You are on page 1of 23

Thin-Walled Structures Vol. 30, Nos 14, pp.

111133, 1998
1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0263-8231/98 $19.00 + 0.00
PII: S0263-8231(97)00015-3

The Design Analysis of Light Structures with


Combined Aluminium/Steel Sections
G. T. Taylora, M. Macdonaldb and J. Rhodesc
a

Department of Energy and Environmental Technology, Glasgow Caledonian University,


Glasgow, UK
b
Department of Engineering, Glasgow Caledonian University, Glasgow, UK
c
Department of Mechanical Engineering, University of Strathclyde, Glasgow, UK

ABSTRACT
Extruded aluminium hollow box sections with steel section inserts are
riveted together to form a combined member increasingly used as the
basis for the production of skeletal structures ranging from shelters on
transport routes to larger covered walkways at airports, hospitals and
industrial compounds. At these locations the structures have to be functional, safe and attractive. The paper reports on research work carried
out on the stress and deflection behaviour of a typical riveted combined
member investigated using the finite element method and supported by
experimentation. The effect of the fit between the steel section and the
aluminium section is given particular attention. The paper goes on to
report the results of an examination of the dynamic response of light
aluminium/steel structures, where the dynamic analysis is carried out
using the finite element method and long walkway shelters, upwards of
300 m, are a particular focus. The results of a parametric study are
presented and combined with wind excitation criteria design curves are
proposed. Conclusions are drawn on the engineering value of the use of
aluminium/steel combined members including the implications for the
design of light aluminium/steel structures. 1998 Elsevier Science Ltd.
All rights reserved

1 INTRODUCTION
The last decade has seen an increasing market for light skeletal structures
such as long, covered walkways, simple bus shelters and larger bus stations
constructed from extruded aluminium sections with steel section inserts
111

112

G. T. Taylor, M. Macdonald, J. Rhodes

riveted together. This type of construction (Fig. 1) allows significant ease


of manufacture, for example, all elements are easily manhandled with no
cranage required and the pre-fabricated panels are easily erected on site.
The structure requires no maintenance while remaining very attractive
throughout its life. These characteristics collectively result in a product that
is very competitive in the market place.
The primary strength of these structures is derived from inserting a thin,
cold formed steel plain channel into an extruded thin walled, aluminium
box-section, each component then riveted together at particular intervals
along their length resulting in a combined member.
Many of the structures manufactured in Glasgow now require a structural
integrity check to British Standards criteria. This requires consideration of
snow and wind loading,1,2 and the determination of the stress response of
the structure.3,4
The design and stress analysis of complete structures can only be effectively carried out by finite element (FE) analysis and only beam elements
can be used to represent the combined members purely on economical
grounds. The steel and aluminium beam elements are tied together in the

Fig. 1. Commutaport product. (Courtesy: Southside Engineering Ltd).

The design analysis of light structures with combined aluminium/steel sections

113

FE model at nodes located at rivet points. The aluminium and polycarbonate


panels are represented by plate elements.
Initial research work was undertaken where a typical aluminium/steel
combined member was investigated as a cantilever modelled by beam
elements and then more rigorously by plate elements. This early work was
reported in Taylor.5
The research work has continued with a more detailed and comprehensive
study using rigorous finite element models. The development of a best
finite element model representation of the physical problem is briefly
described and the significant results of the study are presented in this paper.
Complementary to this work an experimental investigation was carried out,
initially focusing on a series of load capacity tests, followed by a stress
investigation of a combined member where strain gauge technology was
applied. The results of the load capacity tests are also presented herein.

2 LOAD CAPACITY TESTS OF COMBINED MEMBER


2.1 Representation of structural member
A typical combined member was represented by an encastre/free end cantilever of length 1.26 m. The load was applied at the free end of the cantilever,
acting about the major axes of the sections. The cross-section of the structural member is shown in Fig. 2 and it should be noted that tests were
carried out on riveted and unriveted sections to establish if riveting had
any effect on the member performance.
2.2 Load capacity test equipment
A TiniusOlsen Universal Testing Machine, which has a 90 t load capacity,
was used to test the cantilever to failure. The applied load was measured
by means of a sensitive load cell within the machine table which was also
connected to a plotter from which a graph of loaddeflection for the cantilever was produced signifying the point at which failure of the cantilever
occurred and hence the load capacity of the structural member. A schematic
diagram of the experimental set-up used in the investigation is shown in
Fig. 3.
The results are shown in Table 1. Other tests were carried out on combined members where the steel channel inside the aluminium hollow box
section was a snug fit and on others there was a loose fit to find out the
effect that the fit had on the load capacity of the combined member. Macdonald and Macdonald et al.6,7 give a more comprehensive description of

Fig. 2. Arrangement of riveted aluminium/steel combined section gallows.

114
G. T. Taylor, M. Macdonald, J. Rhodes

The design analysis of light structures with combined aluminium/steel sections

115

Fig. 3. Experimental investigation schematic layout.


TABLE 1
Experimental investigation and design code results
Specimen configuration

Experiment, P(kN)

Design Code P(kN)

3.78

3.63 (BS5950)

3.84

3.80 (BS8118)

10.25

7.43

9.70

7.43

1.74

0.8

1.76

1.71

4.10

2.51

the above experiments. A schematic representation of snug fit and loose fit
is referred to in Section 4.2.3.
2.3 Design code approach
To obtain a load capacity for the cold formed steel channel for comparison
with that obtained from the load tests, BS5950, Part 5, Section 33 was
referred to. To determine a load capacity for the aluminium hollow box

116

G. T. Taylor, M. Macdonald, J. Rhodes

section, BS8118, Part 1, Chap. 44 was referred to. The results obtained
from the codes are shown in Table 1. Further details of the above analyses
can be found in 6,7 along with other theoretical approaches to the problem.
2.4 Member load capacity
The load capacity tests of the combined members, riveted and unriveted
showed that the failure, as expected, occurred at the built-in end where a
plastic hinge mechanism formed in the webs and flanges of the aluminium
box section, and the steel channel failed due to twisting and warping effects.
The difference in load capacity of riveted and unriveted members was found
to be of the order of 5%. However, it should be noted here that the unriveted
combined member had a snug fit and the riveted combined member had a
loose fit. A combined member where there was a loose fit with no rivet
was not tested, thus a correlation of load capacity between this and a loose
fit riveted member cannot be made. It is thus concluded that riveted combined members, where a snug fit existed, produced an increase in load
capacity of the order of 5% over loose fitting members. The design codes
produced load capacities that were conservative as they provide elastic solutions only with no allowance for plasticity. 6,7 show other theoretical
approaches to obtain a load capacity which compare much more favourably
with the test results.

3 FINITE ELEMENT INVESTIGATION OF COMBINED MEMBER


3.1 Introduction
The finite element software used for all FE analysis was PAFEC-FE8
operating on Hewlett Packard workstations. A typical beam/plate finite
element model of a light structure is shown in Fig. 4 and Fig. 5 shows how a
combined member is modelled in this context. The generalised constraint
facility of the PAFEC software is used to tie the aluminium and steel
beam element nodes together. The nodes hence take the same magnitude
of translational deflection in any solution. Plate elements representing wall
and roof panels are connected to the beam element skeletal structure in the
usual way.
The detailed stress/deflection analysis of the cantilevered combined member involved rigorously modelling the combined member using plate
elements type 44210.8

The design analysis of light structures with combined aluminium/steel sections

117

Fig. 4. Finite element model of light structure.

3.2 Convergence study of plate FE model


A convergence study was run for each of the primary members wherein a
series of models with increasing mesh density near to the fixed end was
generated. From this study a suitable mesh was generated that would ensure
accurate stress and deflection values.
Having generated a suitable mesh for the aluminium box section the
process was repeated for the steel channel. The FE models were then combined to represent the combined member. A typical plate FE model is shown
in Fig. 6 where the aluminium box is cut away on the front web and flange
to expose the channel.
3.3 Development of best plate FE model representation of
engineering problem
The two members must satisfactorily interact at their interface, a suitable
representation of the rivet connections must be made and restraint and load
application have to be considered.

118

G. T. Taylor, M. Macdonald, J. Rhodes

Fig. 5. Riveted aluminium/steel combined member. Beam element representation.

Loads were applied at the cantilever free end as discrete point loads and
a planer restraint applied at the encastre end. A single row of rivets, centrally located on the web, is modelled by tying corresponding steel and
aluminium nodes together at rivet locations by using a generalised constraint. This allows specification of relationships between degrees of freedom at node pairs and the selection of appropriate freedoms reflecting the
restraining effect of the rivet.
The interaction of the steel and aluminium sections at their interface
provided a major modelling problem. Early models used discrete generalised constraints to specify the interactive behaviour. This, however, does

The design analysis of light structures with combined aluminium/steel sections

119

Fig. 6. Plate FE model of combined member.

not allow the model to reflect the independent displacement behaviour of


each member. Differential displacement occurs as the steel channel twists
resulting in it apparently ghosting through the box section. PAFEC provides a GAPS module which serves to identify the physical gap and/or
contact between the aluminium box and the steel channel. A number of
different models were generated and solutions examined. The generalised
constraint was confirmed as the most economical representation of rivets
and the GAPS option applied selectively across the interface was used. The
GAPS restraint option, applied across nodes in PAFEC, was invoked at all
nodes lying in a transverse planer section located at the top of the cantilever
and at each rivet location.
3.4 Analysis of cantilevered combined member
The physical problems considered were cantilevers of length 1.26 m and
one of length 1.84 m with one end fixed and one end free, the first being
consistent with test specimens of section 2. A load of 1 kN was applied at
the free end giving bending about the major axis.
A series of cantilever problems was modelled where the web depth of
the steel channel was varied from 92 mm (loose fit) to 94 mm (snug fit).

120

G. T. Taylor, M. Macdonald, J. Rhodes

This series of solutions was repeated without rivets to establish the effect
of rivets and a snug fit.
Further loose fit models were developed to include the effect of no contact between steel flange tips and aluminium web.
3.4.1 Load capacity of combined members
The maximum stress profiles on the inside surface of the aluminium and
steel sections are shown in Figs 710. The Von Mises equivalent stress is
shown versus the distance around the section in the fixed end region of the
cantilever where the highest stresses occur. The very nature of the problem
gives rise to a complex stress problem. Combined members where the steel
channel, either fitting snugly into the aluminium box section or not, will
behave very similarly as long as they are riveted together. It appears that
a snug fit does not change stress patterns around the sections and magnitudes of stress are not significantly different. This suggests that for riveted
sections no significant load capacity gain is made on fitting the steel section
snugly into the box section. Rivets being excluded from the combined beam
does have a clear effect on the stress patterns around each section. The
magnitudes of peak stress are not significantly different but show interesting
variation. Stresses in the aluminium reduce on riveting and in the steel
show an increase. The steel channel is obviously brought into action sooner
in a riveted combination. The maximum peak stress in steel increases by
about 5% although other peak stresses increase by a larger margin. The
aluminium maximum peak stress reduces by about 12%. It could be interpreted from this that the load capacity of the combined beam on riveting
and considering the aluminium member as the critical component, will
increase by about 12%. Here a higher yield value of steel over aluminium
is assumed.
Deflections of the combined member are no less complex than stresses
as the sections translate and rotate in two dimensions. For a riveted combined member where the steel is snug with the box section, the cantilever
tip deflections vary from those where the steel has a loose fit by less than
5%. Again the loose fit and the snug fit apparently making no significant
difference to load capacity. The deflections for a non riveted section versus
a riveted combined section (snug fit) shows the deflections to differ by
34%.
3.4.2 Correlation of plate finite elements and beam elements
Since the design of the light skeletal structures considered in this study is
based on beam element analysis it was an objective of the study to compare
beam element results with those from plate finite elements. Tables 2 and

The design analysis of light structures with combined aluminium/steel sections

121

TABLE 2
Beam element results
Combined section length

Free end deflection

Max. bending stress (MN/m2)

1.84 m
1.26 m

4.9 mm
2.68 mm

Steel: 39.9; Al: 15


Steel: 49.84; Al: 17.38

3 present the results obtained for combined section beams only although
analyses of individual members were also done. The results of the analysis
of individual members5 suggests that beam stresses should be increased by
about 25% for the aluminium box section and about 11% for the steel
channel. This however does not apply to members combined by riveting.
Examination of the results for combined members suggests that no adjustment of beam stresses are necessary.
An experimental investigation of a combined member using strain gauge
technology was carried out and the results for a 1.84 m cantilever5 show
that finite element and gauge stresses were in general agreement. However,
a more comprehensively gauged combined beam is required to be tested
for a more detailed correlation.

4 DYNAMIC ANALYSIS OF LIGHT ALUMINIUM/STEEL


STRUCTURES
4.1 Introduction
The design of structures, taking into consideration the criteria associated
with the dynamic response to wind excitation, is a specialised and highly
complex task. The most significant guide for designers available at present
is that produced by ESDU International plc9. This is, however, apparently
limited to particular types of structures and requires considerable interpretation in its use.
TABLE 3
Plate element results
Combined section length

Free end deflection


(averaged)

Max. bending stress (MN/m2)


(averaged)

1.84 m
1.26 m

St: 4.12 mm; Al: 3.98 mm


St: 2.875 mm; Al: 2.858 mm

Steel: 40; Al: 16


Steel: 45; Al: 17

Fig. 7. Stress distributions snug fit + rivet.

122
G. T. Taylor, M. Macdonald, J. Rhodes

Fig. 8. Stress distributions loose fit + rivet.

The design analysis of light structures with combined aluminium/steel sections

123

Fig. 9. Stress distributions snug fit no rivet.

124
G. T. Taylor, M. Macdonald, J. Rhodes

Fig. 10. Stress distributions loose fit no rivet.

The design analysis of light structures with combined aluminium/steel sections

125

126

G. T. Taylor, M. Macdonald, J. Rhodes

In this study a modal analysis is required to be performed to predict


natural frequencies and mode shapes of typical structures. The criteria
invoked from the ESDU design guide requires the fundamental natural frequency to be identified.
The study focuses on cantilevered gallows type walkways which can be
constructed for any length without a break in the length. They also may
have roof widths varying from 1 to 2 m. The height of the gallows column
is normally constant at 2 m with a gradual rake on the roof for drainage
purposes. Over its total length, a walkway will be constructed in sections
(bays) one bay is the distance between consecutive gallows; in this study
the length of each bay of the walkway is 4.9 m. The roof sections are also
constructed per bay length and is a fabrication of aluminium box or Tsections covered either by aluminium plate or polycarbonate sheet. Side
walls are constructed from T-sections with polycarbonate or toughened
glass panel inserts.
4.2 Beam/plate finite element models
The finite element software package used was the ANSYS software. The
choice of software was purely based on the expertise of the researcher.
PAFEC could equally have been used to produce the dynamics results. The
modelling of the light structure in question and combined members is as
described in section 3.1 and as shown in Fig. 4 and Fig. 5.
For extremely long structures, upwards of 300 m, a replica finite element
model requires a very large number of elements which is undesirable.
Thus the initial part of the study focused on finding an idealised structure
which reduced the number of elements significantly but was accurate in the
prediction of natural frequencies. Three issues were considered:
(1) the need to include the side wall in the model;
(2) the reduction or idealisation of the roof section to an equivalent plate;
(3) the idealisation of the side wall to an equivalent plate.

4.2.1 Model of complete structure without side wall


Finite element models of complete structures were created, one with and
one without the side wall. The results presented in Table 4 show that the
difference between the fundamental natural frequencies of the two structure
configurations is 11.3% for 1 Bay and 6.8% for 2 Bay. This indicates that
the side wall cannot be neglected in a best idealised finite element representation.

The design analysis of light structures with combined aluminium/steel sections

127

TABLE 4
Side wall effect natural frequencies
Shelter with side panel
MODE

1 Bay

1
2
3
4

2.518
4.747
6.514
11.289

Shelter with no side panel

2 Bay

1 Bay

2 Bay

2.240
3.359
5.868
6.650

2.263
2.543
5.299
6.028

2.098
2.308
3.629
5.917

4.2.2 Determination of an equivalent plate for roof sections


The roof of the walkway shelter is of hollow-box sections and T-sections
in a rectangular criss-cross configuration. These are covered by 12.5 mm
sheet aluminium. This section is idealised as one single, plate with an equivalent stiffness and mass reducing the required number of finite elements
from 30 to 8. Over a full-length structure upwards of 300 m this leads to
a significant economy in a dynamics analysis.
Comparison of the natural frequencies between the model with original
actual roof and that with the equivalent roof plate is shown in Table 5.
In a similar way, the actual side wall was idealised and an equivalent
plate with appropriate thickness and density determined thus making further
reductions in finite element numbers.
The theoretical basis on which the roof and wall equivalent plates are
derived is given in Ref.10.
Comparison of the natural frequencies between the actual structure model
(original roof and side wall configurations) and the best model (equivalent
TABLE 5
Equivalent roof effect natural frequencies
1 Bay
MODE

1
2
3
4

Original
model

2.518
4.747
6.514
11.289

With
equivalent
roof
2.515
4.920
6.410
11.230

2 Bays

4 Bays

Original
model

With
equivalent
roof

Original
model

With
equivalent
roof

2.240
3.359
5.868
6.650

2.224
3.446
6.064
6.876

2.085
2.582
4.979
5.634

2.078
2.615
5.291
5.831

128

G. T. Taylor, M. Macdonald, J. Rhodes

roof and side plate) is given in Table 6. It can be seen that extremely good
results are obtained.
4.2.3 Proposed design guides
Having established a convenient idealised best finite element model predicting accurate value of natural frequencies a parameter study was then
undertaken to establish an aid to the design of cantilevered walkways. Varying structure length and roof width, natural frequencies were obtained from
many computer runs. Fig. 11 presents the results of this study as a design
guide where linear interpolation for roof widths not shown is valid as shown
in Fig. 12. The best finite element models did not incorporate snug and
loose fit configurations, shown in Fig. 13, as was considered in the static
stress analysis.
4.2.4 Use of ESDU wind engineering guide
The nature of the frequency of wind loading is that it mainly affects structures with low natural frequencies. The worst possible case obviously happens in situations of extreme weather conditions. Structures susceptible to
wind excitation are usually extremely flexible with very low natural frequencies.
The ESDU Wind Engineering design guide8 is used in this study to determine the stability of the structures under wind excitation.
From the ESDU guide, the criterion for the structures not to be susceptible to wind excitation is:
fn
30
s
TABLE 6
Actual structure versus idealised structure natural frequencies
1 Bay
MODE

1
2
3
4

Original
model

2.518
4.747
6.514
11.289

With
equivalent
roof and
side plate
2.515
4.944
6.351
11.329

2 Bays

4 Bays

Original
model

With
equivalent
roof and
side plate

Original
model

With
equivalent
roof and
side plate

2.240
3.359
5.868
6.650

2.238
3.454
6.022
6.827

2.085
2.582
4.979
5.634

2.078
2.615
5.291
5.831

The design analysis of light structures with combined aluminium/steel sections

129

where s is the damping ratio; the damping ratio for the structures considered here was selected as 0.005 from the guide and fn is the fundamental frequency.
Thus, fn 2.12 is the critical value derived from the criterion.
The criterion noted here is shown on Fig. 11 as a line at frequency 2.12.
This allows selection of structure lengths to avoid wind excitation problems.
For example structures with roof widths up to 1.65 m at any length would
not be susceptible to excitation problems. Between 1.65 and 1.9 m the selection of a suitable length of structure becomes important between 10 m and
about 120 m. This, in effect, defines the bounds wherein selection of dimensions is important.
The relevance of this criterion to the type of structure under consideration
requires to be investigated. The criterion was selected on the basis of matching the walkway with the class of structure for which the criterion was
established. However, the class of structure may include a broad range of
structure configurations and there is thus doubt as to the appropriateness
of the criterion to the walkway. Research work adopting experimental
approaches is now being undertaken to establish the damping characteristics
of the light structures of this study and to thus identify more appropriate
values of damping ratio. An alternative criterion derived using a different
approach from that of ESDU is also being considered.
5 CONCLUSIONS
A significant number of problem parameters are involved in the investigation of the combined member studied here. Good quality control of the
experimental and finite element work is essential. The testing of the combined member highlighted this point as the specimens supplied by the
manufacturer were not always of consistent dimensions. Results of the
investigation show that members riveted together, either with the steel fitting snugly inside the box or not behave almost identically which was also
verified by the tests. The effect of a more significant size of gap could be
further investigated. Riveted and unriveted combined members do appear
to behave differently. The tests and the FE results show an increase in load
capacity for riveted members of about 510% is gained. Beam finite
element models are adequate for design analysis. Beam finite element
stresses for individual members should be factored by about 25% for the
aluminium box and 11% for the steel channel. Beam stresses for a combined
member are sufficiently close to those of a more rigorous plate element
solution and a factor on stresses is not necessary.
Economic finite element models for the dynamic analysis of long, cantilever walkway structures constructed of aluminium/steel members have been
established and shown to give excellent results.

Fig. 11. Design curves for cantilevered structures.

130
G. T. Taylor, M. Macdonald, J. Rhodes

Fig. 12. Variation of natural frequency and roof width.

The design analysis of light structures with combined aluminium/steel sections

131

132

G. T. Taylor, M. Macdonald, J. Rhodes

Fig. 13. Schematic representation of snug and loose fits.

Design curves as an aid to the designer of long walkway structures have


been produced based on a study of the main parameters. They include the
ESDU criterion for structures susceptible to wind excitation. Design curves
based on finite element analysis can be produced for other similar type
structures and is the preferred way forward.
The ESDU criterion involved herein requires to be examined as to its
appropriateness for the walkway type structures focused on in this study.
Appropriate research is being undertaken.

REFERENCES
1. BS CP3, Parts 1 and 2, Code of Basic Data for the Design of Buildings, Ch.V,
Loading: Dead, Imposed and Wind Loads. BSI, London, 1967 and 1972.

The design analysis of light structures with combined aluminium/steel sections

133

2. BS5950, Part 1, Loading for Buildings. Part 2, Code of Practise for Wind
Loads. BSI, London, 1990.
3. BS5950, Part 5, Code of Practice for the Design of Cold Formed Sections.
BSI, London 1990.
4. BS8118, Part 1, The Structural Use of Aluminium. BSI, London 1990.
5. Taylor, G. T., The design by finite elements of riveted aluminium/steel composite structures. Proceedings of the 7th World Congress on Finite Element
Methods, Monte Carlo, 1993, pp. 330335.
6. Macdonald, M. Bending of a thin-walled combined section beam. MSc thesis,
University of Strathclyde, Glasgow, 1993.
7. Macdonald, M., et al., The analysis of riveted aluminium/steel combined members under bending action. Proceedings of CMT 96 Computational Methods
and Testing for Structural Integrity, Kuala Lumpur, 1996.
8. PAFEC Software, Strelley Hall, Nottingham, England.
9. ESDU Engineering Data, Wind Engineering, Sub-series Vol. 3a. ESDU International plc, December 1989.
10. Taylor, G. T. and Wu, P., Aluminium/steel structures subjected to wind excitation. Proceedings of CMT 96 Computational Methods and Testing for Structural Integrity, Kuala Lumpur, 1996.

You might also like