You are on page 1of 6

Insights and Innovations in Structural Engineering, Mechanics and Computation – Zingoni (Ed.

)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-02927-9

Design and analysis of structural steel tower assembly subject to high


axial and transverse stresses using BS EN1993-1-1 and BS EN1993-1-5

Maja Wilson & John Anderson


SMEC South Africa, Cape Town, South Africa

ABSTRACT:  This paper describes the process developed to analyze and design a plated structural steel
tower used to support the stay cables on a curved cable stayed rail bridge. The tower’s internal steel
assembly was relatively complex because of the varying vertical and transverse geometry of each cable
anchorage. These anchorages subject the plated steel sections to both high axial and transverse stresses.
The accurate modelling of the ultimate and serviceability limit state behaviour was therefore critical. For
this reason a detailed three dimensional Finite Element Model (FEM) was used to verify the design. This
paper describes the practical implementation of the methodology provided by BS EN1993-1-1:2005 and
BS EN1993-1-5:2006 for the design of a steel plated section using hand calculations and using a FEM.
When checking a design using a FEM, the loss in stiffness due to a deformed and deflected shape can
be modelled in a non-linear analysis. The deformed shape of the structure (obtained from a previous
analysis) can be used as the initial imperfect shape to give a reduced section resistance (Cl. 5.3.5 of Hendy
& Murphy, 2007). The paper describes how imperfections applied to the model were therefore used to
account for both geometric imperfections and residual stresses.
The paper also describes how the critical buckling analysis was performed on the tower anchorage
using the same three dimensional, finite element shell model used for the ultimate limit state design check.
This buckling analysis was performed in two parts: firstly as a linear eigenvalue buckling analysis; and
then secondly a geometrically nonlinear analysis.
The aim of the paper is to document benefits and challenges of the chosen methodology for the design
of irregular steel plated section subject to both axial and transverse stresses. It hopes to provide some
practical insight into the use of BS EN1993-1-1:2005 and BS EN1993-1-5:2006 for this purpose.

1  InTroduction

The use of structural steel towers on cable stay


bridges offers certain advantages. A more compact
structure is possible and there are often greater
architectural opportunities. The disadvantage can Figure 1.  Bridge elevation.
be the complex design process required for irregu-
lar shaped plated sections that carry both vertical
compressive stresses and transverse tensile stresses. Above the composite section is a 15.4 m high
This paper describes the necessary analysis and anchorage assembly that is constructed using
design process developed for a particular struc- plated steel sections.
tural steel tower assembly on a cable stayed bridge A steel section was chosen because the tower
in Sydney, Australia. width is limited by the fact it sits centrally on the
The cable stayed bridge in question is a rail deck between two rail lines. The lower steel con-
bridge with a span configuration of 76.9 m, 131 crete composite section provides the necessary
m and 61.1 m with a constant horizontal radius of stiffness to limit the tower deflection and any asso-
402.75 m. ciated second order effects.
The towers supporting the stay cables are in Because the deck is on a horizontal radius the
simple terms fabricated steel boxes. The box cross tower is subjected to out-of-plane bending moments.
section tapers from 6.0 m long and 2.3 m wide at In simple terms the cable stays, which do not sit in a
the base to 4.0 m long and 1.8 m wide at the top. single plane because of the deck’s horizontal radius,
The lower 13.4 m of the tower section is filled with pull the top of the tower towards the inside of the
concrete and it acts as a composite structure. curve. The towers are also subject to longitudinal

1178
Figure  3.  Isometric view of upper structural steel
assembly.

The FEM was used to check both the Ultimate


Limit State (ULS) capacity and the critical buck-
ling capacity of the steel tower assembly.
In reviewing this approach this paper aims to
document the benefits and challenges of using
each of the two outlined methodologies for the
design of irregular steel plated sections subject to
complex design stress profiles.

3  Design to BS EN1993-1-5: 2006

The structural steel tower assembly is a non-com-


pact Class 4, non-standard plated section subject
to significant transverse and longitudinal bend-
ing stresses. The section it is therefore unable to
attain its full yield stress under loading because of
Figure  2.  Isometric view of upper structural steel the onset of local buckling. BS EN1993-1-5:2006
assembly. provides two design methods to account for this
fact. The first method uses effective section derived
bending moments when the main span of the bridge from clause 4. The second method uses reduced
is fully loaded. Significant biaxial bending moments stress limits as defined by clause 10.
are therefore generated in the tower section. Although it’s not specifically stated in the code,
There are 16 stay cables per tower and the maxi- the first method cannot be used where there is a
mum serviceability forces within each ranges from uniform transverse direct stress accompanying the
5.0 MN to 8.3 MN. longitudinal stress. This effectively rules out this
method, given the cable anchorages induces both
longitudinal and transverse stresses in the tower.
2  Design approach The reduced stress method is therefore used. It
should be noted that the required application of
The towers are designed in accordance BS EN1993- the method is not clearly outlined in the code and
1-1:2006, Eurocode 3: Design of steel structures— the authors had to make use of additional guides1
Part 1-1: General rules and rules for buildings, and and commentaries2 to glean the correct approach.
BS  EN1993-2:2006, Eurocode 3: Design of steel The reduced stress method determines the over-
structures—Part 2: Steel bridges. In the first stage of all slenderness of each plate element under the
the design process the member’s cross section resist- applied stresses. From the slenderness value final
ance is checked making provision for local buck- reduction factors are calculated to check the allow-
ling effects over a finite length. The necessary plate able stresses in the plates. Generally both plate-like
buckling checks make reference to BS  EN1993-1- and column-like buckling must be checked.
5:2006, Eurocode 3: Design of steel structures—Part The reduction factors for direct stresses and
1-5: General rules—Plated structural elements. In shear stresses are applied to the cross section check
the second design stage the overall member stability using the design values of the material properties
is checked using BS EN1993-2:2006, clause 6.3. as shown in the below formula.
To complement the above mentioned design
process it was decided to verify the design using a 1
Hendy and Murphy (2007) Designers’ Guide to
detailed three dimensional Finite Element Model EN1993-2.
(FEM). The methodology followed for using the 2
Johansson, Maquoi, Sedlacek, Müller, and Beg (2007)
FEM for the design check as in accordance with Commentary and Worked examples to EN 1993-1-5
BS EN1993-1-5:2006 and BS EN1993-1-1:2005. “Plated Structural Elements”.

1179
In review, the above mentioned procedure was
automated as far as possible but it was judged a
relatively time consuming process to consider the
buckling permutations in conjunction with the
various stress state permutations. The method was
judged relatively conservative in that it could not
take account of the fact that the cable anchorages
effectively braced certain panel sections.

4  Finite Element Analysis

4.1  LUSAS model


The finite element analysis of the tower was under-
Figure  4.  Anchor detail with fleshing: front plate taken using LUSAS engineering analysis software.
removed for clarity. A local shell model of the top section of the tower
was developed that included all of the transverse
2 2
 σ x , Ed   σ z , Ed   σ x , Ed  and longitudinal stiffeners and cable stay anchor
  +  −  plates assemblies.
 ρcx f y / γ M1   ρcz f y / γ M1   ρcx f y / γ M1  The setting out geometry of the stay cables and
2
 σ z , Ed   τ Ed  their associated bearing and anchor plates was
  + 3   ≤ 1. important in order to accurately model the load
 ρcz f y / γ M1   Xv fy / γ M1  effects on the tower. To achieve the required accu-
racy the setting out geometry was first calculated
where: numerically using 3D vectors. This proved useful
when it came to the verification of the model.
• ρcx is the plate buckling factor for λ taking
As with any FEM model the defined bound-
account of the interaction between plate like
ary conditions play an important role in ensuring
behavior and column like behavior in the direc-
the analysis was representative. In this instance
tion of σx;
the supports defined at the base of the model had
• ρcz is the plate buckling factor for λ taking
to simulate the tower’s transition to a composite
account of the interaction between plate like
section. Using spring supports, moment continu-
behavior and column like behavior in the direc-
ity and translational fixity was therefore provided
tion of σz;
along all sides of the tower’s bottom perimeter.
• Xv is the plate buckling factor for λ for shear
A key finding during the design iterations was
stresses.
that the provision of rigid transverse supports cre-
The longitudinal face of the tower is stiffened ated peak von Mises stresses in the bottom corners
with two vertical T-shaped stiffeners. These stiff- of the tower box section. By reducing the spring
eners are intersected by simple transverse plate stiffness of the transverse supports, the base of
stiffeners at regular intervals. Because of this the tower was not forced to transition from a dis-
arrangement it was necessary to use a staged torted box section to a perfectly aligned section. A
approach in checking the sections. limit for the transverse displacement of 2 mm was
Firstly the unstiffened sub-panels of various set after interrogation of the global tower model.
thickness were checked for the applied stress at Although somewhat conservative, rigid vertical
each level. Thereafter various permutations of supports were provided.
stiffened panels were checked using the gross sec- The applied design loads were the maximum
tion properties of those panels. These permuta- Ultimate Limit State (ULS) cable forces extracted
tions included the three stiffened panels between from the global model (which was modelled in
the two adjacent transverse stiffeners. It also con- Bentley’s RM).
sidered buckling modes that extended across mul- LUSAS provides the option of using plate or
tiple transverse stiffeners. shell elements that are either thin or thick. Plates
As the reduced stress method is used no specific are 2-dimensional whereas shells are 3-dimentional
check for web breathing is required. This is a phe- elements. Thick elements include transverse shear
nomenon where geometric imperfections in plated effects whereas thin elements exclude transverse
section grow under load and then reduce again shear deformations. Thick shell elements are
when the load is removed. It can lead to fatigue required for geometric nonlinear analyses and were
damage at plate boundaries. hence used in this instance.

1180
cally, the Total Lagrangian is considered to be a
more robust formulation, which is used to cope
with substantial load increments. This formulation
importantly applies the load in the same direction
as it was initially prescribed.

4.2  Design process


General advice on the use of finite element meth-
ods of analysis can be found in Annex C of
BS  EN1993-1-5:2006 whilst more specific advice
on imperfections and how to model them can be
found in BS EN1993-1-1:2005 clause 5.3.
As described previously, the design code allows
for the local buckling effects in a section by consid-
ering reduced section properties or reduced stress
limitations. However, when checking a design using
a FEM, the loss in stiffness due to a deformed and
deflected shape can be modelled in a non-linear
analysis. The deformed shape of the structure
Figure 5.  Elevation of steel tower modelled in LUSAS. (obtained from a previous analysis) can be used as
the initial imperfect shape to give a reduced section
resistance (Cl. 5.3.5 of Hendy & Murphy, 2007.)
A useful feature of LUSAS is that it allows the size Imperfections comprise of geometric imperfec-
of the mesh generated on the thick shell elements to tions and residual stresses. Geometric imperfections
be varied. The size of the mesh can have an adverse are used to describe any deviation of the constructed
effect on the accuracy of the results, especially in a structure to what is shown on fabrication drawings.
buckling analysis. For this reason the sensitivity of Hence, these are dimensions that have arisen dur-
the analysis results was checked for successive refine- ing fabrication and erection. These imperfections
ments in the mesh size. The stresses at key nodes can include a lack of verticality, lack of straight-
were plotted against the model mesh size in order to ness, lack of fitup and joint eccentricities. Residual
determine at what point the results became consist- stresses are mainly caused by welding and can lead
ent. The mesh size at which the results sufficiently to yielding at lower applied loads than predicted
converged was deemed to be accurate enough whilst from stress analysis ignoring these effects. The effects
still allowing the model to run relatively quickly. of residual stresses can be modelled by additional
As mentioned the design process included a equivalent geometric imperfections. The geometric
nonlinear analysis. An important decision becomes imperfections discussed below therefore cover both
the choice of control options to define the type of geometric imperfections and residual stresses.
geometric nonlinearity and solution control to To take into consideration the geometric and
be used in the analysis. Geometric nonlinearity is structural imperfections, the model was run with
accounted for using four basic formulations. These an initial deflected shape in a non-linear analysis.
formulations are the: This initial imperfection magnifies the load effects
• Total Lagrangian—A strain formulation that associated with the geometric non-linearity of
has its reference as the initial configuration at the system. The shape of the imperfection should
the start of the analysis. be derived from the shape of the elastic buckling
• Updated Lagrangian—A strain formulation that mode of failure. Given the tower is a cantilever the
has its reference as the end of the last converged required initial lateral deflection was created by
increment. applying a lateral force at the tip of the cantilever.
• Eulerian—A strain formulation that has its ref- The global initial sway imperfection of the tower
erence as the current configuration. was calculated from BS EN1993-1-1:2005, clause
• Co-rotational—Form of geometric nonlinearity 5.3.2, equation 5.5, which correlates to a transverse
in which large displacement effects are related displacement of 42.2 mm at the top of the tower.
to a set of axes that follow and rotate with the
element. 4.3  Model results
In this FEM Total Lagrangian solution con- The analysis model was run twice; first in a linear
trol was selected because it is the only formulation analysis and second in a non-linear analysis with
that is compatible with thick shell elements. Typi- an initial imperfection of 42.2  mm as shown in

1181
Figure  6.  Deflected shape which is used as the initial
imperfection for the nonlinear analysis. Figure 8.  Eigen mode for EV 5 = 6.75.

In the linear eigenvalue buckling analysis, the


first relevant eigen mode has an eigenvalue of 6.75.
This means that should the applied load be fac-
tored by 6.75 then the structure will buckle with
the specific deformed shape as shown in Figure 8.
However, the linear eigenvalue buckling proce-
dure computes the stress state according to a linear
elastic procedure and the effects of deformations
are ignored. Futhermore, it does not take into con-
sideration any initial imperfection within the struc-
ture. It is therefore important to perform a second
buckling analysis using a Geometrically Nonlinear
Figure 7.  Von Mises Stress plot for the linear analysis. (GNL) analysis.
The GNL analysis takes of account of pre-buck-
ling displacements of the structure and provides a
Figure 6. In both analyses the applied load was the
complete response of the structure at all stages of
ULS cable forces and gravity.
the analysis, providing a complete deformation his-
When comparing the results it could be seen
tory. Furthermore, a relevant initial imperfection
that running the tower steel plated assembly model
can be applied to the structure to encourage the
in a non-linear analysis with an initial imperfection
buckling behaviour.
resulted in an increase of stress of 5 MPa (approxi-
In this instance, the deformed shape from the
mately 2%).
critical eigen mode from the eigenvalue analysis
(shown in Figure 8) was applied to the GNL analy-
sis as the initial imperfection. The magnitude of
5  BUCKLING Analysis
the maximum imperfection at mid-panel was cal-
culated as being 6.7 mm. This magnitude takes into
5.1  Design process
consideration both geometric and structural imper-
A critical buckling analysis was also performed fections (residual stresses) in accordance with BS
on the steel plated tower assembly using the same EN1993-1-1:2005. For sub-panel buckling, such as
local analysis model. The applied loads were the this one, the recommended maximum imperfection
ULS maximum cable forces and gravity. in the panel is ℓ/200, in accordance with Table C.2
The buckling analysis was performed in two and Figure C.1 of BS EN1993-1-5: 2006, whereby ℓ
parts: firstly by means of a linear eigenvalue (EV) is the smaller of the two dimensions of the panel.
buckling analysis; and secondly using a geometri- A force-displacement graph can then be used to
cally nonlinear analysis. determine the point at which structural stiffness is

1182
stay anchor had to be ignored in the hand calcu-
lations. The impact of the stay tube penetrations
through the tower walls could also not be discerned
using hand methods.
The hand ccalculations required to ensure code
compliance are relatively labour intensive given
that different buckling modes must be tested to
ensure that the most critical element is being evalu-
ated and designed appropriately.
Conversely, a FEM offers the opportunity to
model the exact structure even when the geometry
is complex and the applied loading unique. The
guidance provided in Annex C of BS  EN1993-1-
Figure  9. Graph showing the deformation history 5:2006 on how to design the steel plates sections
(pre- and post-buckling) of a node at mid-panel during using a FEM is therefore valuable. Although the
increasing loads. modelling process can be time consuming and
fraught with pit falls, once the model is complete,
lost and the panel buckles. In the graph below, it design changes and modifications are quickly and
can be seen that the panel buckles at a load factor easily implemented. This makes iterating the design
of approximately 6.5. As would be expected, this is relative simple.
less than in the linear eigenvalue analysis. The greatest drawback to the FEM analysis is
The EV analysis was run to 30 EVs with a maxi- the lack of transparency in the computing and it
mum load factor of 15.45 and no global tower is very easy to make fundamental errors. Calcula-
buckle eigen modes resulted within the analysis. tions of some nature are therefore always required
to check and test the results of a FEM. In critical
elements such as the tower of a large cable stayed
5.2  Full tower buckling analysis bridge, it is recommended that both methodologies
The LUSAS model of the tower anchorage was outlined above are adopted. This enables the veri-
then extended so as to include the full tower. A fication of one method against the other. It also
linear eigenvalue analysis was performed on this gives the designer a high degree of confidence that
model so as to confirm that global tower buckling the results obtained in the FEM are correct and
is less critical than local panel buckling. possible economies valid.
The lower half of the tower is composite and In the design of this steel tower assembly,
was therefore modelled as a steel outer shell com- there was good correlation between the critical
prising of shell elements, which also form the con- local buckling modes identified between the two
crete volume infill. adopted methodologies. The FEM analysis process
The eigen modes produced were similar to those highlighted the conservatism of using the reduced
from the first buckling analysis, which indicates that stress method in this instance for the reasons men-
global tower buckling occurs at a higher load factor tioned above.
and is hence less critical than local panel buckling.
References
6  CONCLUSION
[1] Hendy, C.R. and Murphy, C.J. (2007) Designers’
Guide to EN1993-2, London: Thomas Telford.
The BS EN codes referenced in this paper undoubt- [2] Johansson, B., Maquoi, R., Sedlacek, G., Müller, C.
edly provide an effective means of designing rec- and Beg, D. (2007) Commentary and Worked exam-
tangular steel plated structures. However, when ples to EN 1993-1-5 “Plated Structural Elements”, 1st
both the geometry and loading is irregular, con- edition, Luxembourg: European Communities.
servative assumptions must be made. This can lead [3] Modeller Reference Manual (2011) London:
to a design that is safe but uneconomical. In this LUSAS.
instance the stiffening effects of the diagonal cable

1183

You might also like