M.Sc.

Lecture Note

“Stability of Structures" with application to practical design
by S.L. Chan
September 2004, Department of Civil & Structural Engineering, HKPolyU

1

Chapter1, MSc Lecture Note, 2004-2005 Professor SL Chan, HKPolyU

CHAPTER 1 Behaviour, analysis and design of steel frames
1.1 Introduction The primary objective of a structure is to carry and transmit loads to foundations safely and to perform well in a serviceable condition. The old fashioned allowable stress design method controls the maximum stress of a structure below the allowable stress of the material, generally taken as the characteristic yield stress divided by a factor of safety. Additional checks will be imposed for checking against instability and buckling. In the more recently introduced and widely accepted limit state design concept, a structure is considered to reach a limit state when it is unable to fulfil the requirements for its functional uses. The inability of the structure or its components to meet these performance requirements can be considered as a structural failure at ultimate limit state and as non-structural failure at serviceability limit state, which comprises of a number of criteria related to safety and serviceability. During the service life of a structure, the structure should have an acceptable low probability of any kind of collapse or damage locally or globally and should not deflect to such an extent that the occupants feel uncomfortable or that the finishes or cladding are damaged, in addition, the fire resistance of the structure must allow sufficient time for the occupants to escape in case of fire and to prevent rapid spread of fire. The corrosion resistance should enable the structure to last for the designed life and the vibration should not be excessive etc. These requirements are covered in several limit states detailed in the design codes and all these limit states are needed to be fulfilled for a generally satisfactory structure. This relationship can be stated in a general term as,

φR ≥λF
in which φ is the material factor, R is the resistance of the structure, λ is the load factor and F is the external load.

Collapse load and the ultimate load-carrying capacity are related to the safety consideration for the design of a structure. Due to the importance of ultimate strength 2
Chapter1, MSc Lecture Note, 2004-2005 Professor SL Chan, HKPolyU

limit state, the safety margin or factor of safety is generally larger than other limit states and greater than 1. This book is mainly concerned with the limit analysis and design of a structure against collapse and under the action of static and cyclic loads. Obviously, the theory in this book can be applied equally well to the serviceability deflection limit state at a lower load level with a less severe degree of non-linearity. A structure or its components are, in general, three dimensional. However, most structures and their components can be idealized as one or two dimensional for practical purposes. A plate or shell has lateral dimensions much larger than its thickness. The beamcolumn is one dimensional structural member with a longitudinal length much larger than its transverse dimensions. This includes the widely used structural elements such as I, channel, angle, hollow sections and others. These types of structural members are commonly used in steel and metal structures due to their structural efficiency, relatively light weight, ductility and ease of fabrication. They can be manufactured as hot-rolled, cold-formed or intruded for some types of alloy like aluminium. The cross sections of compressive members are generally thin-walled and seldom solid because of their larger flexural stiffness for the same weight of material. In general, the member is under the action of axial force, bending and torsional moments and shears. Most members are arranged to be under a dominant action from axial force and bending moments and essentially free from torsional moment, although this condition may not be attainable in all structures. In engineering practice, a developer requesting for a project to be undertaken for a specific purpose will contact the architect. Normally, several options are then worked out. Structural feasibility will also be considered and the structural engineer will be consulted at this preliminary stage for some types of structures of considerable size. The structural engineer will estimate loading, and propose suitable framing systems. After consideration under the constraints of finance, environment, construction period, aesthetics and other relevant factors, an option will be chosen and a more detailed design will be worked out. The complete process may be iterated many times until all parties are satisfied with the final arrangement upon which the contract documents and specifications will be prepared. In the course of this option selection process, safety consideration plays an important role and may often affect the final solution. For example, in regions of seismic risk, the option of using steel is preferred and very often the management team requires a careful study on the safety of the 3
Chapter1, MSc Lecture Note, 2004-2005 Professor SL Chan, HKPolyU

structure under severe loads. Obviously the consideration of safety indirectly affects the architectural details. In the past, structural engineers normally adopted linear analysis from which the result is employed for individual member design based on formulae in the design codes. In a linear analysis, the structural member and thus its composing structure is assumed to have a constant stiffness independent of the deformations and magnitude of member forces. Mathematically, this condition can be stipulated as, Ku=F in which K is the element stiffness matrix, u is the displacement vector and F is the force vector of a structural system. In a linear system, the stiffness K is independent of the force and displacement. In the solution for the stiffness above Equation, the determination of external forces or loads and of the stiffness properties of the structural system is essential in getting the true behaviour of the structure. In the ultimate limit state design philosophy, they are obtained from a statistical basis. 1.2 Probability of Failure

From a statistical point of view, it will be very uneconomical or even impossible to design a structure without any chance of failure. A structure is allowed to have a certain probability of failure, which is very low and in the order of 10-6. To achieve this objective, the characteristic loads and strength are used in the design. The characteristic load is defined as the load with a certain low percentage being exceeded in the design life of a structure. Similarly, the characteristic stress is defined as the stress value with a certain low probability for the material to have yielding falling below the yield stress. Failure takes place when both these unfavourable conditions for the characteristic load and strength occur simultaneously.

For non-structural failure such as deflection limit state, the requirement is lower because the consequence of reaching the limit state is less serious. Because of this, its design requires an equally demanding level of accuracy since the chance of its occurrence is 4
Chapter1, MSc Lecture Note, 2004-2005 Professor SL Chan, HKPolyU

much higher than the limit state. 1.3 Loads Applied forces acting on a structure can be external or internal. In general, loads are due to the dead weight of the structure and machines, live loads of the occupants and furnishes, wind and earthquake loads from the relative movement of air and earth and internal forces due to shrinkage and temperature change, foundation settlement etc. In design, these loads should be combined in order to produce the most severe loading combination that may occur during the life span of a structure, with an acceptable low probability of actual loads exceeding these combined loads. Load factors are multiplied to the estimated loads and the magnitude for the factor varies from one country to another. 1.3.1 Dead Loads Dead loads are referred to as loads which do not vary with time during the life-span of the structure. These include the self weight of the structure, finishes and some types of permanent equipment such as lift motor. Density of material and estimation of object weight will be used for the computation of dead weight. 1.3.2 Live Loads Live loads are generally due to the occupants or objects that vary in position with time. In some types of live loads, the dynamic effects should be considered and a more detailed dynamic response of the structure should be conducted. In recent years, the wide use of delicate electronic equipment such as computers in offices has made serviceability vibration an important consideration in the design of floor beams which support this equipment. Suddenly applied loads on the floor may introduce vibration that may damage the equipment or information stored in them. It may cause a considerable loss of property and time.

1.3.3

Wind and Earthquake Loads Wind loads are highly localized and national wind codes can provide a

guideline in the determination of equivalent static wind pressure. Consideration will be given to the locations, height and surrounding environment of a building. The return period as well as the 5
Chapter1, MSc Lecture Note, 2004-2005 Professor SL Chan, HKPolyU

The checking of seismic resistance requires a rather different procedure from the conventional design method such as collapse load factor as this cannot be defined in the static case since the rate of this load will also affect structural safety. 1. 2004-2005 Professor SL Chan. the stability and strength of a structure against time needs to be investigated and the response studied. temperature change. the transient response of a structure is required to be traced and the strength.3. In some countries.period taken for averaging the wind speed should also be determined for wind load intensity. MSc Lecture Note. a fifty year return period with three second wind speed average is adopted. the maximum and minimum temperature may be specified in the codes for calculation of thermal strain. The magnitude of these forces is difficult to be standardised and therefore. stability and also the ductility requirements are needed to be met. However. For a general analysis. creep and shrinkage of material and residual stress. dynamic response of a building may also be required to be simulated in a wind tunnel test. a wind tunnel test may be required to determine the wind pressure at different locations of a building. In some cases for the design of tall or large buildings. not commonly covered in detail in codes of practice. In addition. These effects can also be included in an analysis by appropriate modification of initial stress and displacement etc. The consideration of these effects requires the judgement of the engineer. Earthquake loads are due to ground acceleration and the inertia effect of the mass of a building. HKPolyU . Instead.4 Internal Forces Internal forces may be due to foundation settlement. 6 Chapter1. lack of fit. Different countries use a different period for averaging the wind speed and the calculated wind pressure is also dependent on this duration.

2 1. 2004-2005 Professor SL Chan.0 Live Load AISC 0.3 BS5950 0 1.4 1. Further.9 Table 1. In Table 1.0 1.1.3.4 1.4 1. The true collapse load after the formation of a series of plastic hinges is not considered. For conventional design.0 BS5950 1.4 1 2 3 4 1.2 1. the factor of safety can be defined as the ratio between the collapse and the design load factors. their combinations are required to be worked out. the worst combination should be determined from the dead.6 1.1 Load Factors from AISC(1986) and BS5950 (1990) for different load types 1.2 0 0 Wind Load AISC 0. once yielding occurs. the ultimate limit state design concept takes a conservative approximation on assuming the equality of load at the first plastic hinge to the ultimate design load. HKPolyU .5 0.5 Load Factor and Load combinations Once the magnitude and directions of loads are determined.0 0. It can either be a local. In current design practice. The load factor is then multiplied to the design load to give the permissible load of the structure until collapse.2 -0.6 Collapse Load Factor In the ultimate limit state design concept. the live and the wind load. formation of plastic hinges and the presence of axial force in members are ignored in the analysis. the stiffness of the structure changes and thus the original elastic assumption is no longer valid.0 1.1.4 -1. even the variation in structural stiffness due to yielding.3. 7 Chapter1. However. a progressive or a global collapse.2 1.6 0. MSc Lecture Note. Load and resistance factors are multiplied to the material resistance and the load based on which checking is carried out. the loads factors used by the American Institute of Steel Construction (1986) and British Standard 5950 (1990) are listed for a comparison.3 1. Load Combination Dead Load AISC BS5950 1.

Identification of key structural elements. a completely different result can be observed. the member is lengthened and thus a tensile force is developed. Figure 1. of which the failure causes the collapse of a large area. there is no component in the longitudinal direction and this effect due to geometrical change cannot be observed. resulting in a pulling action of the member. 2004-2005 Professor SL Chan. HKPolyU . In a linear analysis. This moves the tip of the cantilever inwards. the deflection is small and the structure is stable.In recent years. the concept of prevention of progressive and disproportionate collapse has been proposed and enforced in practice in some countries. The member stiffness deteriorates when the critical load is approached. This geometrically nonlinear effect will lead to the stiffening of the stiffness matrix for the complete structure when the tensile axial force is present. will be necessary and additional fictitious load on the element will be required to ensure the key member is capable of resisting additional and accidental loads. As the validity of a linear analysis relies on the assumption that the member force is small and thus the material is elastic plus the assumption that. In situations where a structural member is under compression. 8 Chapter1.1 has been analysed and it can be seen that as the vertical deflection of the cantilever increases. the theory is incorrect when the structure is heavily loaded so that it deforms considerably and the composing material yields. Although the structure stiffens after this load. the large deflection in the post-buckling range is unattainable in most practical civil engineering structures and yielding occurs well before such a development of a bending strength in this range. MSc Lecture Note.2 illustrates the softening behaviour of another cantilever subjected to axial force at the end. A stiffening and a softening system are selected for illustration of the error connected with linear analysis. The cantilever shown in Figure 1.

8 1 ∆ L Figure 1. HKPolyU .10 9 8 P v PL EI 2 L-u L 7 6 5 4 3 2 1 0 0 0.1 Elastic Large Deflection Analysis of a Stiffening System 2 1.6 0.6 0.8 1 Deflections Figure 1.4 v L L=4000 mm 2 A=4200 mm 4 I=6482 cm E=200 kN/mm u 2 0.001P P ∆ 0 0 0. MSc Lecture Note.1 Figure 1.5 P Pcr A = 0.2 m 2 2 1 L 0.01 m E = 210x10 N/m 4 I = 8.2 Elastic Buckling of a Softening System Figure 9 Chapter1.5 0.21.2 0.2 0.4 0.33x10 m L = 3. 2004-2005 Professor SL Chan.

HKPolyU .Figure 1.4 Buckling of some members in a full-scale test of a transmission tower 10 Chapter1. MSc Lecture Note.3 A tension stiffening system Figure 1. 2004-2005 Professor SL Chan.

5.5 1.strain curve for a steel bar is shown in Figure 1. Stress σ Proportional Limit Elastic Limit Upper Yield Lower Yield Ultimate Stress Necking Fracture and Failure Strain ε Elastic range Plastic range Strain Hardening Necking and Fracture Figure 1. the material fully recovers upon removal of external stress. Further loading will cause the material to reach the ultimate stress and fracture.5%. strain Figure Idealised stress vs. the material exhibits a strain hardening behaviour.1. HKPolyU . For mild steel further loaded. MSc Lecture Note. 2004-2005 Professor SL Chan.000 kN/cm 2 0 0 40 80 120 160 200 240 (cm) Figure 1. The Young's modulus of elasticity is about 205.4 Material Properties Steel has relatively uniform material properties when compared to concrete. Before loaded beyond the elastic limit.9cm I=9950 cm 2 3 4 S = 723 cm 4 43 52 Locations for Plastic Hinge formation 1 150 1000 cm 5 P (kN) 100 1 750 cm 750 cm σ ys 50 = 25 kN/cm 2 E = 20.5 Idealised stress vs. strain relationship for mild steel relationship for mild steel bar bar 200 2P 2 3 I=6490 cm A=41.6 Rigid Plastic Analysis of a Simple Portal 11 Chapter1.8 cm S = 480 cm 4 2 3 P A=58. The idealised stress .6 Figure 1.000 MPa with a maximum variation of about 2.

Grade 43A steel should be used for relatively thin material. 2004-2005 Professor SL Chan. HKPolyU .and y. concrete. this is not often the case and to prevent brittle fracture. For example. defined as the the stress value at which 95% of the tested samples do not yield.planes. Grade 43B should be used for thicker material and Grade 43C. whilst the use of high grade steel has little or no advantage for structures with behaviour dominated by deflection limit state. The uniaxial yield stress can further be used for prediction of material yielding under bi-directional direct and shear stresses. D or E should be used for even thicker material under cold envirnoments. The characteristic yield stresses for common steel grades are 250. is often used for design. σ 2 + σ 2 . In this Von Mises yield function.and y.The strength of steel depends on the composition of the material. 350 and 450 MPa. MSc Lecture Note. The 95% characteristic yield stress.directions. However. Different steel grades have varied values for yield stress. using the BS5950(1990) system. the yielding shear stress is σ ys 3 . τxy is the shear stress on the plane and σys is the yield stress under uniaxial stress. in which σx and σy are the direct stress in x. choosing a higher steel grade will be advantageous for structures with behaviour critically controlled by ultimate limit state. especially when the structure is loaded under low temperature. Obviously.σ x σ y + 3τ 2 = σ 2 x y xy ys the following equation is employed. 12 Chapter1. It is of interest to note that by substitiuting zero stresses in the x. In addition to higher stress versus weight ratio for steel when compared with its major rival. instability or vibration. The yield criterion generally used for steel material is the Von Mises yield criterion dervied from the maximum distortional energy theory. but the same Young’s modulus of elasticity. which is valid for metal material. different steel grades of the same characteristic yield stress should be used for various wall thicknesses. steel has an advantage in its high ductility and the ductile mode of failure.

stability is an important consideration.5 Plastic Failure In the second case for plastic analysis. of which the collapse load is required to be determined. HKPolyU . the widely used ultimate limit state design philosophy places the factor of safety on the ultimate load capacity of the structure.6 under a set of vertical loads and a small lateral disturbing force will reach its maximum load-carrying capacity at a vertical load of 197 kN at which a plastic mechanism is formed. Furthermore.6 Instability of a structure The design of a member under tensile axial load is relatively simple because the stiffness of a member is stiffened by the presence of axial force in the elastic range. as shown in Figure 1. slender steel structures are currently widely designed and constructed. it may become unstable and failure may be due to buckling. It can be seen that judgement is necessary for the assessment of the importance of geometric and material nonlinearities so that an approximate analysis method can be chosen for design and analysis. a portal shown in Figure 1.1. the so called elasto-plastic buckling. 1. the member will exhibit an elasto-plastic buckling and neither the elastic buckling load nor the plastic squash load is close to the true failure load. When a member is under the action of moment and/or compressive force.5. MSc Lecture Note. In the design and analysis of steel structures. which was an important consideration before 1980's. An engineer should be fully aware of the possible buckling modes in order to account for this or to take appropriate action to prevent the occurrence of a particular type of buckling such as to provide bracing members to reduce the effective length. The failure of the member is then controlled by material yielding. The phenomena and nature for various buckling modes are discussed in this section. 13 Chapter1. material yield or their combinations. Owing to the improvement in material yield stress for steel in the past few decades and the availability of lowpriced personal computers. 2004-2005 Professor SL Chan. has become a relatively unimportant factor for design to date and this trend will prevail in future with the advent in computer technology. In another example when a strut is subjected to a compressive axial load with small disturbance. The cost of computers.

During the service life of a structure. deflection during buckling is nonlinearly proportional to load). There are basically several types of buckling modes which must be carefully considered before a safe design can be produced. Imagine deflection is related to E and buckling is in fact a phenomenon when the structure deflects rapidly under load (Of course. the structure should have an acceptably low probability of any kind of collapse or damage locally or globally and should not deflect to such an extent that 14 Chapter1. 2004-2005 Professor SL Chan. In the limit state design code. unlike linear and first –order deflection due to bending etc. which comprises of a number of criteria related to safety and serviceability. This may be considered as a deteriorating stiffness phenomenon for the structure. The old fashioned allowable stress design method controls the maximum stress of a structure below the allowable stress of the material. HKPolyU . The inability of the structure or its components to meet these performance requirements can be considered as a structural failure at ultimate limit state and as non-structural failure at serviceability limit state. generally taken as the characteristic yield stress divided by a factor of safety. lateral-torsional buckling. local plate buckling and their combinations. Allowable stress design code like BS449 controls the stress σ which may not be the key parameter for failure due to stability which depends more on Young’s modulus E. a structure is considered to reach a limit state when it is unable to fulfil the requirements for its functional uses. FOS for structural collapse ≠ FOS for stress (FOS= Factor of Safety) Example : slender column fails by buckling which depends more on E.Instability refers to the phenomenon that a structure deforms nonproportionally and its stiffness reduces due to the change of geometry and an increasing load. MSc Lecture Note. torsional buckling. These buckling modes for structural members can be classified as Euler buckling.

MSc Lecture Note. Flexural (Euler) buckling 1. Local plate buckling and 4. Torsional buckling 3. 2004-2005 Professor SL Chan. There are 5 main types of buckling in steel structures as. Lateral-torsional buckling 2. the stress is proportional to load and the failure load is equal to the yield stress x area (or plastic or elastic modulus for beams).the occupants feel uncomfortable or that the finishes or cladding are damaged. Types of buckling For tension and very short columns. Snap-through buckling 5. the fire resistance of the structure must allow sufficient time for the occupants to escape in case of fire and to prevent rapid spread of fire. HKPolyU . For members under compressions. The corrosion resistance should enable the structure to last for the designed life and the vibration should not be excessive etc. These requirements are covered in several limit states detailed in the design codes and all these limit states are needed to be fulfilled for a generally satisfactory structure. This is due to buckling. in addition. the failure load is less than this plastic load or squash load. 15 Chapter1.

Euler Column buckling Load Pe Elastic Critical Load Linear Analysis Second-order Elastic Analysis Pc Idealised Test path P δ P δ Deflection The Typical Behaviour of an Imperfect Strut Snap through buckling in shallow roof 16 Chapter1. MSc Lecture Note. 2004-2005 Professor SL Chan. HKPolyU .

HKPolyU . 2004-2005 Professor SL Chan.Cruciform susceptible to torsional buckling Lateral-tosional buckling 17 Chapter1. MSc Lecture Note.

Therefore. Unfortunately. As a result. HKPolyU . the most common form of buckling is the Euler buckling and it is affected by member interaction as a system such that classification of sway or non-sway frames and consideration of stiffness of adjacent members etc are necessary.2000). MSc Lecture Note. 2004-2005 Professor SL Chan. This lecture is devoted to this piece of work which becomes important in many design codes such as the BS5950(1990. a second-order analysis was developed in the past decades to assist in checking the column buckling because of these two reasons. 18 Chapter1.Local buckling Except the Euler buckling. all of them are considered in isolation in the code that their buckling are not much affected by adjacent member influence and their effects are mainly local. simple formulae can be applied to their cases.

When the eccentricity is positive. The true load versus deflection relationship is dependent on the magnitude of initial imperfection. 1993) can be attributed partly to the buckling of members under axial forces. the equation for Euler's buckling load can be obtained analytically as. the direction as well as the magnitude of disturbing force and initial imperfection should be chosen in such a way so that the most severe effect results. This behavioural phenomenon carries a physical implication that. When an imperfect strut or column is subjected to an axial force. it has a tendency to deflect laterally along its length. In some cases.1 Euler Buckling Euler buckling is possibly the most widely encountered and studied buckling mode. 19 Chapter1. For the simplest case of a column under compressive axial load. a stiffening behaviour is observed and vice versa. Pcr = π 2 EI Le 2 in which Le is the effective length of a column. the bifurcation and the plastic collapse loads are only the upper bound solutions to the actual failure load. for design against the worst possible condition. HKPolyU . the post-buckling path may be asymmetric to the primary equilibrium path and the direction of initial imperfection will affect the behaviour of a structure. The frequent collapse cases for scaffolding (Chen et.14159. For practical purposes. Figure 1. The direction of load eccentricity has a significant influence over the pattern of the equilibrium path. al. both in theory and in practice.6. as indicated in Figure 1. MSc Lecture Note. P is the axial force and π is a constant equal to 3.1. 2004-2005 Professor SL Chan.8 for the case of a right angle frame.7 has already shown the load versus deflection path of an imperfect column and it can be seen that the member cannot sustain a load higher than the member squash load (Psq = σys x Area) or Euler's buckling load.

The importance of unstable type of failure of a structure cannot be overlooked. Their design is therefore straight forward.01 e P θ 0. When it occurs. Koiter (1962) 0 -20 -10 0 10 20 Rotation θ (in degrees) Figure 1. by Salvadori and Heller in "Structures in Architecture" When a structure buckles. beams) or wholly in compression. because it may be easier for it than to shorten itself axially. A non-mathematical explanation of buckling is difficult but it can be briefly postulated as follows. the complexity of the formulae in the ultimate steel design codes such as the BS5950 and the American Load and Resistance Factor Design for steel is mainly due to the complicated checking of a member against inelastic buckling. "A slender column shortens when under a small compressive axial load applied at top. The failure is due to instability or buckling. the slender member will possibly choose the bending down action to maximum the lowering down of its position.5 Load P/P cr 1 e/L = 0.1 Formulation Stocky and tensile members fail when the stress exceeds the material yield stress. its stiffness becomes critically small and a vanishing small 20 Chapter1. MSc Lecture Note. it will select the easiest path to lower down its position. whether it is partly (e.1.g. In fact.7 Post-buckling Behaviour of a Right Angle Frame 1. when a slender member is in compression. However. 2004-2005 Professor SL Chan. Further to this. HKPolyU .6. When a larger load is present. In doing so. it fails before the plastic failure load is reached (Plastic Failure Load = Yield Stress x Cross Sectional Area or Plastic Modulus). The tendency to lower down its position is a natural law.01 e/L = -0. the member is said to have buckled".1.5 Large Deflection Theory Analytical Solution e=0. its weight will lower its position.

is subjected to an axial load.disturbing or perturbing force would cause it to have an uncontrolled displacements.1 By Differential Equation A simply supported column shown in Figure 1.6. EI 2 d v +Pv=0 2 d x in which E is the Young's modulus of elasticity. we obtain the following 21 Chapter1. Figure 1. 2004-2005 Professor SL Chan.6. 1.9 The Basic Force versus Displacement relations in a Member Solving the simple differential equation.2 Critical Load of a Column 1. I is the second moment of area. MSc Lecture Note. buckling can be considered as a zero stiffness phenomenon. a plate component or a member. This structure includes a frame. The equilibrium equation in an imaginary and slightly deformed position can be written as. P EI k= Note : A direct checking of the solution is to substitute the solution into the differential equation! Imposing the boundary conditions of zero deflections at the two ends. Therefore.9. HKPolyU . one obtains the solution for v as. v = A sin k x + B cos k x in which k is given by.1. v is the lateral deflection in the y-axis and x is the distance along the member length.1. we focus on the buckling of a strut.2. In this chapter.

In the second condition. it is not a good approach in general and more complex problems.2 The Principle of Minimum Potential Energy As an alternative method to solve a buckling problem. In the first condition. it can only be applied to problems of simple geometry such as beam-columns. the Euler's load is given by. 2004-2005 Professor SL Chan. The vanishing of the first variation of the total potential energy functional implies the satisfaction of the equilibrium condition. kl is equal to nΠ in which n is an integer. the principle of minimum potential energy can be used. 1. However. Thus. A sin (k L) = 0 we have either a = 0 or sin(kL) = 0. it is extremely difficult or even impossible to write down a set of differential equilibrium equation for a frame and therefore cannot be used in such types of structures.1. it implies that the member will not deflect and represents a higher energy path. The vanishing of the second variation of the energy functional means the structural system is in the state of neutral equilibrium. MSc Lecture Note. n is equal to 1. Its application to more complex problems of inelastic buckling analysis of angle beamcolumns was also conducted. P= Note : π2E I L 2 Although the present approach to find the buckling load is straight forward.2.equation. For the minimum or the first mode buckling load. Thus. A mathematical technique of Variational Calculus is often used to minimise a functional like : 22 Chapter1. HKPolyU . δ Π=0 for equilibrium >0 For Stable Equilibrium For Neutral Equilibrium For Unstable Equilibrium δ2Π=0 <0 An energy function is an expression for the energy.6.

we put. Thus. . MSc Lecture Note. We can only have the expression or a set of equations governing the instability condition. 2004-2005 Professor SL Chan. To derive the Euler buckling load of a column. we cannot obtain the solution directly. we obtain the exact value for the Euler buckling load as in previous case. That is.n ⎛ dy d 2 y d y⎞ I = f ⎜ y.. n ⎟ ⎜ dx ⎟ dx ⎠ dx ⎝ Note that.. the energy functional of the column can be written as (Tension +ve). P=- π2E I L 2 (Compression) Π= E I A2 π 4 L P A2 π 2 L + 2 L4 2 L2 δ2Π= E Iπ 2 L 2 +P Note : The present energy approach will be found to be very general and can be applied in 23 Chapter1.. the assumption of a half sine curve will satisfy the deflected form the column. . HKPolyU . 2 2 1 L ⎛ ⎛ d2 v ⎞ ⎛ dv ⎞ ⎞ ⎜E I ⎜ Π = ∫O ⎟ + P ⎜ ⎟ ⎟ dx 2 ⎜ ⎜ dx 2 ⎟ ⎝ dx ⎠ ⎟ ⎠ ⎝ ⎝ ⎠ In the case of a simply supported column. after minimisation. 2 . ⎛π x ⎞ v = A sin ⎜ ⎟ ⎝ L ⎠ into above equation and after differentiation and integration.

Generality. In real cases. However. vo = Co sin πx L The change in curvature is then given by. Unfortunately. Zmax is the distance of the extreme fibre from the centroidal axis. Their major difference is on the assumption of the initial imperfection.3 The Perry-Robertson formula for design of struts In most of the design codes. it does not directly give us the solution. the maximum mid-span deflection is taken as the member length/1000 whilst in the BS449 code. In fact. To this end. it is taken as follows. HKPolyU . This effect is included in the member strength determination by assuming a midspan initial deflection.1. depending on the number of coefficients (A1 to An). Co is the amplitude of the initial imperfection. MSc Lecture Note. the members are not perfectly straight nor the loads are perfectly concentric. ⎛ L η = ZMAX CO = 0. Co and the imperfection shape as. 1 1 d 2 v d 2 vo M . it gives us a set of equations of order 1 to thousands. the Perry-Robertson formula is used for the design of struts. we will face with a similar problem in the precious section in finding a correct displacement function to satisfy all the boundary conditions of a complex problem. 24 Chapter1.3 ⎜ 2 ⎜ 100 r r ⎝ ⎞ ⎟ ⎟ ⎠ 2 Where L is the length and r is the radius of gyration.6. we have. To allow the method to be used in a general problem without the need of having a correct displacement function to satisfy the boundary conditions. 1. the displacement is assumed to be in the form of a polynomial function and the coefficients (A1 to An) will then be expressed in terms of the degrees of freedom.Pv = = = R R o dx 2 dx 2 EI EI Put n2=P/EI and symbolise d()/dx as a superdot. In the American codes. the finite element method of analysis is commonly utilised.complicated problems. 2004-2005 Professor SL Chan. it will be further elaborated later. it is very tedious for us to write down the energy functional for each particular problem.

πx ⎛π ⎞ 2 && + n 2 v = &&o = . we use the following boundary conditions.P ⎦ To limit the maximum stress to a particular value such as the yield stress (limit state) or its value divided by a certain factor of safety (allowable stress). x=L/2 and Mmax is given by. ⎛π ⎞ Co ⎜ ⎟ ⎝ L ⎠ sin π x v = A sinnx + B cosnx 2 L 2 ⎛π ⎞ n -⎜ ⎟ ⎝L⎠ 2 2 To find A and B in the above equation. For maximum moment. thus B=0.Co ⎜ ⎟ sin v v L ⎝L⎠ Solving. MSc Lecture Note. thus A = 0. v = 0.P ymax ⎡C ⎤ = . we have. At x =0. 25 Chapter1.n2 ⎝L⎠ πx C = O Pcr sin L Pcr . HKPolyU .P in which Pcr is the Euler's buckling load and equal to π2EI/L2. v=0.P ⎢ o Pcr ⎥ ⎣ Pcr . 2 ⎛π ⎞ Co ⎜ ⎟ ⎝ L ⎠ sin π x v= 2 L ⎛π ⎞ ⎜ ⎟ . we have the combined stress as. M max = . At x = L. 2004-2005 Professor SL Chan.

4 1.1.6.P I Exercise Derive the following equation for permissible stress of struts in BS449 (1990) p.4. we focus on individual member behaviour.⎨⎜ ⎟ . in contract to the finite element method where the solution is approximate that the displacement function is assumed. Hence write a simple program to generate and plot the design curves for tables 17 in the code for a particular section and steel grade.1. 2004-2005 Professor SL Chan. HKPolyU . 1. the element stiffness matrix for a second order analysis of frames must account for the effect. 26 Chapter1. From the equations. the stability effect will be allowed for. This type of analysis is accepted by the Australia AS4100-1990 and is currently under active study by the drafters of various national codes such as the American and the European codes.102. we realise a frame composed of an assemblage of members will have a stiffness dependent on the load. we note that the member capacity and thus its resistance against external loads are dependent on the magnitude of the axial load (and also the moment to be described in later section). Element Stiffness Matrices In the above analysis and design. One of the basic but not sufficient requirements is that.1 The stability function A method of using the element stiffness matrix allowing for the presence of axial force is the method of stability function (or the s. To this end. All these necessiate the development of element stiffness matrix allowing for the effects of axial force. Analysis of structures allowing for the effects will be called the second order analysis.c function). where σcr is the critical stress equal to Pcr/Area. Plot also the curves for the Euler's buckling load and the squash load in your diagram.6. 1.7 pc = σ ys + (η + 1) σ cr 2 ⎫ ⎧ ⎛ σ ys + (η + 1) σ cr ⎞2 ⎪ ⎪ . Thus. The element matrix is derived from the differential equilibrium equation directly and thus the stiffness matrix is exact. Advanced analysis is defined as a method of analysis which is capable of representing the limit state behaviour associated with specification member capacity checks to an extent such that the analysis alone is sufficient to verify the adequacy of both member and overall structural system performance and strength.σ max = P M max Zmax + ≤ σ allow A I P C P = + o Pcr Zmax A Pcr .σ ys σ cr ⎬ 2 ⎪ ⎪⎝ ⎠ ⎭ ⎩ Appendix B. MSc Lecture Note. the geometrically nonlinear analysis or the P-delta analysis. Note that the widely used analysis computer program is based on the linear assumption that the effects are ignored. in the advanced analysis.

x ⎞ .φ sin φ c1 = 27 Chapter1.The deflection of a beam under an end moment and an axial load. is given by (see p. EI ⎡ c1 c2 ⎤ ⎡ M a ⎤ ⎢ ⎥=⎢ ⎥ L ⎢c2 c1⎥ ⎢M b ⎥ ⎣ ⎦ ⎣ ⎦ φ (sin φ . ⎛ sinkx x ⎞ M a ⎛ sink(L.φ sin φ where φ (φ .cos φ ) . Timoshenko and Gere).φ cos φ ) 2(1 .x) L.e.12. ⎛ sinkx x ⎞ . 2004-2005 Professor SL Chan.⎟ v = Mb ⎜ P ⎝ sinkL L ⎠ The deflection due to the moment at the other end can be obtained by replacing Mb by Ma and x by (L-x).sin φ ) c2 = 2(1 . MSc Lecture Note.cos φ ) .e. P. x = f F). The final total deflection can be evaluated by superimposition as.⎟+ v = Mb ⎜ ⎜ ⎟ L ⎠ P ⎝ sinkL L ⎠ P ⎝ sinkl Ma L ψ + M b L φ 3 EI 6 EI M b L ψ + Ma L φ θb = 3 EI 6 EI θa = The nodal rotations can therefore be calculated as.kx=F) can be obtained by inversing the f and can finally be obtained as. (Derive by yourself) ⎡ψ φ ⎤ ⎡θ a ⎤ ⎥ ⎡Ma ⎤ ⎢ ⎢ ⎥ L ⎢ 3 6⎥ ⎢ ⎥ ⎢ ⎥= ⎢ ⎥⎢ ⎥ ⎢ ⎥ EI ⎢ φ ψ ⎥ ⎢ ⎥ ⎢θ b ⎥ ⎣M b ⎦ ⎣ ⎦ ⎢ 6 3⎥⎢ ⎥ ⎦ ⎣ The above is the flexiblity matrix (i. Where ψ = and φ = In matrix form as. The stiffness matrix (i. HKPolyU .

5 Element Matrices by the Principle of Total Potential Energy The assume sine function for the problem of a simply supported column is not valid in a general problem. because a general beam-column has a deflected shape other than a half sine curve. for a strut against the axial load. determinate the relative stiffness of a pin-fix strut under an axial load equal to 0. c1 and c2. (b) Derive the stiffness coefficients for a strut without any axial load and with a tensile load.Note that the above stiffness equations are only valid for compressive axial load case.5 of the Euler's buckling load and under no axial load in which case c1=4 and c2=2. a cantilever strut has a deflected shape in the following form. The most commonly used shape function for deflection is cubic for the lateral deflection and v = a 0 + a1 x + a 2 x 2 + a 3 x 3 linear for axial lengthening.1. Note that the expression satisfies the boundary conditions of zero deflection and slope at x = 0 and deflection equal to A at x = L. Thus. u = b0 + b1 x 28 Chapter1. ⎛ ⎡ x ⎤⎞ v = A ⎜ 1 . The coefficients in tensile and zero load cases can also be derived in the same procedure. 1. Hence. Exercise (a) Plot the stiffness coefficients. For example. MSc Lecture Note.cos ⎢π ⎟ ⎜ ⎟ ⎣ 2L⎥ ⎠ ⎦ ⎝ in which the coefficient A is the lateral dip deflection.6. HKPolyU . P. 2004-2005 Professor SL Chan.

we have the following boundary conditions as.2 ρ 1) ρ ρ ρ 2 L (3 . vol is the volume of the continuum.2 ρ 2 ) ρ . the energy expression for a continuum is first expressed as. 29 Chapter1. Expressing the coefficients a1 to a4 in terms of the degrees of freedom y1. [F] and [r] are the applied force and the corresponding displacement vectors. Note that the right hand rule has been used to denote the rotation. ended nodes. we obtain an alternative expression as. Expressing the strain in terms of the linear (εL) and the nonlinear strain tensor (εN). dy = θ z1 dx dy = θ z2 dx WHEN x = L : y = v2 . 2004-2005 Professor SL Chan. θz1 and θz2. MSc Lecture Note. [σ] and [ε] are the stress and strain tensor 1 Π = ∫ vol [σ ]T [ε ] dv + [F ]T [r] 2 1 = ∫ vol [ε ]T [D] [ε ] dv + [F ]T [r] 2 respectively.ρ 1 ρ L⎥ 2 ⎦ 1 1 2 ⎣ where ρ1 = 1 - x x . Thus. we have.For lateral deflection. which are the deflections and rotations at the two WHEN x = 0 : y = v1 . in which Π is the total potential energy functional. y2. ρ2 = L L To obtain the element stiffness matrix for a general thin-walled beam-column. 2 2 2 2 ⎤ ⎡ f 3 = ⎢(3 . we now have 4 coefficients and can be expressed in terms of the degrees of freedom at the two ends of an element. HKPolyU .

and z. [ε ] = [ε ]L + [ε ]N Note that [σ ] = [D] [ε L] 1 Π = ∫ vol [ε L ]T [D] [ε L ] + 2[σ L ]T [ε N ] dv1 + [F ]T [r] 2 ( ) For the present studies. 2004-2005 Professor SL Chan. we ignore those stresses acting perpendicularly to the longitudinal member axis and shears in the directions parallel to the x-axis (i. HKPolyU . 30 Chapter1. To this. we obtain. only the following terms are considered to be non-zero. vi and wi are the displacements of an arbitrary point on the cross section in the x. only the beam-column element is considered. the y. Thus.e.Substituting and ignoring the higher order term. 2 2 2 ⎛ ∂ u i ⎞ 1 ⎡⎛ ∂ u i ⎞ ⎛ ∂ vi ⎞ ⎛ ∂ w i ⎞ ⎤ ⎟ + ⎢⎜ ⎟ +⎜ ⎟ +⎜ ⎟⎥ ε xx = ⎜ ⎝ ∂ x ⎠L 2 ⎢⎝ ∂ x ⎠ ⎝ ∂ x ⎠ ⎝ ∂ x ⎠ ⎥ N ⎣ ⎦ ui wi i i i i i i ε xz = ⎨ ⎢ ⎥ + ⎢ ∂ x ⎥ ⎬ + 2 ⎨ ∂x ∂z + ∂x ∂z + ∂x ∂z ⎬ 2⎪ ⎣∂z⎦ ⎣ ⎦ ⎪L ⎩ ⎭N ⎩ ⎭ 1 ⎧ ⎡ ∂ u i ⎤ ⎡ ∂ vi ⎤ ⎪ + ⎨ 2 ⎪ ⎢ ∂ y ⎥ ⎢∂ x ⎥ ⎦ ⎦ ⎣ ⎩⎣ ⎫ 1 ⎧ ∂u ∂u ∂v ∂v ∂w ∂w ⎫ ⎪ i i i + i i + i ⎬ ⎬ + ⎨ ∂x ∂y ⎭ N ⎪ L 2 ⎩ ∂x ∂y ∂x ∂y ⎭ 1 ⎧ ⎡∂ ⎪ ⎤ ⎡∂ ⎪ ⎤⎫ 1 ⎧ ∂u ∂u ∂v ∂v ∂w ∂w ⎫ ε xy = in which the subscripts L and N represent the linear and nonlinear components and ui. MSc Lecture Note. y and z directions respectively.axes).

If these displacements are not used in the derivation.M y1 ρ 1 + M y 2 ρ 2) + (M z1 ρ 1 . The reason for the selection of centroidal axial lengthening and the shear centre lateral deflections is due to the fact that these displacements are not coupled. ω = ∫s ρ ds 0 where ρ is the distance between the point and the shear centre. and the shear centre displacements.M z 2 ρ 2) + Mω A I yy Izz Iω τxy = τxz = Vy A Vz A 31 Chapter1. HKPolyU . v and w. ui. This coordinate system is called the principal coordinate.θ x z in which ω the sectorial coordinate given by. This means that our assumption that the shape function for the displacement along a principal axis does not depend on the displacements in other principal axes is valid. The corresponding stresses for are given by. the displacement function for the deflections along the any orthogonal axes will be coupled.y wi = w + θ x y vi = v . MSc Lecture Note. 2004-2005 Professor SL Chan.For small but finite deformation. about the shear centre axis as follows. u. ∂v ∂w ∂θ x -z -ω ∂x ∂x ∂x ui = u . vi and wi can be expressed in terms of the axial displacement at centroid. This important point has been overlooked by many other researchers. σ xx = P z y ω + (. and the angle of rotation θx.

M 0 L L ∂ ∂v ∂θ ∂ v dx + ∫ (M z1 ρ 1 . P is the axial force along the centroidal axis. 2004-2005 Professor SL Chan. Shear due to warping is also ignored for simplicity. we obtain the general expression for the energy functional of a general and asymmetric section as. 2 Iω = ∫ A ω d A Substituting. A is the cross sectional area.M z 2 ρ 2 ) x dx ρ1 + M y 2 ρ 2 ) θ x y1 ∂x ∂x 0 ∂x ∂x 32 Chapter1. 2 Iz = ∫ A y dA 2 Iy = ∫A z d A r1 = 2 1 2 2 ∫A (y + z ) d A A and Iω is the warping section constant given by. z and y are the co-ordinates with respect to the centroid. Vz are the shear centre in the y and z directions acting on the place perpendicular to the x-axis. HKPolyU .in which the subscripts 1 and 2 represent the near and the far ends of the element. Iy and Iz are the second moments of area about the principal axes given by. 2 2 2 2 2 L ⎛ ∂2 v ⎞ ⎛ ∂2 w ⎞ ⎛ ∂ 2θ x ⎞ 1 ⎡ ⎛∂u⎞ ⎛ ∂θ x ⎞ ⎤ Π = ∫ ⎢EA ⎜ ⎟ + EIz ⎜ 2 ⎟ + EI y ⎜ ⎜∂ ⎟ ⎜ ∂ 2 ⎟ + EIω ⎜ ∂ 2 ⎟ + GJ ⎜ ∂ x ⎟ ⎥ dx ⎟ ⎜ ⎟ 2 0 ⎢ ⎝∂x⎠ ⎝ ⎠⎥ ⎝ x ⎠ ⎝ x ⎠ ⎝ x ⎠ ⎦ ⎣ +∫ 0 L P ⎡⎛ ∂ v ⎞ ⎛ ∂ w ⎞ 2 ⎛ ∂θ x ⎞ ⎟ ⎟ + r1 ⎜ ⎟ +⎜ ⎢⎜ 2 ⎢⎝ ∂ x ⎠ ⎝ ∂ x ⎠ ⎝ ∂x ⎠ ⎣ 2 2 2 ⎤ ⎥ dx ⎥ ⎦ ∫ (. Vx. MSc Lecture Note. My and Mz are the moments about the principal axes and Mω is the Bimoment.

∫ ⎢τ ⎣ 0 L L ⎡ xy ⎛ ∂w ∂u ∂v⎞ ⎛ ∂ v ∂ u ∂ w ⎞⎤ + ⎜θ x ⎟ . [r]. Also. the section is assumed to be symmetrical that the asymmetrical terms are all ignored. in the above derivation. the orthogonality conditions for the generalised principal coordinates are such that the following terms are zero. Instead.τ xz ⎜θ x ⎟⎥ dx ⎝ ∂x ∂x ∂x⎠ ⎝ ∂ x ∂ x ∂ x ⎠⎦ 2 ⎛ ∂θ x ⎞ T ∫ Mω β ω ⎜ ∂ x ⎟ dx + [F ] [r] ⎝ ⎠ 0 in which J is the St. 2004-2005 Professor SL Chan. the moment term can be ignored etc. we do not use a single and close formed expression for the displacement which may not be possible for complex problems. That is. HKPolyU . For the case of an I-beam. the terms for asymmetry and the axial force can be ignored. For the case of the buckling of an angle column. MSc Lecture Note. v and w. u. ∫ xdA = ∫ ydA = ∫ xydA = ∫ x ω dA = ∫ y ω dA = 0 The formulation described in the above represents the energy functional for a general asymmetric thin-walled beam-column. as follows. Summary for the derivation of an element matrix by the finite element method 33 Chapter1. we obtain the expression for the energy functional in terms of nodal degrees of freedom. Note. In the context of the finite element analysis. Thus substituting the cubic displacement function for the lateral displacement. Venant torsion constant and r1 is the polar radius of gyration about the shear centre given by. linear function for axial lengthening. [r] = [ u1 v1 w1 θ x1 θ y1 θ z1 u 2 v2 w 2 θ x 2 θ y 2 θ z 2 ] T 2 ∂ Π = [k L] + λ [k G ] ∂ xi ∂ x j in which kL and kG are the linear and geometric stiffness matrices for a general thin-walled beam-column and λ is the load intensity parameter (see next page). in the above. we use the cubic polynomial for the lateral deflection to represent the deflected shape of the displacement and linear for axial lengthening. that.

v = v 2 . we write down the energy term corresponding to bending. For example. we use cubic polynomial which has also 4 coefficients so that we can solve the coefficients. Depending on the nodal degree of freedom for an element.v1 + v2) 2θ z1 + θ z 2 2 L L 34 Chapter1. Thus. MSc Lecture Note. for a beam element. dv = θ z1 dx For x = L. write down a polynomial for the deflection of the element. v = a 0 + a1 x + a 2 x 2 + a 3 x 3 For x = 0. Write down the energy functional for the particular type of member. Thus. for simplicity and demonstrative purpose.1. HKPolyU . 2004-2005 Professor SL Chan. If we have 4 degrees of freedom. 2.v2) L 3 + θ z1 + θ z 2 L 2 a2 = 3(. v = v1 . the energy functional corresponding to the Π= ⎛ d2 v ⎞ 1 P ⎛ dy ⎞ EI⎜ 2 ⎟ dx + ∫ ⎜ ⎟ ∫ ⎜ dx ⎟ 2 0 ⎝ dx ⎠ 20 ⎝ ⎠ L L 2 dx + M θ 1 1 + M 2 θ 2 + F1 v1 + F2 v2 action of an axial force is selected and can be expressed as. dv = θz2 dx a 0 = v1 a1 = θ z 1 a3 = 2( v1 .

L ⎢ 10 30 10 15 ⎥ ⎦ ⎣ L⎥ ⎦ Note that the coefficients are given by.x )⎥ ⎢ v = ⎢(1 . x j are the nodal degrees of freedom 35 Chapter1. ⎡ 12 ⎢ 3 ⎢ L ⎢ ⎢ 6 ⎢ 2 ⎢ L [k l + k G ] = EIz ⎢ ⎢ 12 ⎢. MSc Lecture Note. 2 ∂ Π .3 ⎢ L ⎢ ⎢ 6 ⎢ 2 ⎣ L 6 L 2 - 12 L 2 4 L 6 L 2 - 6 L 2 12 L 3 2 L 6 L 2 6⎤ 1 6L 1⎤ ⎡ 6L 2⎥ ⎢ 5 L ⎥ 10 5 10 ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ 2⎥ ⎢ 1 2L . in terms of the nodal degree of freedom into the energy functional. ⎡ v1⎤ ⎢ ⎥ 3 2 3 x 2 2 x3 2 x 2 x3 ⎡ 3 x 2 2 x3 ⎤ ⎢θ z 2⎥ + 2 ) ( 2 . v.1 . HKPolyU .3 ) ( x 2 . After differentiating the functional with respect to the degree of freedom two by two. after solving.L ⎥ ⎢ 10 15 L⎥ 10 30 ⎥ ⎥ ⎥ ⎢ ⎥+P⎢ ⎥ 6⎥ 6L 1⎥ ⎢. we obtain the functional in terms of the nodal degree of freedom.And.6 L . Substituting the displacement function. ∂ xi ∂ x j k ij = k ji = x i .2⎥ ⎢ 5 10 5 10 ⎥ L ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ 1 2L⎥ 4⎥ ⎢ 1 . we obtain the stiffness matrix as follows.1 .2 + 3 ) (x⎥ L L L ⎦ ⎢ v2 ⎥ L L L L L ⎣ ⎢ ⎥ ⎣θ z 2⎦ 3. 2004-2005 Professor SL Chan.

some researchers do not consider the beam-column element as a finite element which implies that the exact expression for the deflection cannot be obtained but approximated by a series of approximate functions such as cubic polynomial. λ.1 Using one single cubic polynomial as a displacement function.e. Exercises 1. | kL + λ kG | = 0 To this. MSc Lecture Note. calculate the buckling load of a simply supported column with a pinned joint at one end and a roller joint at the other end. in the present studies of buckling problems. HKPolyU . 36 Chapter1. is far from cubic and therefore we need to use at least 2 elements to obtain a solution with an accuracy of less than 1%. 2004-2005 Professor SL Chan. a computer program has been developed by the lecturer to calculate the value of the load factor. we can easily see the nonlinear solution. However. say. EI = M1(1-x/L)+M2(x/L)}. That is. Because the cubic Hermite function for lateral deflection represents the exact linear solution for the bending for a beam {i. for the semi-indefinite condition of the eigenvalue.The condition for the structural system to become unstable is the vanishing of the determinant of the matrix. which is the half sine curve for a simply supported strut. What is the percentage error for such a solution? Can we use 1 element to represent 1 member in a nonlinear analysis if we need a solution of less than.2 Determine also the buckling load of a cantilever by using 1 element. 10% ? 1.

6. = vo ⎜ ⎟ cos . Let θ = θo sin πx ∂θ πx d 2θ πx ⎛π⎞ ⎛π⎞ = θo ⎜ ⎟ cos . L dx L dx L ⎝L⎠ ⎝L⎠ 2 πx d 2 v ⎛π ⎞ v = vo sin . 2 = −θo ⎜ ⎟ sin . MSc Lecture Note.1. 2 = −v L dx L dx ⎝L⎠ πx dv π ⎛π ⎞ ⎜ ⎟ sin L ⎝L⎠ 2 So. it may deflect and bend about its minor axis when the applied moment is close to its critical moment.2 Lateral Buckling of Beams When a member is under the predominant action of bending moment about its principal major axis. HKPolyU . 2004-2005 Professor SL Chan. 2 2 L L ⎡ ⎡ ⎛π ⎞ π⎤ πx ⎤ 1 1 ⎛π ⎞ π = ∫ EI z ⎢− vo ⎜ ⎟ sin ⎥ dx + ∫ GJ ⎢θ o ⎜ ⎟ cos ⎥ dx L⎥ 2o L⎦ 2o ⎝L⎠ ⎣ ⎝L⎠ ⎢ ⎣ ⎦ 2 ⎛ ⎛π ⎞ πx ⎞ ⎛ ⎛ π ⎞ πx ⎞ + M ∫ ⎜θ o ⎜ ⎟ cos ⎟ ⎜ vo ⎜ ⎟ cos ⎟dx ⎜ ⎟⎜ L ⎠⎝ ⎝L⎠ L⎟ ⎝L⎠ ⎠ o⎝ L 1 2⎛ π ⎞ = EI z vo ⎜ ⎟ 2 ⎝L⎠ 4L 1 2⎛π ⎞ ∫ sin L dx + 2o GJθ o ⎜ L ⎟ ⎝ ⎠ o 2 2L π 2L ∫ cos o 2 π L dx ⎛π ⎞ + Mθ oVo ⎜ ⎟ ⎝L⎠ ∫ cos o 2 π L dx Since cos 2 x = 1 − 2 sin 2 x 1 − cos 2 x 2 sin x = 2 2 cos 2 x − 1 = cos 2 x πx L L sin 2 dx = ⎫ L 2 ⎪ ∫ ⎪ o ⎬⇒ ⎪ L L 2 πx dx = ⎪ ∫ cos ⎭ L 2 o 37 Chapter1. Ex. Given 2 L ∂θ ∂v 1 L ⎛ d2v ⎞ 1 L ⎛ ∂θ ⎞ dx π = ∫ EI z ⎜ 2 ⎟ dx + ∫ GJ⎜ x ⎟ dx + M ∫ 2 0 ⎜ dx ⎟ 2 0 ⎝ ∂x ⎠ O ∂x ∂x ⎝ ⎠ 2 To determine Mcr under uniform bending.

1987a). The close form general solution for the elastic buckling moment can then be obtained as. MSc Lecture Note. ⎡ ∂ 2π ⎢ 2 ⎢ ∂vo δ 2π = ⎢ ⎢ 2 ⎢ ∂π ⎢ ∂θ o ∂vo ⎣ ∂ 2π ⎤ ⎡ ⎛ π ⎞2 ⎥ ∂vo ∂θ o ⎥ 2 EI ⎜ ⎟ 1 ⎛π ⎞ ⎢ z ⎥ = ⎜ ⎟ ⎢ ⎝L⎠ 2⎝ L⎠ ⎢ ∂ 2π ⎥ ⎢ M ⎥ ⎣ 2 ∂θ o ⎥ ⎦ 2 2 ⎧ ⎫ ⎪ ⎛π⎞ 2⎪ ⎨GJ x EI z ⎜ ⎟ − M ⎬ = 0 ⎪ ⎪ ⎝L⎠ ⎩ ⎭ ⎤ M⎥ ⎥ ⎥ GJ ⎥ ⎦ π ⎛π⎞ ⎛π⎞ M = GJ EI z ⎜ ⎟ ⇒ M cr = GJ EI z ⎜ ⎟ L ⎝L⎠ ⎝L⎠ 2 2 1⎛ π⎞ det = 0 ⇒ ⎜ ⎟ 2⎝ L⎠ The elastic buckling moment can be obtained by the solution of the differential equilibrium equation (Trahair. 2004-2005 Professor SL Chan. HKPolyU . 1965) or the energy equation using the principle of total potential energy (Chan and Kitipornchai. 38 Chapter1.1 1 ⎛π ⎞ L ⎛π ⎞ 1 2⎛π ⎞ 1 ⇒ π = EI z vo ⎜ ⎟ + GJθ o2 ⎜ ⎟ + Mπ o vo ⎜ ⎟ 2 ⎝L⎠ 2 2 ⎝L⎠ 2 ⎝L⎠ 2 4 2 2 4 ⎧ L ∂ 2π ⎛π ⎞ = EI z ⎜ ⎟ ⎪ 2 2 ∂Vo ⎝L⎠ ⎪ ⎪ ⎪ 2 L ⎛π ⎞ ∂ 2π ⎪ = ⎜ ⎟ GJ ⎪ 2 Q⎨ 2⎝L⎠ ∂θ o ⎪ ⎪ 2 ⎪ ∂ 2π ∂ 2π ⎛ 1 ⎞⎛ π ⎞ = = M ⎜ ⎟⎜ ⎟ ⎪ ∂vo ∂θ o ∂θ o ∂vo ⎝ 2 ⎠⎝ L ⎠ ⎪ ⎪ ⎩ So.

6 . The elastic buckling load of a thin-walled member with asymmetric cross-section can be obtained as the root of the following cubic equation. m= 1 < 2. a small disturbing force to initiate lateral deflection is necessary and the load versus deflection path can then be plotted using an incremental .5 and β = M L 0.P y ) = 0 39 Chapter1. 2 2 ( P .P z ) (r 2 + y o + z o ) p 2 2 .P y ) ( P .P z ) . For non-bifurcation type of analysis.Py o ( P . This mode of buckling may be found in short and open section such as angle and cruciform.3 Torsional Buckling A short thin-walled column under axial force may buckle and twist torsionally. Figure in Chapter 1 sketches the lateral buckling analysis of a cantilever beam of rectangular cross section. MSc Lecture Note. 2004-2005 Professor SL Chan. Iω is warping section constant. L is the effective length of the beam and m is a coefficient to account for the effect of moment gradient given empirically by. which is used as the supporting base of space frames shown in Figure in Chapter 1. G is shear modulus of elasticity.P z o ( P . The elasto-plastic buckling load was also predicted by Chan (1989) using the finite element approach which compared well with the test results by Usami (1971).iterative method which is discussed in the next Chapter. HKPolyU . 1.P x ) ( P . Kitipornchai and Lee (1984) conducted experiments on the elasto-plastic buckling behaviour of angle with varying length and Chan and Kitipornchai (1987b) proposed a finite element approach to the elastic buckling analysis of angle struts.4 β MS in which ML and MS are the larger and smaller nodal moments applied to the two ends of a beam respectively. J is torsional section constant.0. Short thin-walled columns under axial force may twist torsionally before reaching the squash load.M cr = m π L EI y GJ 1 + π 2 EIω GJL 2 in which Iy is second moment of area about the minor principal axis.6.

yo and zo are the coordinates of the centroid about the axes parallel to the principal axes and passing through the shear centre and rp is the polar radius of gyration given by. Most hotrolled sections are not susceptible to this type of buckling whilst built-up sections such as girders require careful consideration of the plate stability. the cross section is assumed to remain undeformed when loaded to its ultimate strength. Using this assumption. Px is the torsional buckling load about the shear centre axis. The overall 40 Chapter1. depending on the boundary condition of the plate being considered. the cross section of the member may deform and exhibit local plate buckling before the member yields or buckles in any other modes previously described. In the design of structural elements susceptible to local buckling.6. For the majority of hotrolled sections. 2004-2005 Professor SL Chan. National codes for design of steel and metal structures have clauses to prevent this or to allow for a reduction in cross sectional capacity when the plate width versus thickness ratio exceeds a certain value. the effective stress or the effective width concept is normally adopted. For the effective width method.4 Local Buckling In the conventional theory of thin-walled beams. material lying outside the width limit of the composing plate elements making up the section is ignored. MSc Lecture Note.in which Py and Pz are the Euler flexural buckling loads of the strut about the y. this assumption is valid and the result obtained on this basis is accurate. In the effective stress method. Figure in Chapter 1 shows the local plate buckling simulated by the large deflection finite shell element computer program. When the thickness of the plate components making up a section is thin. for the top flange of a square hollow section (SHS) under moment. 1. Part 1 (1990) adopt this approach for hot-rolled sections. Ix + Iy A rp = in which Iy and Iz are respectively the second moments of area about the two principal axes respectively and A is the cross sectional area. HKPolyU . NAFSHELL (1995). Codes such as BS5950.axes. the permissible stress of the structural element is reduced to limit the permissible stress in the section. theories developed for analysis of thin-walled beam-columns can be simplified and the time required for the analysis of this type of structures is greatly reduced.and the z.

Considering the beam-column element under an axial force as shown in Figure 1. this concept is used. Et π 12(1 -ν 2) b2 3 2 9. I is the second moment of area. the differential equilibrium equation can be written as. It affects the state of stress as well as the stiffness of the member. x is the distance along the member.resistance of the section is then lowered due to the ignorance of this material and therefore.7 The P-δ Effect The P-δ effect is referred to as the second-order effect due to the deflection along a member and the axial force. 2004-2005 41 Professor SL Chan.9. MSc Lecture Note. A plate component may buckle by compressive edge loads or by shear loads acting along the edges of the plate. HKPolyU . Narayanan (1985) edited a series of recent research works in the area of plate buckling in various types of steel structures. + d v = .M1 2 L dx 2 EI in which E is the Young’s modulus of elasticity. To formulate the P-δ effect and to obtain the correct relationship between the bending coefficients under the action of axial force. Pcr = and the shear buckling load by. In AS4100(1990). v is the lateral deflection in y-axis. Chapter1.Pv + M1 M 2 x. The lowest buckling load acting perpendicularly on the edge of a simple supported square plate is given by. the overall sectional capacity is reduced to cater for the effect of local plate buckling.35π 2 E Ps = 12(1 -ν 2) b2 in which b is the width of the square and υ is the poisson's ratio. resort can be made to the equilibrium equation for a beam-column element. Its careful consideration is important for buckling analysis and design of slender skeletal structures. M1 and M2 are nodal moments at the left and right ends and P is the axial force. 1. Note that positive nodal moment is assumed by the right hand rule using the z-axis pointing outside the plane of this paper.

M2.⎡ sin( µ L. c1 = φ (sin φφ cos φ ) 2(1cos φ )φ sin φ c2 = φ (φ sin φ ) 2(1cos φ )φ sin φ 42 Chapter1. L is the member length. E is Young's modulus of elasticity. 2004-2005 Professor SL Chan. MSc Lecture Note. I is the second moment of area about the bending axis and c1 and c2 are the stiffness coefficients given by the followings.µ x) L. HKPolyU . θ1 and θ2 are the nodal moments and rotations at the two ends of the member. d v1 dv |x = 0 = θ 1 . equation as. 1 |x = L = θ 2 .x ⎤ M 2 ⎡ sin µ x x ⎤ v = M1 ⎢ P ⎣ sin µ L L ⎥ P ⎢ sin µ L L ⎥ ⎦ ⎣ ⎦ in which µ= P EI Differentiating above equation with respect to θ1 and θ2. we have the stiffness dx dx EI (c1θ 1 + c2θ 2 ) L EI M 2 = (c2 θ 1 + c1θ 2 ) L M1 = in which M1. and expressing the rotations at two ends as the nodal rotations as.

Figure 1. the differential Equation can be solved directly as in the case for linear analysis. P.10 shows the variation of axial force and the stability functions. c1 = 4 .PL EI 2 For the zero load case. HKPolyU . 2004-2005 Professor SL Chan. c1 = ψ (sinh ψψ coshψ ) 2(coshψ 1)ψ sinh ψ c2 = ψ (ψ sinh ψ ) 2(coshψ 1)ψ sinh ψ in which ψ = . is reversed. we have the stability function for c1 and c2 as follows. It can be seen that the conventional linear analysis of using c1 and c2 is only correct for the condition of zero axial load. The cubic displacement function for displacement will be obtained and the stability coefficients can be determined directly as. The expressions allow for the variation of the stiffness of a 43 Chapter1. MSc Lecture Note.in which φ = PL EI 2 Repeating the procedure for the tensile load case in which the sign of axial force. which assumes that member is under predominant axial load and small end moment. since it considers correctly the second-order effect due to the element deflection. c2 = 2 The above procedure is an exact means of incorporating the P-δ effect into an element under the Timoshenko beam-column theory. c1 and c2.

The P-δ effect can also be transformed to the P-∆ effect by using several elements per member. the transformation matrix is updated and the effect of large deflections is then included. we can update the geometry of a structure by the updated Lagrangian formuation as. provided that their effects are prevented or insignificant.member in the presence of axial force and they can be used to revise the stiffness of the global stiffness matrix comprising hundreds of elements as well as predicting the buckling of a single member.9 Design of Steel Structures In practical design of steel structures. both these methods for including the effect of large deflections yield the same result (Zienkiewicz [1977]. With proper formulations. 2004-2005 Professor SL Chan. Bathe [1982]). 1. this effect can be included through element formulation so that the coordinate system remains unchanged for the complete loading history. 1. Using more elements for a member. Based on the new or updated geometry. MSc Lecture Note. This effect may become important when the deflection and/or the conjugate force is large such as in the case of a building under a heavy mass at the roof and a lateral wind load. HKPolyU . not all the buckling modes are required to be included in an analysis. To include this effect in an analysis. generally termed as total Lagrangian formulation.8 The P-∆ Effect When a structure deforms grossly. x i+1 = x i + u i in which xi and ui are the coordinates and deflections at step “i” respectively. the original geometry can no longer be employed for the formulation of the transformation matrix simply because the coordinates have been varied. more nodes for geometry update can be introduced along the member and thus the effect of axial force can be included into the member stiffness. at the expense of heavier computation time. Beams with full restraints along the length do not require the determination of the effects of 44 Chapter1. Alternatively.

Framing is also aimed for the symmetrical and balanced layout in order to ensure stability and ductility in resisting dynamic forces. different levels of analysis with various degrees of refinement can be used to design the structure. may not necessarily be directly related to the manual manipulation work involved due to the availability of powerful personal computers. In practical design of steel buildings. the effective length factor or the K-factor. A column of I or box section and of sufficient torsional stiffness is not susceptible to torsional buckling. 2004-2005 Professor SL Chan. To aid the designer and reader to fully appreciate the nature. When the plate components are thicker than a certain breadth versus thickness ratio. needs to be considered in the analysis and design. ductility. In dealing with this type of design and to obtain a better insight into the dynamic and collapse response of a structure. however. limitation and accuracy of these methods. Unfortunately. in addition to strength and stability. the second-order effect normally present is due to axial forces in columns or due to the vertical loads. The ductility factor. 1. the analysis should be equipped with the ability to carry out inelastic and large deflection analysis. HKPolyU . they are summarized below with a cross reference on the generalized load versus deflection path for a structure depicted in Figure in 45 Chapter1. local plate buckling will not occur before the structure reaches the plastic moment.lateral instability. In the conventional linear design and analysis of steel structures. MSc Lecture Note. is introduced to the design procedure for determining the permissible stress. The column strength is very much dependent on the member slenderness ratio and also the side sway of the whole building. is determined.10 Types of Analysis Although the actual behaviour of a structure is nonlinear and complicated. the linearized response of a structure is used to construct the bending moment diagram of the structure from which the strength and stability of each member is checked and designed. defined as the ratio of the displacement at the instance of collapse to displacement at the onset of yielding. As many practical columns are of modest to high slenderness that invalidates the assumption of an analysis considering only material yielding. this slenderness depends on the rigidity of the connections as well as the rigidity of the connecting members to this column being designed and therefore a rather complicated formula is used in different national design codes. These analysis techniques require different degrees of complexity which. In design of structures against extreme and dynamic loads.

Mu. A special feature or convenience about this method of analysis is on the validity of the principle of superposition of different load cases. Thus. As this straight line by itself does not provide any information on stability and the true capacity of the structure. as shown in the line coinciding with the initial slope of any analysis in Figure. In BS5950 (1990). the designer must employ other means to check the safety of a structure such as the member capacity check based on an assumed K-factor for columns and to limit the maximum stress in the structure. non-sway and partial sway frames. 2004-2005 Professor SL Chan. which has been used for decades. the relationship between the applied load and the deflection at any point in a structure is a straight line.Chapter 1. In order to account for any nonlinear effect that has not been allowed for in a linear analysis. the amplification factor is used to increase the bending moment of a column obtained from a linear analysis.1 Linear Analysis Linear analysis. can be estimated from a linear analysis by the following M u = B1 M nt + B2 M lt formula.10. the slenderness ratio of a column in a multi-storey frame is determined from different charts for sway. Thus. 1. In the LFRD (1986) specification. B1 and B2 are the magnification factors corresponding respectively to the P-δ and the P-∆ effects and given in detail in the specification. assumes the deflection is proportional to the applied force. 46 Chapter1. the following factor can be used to amplify the moment in a sway frame. in which Mnt and Mlt are the maximum first-order elastic moments in the member calculated on the assumptions of "lateral translation" and "without translation". one can superimpose the force diagram due to various loads in order to obtain the final force distribution in the structure. the second-order elastic moment. MSc Lecture Note. HKPolyU . Alternatively.

amplification factor = λ cr λ cr . the determination of the effective length factor is difficult to judge. The solution methods for the above problem are the methods of bisection. an eigenvalue problem is formulated by a standard procedure as.1 in which λcr is the elastic buckling load factor which multiplied to the design or applied load vector gives the buckling load vector for the structural system. | K L + λ cr K G | = 0 in which KL and KG are the linear and the geometric stiffness matrices respectively. A detailed discussion of these methods are given by Bathe (1982). Only the lowest or 47 Chapter1. In many cases. There exist a number of numerical methods for the solution of above equation and the computation of eigen vectors which represent the buckling mode shape. is termed as the elastic buckling factor. In many cases. For a general stiffness matrix of a structural system with n degrees of freedom. 1. which causes the structural system to buckle elastically with the structural geometry assumed unchanged until buckling occurs. It can be seen that the application of the above formula becomes tedious when a large number of columns are to be designed and checked. In the bifurcation type of analysis. λcr.10. Furthermore. MSc Lecture Note. The situation will be more complicated when the member is slant or in cases where the member is under non-uniform axial force. the stiffness of a slender member depends on the sign and magnitude of initial load in the member. interpolation and the sub-space iteration schemes.2 Elastic Bifurcation Analysis The buckling load factor. the effective length is difficult to assess by judgement. This factor can be obtained by an eigenvalue analysis detailed in next section. 2004-2005 Professor SL Chan. there exist n roots for the bifurcation load and they represent n possible buckling modes. but this effect of change of member stiffness due to external loads via the initial stress in members is ignored in a linear analysis. HKPolyU .

the lowest few modes are usually of practical interest. i = 1. however. The bifurcation analysis is generally used to determine the exact effective length factor K for a member in a structural system. n p for the jth cycle ⎟ ⎠ in which i M is the moment due to an initial set of applied loads and at node i. Manual methods based on the energy and the equilibrium considerations for the determination of critical loads have been presented by Timoshenko and Gere (1961). Horne (1949) developed a kinematic energy approach for determining the plastic failure load of a steel frame. In this approach. a hinge is inserted to this node by reducing the stiffness of end spring in the element to zero and the 48 Chapter1.6. 1. Mp is the plastic moment capacity of the section at node i. ⎛ ∆ λ j = min ⎜ ⎜ ⎝ i Mp iM ⎞ ⎟ . Although the method can only provide an upper bound solution to the stability of a frame.75 for uncladed frames or claded frames with inclusion of wall panel stiffening in analysis. it can be used as an approximation to the instability stage. 2004-2005 Professor SL Chan. a linear analysis is carried out first to search for the location for the formation of the first plastic hinge under a set of increasing loads acting upon a structure.3 First Order Elastic-Perfectly Plastic Analysis of rigid-jointed frames In the plastic analysis of steel frames. In the computer analysis. elements with end springs to simulate the softening of a section stiffness are used to model the members in a framed structure. When the possibility of formation of a plastic hinge along a member exists. np is the total number of possible plastic hinge locations simulated by element springs . Mathematically the load factor leading to the formation of the first plastic hinge analysis can be computed by the following simple equation.10.6 for clad frames ignoring the stiffening effect from wall panels and 5. which was subsequently modified by others. Chajes (1974) and Brush and Almroth (1975). BS5950 (1990) requires the elastic buckling load factor to be larger than 4. Once the location of the first plastic hinge is spotted. becomes tedious when the structure size is large. HKPolyU . Its application to a simple portal has been shown in Figure 1. This method ignores the buckling and the large deflection effect until collapse. MSc Lecture Note. several elements are needed to model each member. The manual approach.

In many practical structures.analysis is continued based on the modified structure with this new plastic hinge. A very small force will cause infinite displacement. HKPolyU . MSc Lecture Note. the geometric or the material nonlinearity does not dominate its failure. neither the bifurcation nor the plastic load factors are close to the true collapse load of a structure.10. 49 Chapter1. 2004-2005 Professor SL Chan. In this mechanism. The second incremental load factor. Rankine (1863) used a reciprocal additive formula for computation of collapse load which was extended to frameworks by Merchant (1954) as. The plastic and the elastic buckling factors can be combined to check the elastic and plastic stability of a frame. which can be sensed by the semi-definite condition or zero determinant of the stiffness matrix and the final collapse load factor is determined as the sum of all the incremental load factors as. deflections can occur without creation of strain energy. ∆λ2. BS5950 (1990) specifies the range of these factors for steel frames in a very simple and empirical way and it can be used in conjunction with the elastic bifurcation load for a quick and approximate check for structural stability. 1. is calculated by repeating the above process.4 The Merchant Rankine's formula for collapse load of rigid-jointed frames Unless a structure is very slender or very stocky. Empirical formula has been developed to combine these two failure load factors to approximate the collapse load. λp = ∑ ∆ λi i =1 j in which Aj@ is the total number of plastic hinges formed before a plastic mechanism is formed and λp is the plastic collapse load factor for the rigid-jointed frame. This procedure is repeated until a mechanism is detected.

In a bifurcation point. one must distinguish between the limit and the bifurcation point which requires a different numerical treatment in tracing the equilibrium path by computer. material yielding is not considered in the second-order elastic analysis. The checking of member Chapter1. Figure in Chapter 1 shows these effects in a frame. Although modified formulae are proposed to refine the equation for a better approximation of the collapse load through the elastic bifurcation load and the plastic collapse load. HKPolyU . In a nonlinear analysis. 1. the structure is assumed to be undeformed in all except the loaded directions. For the stability theory involving large deflection effects. This can either be a bifurcation point or a limit point as shown in Figure 1. the plastic and the elastic buckling load Although the Merchant-Rankine formula is simple to use. Their effects are due to deflections in the members and in the structures respectively and difficult to be rationally accounted for in a linear analysis for all cases. = 1 λp + 1 λ cr in which λu. the effects of initial imperfection. A limit point refers to the turning point for the equilibrium path of a structure. These P-δ and P-∆ effects may be important for slender structures in practice and their existence illustrated previously. MSc Lecture Note.10.11. it takes no account of many important factors such as initial imperfection. Second-order elastic analysis implies an analysis allowing for the secondorder effects due to change of geometry and initial stress in members. they are mostly empirical and lack a rational consideration of the effect of member initial imperfection and residual stress. λp and λcr are the ultimate. which includes the P-δ and the P-∆ effects. residual stress.1 λu factors respectively. As implied by its name. 2004-2005 50 Professor SL Chan. the P-δ and the P-∆ effects can all be included in a rigorous analysis. This can be considered as the transition point from stable to unstable equilibrium states or vice versa. the critical point refers to the semi-definite condition for the tangent of the equilibrium path or the tangent stiffness matrix of the tangent stiffness matrix. the sway P-∆ effect and no rigid mathematical proof is yet available for verification of the formula. In principle.5 Second-order Elastic Analysis A relatively more complicated approach for accurate analysis of structures involves the tracing of the equilibrium or the load versus deflection path.

The inclusion of material nonlinearity can either be carried out by the lump or by the fibre plasticity models. yielding is assumed to spread along the element length and across the section. The error for a second-order inelastic analysis may come from the approximation for the material model. MSc Lecture Note. 1989). As shown in the idealized load versus deflection response curves in Figure. the second-order inelastic analysis may give an error in predicting the true collapse load. As shown in Figure. generally termed as the plastic hinge (Yau and Chan. residual stress and initial imperfection in the members but improvement can be achieved by refining the model. The process of determining the effective length of the member by a formula in a design code in order to calculate the second-order moments can. However.strength relies on the application of design formula whilst the second-order effects such as the P∆ and P-δ effects are included automatically in the analysis. the member capacity check can be carried out by superposition of the computed nodal moments at the two ends of a member obtained directly from the second-order elastic analysis and the moment along the member due to member load and allowing for the P-δ effect. Generally speaking. They deviate when the load is at a level causing material yield and thus the second-order elastic analysis over-predicts the collapse load of a structure. For example. the path for second-order elastic analysis follows the exact equilibrium path before yielding. 2004-2005 Professor SL Chan.6 Second-order Inelastic Analysis If the effect of material yielding is included in the above second-order elastic analysis it becomes the second-order inelastic analysis. the lump plasticity model is more efficient and easier to apply whilst the fibre model is more accurate and considered to be exact. a plastic hinge concept of reducing the stiffness at a particular location with zero length may not be able to completely represent the true yielding pattern of the structure.10. As a result. if 51 Chapter1. The error is difficult to quantify but is problem dependent. the difference between the second-order inelastic analysis and the advanced analysis is expected. and numerical integration is normally employed to evaluate the element stiffness of a partially yielded section (Chan. In the lump plasticity model. therefore. In the fibre model. However. the amplification factors or equivalent become unnecessary. 1994). be skipped and the accuracy and saving in manipulation effort can be achieved. In many cases yielding is spread over such a distance that the plastic hinge cannot sufficiently capture the true behaviour of the frame. yielding is assumed to be concentrated in a small region of zero length. HKPolyU . For example. 1. the second-order inelastic analysis is not yet fully qualified for the requirement of an advanced analysis which can be used for direct design.

the phenomenon of resonance and ductility of a structure cannot be revealed in this static approach. the structure is allowed to undergo a certain degree of damage and plasticity in structures is normally allowed. For instance. input information and the model must be precise and accurate in order to simulate the structural behaviour reliably. Energy absorption capacity must be adequate and fracture failure should be prevented. This so-called transient response analysis plots the stress and deflection of the complete structure against time using various loadings or earthquake history. The Northridge earthquake in the U. 1. MSc Lecture Note. The widely used equivalent load approach to transform the dynamic force to a set of static forces in many cases cannot reflect the true behaviour of a structure. allowing for various sources of non-linearities such as the material yielding. It is economically unjustified to design a structure to remain in its elastic stage under severe load. Stiffness is the 52 Chapter1. environment. the structure should be repairable after an earthquake strike.S.10.7 Dynamic and Cyclic Analysis Wind and seismic loads are time dependent and the response of a structure under this loading type is dynamic in nature. in 1994 and the Hyogoken-Nanbu earthquake near Kobe city in Japan in 1995 revealed the design code at that period was still not fully sufficient to resist even a moderate earthquake attack. the idealization may become inaccurate and invalid. is to trace the complete equilibrium path. although the analysis method by itself is accurate. The sequence for formation of plastic hinges must be examined and controlled so that the structure remains stable. and is required for consideration. Generally speaking. The modal analysis and the spectrum analysis (Clough and Penzien 1993) is an improved method to incorporating these dynamic effects into the analysis. collapse should simultaneously be prevented even during a severe earthquake. large deflection and buckling. In the design of structures against extreme and dynamic loads.the yielding zone in a beam is distributed over a large area and only an element with a plastic hinge allowed to form at its ends is used to model each member. in addition to strength and stiffness.A. For a less severe load. HKPolyU . provided that it does not collapse. however. 2004-2005 Professor SL Chan. strength must be sufficient for stress reserve in a structure to resist loads before yielding. A more natural and accurate analysis but expensive in terms of computer time. Depending on the severity of the applied load expressed as years of return period. loading sequence is also an important parameter for analysis and ductility. It must be emphasised that.

The most refined Advanced Method of Analysis includes all the second-order effects and also material properties and residual stress. residual stress. 2004-2005 Professor SL Chan. The geometrically non-linear or the second-order elastic analysis ignores the effects of material yielding and thus it can only follow the equilibrium path of a real structure before yielding. It can be an improved or a refined second-order plastic-hinge analysis or direct plastic zone analysis. member initial imperfection and/or semi-rigid joint stiffness is included in the method. erection procedure and interaction with foundations. The effect of initial imperfections is also encountered for in the analysis and a practical design can be directly carried out on the basis of the stress computed by this type of rigorous analysis. At the present stage of development. HKPolyU .rigidity of the structure against external loads and its inadequacy leads to buckling and excessive deformation. the curves generated by this type of analysis follow closely to the actual behaviour of a structure. 53 Chapter1. MSc Lecture Note. In other words. only the section capacity check is needed for member design.10.8 Advanced Analysis Advanced Analysis combines the effects of material yielding and geometrical change with sufficient accuracy for practical purposes so that member capacity checks can be avoided. Ductility is the character of a structure to tolerate large deflection before fracture and its consideration is important in assessing the performance of a structure under dynamic loads. For an idealized model. The second-order inelastic analysis includes the effects of material yielding but can be distinguished from the most sophisticated method of analysis. initial member imperfection. out-of-plane buckling such as lateral torsional buckling should be prevented when using this type of analysis and design. 1. the Advanced Analysis. As such. by the fact that no detailed consideration on the accuracy of the plasticity model. This is due to the fact that these buckling modes may not be easily conducted in the global analysis for the complete structure and they have relatively less interaction with the global structure. as indicated in Figure. Advanced Analysis can be defined as any analysis method that does not require specification member capacity checks. The ductility factor can only be evaluated accurately by a complete second-order plastic analysis.

2nd edition. (1949).D.21. and Lee. N. 54 Chapter1. Englewood Cliffs. Department of Structural Engineering.11 REFERENCES American Institute of Steel Construction (1986). (1987b).L. Research Report CE-STR-93-26. 27. Sydney. 1989. Journal of Structural Engineering. plates and shells. pp.L. Construction safety of temporary structures. Chan. 1993. Structural use of steel in building. S. Vol. Chan. S. S. Chicago. Prentice-Hall Inc. no. Structural Engineer. Contribution to The design of steel frames by Baker.F. and Almroth.J. University of Queensland. D. Journal of Structural Engineering. specification for structural steel buildings. K. British Standards Institution (1990). Department of Civil Engineering. (1994).1. A.W.99-106. Chen. Geometric nonlinear analysis of asymmetric thinwalled beam-columns. S. (1982). McGraw-Hill.243-254. 113(4). Civil Engineering Series. pp. (1989). W. M.F. 11. School of Civil Engineering. Journal of Engineering Structures. (1987). pp. (1993). S. W. Civil Engineering and Engineering Mechanics Series. and Kitipornchai.23-30. Bathe. ASCE. Inelastic Post-Buckling Analysis of Tubular Beam-Columns and Frames. and Kitipornchai. S. J.W. Chan..O.721-739. AISC. S. Purdue University. 9.O. Research Report CE54. pp. (1987a). A. H. MSc Lecture Note. Prentice-Hall Inc.L. B. Finite element procedures in engineering analysis. (1975). J.L.. Second-order inelastic analysis of steel frames by personal computers. Inc. AS-4100. Principle of structural stability theory. Nonlinear analysis of structures composed of thin-walled beam-columns.J. Chan. S. Inelastic experiments on angles and tee struts. (1974). Brush. Australia. Load and resistance factor design. U.R. and Penzien. R. Australian Standard for Steel Structures (1990). Buckling of bars. Chajes. and Chan. Australia for partial fulfilment of the Degree of Doctor of Philosophy. vol. Pan. BS5950. (1984).J. Tsong. pp. Nonlinear finite element analysis of angle and tee beam-columns.. 2004-2005 Professor SL Chan.2. Horne.L. Clough.F.. Journal of Engineering Structures. N. McGraw-Hill. Englewood Cliffs. Chen. and Yen.K. Part 1. HKPolyU . 421 Kitipornchai. pp. University of Queensland. Dynamics of Structures. 33. thesis submitted to the Department of Civil Engineering.

MSc Lecture Note. 2nd edition.S. W. Stability of I-beam with elastic end restraints. pp. Elsevier Applied Science. N. The Structural Engineer. Louis. 2004-2005 Professor SL Chan.J. O. Timoshenko. C. (1977). Charles Griffin and Comp. Dissertation presented to Washington University at St. Australia. Plated structures . T. J. Structures in Architecture. pp.2803-2819. (1965). University of Sydney.185-190.M. and Chan.stability and strength.C. HKPolyU . R. pp. (1961). Nonlinear analysis of shells. S.RXXX. (1994). and Heller. And Chan. Rankine. S. “Out-of-plane advanced analysis of steel structures”. The failure load of rigidly jointed frameworks as influenced by stability. 38.Y. McGraw-Hill.. S. R. 3rd Edition. (1863). Englewood Cliffs. (1985). Mo. 32. Yau.157 Trahair. Research Report No. Journal of the Institution of Engineers. 1993. Usami. Trahair. User's Reference Manual. M. Restrained single-angle columns under biaxial bending.P.L. and Gere.S. Narayanan. Prentice-Hall.. Theory of elastic stability. (2002). 2nd edition. London. N. Inelastic and stability analysis of flexibly connected steel frames by the spring-in-series model. W.Y.Merchant. New York. The Finite Element Procedure. (1954). 55 Chapter1. (1971). NJ. 1963. N.L. McGraw-Hill. Zienkiewics. for partial fulfilment of the requirements for the degree of Doctor of Science.M. NAFSHELL. A manual of civil engineering. ASCE. Salvadori. Journal of Structural Engineering.