You are on page 1of 28

J. Construct. Steel Res. Vol. 42, No. 1, pp.

2 1 4 8 , 1997
© 1997 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
PII: S0143-974X(97)00~5-9 0143-974X/97 $17.00 + 0.00
ELSEVIER

Advanced Analysis and Design of Spatial Structures

J. Y. Richard Liew, N. M. Punniyakotty & N. E. Shanmugam


National University of Singapore, Department of Civil Engineering, 10 Kent Ridge
Crescent, Singapore 119260

(Received 4 March 1996; revised version received 28 November 1996;


accepted 16 December 1996)

ABSTRACT

Modern limit-state design codes are based on limits of structural resistance.


To determine the 'true' ultimate load-carrying capacity of spatial structures,
an advanced analysis method which considers the interaction of actual
behaviour of individual members with that of the structure is required. In
the present work, a large-displacement inelastic analysis technique has been
adopted to compute the maximum strength of spatial structures considering
both member and structure instability. The actual behaviour of individual
members in a spatial structure is depicted in the form of an inelastic strut
model considering member initial imperfections as 'enlarged' out-of-
straightness. The maximum strength of the strut is computed based on a
member with 'equivalent out-of-straightness' so as to achieve the specifi-
cation's strength for an axially loaded column. The results obtained by the
strut model are shown to agree well with those determined using plastic-zone
analysis. The nonlinear equilibrium equations resulting from geometrical and
material nonlinearities are solved using an incremental-iterative numerical
scheme based on generalised displacement control method. The effectiveness
of the proposed advanced analysis over the conventional analysis~design
approach is demonstrated by application to several space truss problems.
The design implications associated with the use of the advanced analysis are
discussed. © 1997 Elsevier Science Ltd.

NOTATION

A Area of cross-section of a member


A~ Effective area of cross-section of a tension member
D Outer diameter of a tubular member
E,E~ Modulus of elasticity and plasticity
{F~_,} -- Vector of internal element forces summed at each node of

21
22 J. Y. R. Liew et al.

the structure up to the (j-1)th iteration of ith load


increment
GSP Generalised stiffness parameter
GSP i Generalised stiffness parameter at ith load increment
I Second moment of area
[ K i-1] Tangent stiffness matrix of the structure at the beginning of
the load increment i
[Kj-1] Tangent stiffness matrix formed at the beginning of the jth
iteration based on the known element details at ( j - 1 ) t h
iteration of ith load increment
L, L o Chord length and arc length of a member
Full plastic moment capacity
Reduced plastic moment capacity in the presence of axial
force P
P Applied axial force
Pe Euler buckling load capacity of a strut member
Pmax Axial load capacity of a strut member as per design
specifications
Py Axial load capacity at full yielding of cross-section
Pul Axial load at which unloading begins in a strut member
{Pj} Vector of total external nodal loads on the structure at jth
iteration of ith load increment
{P} Reference load vector on the structure
Y Radius of gyration
{R}-l} Vector of unbalanced forces during ( j - 1)th iteration of ith
load increment
t Wall thickness of a tubular member
W Lateral deflection at a distance x from one end of the strut
member
Z Plastic section modulus
8o Member initial out-of-straightness magnitude at mid-span
6 Member mid-span deflection
{6U~} Iterative displacement vector at the structure level in jth
iteration of ith load increment
Iterative tangential displacement vector at the structure level
in jth iteration of ith load increment
(6u3} Iterative residual displacement vector at the structure level
in jth iteration of ith load increment
8Al Initial iterative load parameter specified as input
Iterative load parameter for the jth iteration of ith load
increment
Ae Total elastic shortening in a strut
Advanced analysis and design of spatial structures 23

a a
Member axial shortening due to compression
age, Member axial shortening due to change of geometry in the
elastic and plastic range
a Member axial shortening before elastic unloading
max

Ap Total member axial shortening in the plastic range


Ay Member elongation at yield load
{APi} Incremental load vector on the structure during the ith load
increment
{APi} Incremental displacement vector of the structure during the
ith load increment
PE Specified energy tolerance
O'y Design yield strength of member material

1 INTRODUCTION

Spatial structures having rigidly connected joints are more complicated to


analyse because each joint consists of at least 6 degrees of freedom, and the
members are subjected to combined actions of axial force, bending, torsion
and shears. However, rigidly jointed structures have the advantage of less
member density because they are stiffer and require smaller member sizes to
satisfy the design limit states. Experimental studies have shown that actual
spatial structures generally fall between the rigid jointed and pin-jointed ana-
lytical models, even though the actual connections may appear to have been
pinned or rigid. Actual joint rotational stiffness is difficult to quantify for
design, because it depends on many factors such as installation sequences,
lack of fit, types of nodes and their configurations, member and system imper-
fections, etc. If the joint stiffness cannot be quantified for analysis, the assump-
tion of pin-jointed members should be used because it would lead to a con-
servative design and safe estimate of deflection.
For structures with higher rigidity against instability, such as double layer
grids and deep domes, it is a common practice to assume that the structure
behaviour is linearly elastic. Elastic first-order analysis of such structures is
straightforward, commercial programs are available and approximate methods
can be applied for manual computations. However, to ensure member stability,
the forces obtained from the analysis must be checked against the specification
of member equations. These equations consider stability of an individual mem-
ber, and they implicitly include member imperfection effects, such as residual
stresses and initial out-of-straightness. However, there is no theoretical reason
to suppose that the capacity determined through member capacity equations
for the most critical member in a general, especially a slender spatial structure,
represents the limit load of the structure. To compute the true limiting strength
24 ]. Y. R. Liew et al.

of such structures, an advanced analysis, which can capture the interaction


effect of local buckling of members and overall buckling of structure, is
required. Advanced analysis refers to an analysis for all relevant load combi-
nations of which specification limit states checks are satisfied implicitly and
thus explicit verification of design limit states are not required. Work on
advanced analysis of two-dimensional semi-rigid frames has been reported
recently [1,2]. Educational software for advanced analysis and design is also
available in book form [3,4].
Advanced analysis is useful for slender spatial structures which are suscep-
tible to geometric instability and member buckling. This is particularly true
for single-layer reticulated shallow domes, or for domes with small rise-to-
span ratio, in which the structures may exhibit several types of instability
associated with member buckling, local or dimple buckling at a joint, and
interaction of local and overall instability. Substantial discrepancies have been
observed between the structural behaviour as predicted by linearly elastic
analysis and the actual behaviour [5,6].
Murtha-Smith [7] presented the chronological development of various ana-
lytical strut models, and highlighted the limitations of their use in actual design
practice. The basic problem of these analytical models is that they do not
satisfy the code requirement for member capacity checks, which therefore
deters one from adopting them for use in actual design. In the present work, an
inelastic large-displacement advanced analysis, which is capable of capturing
individual member strength and overall buckling of structure, is presented.
The strut model is based on a physical analogy that one strut element is used
to represent each truss member in a spatial structure, and the strut capacity
is made to conform with the specification's member capacity equation. A
numerical scheme using generalised displacement control algorithm has been
adopted for solving the nonlinear equations. This numerical scheme has been
shown to be capable of handling many kinds of instability problems associated
with slender spatial structures.

2 ELEMENT MODELS

2.1 General description

The present model is based on a physical analogy that one strut element is used
to represent each truss member in a spatial structure, and the strut capacity is
made to conform with the specification's member capacity equation. The main
objective is to minimise computing time. Further, one element per member
approach is oriented towards conventional design philosophy since typical
elements in a numerical model correspond to components to be checked in
Advanced analysis and design of spatial structures 25

conventional design. However, compared to conventional numerical analysis


based on updated Lagrangian formulation of so-called co-rotational formu-
lation, the element model implies a coarse element representation because
each element between the nodal points is modelled by only one element.
Therefore, it is necessary to take into account geometric nonlinearities at the
local element level in order to capture the effect of member buckling.
The present formulation handles large displacements on local and global
levels. On the local level, large deflection is incorporated when a plastic hinge
is formed at the mid-span of the element, in which the element is assumed
to be rigid plastic; whereas the global effects are taken in by updating nodal
point coordinates. Thus, a total Lagrangian formulation is implemented on the
element level. However, the program does not imply a complete total Lagrang-
ian formulation since the element reference axes are updated based on the
deformed geometry at each load step.

2.2 Strut model

In the present analytical model, the member initial out-of-straightness is


assumed in the form of a half-sine wave. The load-displacement curve of a
completely elastic strut with a sinusoidal initial bow is combined with that of
a plastic-hinge curve to form an elastic-perfectly plastic strut model. This
concept was originally proposed by Papadrakakis [8] and was further extended
by Maheeb [9] for struts with various types of cross-section for planar bracing
systems. The present study extends Maheeb's approach to three-dimensional
systems.
The behaviour of a strut under the action of axial load shows three distinct
characteristics, as shown in Fig. 1. They are discussed in the following subsec-
tions.

2.2.1 Elastic pre-buckling curve


The first stage of the load-displacement curve (curve AB in Fig. 1) is associa-
ted with the compressive loading of an initially imperfect column that buckled
at point B. During this stage of loading, the total elastic axial shortening, Ae,
due to an applied axial force P consists of two components:

Ae = A a + A~. (1)

The first component, Aa, is due to the axial deformation. The second compo-
nent, Ag, is associated with the change of geometry due to bow in the strut.
With reference to Fig. 2, these two components may be written as:

PLo
Aa - (2)
EA
26 J. Y. R. Liew et al.

Axial Load, P

Py 7 -p - ~ ~ ~, ~a~ p
.x,C"--

.ineor~///.~astic Curve(exact}

PE ---~-'~8~----_/ElasticLoadingCurve(exact)
7 t

Prnox

Put -//,- 2
Ao Af /~b Ab Amox
Axial Shortening,A
Fig. 1. Axial load-displacement curve for a strut member.

L L _1
V
L_.. ~ _ _ ~ 1~ I~° P

Fig. 2. Imperfect strut in the elastic range.

Ag = Lo - L (3)

where L is the member's chord length, and L0 is the member's arc length
given by:

Lo= 1+ (4)

in which w is the lateral deflection at a distance x from one end of the member
as shown in Fig. 2.
For a strut with an initial out-of-straightness in the form of a half-sine wave
and maximum deflection 30 at the mid-span, the total lateral displacement
including the second-order displacement due to the axial force effect may be
written as [ 10]:
Advanced analysis and design of spatial structures 27

w(x)_6o ['rrx]
p sin ~ - (51
1-
Pc

where Pe is the Euler buckling load of the strut. The member's arc length Lo
can be computed by substituting eqn (5) into eqn (4) and perform the inte-
gration using a parabolic approximation, which can be shown to compare well
with the numerical integration [9]. With chord length equal to L and total
mid-span lateral deflection equal to 8, eqn (4) may be approximated as:

Lo=L l + 3\Loj J (6)

in which the mid-span deflection, 6, can be computed by letting x = L/2 in


eqn (5) as:

6 = w(x = L/2) - p. (7)


1-
Pe

Substituting eqn (6) into eqn (3),

(1) 2_(287 (8)


1 + 3\Lo/

From eqn (1), the total elastic axial shortening can be computed as follows:

Ae=Lo ~ + 2]28 : " (9)


1+ 5

Substituting 6 from eqn (7) into eqn (9), the axial force-shortening relation-
ship of a strut in the elastic range can be computed as:

Ae=Lo ~ + 1 - (8 3o )"2 (10)


1 + 3 Lo(1 - P/Pe
28 J. Y. R. Liew et al.

2.2.2 Plastic p o s t - b u c k l i n g curve


The curve BC in Fig. 1 is characterised by a decreasing axial load
accompanied by column shortening after the axial load P reaches its maximum
strength, Pmax • The shortening is primarily due to lateral deflection of the
member, which facilitates the through-section and along-length plastification
of the member at mid-span caused by the P - 8 effect. Once the strut begins
to buckle, yielding develops rapidly, and the axial force decreases with the
increase in axial shortening and the latter is related to the lateral deflection.
The present model is developed by assuming that a plastic hinge is formed
at the member mid-span as shown in Fig. 3.
Neglecting the flexural deformation due to the elastic bending of two halves
in comparison with those due to the plastic rotation at the mid-span, the lateral
deflection at mid-span, 8, can be written as:

8 - m _ MpJMpZ (11)
P P/Py A

where Mp = full plastic moment capacity, Mpc = reduced plastic moment


capacity in the presence of axial force P, Py = axial load capacity at full yield,
A = cross-sectional area and Z = plastic section modulus of the cross-section.
From Fig. 3, the axial shortening due to geometrical change, Apg, can be
derived as:

\G/]" (12)

By adding the axial deformation Aa with Ag, the total axial shortening, Ap,
in the plastic range is given as:

I_ [. •

P
I- P

Fig. 3. Strut in the perfectly plastic range.


Advanced analysis and design of spatial structures 29

A p = A a + A g= Lo EA + 1 - 1 - \Loo] " (13)

Substituting 3 from eqn (11) into eqn (13), the axial force-shortening
relationship of a strut in the plastic range can be computed as:

,p j1 (14)

2.2.3 Elastic post-buckling unloading curve


Curve CD in Fig. 1 is due to elastic unloading of the strut at the post-collapse
range. Unloading occurs when the strut member is stretched due to the change
in sign of the incremental force. If unloading occurs in the elastic range,
the same elastic pre-buckling curve is followed; otherwise, the post-buckling
unloading curve is developed based on the total axial shortening, Amax, and
the axial load, Pu~, at the unloading point as shown in Fig. 1. Letting Ae =
mmax and P = Pu~, the equivalent member out-of-straightness magnitude 60 can
be computed from eqn (10). When the 60 value is known, the elastic post-
buckling unloading curve can be generated using eqn (10).

2.2.4 Plastic hinge yield surface


A plastic hinge is inserted at the mid-span of the member when the cross-
sectional forces, computed from a second-order analysis, reach the section
plastic strength. Once a plastic hinge is formed, the cross-section forces are
assumed to move on the plastic strength surface. The expressions for Mpc/Mp
in eqn (11) are dependent on the type of cross-sections. Plastic strength equa-
tions, which represent the relationship between the axial force and moment
on the yield surface, can be found in [11] for typical doubly symmetric cross-
sections as given below:
For rectangular section:

Mpc/Mp = 1.0 - (p/py)2. (15)

For circular tubular section:

Mpc/Mp = cos[( qr/2)(P/Py)]. (16)

For I- or H-section bending about the weak axis:


30 J. Y. R. Liew et al.

Mpc/Mp = 1.0 for 0 --< P --< 0.4Py (17)


MpdMp = 1.1911.0 - (p/py)2] for 0.4Py < P _< Py.

For I- or H-section bending about the strong axis:

MpdMp = 1.0 for 0 --< P -< 0.15Py (18)


MpdMp= 1.1811.0 - P/Py]for 0.15Py < P --< Py.

2.2.5 Design implementation


The load-carrying capacity of an axially load strut can be made to conform
with the code requirement for the design of columns. The code formulae for
axially compressed columns account implicitly for the effects of member
imperfections such as initial out-of-straightness and residual stresses. There-
fore, the strut model, with its ultimate load computed from the design code,
would satisfy the code's intention for column design including the member
imperfection effects. One major advantage of such approach is that explicit
modelling of member imperfections can be avoided in the analysis.
In the present formulation, the multiple column curves given in BS5950
(1990) [12] have been adopted to compute the maximum strength of strut
members depending on the types of cross-sections. The procedure to obtain
the axial force-deformation relation of a strut is given below:
(1) From the known material, cross-sectional properties and length of a
strut member, the axial load capacity, Pmax, is computed based on
BS5950 (1990) [12].
(2) Using the plastic post-buckling formulas eqn (14), compute the total
axial shortening Ap corresponding to P = Pmax.
(3) At P = Pmax the elastic pre-buckling curve intersects the plastic post-
buckling curve, and the total axial elastic shortening Ae from eqn (10)
is equal to Ap from eqn (14). The corresponding member initial out-
of-straightness magnitude, 30, can be back calculated from eqn (10) by
letting P = Pmax and me = m p .
(4) Given the value of 60, the elastic pre-buckling curve can be generated
for various values of P up to Pmax using eqn (10).
(5) When the strut is deformed beyond Ap which is shown as Ab in Fig.
1, the strut is in the plastic post-buckling range. Equation (14) should
be used to compute the axial force-shortening relationship.
(6) When the incremental axial displacement shows a reverse sign,
unloading from the strut curve is expected. The unloading curve from
the plastic post-buckling curve is assumed to follow the elastic pre-
Advanced analysis and design of spatial structures 31

buckling curve with member imperfection magnitude computed based


on initial shortening value, Af, as shown in Fig. 1. The procedure has
been described in section 2.2.3.

2.2.6 Verification of the strut model


Sugimoto and Chen [13] analysed the load--deformation behaviour of tubular
members using the finite segment method based on generalised cyclic stress-
strain relationships in the form of moment-axial force-curvature and axial
force-moment-axial strain relations and an automatic load control technique.
Chan and Kitipornchai [14] also developed theoretical axial load-shortening
curves for pin-ended tubular struts based on the fundamental stress-strain
relationship of the material and compared with those of Sugimoto and Chen
[13]. Sugimoto and Chen carried out their analyses on three pin-ended col-
umns with maximum initial out-of-straightness magnitude of 0.001 L at the
mid-span. The column member was divided into 13 finite segments along half
of the length. Tubular sections were used with outer diameter D = 114.3 mm,
wall thickness t = 2.4 mm, elastic modulus E = 2 × 105 MPa, and design
strength O-y = 248 MPa. The column slenderness ratios used in the study are
80, 120 and 160.
Figure 4 shows the comparison between these results [13] and those
obtained by the present strut model based on the same maximum strengths
(Pmax) obtained by Sugimoto and Chen [13]. It is seen from the figure that
there is a general agreement between the load-displacement results except in

1.0
P b L q P

0.0
~ . - Sugirnoto& Chert(1985)
.. PresentModel
0.6
P
Py
O.l,

0.2

0.0
0.000 0.002 0.00/, 0.006
Axiat Strtzin,AlL
Fig. 4. Comparison of strut curves with refined model.
32 J. Y. R. Liew et al.

the post-buckling range. This is because the proposed strut model is based on
member with 'enlarged' initial imperfection and elastic-perfectly plastic hinge
behaviour whereas in the Sugimoto and Chen model the cross-sectional behav-
iour is elasto-plastic with gradual plastification and spread-of-plasticity within
a finite length. The 'enlarged' imperfection approach adopted in the proposed
model makes the load-displacement curve more flexible towards the inelastic
region. However, the inaccuracy due to such approximation should be accept-
able for design purposes.

2.3 Tie model

The axial load-displacement relationship of a tie member is assumed to be


linearly-elastic with strain-hardening when the applied tensile force exceeds
the yield load Py as shown in Fig. 5. Assuming axial shortening as positive,
the governing equations may be written as:

P- E IAI for IA[ -< lay (19)

EAe
p- ~- my -- ~ ( I A I - Imyl) for IAI>IAyl (20)
where Py : Aetry is yield load, my is the axial elongation due to the tensile
force equal to Py, Ae is the effective area under tension, O-y is the material
yield strength, E is the modulus of elasticity, and Ep = E/IO00 is the strain-
hardening parameter.

Axial Load, P

ell

/ LIE.

Ay Axial elongation. A
Fig. 5. Axial load-displacement curve for a tie member.
Advanced analysis and design of spatial structures 33

3 SOLUTION ALGORITHM

3.1 Incremental-iterative equilibrium equation

Based on the updated Lagrangian formulation, the incremental equilibrium


equation of the structure can be written as:

[K i- I]{AU i} = { A p i} (21)

where [Ki-l] is the tangent stiffness matrix of the structure obtained at the
beginning of the load increment i and {AUi} is the incremental displacement
vector of the structure in response to a load increment of {APi} during the ith
load increment. Equation (21) can be solved by a simple incremental tech-
nique. The major deficiency of this technique is the risk for drift off from the
exact solution path and the difficulty in bypassing the limit points on the load-
displacement path. Correction for this uncertainty is taken care of by including
equilibrium iterations in the load vector at every load increment by means of
the generalised displacement control method.

3.2 Generalised displacement control method

The generalised displacement control (GDC) method originally proposed by


Yang and Shieh [15] is selected to solve the incremental equilibrium equation
given by eqn (21) considering its numerical stability near all types of limit
points. With the superscript i denoting the current load increment step and the
subscript j denoting the current iteration number, eqn (21) may be rewritten as:

(22)

where [Kj-d is the tangent stiffness matrix formed at the beginning of the
jth iteration based on the known element details at the (j-1)th iteration,
{rU)} is the iterative displacement vector obtained for the jth iteration, {Pj}
is the total external nodal loads applied on the structure at jth iteration, and
{~_ ~} is the internal element forces summed at each node of the structure up
to the (j-1)th iteration--all during the ith load increment. For the case of
proportional loading with a reference load vector {P} applied through a scalar
iterative load parameter o-Aj, eqn (22) can be further modified as:

(23)

where {Rj_ ~} denotes a vector of unbalanced forces given as:


34 J. Y. R. Liew et al.

{R~_,} : {Pj_~} - { ~ _ 1 } . (24)

The iterative displacement vector {6U~} in eqn (23) can be decomposed into
two parts as:

{6v~} = 6~{60j} + {au-j} (25)

where {80}} and {6U-}} are, respectively, the tangential and residual iterative
displacement vectors expressed as:

{60~}: [K~_,]-,{p} (26)

{6U3} [K~_I] {gj--1}.


=
-1 i
(27)

A generalised stiffness parameter (GSP) is introduced to compute the iterat-


ive load parameter until the convergence of incremental equilibrium equation
is achieved. The GSP i for any load increment, i, is defined as the ratio of the
norm of the first iterative displacement vector of the first load increment step
to those at the current load increment step:

{601F{6Ol}
GSP~= {60~- 1}T{68]}" (28)

For the first load increment, the GSP 1 value is equal to one. The first iterat-
ive load parameter of the first load increment is equal to the input value,
6)tl and for the subsequent load increments it can be computed using the
known GSP i value of that particular load increment as:

6A~ = • 6AIIGSPq ''z. (29)

The iterative load parameter, 6)t} for all the subsequent iterations (j > 1) is
computed using the following expression:

(30)

As the GDC method is based on the bounded characteristics of load para-


meter and displacement increments, it is able to bypass both the snap-through
and snap-back limit points without causing any numerical instability.
Advanced analysis and design of spatial structures 35

3.3 Convergence criteria

For each load increment, the equilibrium equation is solved by iterations till
the unbalanced force vector {Rj._~} becomes negligible. This is indirectly achi-
eved by satisfying the following energy criteria:

i T i
l{SU~} {R~-I}I
(31)

in which ~ is a user-specified tolerance. It can be a value between 1 x 10 -6


and 1 x 10 -1°. To terminate the iteration for non-converging and slow-
converging systems, a maximum number of iterations per load step is also
imposed.

4 NUMERICAL EXAMPLES

4.1 Two-bar truss system

This numerical example is intended to demonstrate the effect of member


imperfections on the limit load of the structure and to compare the limit load
predicted by the advanced analysis with the conventional elastic analysis. A
two-bar truss system, shown in Fig. 6, consists of square bars of 0.254 m x
0.254 m size with member slenderness ratio L/r of 150. The modulus of elas-
ticity of material is E = 2.06 × 105 N/mm 2 and yield strength, O'y = 235 N / m m 2.
The truss supports are restrained against translations. The structure is subjected
to a concentrated downward load, P, applied at the crown joint resulting in
compressive forces in both the members.
The structure is analysed using, (1) large-displacement elastic analysis in
which the axial force-shortening relationship of the members is assumed to

P
I
-~.695 rn

L 10.977rn _J_ 10.977m _J


i 7- i
Fig. 6. Two-bar truss system.
36 J. Y. R. Liew et al.

5000
l Elastic/ z r ~ t'6181

2500

0 A E
0.000 0.005 0.010 0.015 0.020 0.025
Axial shortening (m)

Fig. 7. Axial load-shortening curves for two-bar truss members.

be linearly elastic as shown by the dashed line in Fig. 7, and (2) advanced
analysis in which the axial force-shortening relationship of the members is
assumed to be nonlinear as shown by the solid curve in Fig. 7. The actual
nonlinear behaviour of these strut members is obtained using the strut model
proposed earlier in section 2.2. The strut capacity, P . . . . is computed as
4618 kN using the column curve 'b' of BS5950 [12]. The load versus vertical
displacement curves at the crown joint of the two-bar system obtained by the
two analyses are shown in Fig. 8.
For the advanced analysis, the load-shortening behaviour of the members
corresponding to different stages of loading are labelled as A, B, C, D and E
in Figs 7 and 8. The limit load predicted by the large-displacement analysis
assuming linearly elastic member behaviour is 1313 kN, whereas the consider-
ation of member nonlinearity by the advanced analysis yields a limit load of
445 kN. The advanced analysis shows that the structure attains its limit load
at point B (Fig. 8) before full member capacity of 4618 kN corresponding to

1500
1000 Elasticlarge
displacement /
500
alysis /
"~ o.
~ -SO0
-1000

-1500
80 o'.s 1'.o 1.s
Vertical displacement(m)
Fig. 8. Load-displacement curves for the crown joint of two-bar truss.
Advanced analysis and design of spatial structures 37

point C (Fig. 7) is reached. The failure of the two-bar system is due to the
geometric instability, with the members attaining their maximum strength at
the post-collapse range corresponding to point C in Fig. 7. Point D in Fig. 8
corresponds to the case where the crown joint of the truss is deflected up to
the support level. The deflection beyond point D will lead to unloading of the
member at the post-buckling range as indicated by curve DE in Fig. 7.
If one adopts an elastic analysis and applies the conventional member
capacity check approach to determine the limit load of the structure, the sum-
mation of vertical components of individual member capacity of 4618 kN
works out to be 584 kN. As this value is well below the buckling load of
1313 kN given by the elastic large-displacement analysis, it would be treated
as the limit load capacity of the truss. This way, the limit load of 584 kN
obtained by conventional analysis and design overestimates by about 30%
from the true limit load of 445 kN obtained by the advanced analysis. The
performance of advanced analysis reveals that the consideration of member
imperfections during the geometric nonlinear analysis triggers the structure to
attain its unstable condition at an earlier stage.

4.2 S p a c e d o m e s y s t e m

This numerical example is presented to demonstrate the effect of external


loading on the limit load of the structural system. Figure 9 shows a spatial
truss system consisting of 24 members to form a star-shaped structure. The
member cross-sectional properties are: A = 10 mm2; 1 = 41.7 mm4; E = 2.034
× 105 N/mm 2, O-y = 400 N/mm 2 and Ep = E/1000. All the supports are
restrained against translations and the remaining nodes are free to translate in
the space. The slenderness ratios (L/r) for members 1-6 and 13-24 are 123
and 155, respectively.

18V17 PLAN

t 62.16mm
• Supports
ELEVATION

Fig. 9. Star truss system.


38 J. Y. R. Liew et al.

_ 633

~oo j X

o g 1'o 1s
Vertical displocementat crown (ram)
Fig. 10. Load-displacement curves for the crown joint of star truss under load case 1.

The space truss is analysed for two load cases, (1) a single downward
concentrated load applied at the crown joint and (2) downward concentrated
loads applied at all the seven unrestrained joints. For the advanced analysis,
the axial load-shortening curves for the strut members are obtained based on
the column curve 'b' of BS5950 [12] and the tension members are modelled
as linearly elastic with strain-hardening defined by the plastic modulus Ep. The
structure is analysed by both elastic large-displacement analysis and advanced
analysis, and the relationships between the applied load, P, and the vertical
displacement at the crown joint of the structure are shown in Figs l0 and I 1
for load cases 1 and 2, respectively.
The axial load-shortening curves for the truss members are shown in Figs
12(a) and (b) for load case 1 and in Figs 13(a)-(c) for load case 2. The one-
to-one correspondence between the behaviours of structure and members are
labelled as A, B, C and D in Figs 10 and 12(a) and (b) for load case 1 and
as A, B and C in Figs I I and 13(a)-(c) for load case 2. For the load case
l, the ultimate load capacity of the structure predicted by the elastic large

2000.
~ 1500- : ~ 15/,3
"Loadedjoints . ~ "
-~ 1000" ~ x-Etosticlarge
/ displocement
-- / anolysis
.~-
'I:
500- /~Advonced onolysis

o ~ ~ ~ ~ lo
Vertical displacementat crown (ram)
Fig. 11. Load-displacement curves for the crown joint of star truss under load case 2.
Advanced analysis and design of spatial structures 39

800. .... ~7/.t,


1500
600.
II
0 ./L
I
~-members
9{-861~ -758} I to 6 LO0.

) -1500- /"C(-! 'I) --_=


members~ / a
7 to 12 "V 200,
"~ -3000- i I B(157)
/
I C(136)
J 0(51)
-LSO0
-0.8 -o'.L o.o o:~ o.B 0 O0 ~2 o'.~ 0'.6
Axid shortening {mm) ~ot shortening (ram)
(Q) {b)
Fig. 12. Axial load-shorteningcurves for star truss members. (a) 1-12; (b) 13-24 under load
case 1.

displacement analysis and advanced analysis are 633 and 372 N, respectively.
The corresponding limit loads for load case 2 are 10,801 N (7 × 1543 N) and
1659 N (7 × 237 N), respectively. By observing the label B in Figs 10, 11,
12(a) and 13(c), it can be seen that the limit load of the truss corresponds to
the geometric instability of the system and is not due to any member instability
under load case 1 whereas it corresponds to the instability of supporting mem-
bers 13-24 under load case 2. As per the conventional design, the vertical
components of axial capacity of six top members 1-6 (1132 N) yields the
limit load of the truss as 543 N under load case 1 as against 372 N obtained
by advanced analysis making an overestimation of about 45%. Similarly, by
considering the vertical components of 12 supporting members 13-24 (744 N),
the limit load of the truss by conventional design works out to be 1756 N as
against 1659 N obtained by advanced analysis making an overestimation of
only 6%. This shows that if the limit load of the structure is governed by the
instability of individual members and not due to the accelerated geometric
instability of the system due to the inclusion of member imperfections, then
the overestimation of limit load by conventional design is at an acceptable
level.
The analyses show that the structure can carry more load if the loads are
distributed evenly to all the unrestrained nodes rather than concentrated at the
crown node alone. The advanced analysis shows that the structure can carry
about three times more load as load case 2 (1659 N) than load case 1 (372 N),
whereas the conventional design (543 and 1756 N) shows about two times
increase in the total applied load.
The reason for the increase in load-carrying capacity due to the change of
loading conditions from load case 1 to 2 can be explained as below:
(1) Comparison of axial load-shortening curves for members 1-6 as shown
in Fig. 12(a) with Fig. 13(a) indicates that the maximum compressive
40 J. Y. R. Liew et al.

1200.
....

== 800" I
J

I
I

;B15031
•~ /,00-
/~(/3L71
'A
0
0.0 o:1 0".2 o.3 o:L 0.5
Axial shortening(ram)
(o)

1200 ...._.. -,K1~38


z js
800- i I
I
I
z~ iI
.~ ~oo-
~Ci2(353)
83)
0
0:0 o;1 0;2 o:3 o:L o.s
Axial shortening(mrnl
(b)

800

600-
"d
tOO. t

200- /

0
o.o o;5 fo ~:5 2.o
Axial shortening (ram)
(c)
Fig. 13. Axial load-shortening curves for star truss members. (a) 1~6; (b) 7-12; (c) 13-24
under load case 2.

loads in members 1-6 reduced by more than 50% from load case 1 to
2, and the members exhibit unloading in the elastic range under load
case 2 instead of attaining the maximum strength and loading into the
post-buckling range as under load case 1.
(2) Figures 12(a) and 13(b) show that load case 1 creates tensile force in
members 7-12 whereas load case 2 creates only compressive force.
(3) The supporting members (members 13-24) are subjected to compress-
ive force under both the load cases. However, corresponding to the
Advanced analysis and design of spatial structures 41

21 ,,, 22 23

18

Fig. 14. Deformed configuration at the maximum load under load case 1.

label B at which the structure reaches its limit load, the magnitude of
axial force is only about 20% under load case 1 than under load case
2 as indicated in Figs 12(b) and 13(c).
The above observations show that the concentrated load under load case 1
causes an instability failure due to a snap-through of the crown joint, as shown
by the deformed structural configuration at the maximum load in Fig. 14. The
distributed loads under load case 2 impede the snap-through behaviour of the
crown joint and stress the horizontal members 7-12 in compression instead
of tension causing all the 12 supporting members to be stressed to their indi-
vidual axial capacity in comparison with load case 1 in which all the forces
are concentrated mainly to the top six inclined members, 1-6. These combined
effects lead to an enhancement of the limit load of the structure under load
case 2. The maximum load under load case 2 is reached due to the instability
of the supporting members 13-24 as shown by the deformed configuration in
Fig. 15. Thus, the advanced analysis enables a better understanding of the
role of each individual member in the structure and provides an accurate pre-
diction of the member and system collapse behaviour. The use of elastic large-

S 15 -

Fig. 15. Deformed configuration at the maximum load under load case 2.
42 J. Y. R. Liew et al.

displacement analysis may sometimes lead to an unconservative estimate of


the system strength and stability.

4.3 Twelve-bar space truss system

This numerical example is presented to demonstrate the effect of member


slenderness on the limit load of the structural system. A spatial truss system
consisting of 12 tubular members is shown in Fig. 16. Tubular members of
different diameters, namely CS-1 and CS-2, are used in the analysis to demon-
strate the effect of member slenderness on the overall behaviour of the struc-
ture. The CS-1 section consists of an outer diameter D = 193.7 mm, wall
thickness t = 10 mm, area of cross-section A = 5770 m m 2 and moment of
inertia I = 2442 × 104 m m 4. The corresponding properties for CS-2 section
are D = 168.3 mm, t --- 5 mm, A = 2570 m m 2 and I = 856 × 104 m m 4. The
modulus of elasticity for material is E = 205 kN/mm 2 and yield strength, try
= 275 N / m m 2. All the supports are restrained against translations and the
remaining nodes are free to translate in the space. The slenderness ratios (L/r)
of supporting members 1, 2, 11 and 12 are, respectively, 65 and 58 with CS-
1 and CS-2 member properties. The spatial truss is analysed with a load P
applied at the central joint marked as 'A' and concentrated load 1.5P at the
two end joints. The structure is analysed by elastic large-displacement analysis
and advanced analysis, and the relationship between the applied load, P, at
joint A and the vertical displacement, W, at joint A, and horizontal and vertical
displacements U and W at joint B are presented in Figs 17(a)-(c) and 18(a)-
(c), respectively. For the advanced analysis, the axial load-shortening curves
for the tubular strut members are obtained based on column curve 'a' of
BS5950 [12].
Figure 19(a)-(d) shows the axial load-shortening curves of individual mem-
bers at different stages of loading. Similar to previous numerical examples,

~1.0P/1.SP
iO00

iO00 ELEVATION
~11 dimensions ere in mm
IO00~L_ members ere of circdor tubes of,
Type CS-1, 0 = 193:7 mm; t = I0 mm
I. 353~5 ,~ 3539.5
Type CS-2, O= 168.3 mm; t= 5 mm
• S~perts
," Loaded joints

PLAIt
Fig. 16. Twelve-bar truss system.
Advanced analysis and design of spatial structures 43

100O0O

w 80000-

:~
,-~ 6o000-
o.,.
t,0OO0-

20000 -
CS-2-,'
0
-200 -100 0 100 200
gisplocement W at joint A (mm)
(a)
100000 rLse.JP t ~ ~
~ 80000-

o 60000.

-~ ldl000- ' 36030

20000-

0
lOO 2~o 3~o & s~o 6~o 7~o ~o
0is jtncement U at joint B (rnm)
(b)
100000 , tse -P lzs~P

~ 2

o 2000
0isplocement W at joint B (turn)
(c)

Fig. 17. Load-displacement curves for 12-bar truss by elastic large displacement analysis.

the one-to-one correspondence between the behaviours of structure and mem-


bers are labelled as A, B and C in Figs 18(a)-(c)and 19(a)-(d). From Fig. 17,
the limit loads predicted by the elastic large-displacement analysis for struc-
tures with sections CS-1 and CS-2 are 80,980 and 36,030 kN, respectively.
The corresponding .values predicted by the advanced analysis are 1096 and
419 kN as indicated in Fig. 18(a)-(c).
44 J. Y. R. Liew et al.

1200
1SP ,P ILSP 90095)

,¢¢

,oo
- / ~/-CS-1
a-

L,O0 / .~1,191

Oisplocement W at joint I (mm)


(o)
1200"
z
•-" 100o.
,,,c:(

"~ 800.
:e_. ~- CS-1
"G
aL. 600-

~ 200-
""° C1911
01
Oisptocement U at joint B (ram)
(b)

1200
B(1096) tSP |P ~1.5P
i

A
1000
G=
"~ 800
=- 600

LO0
g
:~ 200

0 20 l,O 60 60
Oisplecement W at joint B (mm}
(c}

Fig. 18. Load-displacement curves for 12-bar truss by advanced analysis.

Advanced analysis shows that the limit load of the structure is governed
by the instability of the supporting members as shown by the collapse con-
figuration at maximum load in Fig. 20. The supporting members 1, 2, 11 and
12 reach their maximum strength as indicated by the label B in Figs 18(a)-
(c) and 17(a). The configuration of the structure is such that it is stiff enough
to prevent any joint instability. Hence the limit loads predicted by the
Advanced analysis and design of spatial structures 45

,~oo ,2oo I ...,

I A¢'815121 . .~I.SL
/'00171C(277J
cs-~". . . . . .
I '°" "/
L.,,~,~ /) . - "
"
"--- cs-2
"
"
O ~ PI~/"-"~°B1203)
, ,

0 50 100 150 200 A0 2 ~ 6


Axio[ shortening (mm) Axio[ shortening (mm)
(a) (b)

800. 1500-
58t,
1213
60O
•~ CS-I
lOO0.
p
t,00 I
I s'"" /.8/.
/ 21"8
500- .,,,"a(Lm ..... ~....
200 /0(205).. - "':-~-cs-2
C1/,7 811 ~no~/eTle5 } "---cs 2
0 ¢¢~" C[161 , ,d~..-c1/,ol
A 2 L ~ ~ ~o
Axiut shortening (mm) Axial shortening (mm)
(c] (d)

Fig. 19. Axial load-shortening curves for 12-bar truss members. (a) 1, 2, 11 and 12; (b) 6 and
7; (c) 3, 5, 8 and 10; (d) 4 and 9.

2 9

11

Fig. 20. Deformed configuration at the maximum load.

advanced analysis for structure with member properties CS-1 and CS-2 are
in proportion to the axial capacity of the critical members. In the case of
elastic large-displacement analysis, the structure limit loads are proportional
to the member cross-sectional areas. By computing the limit loads based on
the vertical component of axial capacities of governing members 1, 2, 11 and
46 J. Y. R. Liew et al.

12, the conventional design would yield P values as 1120 and 444 kN as
against 1096 and 419 kN by advanced analysis. The overestimation is only
2%. This strengthens the previous observation made with the star truss
example that if the limit load of the structure is governed by the instability
of individual members and not due to the accelerated geometric instability of
the system due to the inclusion of member imperfections, then the overestim-
ation of limit load by conventional design is at the acceptable level.

5 CONCLUSIONS

A powerful advanced analysis/design method based on inelastic large dis-


placement analysis has been proposed to compute the maximum strength of
spatial truss-structures without the risk of overestimating the maximum
strength of the component members in the structure. The proposed strut model
is made to conform with the code requirement for strut design. This approach
removes, to a large extent, the limitation of the conventional analysis which
requires member capacity checks and the modelling of member initial imper-
fections for representing individual member strength.
From the studies carried out on the example truss systems, the following
conclusions may be drawn:
(1) The performance of advanced analysis with the elastic-rigid plastic
modelling for strut members as per the design specifications predicts
the 'true' limit loads of spatial truss structures.
(2) The conventional structural design of performing elastic large-displace-
ment analysis followed by individual member capacity checks may
sometimes lead to over-prediction of the limit load capacity. If the limit
load of the structure is governed by the instability of individual mem-
bers and not due to the accelerated geometric instability of the system
due to the inclusion of member imperfections, then the overestimation
of limit load by conventional design is acceptable.
(3) The member imperfections, member sizes and the type of load distri-
bution on the structure have appreciable effects over the limit load
capacity of the structure as revealed by the performance of advanced
analysis on the three truss systems.
(4) The advanced analysis, in addition to predicting the limit load of the
structure, helps to identify the load sharing and force distribution mech-
anism of individual members forming the structural system. The identi-
fication of the critical members in the structure enables the engineer to
redesign the system so as to enhance the structural resistance and per-
formance.
Advanced analysis and design of spatial structures 47

The proposed method shows promise as an effective means of accounting


for member initial imperfections without physically altering the frame
geometry. The advanced analysis fulfils most of the design specification
requirements and can be applied for the design of complex spatial structures
in which system instability can be a critical factor for the assessment of the
structural limit states.

ACKNOWLEDGEMENTS

The investigation presented in this paper is part of the research programs on


'computer aided second-order inelastic analysis for frame design' being car-
fled out in the Department of Civil Engineering, National University of Singa-
pore. The work is funded by research grants (RP920651) made available by
the National University of Singapore.

REFERENCES

1. Liew, J. Y. R., White, D. W. and Chen, W. F., Limit-states design of semi-rigid


frames using advanced analysis: Part 1: connection modelling and classification;
Part 2: analysis and design. Journal of Constructional Steel Research, 1993,
26(1), 1-57.
2. Liew, J. Y. R. and Chen, W. F., Implications of using refined plastic hinge analy-
sis for load and resistance factor design. Thin-Walled Structures, 1994, 20(1-4),
17-47.
3. Chert, W. F. and Toma, S., Advanced Analysis in Steel Frames: Theory, Software
and Applications. CRC Press, Boca Raton, FL, 1994.
4. Chen, W. F., Goto, Y. and Liew, J. Y. R., Stability Design of Semi-rigid Frames.
John Wiley and Sons, New York, 1996.
5. Liew, J. Y. R., Punniyakotty, N. M. and Shanmugam, N. E., Inelastic large-
displacement analysis of space structures. MINDEF-NUS Joint R and D Seminar,
Defence Technology Group, Ministry of Defence, Singapore, January 1996, pp.
132-139.
6. Liew, J. Y. R., Shanmugam, N. E. and Punniyakotty, N. M., Advanced analysis
of space structures considering member stability effects. Proceedings of the Asia-
Pacific Conference on Shell and Spatial Structures, ed. T. T. Lien. China Civil
Engineering Society, Beijing, 1996, pp. 228-237.
7. Murtha-Smith, E., Compression member models for space trusses: Review. Jour-
nal of Structural Engineering, ASCE, 1994, 120(8), 2399-2407.
8. Papadrakakis, M., Inelastic post-buckling analysis of trusses. Journal of Struc-
tural Engineering, ASCE, 1983, 109(9), 2129-2147.
9. Maheeb, M. E. Abdel-Ghaffar, Post-failure analysis for steel structures. Ph.D.
thesis, Purdue University, 1992.
10. Chen, W. F. and Lui, E. M., Structural Stability--Theory and Implementation.
Elsevier, New York, 1986.
48 J. Y. R. Liew et al.

11. Chen, W. F. and Sohal, I. S., Plastic Design and Second-order Analysis of Steel
Frames. Springer-Verlag, New York, 1995.
12. BS5950, Structural use of steelwork in buildings. Part 1: Code of practice for
design in simple and continuous construction: hot rolled section. British Stan-
dards Institution, London, 1990.
13. Sugimoto, H. and Chen, W. F., Inelastic post-buckling behaviour of strut mem-
bers. Journal of Structural Engineering, ASCE, 1985, 111(9), 1965-1978.
14. Chan, S. L. and Kitipornchai, S., Inelastic post-buckling behaviour of tubular
struts. Journal of Structural Engineering, ASCE, 1988, 114(5), 1091-1105.
15. Yang, Y. B. and Shieh, M. S., Solution method for nonlinear problems with
multiple critical points. AIAA Journal, 1990, 28(12), 2110-2116.

You might also like