Vous êtes sur la page 1sur 15

Journal of Constructional Steel Research 57 (2001) 12171231



Non-linear behavior and design of steel

S.L. Chan

Department of Civil and Structural Engineering, Hong Kong Polytechnic University, Hung Hom,
Kowloon, Hong Kong
Received 26 March 2001; accepted 13 August 2001

Frames are possibly the most common forms of man-made engineering structures. They are
formed by joining one-dimensional members together. Since the early work of Euler, engineers
started to realize that the strength of a member under compression does not only depend on
the material yield stress, but also on the Youngs modulus of elasticity. The introduction of
steel material and other metals makes the consideration more important because of their relatively slender dimensions. From the vision of the current computer age, this paper is addressed
to a review and summary of the work conducted on the non-linear analysis and design of steel
frames in the past few decades. 2001 Published by Elsevier Science Ltd.
Keywords: Stability analysis; Advanced analysis; Buckling; Structural design; Steel structures; Non-linear
integrated design and analysis

1. Introduction
Early work on stability and buckling of structures are concentrated on member
behavior. Bleich [1], Goodier [2], Vlasov [3] and Timoshenko and Gere [4] are
among the researchers in the study of buckling of one-dimensional members. The
methods of column deflection curves [5], finite difference [6] and finite integral [7]
were employed for solving the differential equilibrium equation for columns and
beams. The RayleighRitz [8] method is based on a correctly assumed deflected
* Tel.: +852-2766-6047; fax: +852-2334-6389.
E-mail address: ceslchan@polyu.edu.hk (S.L. Chan).
0143-974X/01/$ - see front matter 2001 Published by Elsevier Science Ltd.
PII: S 0 1 4 3 - 9 7 4 X ( 0 1 ) 0 0 0 5 0 - 5


S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231

shape and therefore it is again limited to simple problems where the deflected shape
can be defined accurately. Based on the energy method, a discretization concept and
the invention of powerful computers, the finite element received more attention in
the past few decades as a more general and powerful tool for obtaining the equilibrium condition at both the linear and the non-linear ranges. The excellent book by
Chen and Atsuta [9] covers various aspects of numerical methods and analysis and
design of beam-columns under different loading and boundary conditions. Recently,
Lindner [10] presented a summary on the recent work on member design used mainly
in German standard and Euro-code.
The advantage of the finite element method lies mainly in the flexibility of
assigning properties, configuration and boundary and load conditions and consequently a finite element program can be applied to a wide class of problems. Owing
to the solution of a differential equilibrium equation for a beam member in a cubic
form, some researchers do not classify the cubic Hermite element as a type of finite
element which normally approximates the solution in a discrete manner. The linear
frame analysis method is therefore sometimes referred as the matrix method of analysis. However, when a member is under an axial force, the cubic Hermite element is
no longer a valid solution and more elements are needed to approximate the buckled
shape of a column. This paper is addressed to the review of advances, techniques
and theoretical background of the non-linear analysis and design of steel frames with
a context of practical application.

2. Common types of non-linear analysis

Non-linear analysis can be defined as any analysis where linear extrapolation of
stress, load and deflection is invalid. There are several principal types of non-linear
analyses which are summarized graphically in Fig. 1. These methods have their
merits and limitations. With the advent in computer capacity, some of these methods
are no longer an attractive solution methods for engineers to date. However, they
play an important role in the development of stability theory applied to practical
structures and deserve elaboration.

3. Bifurcation analysis
The most simple type is the bifurcation analysis which assumes a sudden interception of a secondary equilibrium path to the primary path and the solution is obtained
by solving the characteristic equation indicated in Fig. 1 and as follows:
KL lKG 0


in which KL is the linear stiffness matrix, l the load factor and KG the geometric
stiffness matrix.
In this eigenvalue analysis, the pre-buckling deformation, initial imperfection and
material yielding are ignored. It represents an upper bound solution and it is generally

S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231

Fig. 1.


General analysis types of framed structures.

not accurate enough for direct practical design, even for the very slender structural
form like steel scaffolding [11,12]. However, its solution is simple and can be easily
included in a vibration analysis software. Also, it can be used to work out the effective length factor [13].

4. Rigid plastic analysis

An opposite type of analysis is the rigid plastic analysis which considers only
material yield and plastic hinge, and instability and large deflection are totally
ignored. It can only be used for frames with negligible geometrical non-linearity,
otherwise additional modification is needed. A mechanism approach of equalizing
the external work done to the internal strain energy for the most critical collapse
mechanism can be used to determine the load factor causing collapse. The method
can be summarized in Fig. 1 and as follows:




in which Fi is the external force, di the conjugate displacement, Mpj the plastic
moment at location j and qj the conjugate rotation.
In the pre-computer age, the location of the most critical mechanism is a rather
complex procedure. The books by Heyman [14] and Neal [15] give a full account
of this method of plastic analysis and design. The use of computers significantly


S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231

simplifies the procedure. A simple computer method is to apply the design load and
construct the bending moment diagram. The ratio of the bending moment to the
plastic moment at any section is computed at all nodes of the frame and the maximum
ratio is then determined. The design load can then be amplified by this ratio so that
the first plastic hinge can be formed by this revised load factor and a plastic hinge
is then inserted into this location or node. The analysis is then repeated incrementally
using the new structure with one plastic hinge after the first increment, or more
plastic hinges after the second increment onwards. The procedure, summarized in
Eq. (3) and in Fig. 1, is continued and repeated until a collapse mechanism is formed
li min

| |
; l



in which l is the incremental load factor, jMp the plastic moment capacity of the
section at node j and jM the bending moment constructed under the current structural plasticity stage. Min | | is an operator for selection of the minimum absolute
value of the ratios inside.

5. Empirical collapse analysis allowing for buckling and plastic hinges

Merchant [16] considered both the effects of material yield and buckling using a
simple and very popular MerchantRankine formula obtained from the elastic buckling and plastic collapse load factors as follows:

lult lp lcr


in which lult, lp and lcr are respectively the collapse, plastic and elastic buckling
load factors.
This approach does not consider the large deflection effect and cannot consider
initial imperfection. Also, it cannot be used to calculate the deflection at various
load stages and its use is relatively inconvenient.

6. Geometrically non-linear analysis

In the non-bifurcation approach of tracing the equilibrium path, much more complex formulations are needed, in addition to the formation of the linear and geometric
stiffness matrices in Eq. (1). This involves the kinematic description of the motion
of a largely deformed element by the total Lagrangian [17], the updated Lagrangian
[18] and the corotational approach [19]. The initial stress in a member can be formulated by the geometric stiffness matrix based on the cubic Hermite displacement
function by Barsoum and Gallagher [20], the stability function by Majid [21], the
weak formulation by Shi and Atluri [22] and the self-equilibrium element of
satisfying equilibrium at discrete points along an element by Chan and Zhou [23].

S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231


Some of these formulations make the direct use of equilibrium equations [21,23],
whilst others use the energy principle [20,22].
Early work on the large deflection and elastic analysis of framework was conducted by Williams [24], Jennings [25], Powell [26], Connor et al. [27] and Rajasekaran
and Murray [28] among others. Hancock [29] used only the secant stiffness for his
large deflection analysis of I-sections. These works were based on small strain but
large deflection assumptions which are valid for steel material. Also, they normally
assume that the rotations in three-dimensional space are vectors that they can be
accumulated. Argyris and Doltsinis [30] proposed a natural formulation for solid
beam-columns and the non-conservative nature of moments at joints were examined.
Yang and McGuire [31] simplified the method and applied it to geometrically nonlinear analysis for frames. Chan and Kitipornchai [32] derived the linear and geometric stiffness matrices and conducted large deflection and bifurcation analysis for
open sections allowing for warping and bimoment. Prokic [33] formulated an element
with warping degree of freedom. However, non-linear frame analysis allowing for
warping and bimoment has not yet developed to a stage of practical use, even for
relatively simple frames.
Oran [19] produced an early work on treating the non-vectorial property of
rotations in three-dimensional space by the joint orientation matrix attached rigidly
to the ends of an element. The member flexural rotations can then be obtained by
the dot product between the member and joint orientation axes. Chan [34] used
another simple approach of avoiding the non-accumulative limitation of rotations in
three-dimensional space by using an incremental secant stiffness procedure. However, in most practical structures, this limitation is generally unimportant since failure
occurs well before attaining this large rotation, of which the limit is generally taken
as 15 degrees.

7. Post-buckling analysis
An extension of the large deflection but pre-buckling analysis is to trace the postbuckling equilibrium path. It is well known that the load control NewtonRaphson
method [35] diverges near the limit point. Suppression of iteration near limit points
is a workable method, but it is inconvenient to use, introduces an unknown amount
of error and is computationally inefficient. Batoz and Dhatt [36] invented a useful
procedure of controlling the advance of the equilibrium path by specifying a displacement increment at a particular degree of freedom instead of the load. Simultaneously,
unlike the argumented spring method [37] or the original form of the arc-length
method proposed by Riks [38], it does not destroy the symmetrical nature of the
tangent stiffness matrix which is highly undesirable in terms of computational
efficiency since symmetry in the stiffness matrix can significantly reduce the computer time and storage. Unfortunately, this displacement method breaks down when
dealing with the snap-back problem, in a way similar to the load control method
breaking down at the limit point. Ramm [39] and Crisfeild [40] developed independently in 1980 the arc-length method for tracing the post-buckling path using the


S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231

arc-distance for the displacement norm to control the advance of the equilibrium
path. Meek and Tan [41] and Chan and Kitipornchai [42] used the arc-length method
to trace the post-buckling equilibrium path of elastic and inelastic frames. Since an
iterative process is targeted at minimizing the error, instead of satisfying the arcdistance, Chan [43] later adopted a simple technique of minimizing the equilibrium
error for displacement norm which, as pointed out by Clarke and Hancock [44], is
more rational and performs equally well with the arc-length method. The implementation of the method is also much more simple since, unlike the arc-length method,
no quadratic equation is needed to be solved and thus avoids the possibility of nonexistence of a real root.
An important consideration for a second-order or large deflection analysis is the
accuracy of element stiffness. The expressions for the geometric stiffness matrix
from the cubic Hermite element and from the stability function are given in a report
by Task Committee on Effective Length [45]. Whilst the cubic Hermite element
gives the exact solution for the differential equilibrium equation of a beam under a
pair of nodal moments and the assumption of linear behavior, the use of a single
element to model a member is adequate for linear frame analysis. However, the
equilibrium equation for a beam-column in the presence of an axial force shows that
the cubic function is inadequate and the displaced shape is in the form of a trigonometric function which is different for tensile and compressive axial loads. The error
for using the cubic Hermite function can be significant. For example, for a column
with one end fixed and other end pinned, the buckling problem can be simplified to
a single degree of freedom. Using the widely used geometric stiffness matrix [45],
an error of 49% will be resulted. Therefore, an unacceptable error will be present
when the axial force in a member is large of, say, more than 40% of the Eulers
buckling load.
A challenge for the researchers in structural engineering is to formulate an element
allowing for the modeling of a member by a single element. This objective is to
allow a linear analysis model for immediate use by a second-order non-linear analysis
without additional model refinement. When considering the software being used by
a practicing engineer during his early stage of using non-linear analysis, he is likely
to try his model for a linear analysis before he goes for a second-order non-linear
analysis. This objective of allowing the consistent use of the model for linear and
non-linear analysis is therefore desired if we really want the non-linear analysis to
be used daily and accepted by the profession. However, whilst the cubic Hermite
element is deficient in this aspect, the form of the stability function used by many
researchers is not satisfactory either. Very often, the stiffness matrix derived from
the stability function neglects the coupling between axial force and bending, implying
that the bowing effect for a member is ignored. This is unacceptable since the use
of the stability function is for accuracy of an element and such an ignorance of the
coupling effect will downgrade the accuracy of the element stiffness to a level similar
to the cubic Hermite element and therefore the use of the stability function is meaningless. Some researchers used a cubic or other element for tracing the equilibrium
path of a shallow arch or toggle frame and claimed that the element is adequate for
buckling analysis. This checking is not valid since the element stiffness is not under

S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231


a high axial force in the selected examples. Using the terminology by Zienkiewicz
[46], the geometrically non-linear analysis can be carried out by an incremental
analysis as follows.
[F] [KL KG K0][u]


in which F and u are, respectively, the incremental force and displacement vectors
and K0 is the large deflection matrix.
The effect of large displacement is included in the last matrix K0 whilst the effect
of initial stress is considered in the second matrix. The use of a large deflection
problem can only validate the accuracy of K0, but not necessarily KG. Note that, the
effect of the large displacement matrix K0 can be included in the analysis by a nodal
coordinate updating process. Kinematic formulations such as the Eulerian formulation and the total and the updated Lagrangian formulations can be used for description of the motion of a largely deflected element with reference to the co-rotational,
undeformed and updated coordinate systems and they will lead to various forms of
total and incremental equilibrium equations. Teh and Clarke [47] studied the merits
and limitations of the co-rotational and the Lagrangian formulations of elastic and
three-dimensional beam elements. Albermani and Kitipornchai [48] developed a formulation for using least elements in modeling with focus on the large displacement
effect. Izzuddin and Elnashai [49] developed a higher order quartic element for
second-order analysis by one element per member in the elastic range. Shi and Atluri
[22] derived an explicit form for the tangent stiffness matrix of slender members
using the virtual work method. Their expressions [22] are accurate and one element
per member is adequate for the buckling analysis of slender frames under high axial
force. Material yield and semi-rigid connections can also be included in their method
by varying the spring stiffness connected to the two ends of their element.
It has been pointed out that testing of an element formulation must be carried out
for structures under high axial force otherwise the accuracy of the element cannot
be checked. Chan and Zhou [23] formulated a finite element which estimates accurately the elastic buckling load of a column under different boundary conditions. In
their formulation, a displacement based and discrete equilibrium finite element is
derived. They further pointed out that the error for Meek and Tan [41] in tracing
the snap-through equilibrium path of the frame with the hexagonal shape on plan is
not due to the vectorial assumption for rotations, but because of the inaccuracy of
the cubic element. Recently, Neuenhofer and Filippou [50] employed a flexibilitybased element with two to four Gausss points to trace the equilibrium path of slender
frames. All these works demonstrate the importance of the accuracy of the element
stiffness for efficient and accurate second-order analysis of slender frames under
high axial force. The cubic Hermite element is certainly inadequate in this aspect,
unless two or more elements are used to model a member. For the refined and
improved elements [22,23], their validity is only limited by the very large rotation
between the element chord joining the two end nodes and the tangent at end nodes
of which the occurrence is very unlikely in practical applications.
All realistic members possess initial imperfection and codes and standards like
the British Standard [51] and the Australia Standard [52] require that an initial imper-


S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231

fection of 0.1% of member length should be assumed in buckling strength determination by formula like the Perry Robertson equation. Interestingly, the research in
this direction is relatively less. This indicates that the above-proposed elements are
inadequate for practical uses unless other remedial measures are undertaken. This
includes the use of several elements to model a member in a segmental manner or the
direct use of a curved element [53,54]. Chan and Zhou [55] extended their element to
allow for initial imperfection and curvature. An originally deformed shape in place
of a perfectly straight element is used in their formulation. Chan and Gu [56] further
modify the stiffness matrix from the stability function to include the effect of initial
imperfection. Using these elements, the buckling strength curves can be produced
by one element per member and compared very well with the design strength curves
in the code by assuming an appropriate value of initial imperfection.
An obvious, alternative and widely used procedure is to make use of a simpler
second-order frame analysis program or other analytical technique for direct computation of effective length and this approach has also been adopted by a number of
researchers including, Yura [57], Duan and Chen [58], Lui et al. [59]. Once a correct
effective length is assessed, the column buckling strength can be determined by the
buckling curve. However, this approach may not be convenient to use, especially in
cases when a column is under a variable axial force and with different sections along
its length, as in the case of a stanchion in a steel building supporting the reactions
of beams at different floors. Most importantly, the approach does not account for
variation in element stiffness which leads to an incorrect construction of the force
and moment diagrams for design. For example, the compression member in a cross
bracing does not take much load and the majority of loads is carried by the tension
member because of varying stiffness for members under compression and in tension.
The use of eigenvalue analysis gives us a buckling load factor for the complete
structure, but it does not provide reliable information for member design which is
a serious limitation and does not meet the requirement for an effective second order,
advanced analysis and non-linear integrated analysis and design (NIDA) which
requires the consideration of varying stiffness against the magnitude of axial force.
The necessary tools for large deflection analysis have been quite well developed.
With the performance and cost of a personal computer, the analysis can be carried
out quite effectively and efficiently in a design office to date. The extension of the
method to inelastic analysis is natural for tracing the complete equilibrium path until
total collapse. Much research in non-linear analysis is targeted at such a design
technique for practical structures. The important feature for this method is the suitable consideration of the practical characteristics of a real steel structure such as the
residual stress, member and frame initial imperfection and out-of-plumbness, varying
element stiffness due to axial force etc. When all these effects are properly accounted
for in an analysis, one can skip the manual member capacity check and only the
sectional capacity check is required. The reserve in strength after the first plastic
moment can be exploited. By the procedure, the effective length is not required to
be assumed which leads to a more economical or a safer design. Advanced analysis
is referred to this type of design method based on the truly ultimate analysis with
allowance of various features in a practical structure. Two common methods used

S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231


by the researchers for conducting a second-order inelastic or an advanced analysis

are the plastic hinge and the plastic zone methods.

8. The plastic zone method

The plastic zone assumes yielding to spread across the section and along the
element. This concept has been used by numerous researchers including Vogel [60],
Chan [61] and Clarke [62]. The essence of the method can be stated in the incremental equilibrium equation as follows and indicated graphically in Fig. 1
[Kp KG K0][lF ] [u]


in which Kp is the elasto-plastic stiffness matrix allowing for material yield. The
above incremental stiffness equation differs from the second-order elastic approach
in the use of this elasto-plastic stiffness matrix instead of the linear stiffness matrix,
KL, in Eq. (5) through a numerical integration approach in sampling element
yielded stiffness.

9. Plastic hinge method

In the plastic hinge method, yielding is concentrated at a small zone modeled by
a flexible spring. When no yielding occurs, the spring stiffness is infinitely large
and, when the plastic moment capacity is reached, the spring stiffness tends to zero.
This process can be considered conceptually by an inclusion of a flexible spring
stiffness in the incremental equilibrium equation in Fig. 1 as follows

spring][lF] [u]


in which Kspring is the spring stiffness for simulation of plastic hinges. Note that when
the spring stiffness is infinite, it has no influence in the stiffness Eq. (7). When the
spring stiffness vanishes due to material yielding, Eq. (7) indicates a smaller or a
diminished stiffness in the associated degree of freedom, which, in some case, refers
to a plastic collapse mechanism.
Note that the linear element stiffness by itself does not consider material yield
here so that KL used is the same as in the second-order elastic analysis in Eq. (3).
Yielding is considered only at the plastic hinge modeled as spring elements Kspring
in Eq. (7).
Powell and Chen [63], King et al. [64] and Chen and Chan [65] employed the
method in tracing the equilibrium path of a structure and compared their results very
well with the plastic zone method in most, but not all, problems. When the spread
of plasticity is rather uniform along a member, the concentrated plastic hinge method
may not truly reflect the behavior of the member. Nevertheless, it appears that the
accuracy of the method is sufficient for practical purposes. Benchmark examples for
calibration of computer output were also produced by Vogel [60] who used a more
refined plastic zone analysis. Toma and Chen [66] summarized several calibration


S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231

frames for checking of accuracy a software for inelastic and advanced analysis of
steel frames using plastic zone method. However, this work is mostly concentrated
either on elastic analysis of semi-rigid frames or elasto-plastic analysis of frames
with rigid joints. Yau and Chan [67] extended this approach and developed a numerical method for efficient and simple analysis of steel frames allowing for both the
effect of plastic hinge and semi-rigid connections. Their work is simple and robust
in dealing with the non-linearity in material yield and semi-rigid connections and
among the earliest research work in this aspect. The classification of semi-rigid connections has been imposed in some codes like the LRFD [68] and the work represents
a possible numerical tool for this type of advanced analysis for semi-rigid steel
frames. Chan and Chui [69] detailed the analysis technique to structural response
under static and dynamic loads, allowing for hysteric loops at plastic hinges. From
the worked examples, their work can be applied equally well in design under dynamic
or cyclic loads. Recently, Chan et al. [70] investigated the response of steel frames
under a non-proportional load, which is noted to be considerably different from the
widely assumed proportional loading condition.
An important point that must be noted here is that some proposed elements to
include the yield functions described in various national standards are unable to
produce an effective second-order or advanced analysis because their stiffness is
inaccurate under compression or tension. This leads to an incorrect construction of
force and moment diagrams and therefore any design using this output is inaccurate.
For example, in the design of a cross bracing system, a compression member shares
a very small load when compared with the tension member. The linear analysis
calculates the compression member to take half the lateral load and the design based
on this analysis result is therefore incorrect. A less accurate second-order analysis
may lead to a similarly incorrect output, perhaps at a smaller degree of error. Whilst
an experienced engineer will intelligently omit the compression member in his design
and assume all lateral load taken by the tension member, it is generally an uneconomical approach, especially when the compressive members are of medium slenderness. Also, when designing a large and complex structure, this procedure is tedious
to apply.
Generally speaking, the plastic zone analysis requires the division of a section
into a mesh and a member into many elements in order to simulate the effect of
spread of plasticity. It is more accurate than the plastic hinge method, but requires
longer computer time. The development of computer technology has permitted the
use of more refined and computer time extensive techniques for refined analysis of
frames. Izzuddin and Elnashai [49] used an adaptive mesh approach in refining an
element where yielding is detected and elastic elements are used for other members
without material yield. Corradi and Poggi [71] and Feng et al. [72] used numerical
integration schemes with different levels of proficiency to sample material properties
across a section in a beam-column. This concept is simple and efficient since the
element matrix is normally formulated by an explicit integration process which is
now converted to a numerical integration procedure in their papers. For typical
frames, the scheme needs only one element to model a member and the division of
a member into many elements is no longer needed.

S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231


The analysis techniques by themselves are not sufficient for practical design since
engineers have to follow the design codes in many countries. Rules with regard to
the control of various parameters such as member initial imperfection, frame outof-plumbness, semi-rigid connections, residual stress, support movement etc. must
therefore be followed. AS4100 [52] is the earliest design code giving such a recommendation. Discussions and reviews on the design and application issues of
second-order and advanced analysis are given by a number of researchers including
the work by Clarke et al [73], White [74] and Chan and Zhou [75], Deierlein [76],
Nethercot [77]. Springfield [78,79] gave an account of the use of the second-order
analysis from practitioners point of view which is valuable for researchers in their
course of development of practical design tools. Recently, Chan [80] implemented
his analysis method for practical structures and proposed a more practical and acceptable design approach, the NIDA procedure which is defined as a design method
without using the effective length concept for strength determination with the load
capacity of the frame limited to the formation of the first plastic hinge. This simple
and conservative approach has been gaining popularity in the authors country.
Indeed, NIDA is similar to the ultimate load advanced analysis where only section
capacity check is sufficient, but confined to the formation of first plastic hinge or
first yield, for the sake of meeting the current practice and the general code requirements. Any other method of direct and indirect determination of effective length is
not therefore under this catalogue. Consequently the second-order elastic or bifurcation method of automatically finding the effective length for the design of the
column strength is not under this classification. He further applied the technique to
the design of a 1300 ton steel frame [80] and other space trusses with the advantages
of the method fully manifested. As the effective length need not be assumed, economical and safe design can be produced since, in general, the effective lengths of
some members are over-estimated whilst others are under-designed. The design time
is greatly reduced since the engineer does not need to assess the effective length of
each member individually and separately. However, his analysis is still based on the
first plastic hinge approach and principally it is an elastic analysis. It is recognized
that the elastic approach is not an economical design due to the ignorance of reserved
strength after first yield or first plastic hinge. Further, the effective length after yielding may not be the same as the elastic value. Although its extension to elasto-plastic
and large deflection collapse analysis has been documented, the method is limited
by the lack of proper design standards and the uncertainty of force re-distribution
for a complicated frame which needs further research, experimental verification and
discussion. Computationally it is no longer a difficult task and stable and reliable
solution methods for dealing with large and complex structures are basically made
available by various researchers.

10. Conclusions
Extensive work on the theory, numerical methods and code development for the
second-order analysis, advanced analysis and the non-linear integrated design and


S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231

analysis have been conducted in the past few decades. Whilst a collapse analysis
with the formation of a series of plastic hinges can be conducted and has been
documented, the use of second-order analysis and design in practice is mainly limited
to the formation of the first plastic hinge. This is possibly due to the lack of explicit
recommendations from national standards, design guides on uses of various parameters for different structural forms and education of engineers to an advanced level
in non-linear structural theory. However, the use of the NIDA avoiding the assumption of effective length for practical structural design has already lead to a substantial
improvement in efficiency, economy, safety and reliability. With the advances in
these techniques and the availability of low-cost and high-performance personal computers, utilization of these sophisticated techniques is possible in a design office to
date. However, further work is needed for extending the design method to practical
structures to allow for various sources of non-linearities, both in the aspects of design
code development and drafting of design guide for practical implementation.

11. Introduction to references

Ref. [9] is a pioneer text giving classical and numerical analysis methods for beamcolumns and frames. The merits and disadvantages of various numerical methods for
traversing the limit point are summarized in Ref. [44], a valuable paper.
In Ref. [50] an innovative and efficient element formulation is formulated by an
unconventional method using the flexibility approach. Ref. [55] is the first paper
giving the formulation of an efficient element permitting the modeling a member
under high axial force by a single element, with allowance of member imperfection
which is unavoidable in real members and mandatory for inclusion in design codes
and standards.
The simple plastic hinge method by King, White and Chen in Ref. [64] provided
the backbone for a simple and powerful analysis tool for uses in an advanced analysis
and design. See also Ref. [66], an outstanding paper, which gives an account of
various calibrating frames for advanced analysis allowing for large displacement and
plastic hinges. Ref. [74] gives a practical plastic hinge design method for steel
frames. The refined studies by Clarke in Ref. [62] using the plastic zone method
were useful in accurately studying the response of steel frames with partial yielding.
Albermani and Kitipornchais work in Ref. [48] on elasto-plastic and large deflection
analysis demonstrated the use of a sophisticated technique for refined analysis of
steel frames. Ref. [77] gives a detailed account on the state-of-the-art in research of
non-linear frame analysis and design. Ref. [73] is a good paper on the advanced
analysis by the strong research group in Sydney University, which makes major
contributions to the first codified clauses for advanced analyses appearing in the
Australian Standards 4100.

S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231


The author acknowledges the financial support of the Hong Kong SAR Government through the research project Static and Dynamic Advanced Analysis for
Design of Steel Structures (B-Q193).
[1] Bleich F. Buckling strength of metal structures. New York: McGraw-Hill, 1952.
[2] Goodier JN. Torsional and flexural buckling of bars of thin-walled open section under compressive
and bending loads. J Appl Mech, ACSE 1942;64.
[3] Vlasov VZ. Thin walled elastic beams, 2nd ed. Washington DC: National Science Foundation, 1961.
[4] Timoshenko SP, Gere JM. Theory of elastic stability, 2nd ed. New York: McGraw-Hill, 1961.
[5] Ellis JS, Jury EJ, Kirk DW. Ultimate capacity of steel columns loaded biaxially EIC-64-BR and
STR2. Trans, Eng Inst Canada 1964;7(2):311.
[6] Vinnakota S, Aoshima Y. Inelastic behaviour of rotationally restrained columns under biaxial bending. Struct Eng 1974;52(7):23555.
[7] Brown PT, Trahair NS. Finite integral solution of differential equations. Civil Eng Trans, Inst Eng,
Sydney, Australia 1968;10(2):1936.
[8] Roberts TM. Second order strains and instability of thin walled bars of open cross -section. Int J
Mech Sci 1981;23:297306.
[9] Chen WF, Atsuta T. Theory of beam-columns. New York: McGraw-Hill, 1977 (vols. 1 and 2).
[10] Lindner J. Stability of structural members. J Construct Steel Res 2000;55:2944.
[11] Peng JL, Pan ADE, Chan SL. Simplified models for analysis and design of modular falsework. J
Construct Steel Res 1998;48(23):189210.
[12] Chan SL, Zhou ZH, Chen WF, Peng JL. Stability analysis of brittle steel scaffolding. In: Chan SL,
editor. Behaviour of flexible joints and analysis of steel frames with semi-rigid connections [special
issue]. J Eng Struct 1995;17(8):56874.
[13] Kirby PA. Continuous multi-storey frames. Introduction to steelwork design to BS5950:part 1, chap.
15. The Steel Construction Institute, 1988.
[14] Heyman J. Plastic design of portal frames. Cambridge, MA: Cambridge University Press, 1957.
[15] Neal BG. The plastic method of structural analysis. London: Chapman and Hall, 1977.
[16] Merchant W. The failure load of rigid jointed frameworks as influenced by stability. Struct Eng
[17] Mallett RH, Marcal PV. Finite element analysis of nonlinear structures. J Struct Div, ASCE
[18] Kitipornchai S, AlBermani FGA. Nonlinear analysis of lattice structures. J Construct Steel Construct
[19] Oran C. Tangent stiffness in space frames. J Struct Eng, ASCE 1973;99(6):9811001.
[20] Barsoum RS, Gallagher RH. Finite element analysis of torsional-flexural stability problems. Int J
Numer Methods Eng 1970;2:33552.
[21] Majid KI. Non-linear structures. London: Butterworths, 1972.
[22] Shi G, Atluri SN. Static and dynamic analysis of space frames with nonlinear flexible connections.
Int J Numer Methods Eng 1989;28:263550.
[23] Chan SL, Zhou ZH. A pointwise equilibrating polynomial (PEP) element for nonlinear analysis of
frames. J Struct Eng, ASCE 1994;120(6):170317.
[24] Williams FW. An approach to the nonlinear behaviout of the memebrs of a rigid jointed plane
framework with finite deflections. Q J Mech Appl Math 1964;17:45169.
[25] Jenning A. Frame analysis including change of geometry. J Struct Div, ASCE 1968;94(3):627.
[26] Powell GH. Theory of nonlinear elastic structures. J Struct Div, ASCE 1969;95(12):2687701.
[27] Connor J, Logcher RD, Chan SC. Nonlinear analysis of elastic framed structures. J Struct Div,
ASCE 1968;94(6):152547.


S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231

[28] Rajasekaran S, Murray DW. Incremental finite element matrices. J Struct Div, ASCE
[29] Hancock G. A matrix non-linear analysis of elastic thin-walled beams. Civil Eng Trans, Inst Eng,
Australia 1974;16(2):16874.
[30] Argyris JH, Doltsinis JS. On the large strain inelastic analysis in natural formulation, part 1, quasistatic problems. Comput Methods Appl Mech Eng 1979;20:21351.
[31] Yang YB, McGuire W. Joint rotation and geometric nonlinear analysis. J Struct Eng
[32] Chan SL, Kitipornchai S. Geometric nonlinear analysis of asymmetric thin-walled beam-columns.
J Eng Struct 1987;9:24354.
[33] Prokic A. New warping function for thin-walled beams II finite element method and applications.
J Struct Eng, ASCE 1996;122(2):144352.
[34] Chan SL. Large deflection kinematic formulations for three dimensional framed structures. Comput
Methods Appl Mech Eng 1992;95:1736.
[35] Livesley RK. Matrix method of structural analysis. New York: Pergamon, 1964199.
[36] Batoz JL, Dhatt G. Incremental displacement algorithms for nonlinear problems. Int J Numer
Methods Eng 1979;14:12627.
[37] Wright EW, Gaylord EH. Analysis of unbraced multi-storey steel rigid frames. J Struct Div, ASCE
ST5 1968;34:114363.
[38] Riks E. An incremental approach to the solution of snapping and buckling problems. Int J Solids
Struct 1979;15:52951.
[39] Ramm E. Strategies for tracing the nonlinear response near limit point. In: Wunderlicj W, Stei E,
Bathe KJ, editors. Nonlinear finite element analysis in structural mechanics. Berlin: Springer; 1981.
p. 6389.
[40] Crisfield MA. A faster incremental/iterative solution procedure that handles snap-through. Comput
Struct 1981;13:5562.
[41] Meek JL, Tan HS. Geometrically nonlinear analysis of space frames by an incremental-iterative
technique. Comput Methods Appl Mech Eng 1984;47:26182.
[42] Chan SL, Kitipornchai S. Geometric nonlinear analysis of asymmetric thin-walled beam-columns.
Journal of Engineering Structures 1987;9:24354.
[43] Chan SL. Geometric and material nonlinear analysis of beam-columns and frames using the minimum residual displacement method. Int J Numer Methods Eng 1988;26:265769.
[44] Clarke MJ, Hancock GJ. A study of incremental-iterative strategies for nonlinear analysis. Int J
Numer Methods Eng 1990;29:136591.
[45] Effective length and notional load approaches for assessing frame stability : application for American
steel design. Task Committee on Effective Length, ASCE, 1997.
[46] Zienkiewicz OC. The finite element method, 3rd ed. New York: McGraw-Hill, 1977.
[47] Teh LH, Clarke MJ. Corotational and Lagrangian formulations for elastic three-dimensional beam
finite element. J Construct Steel Res 1998;48:12344.
[48] Albermani GA, Kitipornchai S. Nonlinear analysis of thin-walled structures using least
element/member. J Struct Engng, ASCE, 116(1), 21534.
[49] Izzuddin BA, Elnashai AS. Adaptive space frame analysis: parts 1 and 2. Proc Civil Eng Struct
Build 1993;99:30326.
[50] Neuenhofer A, Filippou FC. Geometrically nonlinear flexibility-based frame finite element. J Struct
Eng, ASCE 1998;124(6):70411.
[51] British Standards Institution. BS5950, part 1, structural use of steel-work in buildings. BSI, London, 1990.
[52] AS4100-1990. Steel structures. Standards Association of Australia, Sydney, Australia.
[53] Wen RK, Lange J. Curved element for arc buckling analysis. J Struct Eng, ASCE 1981;107(11):
[54] Elisa ZM, Chen KL. Nonlinear shallow curved-beam finite element. J Eng Mech, ASCE
[55] Chan SL, Zhou ZH. Second order analysis of frame using a single imperfect element per member.
J Struct Eng, ASCE 1995;121(6):93945.

S.L. Chan / Journal of Constructional Steel Research 57 (2001) 12171231


[56] Chan SL, Gu JX. Exact tangent stiffness for imperfect beam-column members. J Struct Eng, ASCE
[57] Yura JA. Effective length of columns in unbraced frames. Eng J, AISC 1971;8(2):3742.
[58] Duan L, Chen WF. Effective length factor for columns in braced frames. J Struct Eng, ASCE
[59] Lui E, Leon RT, White DW. Stability analysis of partially restrained frames. In: Effective length
and notional load approaches for assessing frame stability : application for American steel design,
chap. 3. Task Committee on Effective Length, ASCE, 1997.
[60] Vogel U. Calibration frames. Stahlbau 10;1985:17.
[61] Chan SL. Inelastic post-buckling analysis of tubular beam-columns and frames. J Eng Struct
[62] Clarke MJ. Plastic zone analysis of frames. In: Chen WF, Toma S, editors. Advanced analysis of
steel frames. Boca Raton, FL: CRC Press; 1994.
[63] Powell GH, Chen PFS. 3D beam-column element with generalized plastic hinges. J Engng Mech,
ASCE, 1986;112(7):62741.
[64] King WS, White DW, Chen WF. Second-order inelastic analysis methods for steel-frame design. J
Engng Mech, ASCE, 1992;118(2):40828.
[65] Chen WF, Chan SL. Second order inelastic analysis of steel frames by personal computers. J Struct
Eng 1994;21(2):99106.
[66] Toma S, Chen WF. European calibration frames for second-order inelastic analysis. Eng Struct
[67] Yau CY, Chan SL. Inelastic and stability analysis of flexibly connected steel frames by the springsin-series model. Journal of Structural Engineering ASCE 1994;October:280319.
[68] AISC. Load and resistance factor design specification for structural steel buildings. Chicago:
AISC, 1993.
[69] Chan SL, Chui PPT. Non-linear static and cyclic analysis of semi-rigid steel frames. Amsterdam:
Elsevier Science, 2000 (pp. 336).
[70] Chan BHM, Chan SL, Law SS. Nonlinear behaviour of pitched roof steel frames subjected to sequential loads. J Appl Mech Eng 2000;5(4):77394.
[71] Corradi L, Poggi C. A refined finite element model for the analysis of elastic-plastic frames. Int J
Numer Methods Eng 1984;20:215574.
[72] Feng LX, Wong YL, Chan SL. Strength analysis of semi-rigid steel concrete composite frames. J
Construct Steel Res 1999;52(3):26992.
[73] Clarke MJ, Bridge RQ, Hancock GJ, Trahair NS. Advanced analysis of steel building frames. J
Construct Steel Res 1992;23:129.
[74] White DW. Plastic-hinge methods for advanced analysis of steel frames. J Construct Steel Res
[75] Chan SL, Zhou ZH. Non-linear integrated design and analysis of skeletal structures by 1 element
per member. Eng Struct 2000;22:24657.
[76] Deierlein GG. Steel-framed structures. Prog Struct Eng Mater 1997;5(1):107.
[77] Nethercot DA. Frame structures: global performance static and stability behaviour, general report.
J Construct Steel Res 2000;55:10924.
[78] Springfield J. Limits on second order elastic analysis. Proceedings of the Annual Technical Session
on Structural Stability Research Council, 1991. p. 8999.
[79] Springfield J. Frame stability a Canadian engineers approach. In: Narayanana R, editor. Steel
Structures, Advances, Design and Construction, Proceedings of the International Conference on Steel
and Aluminum Structures. Amsterdam: Elsevier; 1987. p. 28998 (Cardiff, UK).
[80] Chan SL. Stability design of steel structures theory and practice. In: Proceedings of International
Conference on Steel and Composite Structures. 2001. p. 4355 (Korea, June).