Condition, Federation Europeen de la Manutention, Version 1.04, 2011.

[5] EN 1998 Eurocode8, Design of structures for earthquake resistance, 2003.

[6] SEISRACKS RFSR-CT-2004-00045, Final Report, Research Program of the Research Fund for
Coal and Steel RTD, may 2007.

[7] ACI 318-05 Building Code Requirements for Structural Concrete, ACI Committee 318,
American Concrete Institute, Farmington Hills, MI, 2005

[8] ETAG 001 Guideline for european technical approval of metal anchors for use in concrete,
EOTA, 2013.

[9] RMI-2008, MH16.1:2008 ANSI-RMI “Specification for the Design, design and Utilization of
Industrial Steel Storage Racks” and Commentary, The Rack Manufacturer Institute, 2008.

[10] FEMA 460 FEMA 460, Seismic Considerations for Steel Storage Racks Located in Areas
Accessible to the Public, Federal Emergency Management Agency, Washington, D.C, 2005.

[11] USGS Open-File Report 01-437 “Earthquake Spectral Response Acceleration Maps”.

[12] NEHRP Recommended Provisions for Seismic Regulations for New Buildings and Other
Structures, Building Seismic Safety Council, 2001.

[13] EN 15629 Steel static storage systems - Specification of storage Equipment, CEN, 2008.

[14] EN1990 – Eurocode: Basis of structural design, CEN, 2002

[15] ECOLEADER Research Program for free access to large scale testing facilities, Research
Project “Seismic Design of Pallet Racking”, 2002.

[16] ASCE/SEI 7-10, Minimum Design Loads for Buildings and Other Structures, American Society
of Civil Engineers, Reston, Virginia, 2010.

[17] ECCS N° 45, Recommended Testing Procedure For Assessing the Behaviour of Structural
Steel Elements under Cyclic Loads, Technical Committee 1 – Structural Safety and Loadings
Technical Working Group 1.3 – Seismic Design, 1986.

[18] D’Annibale G., Fotogram: restituzione di modelli 2D e 3D da sorgenti fotografiche,

D.Flaccovio,Palermo IT, 2003.

[19] Open System for Earthquake Engineering Simulation (OpenSees), Pacific Earthquake
Engineering Research Center, University of California, Berkeley, CA.

[20] Computers and Structures Inc., CSI Analysis Reference Manual for SAP2000 ®, ETABS®, and
SAFE®, Berkeley CA., 2010.

[21] ABAQUS User Manual. Providence, RI, USA: DS SIMULIA Corp; 2009.

[22] PrEN16681 TC 344 WI 00344007 Steel static storage systems — Adjustable pallet racking
systems — Principles for seismic design, CEN, 2013.

119
[23] CEN-TC 344 N 57 Adoption of WI 00344006 “STEEL static storage systems- Racking and
Shelving Systems-terms and definitions” draft resolution, 2006.

[24] FEMA P695, Quantification of Building Seismic Performance Factors, Federal Emergency
Management Agency, Washington, D.C, 2009.

[25] prFEM 10.2.08, Recommendations for the Design of Static Steel Pallet Racking in Seismic
Condition, Federation Europeen de la Manutention, draft, 2005.

[26] FEM 10.2.02, Design of Static Steel Pallet Racking, Federation Europeenne de la Manutention
Section X Equipment et Proceedes de Stockage, 2000, Chairman: CJ Tilburgs

[27] FEM 10.2.03, Specifiers Guidelines, Federation Europeenne de la Manutention Section X

Equipment et Proceedes de Stockage, 2000, Chairman: CJ Tilburgs

[28] FEM 10.2.04, Users Code, Federation Europeenne de la Manutention Section X Equipment et
Proceedes de Stockage, 2001, Chairman: CJ Tilburgs

[29] FEM 10.2.05, Guidelines for working safely with lift trucks in pallet racking installations,
Federation Europeenne de la Manutention Section X Equipment et Proceedes de Stockage,
2001, Chairman: CJ Tilburgs

[30] FEM 10.2.06, Design of Static Steel Shelving, Federation Europeenne de la Manutention
Section X Equipment et Proceedes de Stockage, 2001, Chairman: CJ Tilburgs

[31] FEM 10.3.01, Pallet racking: Tolerances, Deformations and Clearances, 2001, Chairman: CJ
Tilburgs

[32] John A. Blume & Associates, Seismic Design Examples of Industrial Storage Racks, Report
prepared for the Rack Manufacturer’s Institute, San Francisco CA., 1973.

[33] Krawinkler H., Cofie N.G., Astiz M.A. and Kircher C.A., Experimental Study on the Seismic
Behaviour of Industrial Storage Racks, Report No. 41, The John A. Blume Earthquake
Engineering Center, Department of Civil Engineering, Stanford University, Stanford CA.,
1979.

[34] Chen C.K., R.E. Scholl and J.A. Blume, Seismic Study of Industrial Storage Racks, Report
prepared for the National Science Foundation and for the Rack Manufacturers Institute and
Automated Storage and Retrieval Systems (sections of the Material Handling Institute), John
A. Blume & Associates, San Francisco CA., 1980.

[35] Filiatrault, A., Shake-Table Tests of Storage Racks and Contents. Presentation material at the
April 12, 2001 Seismic Safety Commission Hearing on Industrial Storage Racks, San
Francisco, CA, 2001.

[36] Filiatrault, A., and Wanitkorkul, A., Shake-Table Testing of Frazier Industrial Storage Racks,
Report No. CSEE-SEESL-2005-02, Structural Engineering and Earthquake Simulation
Laboratory, Departmental of Civil, Structural and Environmental Engineering, University at
Buffalo, State University of New York, 83 p., 2004.

[37] Baldassino, N. and Bernuzzi C., Analysis and Behaviour of Steel storage Pallet Racks, Thin-
Walled Structures, 37 (4), 277-304, 2000.

[38] Baldassino N. and Zandonini R., Numerical and Experimental Analysis of Base-plate
Connections of Steel Storage Pallet Racks, Proc. of XVIII Conference C.T.A., Venezia, 2001.

[39] Bernuzzi C. and Castiglioni C.A., Experimental Analysis on the Cyclic Behaviour of Beam-to-
Column Joints in Steel Storage Pallet Racks, Thin-Walled Structures, (39), 841-859, 2001.

[40] Castiglioni C.A., Dynamic Tests on Steel Pallet Racks, Costruzioni Metalliche, 55 (3), 35-44,
2003.

120
[41] Degee H., Denoel V., Castiglioni C.A., “Seismic Behaviour of Storage racks made of Thin-
Walled Steel members”, VII European Conference on Structural Dynamics, Eurodyn 2008,
Southampton, July 2008.

[42] Ballio G., Bernuzzi C., Castiglioni C.A., “An approach for the seismic design of steel storage
pallet racks”, Stahlbau, Nov. 1999

[43] Agatino M.R., Bernuzzi C., Castiglioni C.A., “Joints under cyclic reversal loading in steel
storage pallet racks”, Proc. XVIII C.T.A. Conference, Venezia, vol. 2, pag. 105-114,
September 2001.

[44] Castiglioni C.A., N.Panzeri, J.C.Brescianini, P.Carydis, “Shaking table tests on steel pallet
racks”, Proc. STESSA 2003, Napoli, pp. 775-781, June 2003.

[45] Brescianini J.C., Castiglioni C.A., N.Panzeri, “Dynamic experimental tests on steel pallet
racks”, Proceedings of CTA, Genova, pp. 107-116, September 2003.

[46] Castiglioni C.A., “Seismic behaviour of steel storage racks”, Proceedings of the IV Congresso
de Construcao Metalica e Mista, Lisbon, pp 41-62, 4-5 December 2003.

[47] Brescianini J.C., Castiglioni C.A., “Caratterizzazione del comportamento dinamico di

scaffalature metalliche porta-pallets mediante push-over tests”, Proceedings of XXI C.T.A.,
Catania, pp. 275-282, September 2007.

[48] Castiglioni C.A., Calado L., Carydis P., Degee H., Negro P., Rosin I., “Seismic Behaviour of
Steel Storage Racking Systems”, Proceedings of STESSA09, paper 0158, Philadelphia, August
2009.

[49] Castiglioni C.A., Calado L., Carydis P., Degee H., Negro P., Rosin I., “Seismic Behaviour of
Steel Storage Racking Systems”, Proceedings of XXII CTA, Padova, September 2009.

[50] Calado L., Castiglioni C.A., Drei A., “Cyclic tests of beam-upright connections in racking
systems with a new hybrid procedure”, Proc. of STESSA 2012, Chile, paper n. 006, pp. 53-
59, Jan. 2012.

[51] Castiglioni A., Kanyilmaz A., Drei A., Hoffmeister B., Heinemeyer C., Degee H., Braham C.,
Vayas I., Adamakos K., Sesana S., Orsatti B., The “SEISRACKS2” EU-RFCS Research Project
“Seismic Behaviour of Steel Storage Pallet Racking Systems” - Part 1: Project Overview, XXIV
Giornate Italiane della Costruzione in Acciaio, 2013, Torino, Italia.

[52] Castiglioni A., Kanyilmaz A., Drei A., Hoffmeister B., Heinemeyer C., Degee H., Braham C.,
Vayas I., Adamakos K., The “SEISRACKS2” EU-RFCS Research Project “Seismic Behaviour of
Steel Storage Pallet Racking Systems” - Part 2: Experimental Activities, XXIV Giornate
Italiane della Costruzione in Acciaio, 2013, Torino, Italia.

[53] Castiglioni A., Kanyilmaz A., Drei A., Hoffmeister B., Heinemeyer C., Degee H., Braham C.,
Vayas I., Adamakos K., Papadopoulos N., The “SEISRACKS2” EU-RFCS Research Project
“Seismic Behaviour of Steel Storage Pallet Racking Systems” - Part 3: Numerical Activities,
XXIV Giornate Italiane della Costruzione in Acciaio, 2013, Torino, Italia.

[54] Castiglioni A., Kanyilmaz A., Seismic Behaviour of Steel Storage Pallet Racking Systems, 5.
Celik yapilar sempozyumu, Istanbul, Turkey, 2013.

[55] Brambilla G., Chiarelli G.P., “Steel storage pallet racking systems: full scale pushover testing
and numerical analyses”, MSc Thesis, Politecnico di Milano, April 2014.

[56] Castiglioni A., Kanyilmaz A., Angeretti M., Brambilla G., Chiarelli G.P., Bernuzzi C., 2014,
Experimental Results of Full Scale Push Over Tests of Project SEISRACKS2 (Seismic Behaviour
Of Steel Storage Pallet Racking Systems), Second European Conference on Earthquake
Engineering and Seismology, Istanbul 25-29 August 2014.

121
[57] Angeretti M., “Steel storage pallet racking systems in seismic areas: full scale testing and
numerical simulations”, MSc Thesis, Politecnico di Milano, December 2014.

[58] Castiglioni A., Kanyilmaz A., Drei A., Brambilla G., Chiarelli G.P., Hoffmeister B., Heinemeyer
C., Degee H., Braham C., Vayas I., Adamakos K., Sesana S., Orsatti B., The “SEISRACKS2”
EU-RFCS Research Project “Seismic Behaviour of Steel Storage Pallet Racking Systems”,
Costruzioni metalliche, January 2015.

[59] Martens K., Experimental and numerical study of storage racking systems in earthquake
situation, Master dissertation, Faculteit Ingernieurswetenschappen en Architectuur,
Universiteit Gent, Promotor: prof. Hervé Degée, June 2013.

[60] Adamakos K., Vayas I., Tragverhalten von Palettenregalsystemen unter

Erdbebenbeanspruchung, Stahlbau 83, No. 1, p.36-46, Ernst & Sohn, 2014.

[61] Adamakos K., Vayas I., Avgerinou S., Estimation of the behavior factor of steel storage pallet
racks, 4th ECCOMAS Thematic Conference on Computational Methods in Structural
Dynamics and Earthquake Engineering, Kos Island, Greece, 12–14 June 2013

[62] Kanyilmaz A., Castiglioni A., Drei A., Brambilla G., Chiarelli G.P., Angeretti M., “Experimental
seismic assessment of unbraced industrial steel racks”, Journal of Constructional Steel
Research (submitted on Feb. 2015).

[63] Kanyilmaz A., Castiglioni A., Drei A., Brambilla G., Chiarelli G.P., Angeretti M., “Experimental
seismic assessment of unbraced industrial steel racks”, Journal of Constructional Steel
Research (submitted on Feb. 2015).

[64] Research Report No R899, Experimental test on steel storage rack components, The
University of Sydney, School of Civil Engineering, Australia, October 2009.

122
ANNEX FOR CHAPTER 2
Diagrams and tests set up related to the component tests presented in chapter 2 are reported in this
annex. Three types of components tests have been carried out:
1. Joint tests on beam-to-upright connections: in this part the graphics of the measured load
displacement hysteretic behavior observed during beam-to-upright connection test
performed with different payloads are presented. Furthermore, an estimation of friction
coefficient µ is showed.
2. Column-base tests: in this paragraph the results of tests performed on column bases, both
in down aisle and cross aisle direction, are presented. The tests are performed with an
innovative set up respect to those proposed by EN15512 as reported in chapter 2.
3. Substructure tests: the rack is broke down in different substructures individually tested.
Tests set up and diagrams of the measured values are presented for each partner.

JOINT TESTS ON BEAM-TO-UPRIGHT CONNECTIONS

Resulting curves

Figure 115 Measured load displacement hysteresis and determination of friction (Industrial partnerA)

123
Figure 116 Measured load displacement hysteresis and determination of friction (Industrial partnerB)

124
Figure 117 Measured load displacement hysteresis and determination of friction (Industrial partnerC)

125
Figure 118 Measured load displacement hysteresis and determination of friction (Industrial partnerD)

Figure 119 Test results pallets with 50% max service load
126
COLUMN-BASE TESTS in Down-aisle direction

Push over, N=0 Push over, N=-24 kN Push over, N=-48 kN

Figure 120: Test specimen after Test (Product A)

M Push over results M

900 100 0

800
800

600

700

400

600

200

500

400

300
-0.20 -0.10 0.00
-200
0.10 0.20
-400

200

-600

100

-800

0 N=0
-1000

0.00 0.10 0.20 0.30 0.40 Rotation in rad

Rotation in rad
N=0 N = -24 kN N = -48 kN Cyclic Push over

M M
100 0 100 0

800 800

600 600

400 400

200 200

0 0

-0.20 -0.10 0.00

-200
0.10 0.20 -0.20 -0.10 0.00
-200
0.10 0.20
-400 -400

-600 -600

-800 -800

N = -24 kN N = -48 kN
-1000 -1000

Rotation in rad Rotation in rad

Cyclic Push over Cyclic Push over

Figure 121: Moment Rotation curves for loading in down aisle direction (Product A)

Push over, N=0 Push over, N=-24 kN Push over, N=-48 kN

Figure 122: Test specimen after Test (Product B)

127
M Push over results M
900 100 0

800
800

600

700

400

600

200

500

400

300
-0.20 -0.10 0.00
-200
0.10 0.20
-400

200

-600

100

-800

0 N=0
-1000

0.00 0.10 0.20 0.30 0.40 Rotation in rad

Rotation in rad
N=0 N = -24 kN N = -48 kN Cyclic Push over

M M
800 800

600 600

400 400

200 200

0 0

-0.20 -0.10 0.00

-200
0.10 0.20 -0.20 -0.10 0.00
-200
0.10 0.20
-400 -400

-600 -600

-800 -800

N = -24 kN N = -48 kN
-1000 -1000

Rotation in rad Rotation in rad

Cyclic Push over Cyclic Push over

Figure 123: Moment Rotation curves for loading in down aisle direction (Product B)

Push over, N=0 Push over, N=-24 kN Push over, N=-48 kN

Figure 124: Test specimen after Test (Product C)

128
M Push over results M
250 150

200
100

150

50

100

50
-0.20 0.00 0.20 0.40
-50

-100

0.00
-50
0.10 0.20 0.30 0.40
-150

-100 N=0
Rotation in rad -200

Rotation in rad
N=0 N = -24 kN N = -48 kN Cyclic Push over

M M
150 250

200

100

150

50 100

50

-0.20 -0.10 0.00 0.10 0.20

-50
-0.20 -0.10 0.00
-50
0.10 0.20
-100

-100

-150

N = -24 kN N = -48 kN
-150 -200

Rotation in rad Rotation in rad

Cyclic Push over Cyclic Push over

Figure 125: Moment Rotation curves for loading in down aisle direction (Product C)

Push over, N=0 Push over, N=-24 kN Push over, N=-48 kN

Figure 126: Test specimen after Test (Product D)

129
M Push over results M
450 500

400
400

350

300

300

200

250

100

200

150

-0.20 -0.10 0.00

-100
0.10 0.20
100

-200

50

-300

-400

0.00
-50

0.10 0.20 0.30 0.40 N=0

-500

Rotation in rad Rotation in rad

N=0 N = -24 kN N = -48 kN Cyclic Push over

M M
500 400

400

300

300

200

200

100

100

-0.20 -0.10 0.00

-100
0.10 0.20
-0.20 -0.10 0.00
-100
0.10 0.20
-200

-200

-300

-300

-400

-400

N = -24 kN N = -48 kN
-500 -500

Rotation in rad Rotation in rad

Cyclic Push over Cyclic Push over

Figure 127: Moment Rotation curves for loading in down aisle direction (Product D)

130
COLUMN-BASE TESTS in Cross aisle direction

Push over (Pushing) Push over (Pulling) Cyclic

20 120

10

100

0
Force / Displacement

Force / Displacement
0
-10
100 200 300 400 500 80

-20

F_vert 60

F_vert
-30

d_vert d_vert
40

-40

-50

20

-60

-70

0 100 200 300 400 500 600 700

-80 -20

Time in s Time in s

Time history of Push over test (Pushing) Time history of Push over test (Pulling)
100 Moment
80 600

60
500
Force / Displacement

40

400

20

300

200

0-20
500 1000 1500 2000 2500
100

-40
F_vert
0

d_vert
-60

-0.05 0.00
-100
0.05 0.10 0.15 0.20
-80

-200

-100

Time in s -300

-400

Rotation in rad
Time history of cyclic test
Cyclic PO - press PO - Tension

Observations: Very stiff connector. In Monotonic tests resistance is controlled by the stiffness of the section.
In cyclic tests cracks at the the upper bolt (to web) and at the neighbouring notches.
Figure 128: Tests on product A

131
 Push over (Pushing) Push Over (Puling) Cyclic
60 120

F_vert
100

40

d_vert
Force / Displacement

Force / Displacement
80

20

F_vert
60

0
d_vert
0 500 1000 1500 2000 40

-20

20

-40

-60
0 200 400 600 800 1000
-20

-80 -40

Time in s Time in s

Time history of Push over test (Pushing) Time history of Push over test (Pulling)
20 Moment
600

10

500
Force / Displacement

0 500 1000 1500 2000 2500 3000 3500 400

-10

300

-20

200

-30
100

-40

F_vert -0.10 -0.05 0.00 0.05 0.10 0.15

-100

d_vert
-50

-200

-60

Time in s -300

-400

Rotation in rad
Time history of cyclic test
Cyclic PO - press PO - Tension

Observations: When pushing stiffness and resistance is controlled by local buckling of the free edges of the
flanges. When pulling stiffness and resistance is controlled by the bending of the connector. Sliding in holes
occurs when normal forces changes from tension to compression.
Figure 129: Tests on product B

132
 Push over (Pushing) Push Over (Puling) Push over (Cyclic)
15 100

10

80

5
Force / Displacement

Force / Displacement
0 60

0-5
200 400 600 800 1000 1200
40

-10

-15

F_vert 20

F_vert
-20 d_vert d_vert
0

0 200 400 600 800 1000 1200 1400 1600

-25

-30 -20

Time in s Time in s

Time history of Push over test (Pushing) Time history of Push over test (Pulling)
10 Moment
120

0 100 200 300 400 500 600 700 800

Force / Displacement

100

-10

80

-20

60

-30

40

F_vert
-40

d_vert 20

-50

-60
-0.10 -0.05 0.00 0.05 0.10 0.15 0.20
Time in s
-20

-40

Rotation in rad
Time history of Push over test with front side to back
(Pushing) PO - press 2 PO - press PO - Tension

Observations: Stiffness and resistance is dominated by sliding of bolts and bending of connector plate. As the
notches are used as holes for the bolts and thus oversized. The industrial partner confirms that this
connector is not applied in earthquake regions.
Due to the low loading capacity cyclic tests could not be performed.
Figure 130: Tests on product C

133
 Push over (Pushing) Push Over (Puling) Cyclic
10 100

80

0
-10
100 200 300 400 500 600 700
Force / Displacement

Force / Displacement
60

-20

40

-30

20

-40

-50

F_vert F_vert
0 100 200 300 400 500 600 700 800
d_vert d_vert
-20

-60

-40

-70

-80 -60

Time in s Time in s

Time history of Push over test (Pushing) Time history of Push over test (Pulling)
30 Moment
300

20

250
Force / Displacement

10

200

150

0 200 400 600 800 1000 1200

-10 100

F_vert 50

-20

d_vert
0

-0.10 -0.05 0.00 0.05 0.10 0.15 0.20

-30

-50

-40

Time in s -100

-150

Rotation in rad
Time history of cyclic test
Cyclic PO - Push PO - Tension

Observations: When pushing the resistance is controlled by buckling of the free edges of the flanges beeing
in contact with the base-plate. When pulling the resistance is controlled by sliding of the bolts which are
installed in the notches for beam connector. In the cyclic tests the rotation capacity is restricted by cracks in
base-plate structure.
Figure 131: Tests on product D

It must be noticed that higher values of rotational angles in cross aisle plane observed during base
connections tests cannot be achieved in reality when base plates are part of a rack frame, due to the
contribution to stiffness offered by the diagonals of the upright frame.

134
SUBSTRUCTURE TESTS
Test setup
The load is applied by means of an additional beam pushing against steel-stops welded at mid-length
of the pallet-beams.

The test setup for the structures of partners A, B and D is shown on the following picture.

The displacements are measured at 10 different positions, all at the pallet-beam’s height. 3 sensors
measure the longitudinal displacements of the front cross-frame (n°1, 10 and 9) and 3 sensors
measure the displacements of the rear cross-frame (n°4, 5 and 6). The transversal displacements
are measured on each external upright of the front cross-frame (n°2 and 8) and of the rear cross-
frame (n°3 and 7).

Figure 132 Top view of the setup with displacement senor’s positions

The setup for the structure of partners C is different. The tested configuration consists in 2 spans –
1 level in order to preserve the vertical bracing in both longitudinal directions. Displacement sensors
are placed at 7 positions as seen on the following picture:

Figure 133 Top view of the setup of the structure C with displacement senor’s positions

135
Results

Longitudinal displacements pushover test Transversal displacements pushover test

Longitudinal displacements cyclic test with payload Longitudinal displacements cyclic test without payload
Figure 134 Tests on product A

Longitudinal displacements pushover test 2 Transversal displacements pushover test 2

Longitudinal displacements cyclic test with payload Longitudinal displacements cyclic test without payload
Figure 135 Tests on product B

136
 Figure 136 Longitudinal displacements pushover test on product C

Longitudinal displacements pushover test Transversal displacements pushover test 1

Longitudinal displacements cyclic test with payload Longitudinal displacements cyclic test without
payload
Figure 137 Tests on product D

Cross frames tests

Test setup
In order to have the best test/model comparison, the base shear will be applied as suggested in EC8.
This will allow having a setup reproducing what happens in reality.
But for the ease of the test execution the frame will be tested horizontally. Vertical side guides will
be placed on each “beam level” to avoid out-of-plane buckling.
The forces will only be applied on one upright at a time (the upper upright when pushing towards
the floor and the lower upright when lifting up). A hoop will connect 2 plates outside the frame with
2 long screws; this hoop will be useful for the cyclic tests.

The base-plate is rigidly fixed to the upright. In real situation uprights are submitted to an initial
compression due to the load of the merchandized goods stored on the pallets. Transverse loads on
cross-frames, provoked by an earthquake, increase the compression in one upright and decrease the
compression in the other which normally limits the tension effect. The upper upright, in tension, has
been welded to the base-plate to avoid the upright torn from its base-plate by bolt shearing.

137
 The load was applied by only one jack (easier to control – especially later for cyclic) acting on one
beam which itself acts on 2 beams such as to have a triangular linear load distribution:

Figure 138 Test set-up

The only issue is that it is not possible to measure the individual forces. Only the total applied force
is known and we suppose a triangular linear load distribution because the loading system is statically
determined.

The measurements taken during the test are the following:

 displacement transducers are placed at levels 0m, 2m, 4m, 6m and 8m;
 relative displacement transducers are placed between both uprights at levels 2m and 4m;
 rotation of the column on which the cross-frame is fixed;
 a load cell is placed under the application points of the jacks.

Results
The figures show the evolution of the displacement of each beam-level from which the additional
displacement due to the deflection of the supporting column has been subtracted.

Partner A (frame for high seismic zone)

Total force -vs.- displacement [mm] beam-level 8m for Failure modes.

the push and cyclic tests.
Figure 139

138
Partner B (frame for moderate-to-high seismic zone)

Total force -vs.- displacement [mm] at beam-level 8m. Failure modes.

Figure 140

Partner C (frame for moderate seismic zone)

Push downwards

Pull upwards

Total force -vs.- displacement [mm] at beam-level 8m Failure modes.

Figure 141

139
Partner C (frame for high seismic zone)

Total force -vs.- displacement [mm] at beam-level 8m. Failure modes.

Figure 142

Partner D (frame for low seismic zone)

Total force -vs.- displacement [mm] at beam-level 8m. Failure modes.

Figure 143

Partner D (frame for moderate seismic zone)

Figure 144 Total force-vs.-displacement [mm] at 8m– Configuration with bolt without failure.

140
 Total force -vs.- displacement [mm] at 8m – Failure mode of cross-frame D/2 without
Configuration without bolt with failure.
connecting bolts.
Figure 145

Partner D (frame for high seismic zone)

Total force -vs.- displacement [mm] at 8m, Total force -vs.- displacement [mm] at 8m,
Configuration with bolt without failure. Configuration without bolt with failure.
Figure 146

Figure 147 Failure modes of cross-frame D/3 without connecting bolts

141
ANNEX FOR CHAPTER 4
This annex is related to the full scale pushover tests described in chapter 4. Geometry, sectional
properties, top displacements diagrams, q-factor estimation methods and capacity curves are
reported for each partner.

Figure 148 - General rack geometry and pallet configuration

Figure 149 - Rigid base support: rendering plan, rendering top view, on site view

142
IPA1 IPB1 IPC1 IPD1

IPA2 IPB2 IPC2 IPD2

Figure 150 Rack geometry

IPA1 IPB1 IPC1 IPD1

IPA2 IPB2 IPC2 IPD2

Figure 151 Deformed shape summary

143
IPA1 IPB1 IPC1 IPD1

IPA2 IPB2 IPC2 IPD2

Figure 152 Last step deformed shapes of the specimens

Definition A Definition B Definition C

Figure 153 Definitions to estimate yield force and displacements

Method 1 Method 2 Method 3

Figure 154 Methods to estimate behaviour factor – q

144
IPA1 IPB1 IPC1 IPD1

IPA2 IPB2 IPC2 IPD2

Figure 155 q-factor diagram for methods 1 and 2

IPA1 IPB1 IPC1 IPD1

IPA2 IPB2 IPC2 IPD2

Figure 156 q-factor diagram for method 3

Figure 157 Difference between two types of beam-upright joint

145
 TABLES

Specimen
Section Properties Units
A B C D
Upright
2
Gross section area (A) mm 674 686 460 424.7
Section modulus about y axis (Wy,inf,sup) mm3 22430 20100 13438 12426
Material - S420MC S 355 S350 S355MC
Pallet Beam
Gross section area (A) mm2 762 659 588,7 516
3
Section modulus about y axis (Wy,inf,sup) mm 32670 24200 17122 16415.2
Material - S355MC S 275 N/NL S355 S355MC
Upright Bracing
Gross section area (A) mm2 134.3 357.5 159 133
Material - S235JR S 275 S250 S235
Table 35 Unbraced racks, main section properties of elements

Specimen
Section Properties Units
A B C D
Upright
Gross section area (A) mm2 674 536 784 979.6
Section modulus about y axis (Wy,inf,sup) mm3 22430 14300 24308 31714
Material - S420MC S355 S355MC S355MC
Pallet Beam
Gross section area (A) mm2 641 600 645 816
Section modulus about y axis (Wy,inf,sup) mm3 22500 19100 21775 31270
Material - S355MC S275 S235MC S235MC
Transverse Member of Spine Bracing
Gross section area (A) mm2 614 804 900 924
Section modulus about y axis (Wy,inf,sup) mm3 15425 17221 21960 25247
Material S235JRH S235 S235 S235
Upright Bracing
Gross section area (A) mm2 134.3 214.5 207 620.2
Material - S235JR S250 S355 S355
Vertical Bracing
Gross section area (A) mm2 113.1 256 87 684
Material - B450C S350 Fe1860 S355
Horizontal Bracing
Gross section area (A) mm2 200 468.1 290 186.4
Material - S235JR E260 S355 S355
Table 36 Braced racks, main section properties of elements

146
ANNEX FOR CHAPTER 5
In this annex, diagrams of the numerical analyses performed to investigate the influence of sliding
over the pallet beam bending moments are presented. The analyses’ results are presented with
dimensionless axes M/M* and H/Qp, where H is the total applied horizontal load. The ratio H/Qp is
the friction coefficient μ, and the ratio M/M* is actually the Correction Coefficient for Horizontal
Bending. The diagrams present the results step by step, initiating from zero horizontal force until the
end of analysis (sliding of the pallets).
Pallet-Merchandize Interaction

Figure 158 Results for case 15 Figure 159 Results for case 14

Figure 160 Results for case 13 Figure 161 Results for case 11

Figure 162 Results for case 17 Figure 163 Results for case 6

Figure 164 Results for case 5 Figure 165 Results for case 4

Figure 166 Results for case 24 Figure 167 Results for case 23

147
Figure 168 Results for case 22

148
 HOW TO OBTAIN EU PUBLICATIONS

Free publications:
• one copy:
via EU Bookshop (http://bookshop.europa.eu);
• more than one copy or posters/maps:
from the European Union’s representations (http://ec.europa.eu/represent_en.htm);
from the delegations in non-EU countries (http://eeas.europa.eu/delegations/index_en.htm);
by contacting the Europe Direct service (http://europa.eu/europedirect/index_en.htm) or
calling 00 800 6 7 8 9 10 11 (freephone number from anywhere in the EU) (*).
(*) The information given is free, as are most calls (though some operators, phone boxes or hotels may charge you).

Priced publications:
• via EU Bookshop (http://bookshop.europa.eu).
 KI-NA-27583-EN-N
The project focuses on the seismic design of static steel pallet racks, which are
widely adopted in warehouses. Despite their lightness, racks can be loaded with
tons of valuable goods, a live load by far higher than the self-weight, opposite
to what happens in usual civil engineering structures. The loss of these goods
during an earthquake may represent, for the owner, a very large economic loss,
much larger than the cost of the whole rack on which the goods are stored or
of the cost for its seismic upgrade. Hence, solution of the problems connected
with safe and reliable design of steel storage racks in seismic areas has a very
large economic impact. The objective of the SEISRACKS2 project is to increase
knowledge on actual structural behaviour and ductility of steel pallet racks,
and to assess design rules for earthquake conditions by full-scale testing and
numerical simulation.

Main outcomes of the research are:

1. Detailed reports on the different aspects investigated;
2. Validation or invalidation of the rules in the current version of FEM 10.2.08;
3. Improvements and extension of the current rules in order to optimize the
seismic behaviour of structures designed according to European rules;
4. Definition of standardized experimental procedures to qualify structural
elements of rack structures to be used in seismic areas;

5. A new software tool for the design of rack structures under seismic loads.

Studies and reports

ISBN 978-92-79-53897-1
doi:10.2777/686466

View publication stats