Sie sind auf Seite 1von 11

KSCE Journal of Civil Engineering (2008) 12(6):379-389 DOI 10.

1007/s12205-008-0379-3

Structural Engineering

www.springer.com/12205

Direct Second-Order Elastic Analysis for Steel Frame Design


C.K. Iu*, W.F. Chen**, S.L. Chan***, and T.W. Ma****
Received March 29, 2008/Accepted July 13, 2008

Abstract
The traditional structural design procedure, especially for the large-scale and complex structures, is time consuming and inefficient. This is due primarily to the fact that the traditional design takes the second-order effects indirectly by virtue of design specifications for every member instead of system analysis for a whole structure. Consequently, the complicated and tedious design procedures are inevitably necessary to consider the second-order effects for the member level in design specification. They are twofold in general: 1) Flexural buckling due to P- effect, i.e. effective length. 2) Sway effect due to P- effect, i.e. magnification factor. In this study, a new system design concept based on the second-order elastic analysis is presented, in which the second-order effects are taken into account directly in the system analysis, and also to avoid the tedious member-by-member stability check. The plastic design on the basis of this integrated method of direct approach is ignored in this paper for simplicity and clarity, as the only emphasis is placed on the difference between the second-order elastic limit-state design and present system design approach. A practical design example, a 57 m-span dome steel skylight structure, is used to demonstrate the efficiency and effectiveness of the proposed approach. This skylight structure is also designed by the traditional design approach BS5950 for comparison on which the emphasis of aforementioned P- and P- effects is placed. Keywords: traditional design, geometric nonlinearities, P- and P- effects, finite element method

1. Introduction
Owing to the limitation of technology in the past decades, the simple case of flexural buckling of an isolated and individual member was only investigated by means of the theoretical formula and experiments in order to measure the P- effects. Euler (1759) derived a buckling equation to study the P- effect on a simple elastic column theoretically. Later, Timoshenko and Gere (1961) analytically presented a comprehensive study of flexural buckling of a column with different boundary conditions. Using their theoretical solution of a buckling column, P- effect on an ideal column with different support conditions can be evaluated by the imaginary member length, such as effective length or K-factor. Perry-Robertson formula (Ayrton and Perry, 1886; Robertson, 1925) was used to include an initial imperfection, which was extended and modified from an ideal case of a straight member as the Eulers formula (Euler, 1759). Batterman and Johnston (1967) performed a study of the practical imperfections on the column strength under compression, including nonlinear stress-strain properties, residual stress and initial crookedness, by digital computer.

The flexural buckling of a column due to P- effect was only dependent upon whether the column is effectively held in position and restrained in direction according to BS449-1969 without considering the overall frame instability on a continuous column. However, the flexural buckling of a continuous column in a frame should be heavily influenced by frame stability, as discussed by Wood (1974a, 1974b, 1974c). Hence frame buckling due to sway or P- effect is another crucial issue, so elastic frame buckling had received considerable attention of scholars so that the classical analysis methods (Bleich, 1952) are widely available in the past. In the past decades, the structural design is mainly based on the linear elastic numerical analysis. The traditional design procedures to account for the nonlinear flexural and frame buckling due respectively to P- and P- effects have recourse to the empirical design specifications. In the AISC LFRD-1993, the second-order P- and P- effects can be estimated from a firstorder analysis by using the respective B1 and B2 magnification factor to correlate the linear moments to second-order moments based on the results from Kanchanalai (1977) and Bjorhovde et al. (1978). In addition, the design specifications for these insta-

*Research Associate, Dept. of Civil and Environmental Engineering, University of New South Wales, Sydney, NSW2052, Australia (Corresponding Author, E-mail: iu.jerryu@gmail.com) **Professor, Dept. of Civil Engineering, University of Hawaii at Manoa, Honolulu, Hawaii 96822, USA (E-mail: chenwf@eng.hawaii.edu) ***Professor, Dept. of Civil and Structural Engineering, The Hong Kong University of Polytechnic University, Hung Hom, Kowloon, Hong Kong TU712, China (E-mail: ceslchan@polyu.edu.hk) ****Assistant Professor, Dept. of Civil Engineering, University of Hawaii at Manoa, Honolulu, Hawaii 96822, USA (E-mail: tianwei@hawaii.edu) 379

C.K. Iu, W.F. Chen, S.L. Chan, and T.W. Ma

bilities are heavily dependent on both geometry of entire continuous structure as illustrated in reference (King, 2000) and stiffness of the adjacent connected members, such as G-factor method proposed by Bridge and Fraser (1987), which is adopted by AISC-LRFD-1993. The P- effect on a structure can also be approximated by the elastic buckling mode from buckling analysis (Harrison, 1973; Bathe, 1982). Hancock et al. (1995) developed their programs to numerically determine the buckling load and buckling load factor cr of an entire structure based on the buckling analysis approach. This method gives close approximations, when the buckling pattern is the same at each storey of a typical framed structure. And it is conservative, when no horizontal forces and the buckling patterns vary from storey to storey. BS5950-2000 relies on a factor modified from cr in order to amplify the linear moment to be the second-order P moment. Also the simple Merchant-Rankine formula was widely used in the past to compute the elastic buckling load factor cr of a framed structure proposed by Merchant (1954). In short, the elastic buckling load factor cr used in the traditional design allows for the P- effect. According to European Convention for Constructional Steelwork (ECCS) (1984, 1991) recommended that an out-of-plumb of a column, such as with 1/200 times the column height, is incorporated into the modeling for the linear elastic analysis to stimulate the P- effects by first-order elastic analysis. Another method is an equivalent notional load method proposed by Wood et al. (1976a, 1976b), which produces a fictitious horizontal forces applied to the top of the column in the typical frame under consideration by the same principle. As a result, a second-order P moment due to sway effect on a designated column is generated in order to study the frame buckling of a whole structural system. Both methods are also allowed in BS59502000, Eurocode 3-2005 and HKSC-2005 to account for P- effect. In conclusion, the traditional member design requires different design parameters, such as effective length and magnification factor for each member, to incorporate respectively the P- and P- effects on every member. These stability design parameters are approximate and sometimes require the subjective engineers judgment such that it is difficult to determine the buckling load of a member or an entire frame accurately. And even, under some circumstances, the stability design parameters are impossible to be evaluated for the complicated structural forms. Moreover, the design parameters of different members may be various, which heavily depend on the restrained conditions, stiffness of connected members, loading combinations along the member, and also the geometric of entire framed structure, etc. It implies that the traditional design procedures to account for these second-order effects must be carried out member by member. As a result, the traditional design process for the individual member level (AISC-LRFD, 1993; BS5950, 2000; Eurocode 3, 2005; HKSC, 2005) is onerous and time consuming. It is worthy of noting that the traditional design methods commonly refer to the second-order elastic limit-state design in this paper.

Owing to complexity of the traditional design specifications, it is not feasible to assess the global structural behaviour with second-order effects by means of manual calculation. Further, the use of laboratory tests has practical and economical limits to investigate the second-order effects on either a whole structure or an individual member. On the other hand, since the development of computer technology no matter in software and hardware has been so advanced in the recent decade, computational method is feasible to employ the second-order inelastic analysis to design the practical framed structures. Thus, the second-order inelastic analysis was drawn by many scholars attention in 80-90s century, including El-Zanaty and Murray (1983), Goto and Chen (1987) and Chan and Kitipornchai (1987). Later, Ziemian et al. (1992a, 1992b) presented a numerical second-order inelastic analysis for the practical limit state design of the framed structures. Later, Kim and Chen (1996a, 1996b) studied the member and frame buckling behaviour of steel frames by using the second-order inelastic analysis, which was also comprehensively discussed by Chen and Kim (1997). Also Oda and Usami (1997) applied second-order elastic analysis for stability design method without using the effective buckling length. Further, Ioannidis and Raftoyiannis (2005) proposed a simplified nonlinear stability elastic analysis to solve pre-buckling, buckling, post-buckling, bifurcation limit point, as the design of frame members based on the overall stability of the frames often shows the inadequacy of approximate formulas provided by steel codes. So far the nonlinear analysis was well developed in Europe and North American for the practical engineering design. In contrast, the nonlinear analysis for practical design in Asia is just in infant stage. For instance, Chan and Zhou (1998) was one of the earlier developers on this topic in Asia. In the modern age, the structures become more and more slender and enormous for aesthetic and economical reasons. As a natural result, the structures behave more likely nonlinearly and elastically. Hence, an efficient second-order elastic analysis for direct system design is indispensable in this age. The secondorder elastic analysis can play an important role on integrating the incompatibility between system analysis and member design of elastic structure, while this design approach can directly incorporate all second-order effects through its numerical analysis. Meanwhile, the numerical analysis of the present system design approach can also sustain its numerical simplicity and efficiency for all kinds of structural forms, especially feasible for the elastic design of complicated large-scale structures. For example, the countercheck of structural adequacy against stability effects of the Beijing 2008 Olympic game national stadium (Birds Nest) was also implemented by using nonlinear analysis (NIDA, 2007). The objective of this paper is to demonstrate the second-order elastic analysis can directly tackle both P- and P- effects on an elastic structure such that no indirect design specifications and manual calculations for stability or geometric nonlinearities are required. In addition, present system design approach is also helpful to the design of a whole elastic structural system without member-by-member stability design procedures. In Section 2, a
KSCE Journal of Civil Engineering

380

Direct Second-Order Elastic Analysis for Steel Frame Design

basic stiffness formulation of the present second-order elastic analysis is derived, in which the member bowing or P- effect is included. In Section 3, a nonlinear solution procedure is illustrated how to cope with these second-order P- and P- effects such that these nonlinear effects. In Section 4, the traditional design procedures for the simple structures are present. Meantime, these design examples are also carried out by present system design approach, which makes the same design obviously simpler. In Section 5, a real practical design example is used to illustrate the efficiency and effectiveness for the design procedures on the basis of the present system design approach. However, it is worth noting again that if the structure is not prone to second-order effects but material yielding, the member-bymember capacity check is still necessary by present system stability design approach. It is out of the scope of this paper because of simplicity and clarity.

independent twist rotation about the x-axis. EA, EI and GJ are the axial rigidity, flexural rigidity about corresponding axes and torsional rigidity, respectively. P is the axial member load. {dk} and {fk} are the column vectors of the corresponding displacement and external applied force. The applied load {fk} is assumed to be independent of the displacement {dk}. External work done V is a linear function of displacement {dk}. The total potential energy for second-order elastic analysis is equal to the sum of Eqs. (4) and (5) as: = UV (6)

2. Basic Stiffness Formulation


In a second-order elastic analysis using the finite element formulation, it is customary to use the Greens strain tensor on the centroid axis along an element to express the straindisplacement relation in Eq. (1), which consists of nonlinear axial strain due to P- effect or member bowing effect 1/2(/ x). - - t = ----- + -- ----- + -- ------ y ------- z -------x 2 x 2 x x 2 x 2 u 1 v
2

1 w

2v

2w

(1)

in which u, v and w are the axial deformation and transverse displacements in y- and z-direction, respectively. The stressrelated strain t is expressed in Eq. (2) to correlate with the normal stress .

The second-order stiffness formulation can be derived from the total potential strain energy equation of Eq. (6) to include the P- effects, as the second-order term of ( P 2 ) ( dv dx ) 2 dx is L present. The stiffness formulations are then derived by making use of the principle of minimum potential energy. Owing to the inclusion of second-order term, the stiffness formulation constitutes a nonlinear equilibrium equation, which can be solved by using the nonlinear solution procedures as stated in Section 3. Formulations of secant stiffness and tangent stiffness are then derived from the total potential energy function of Eq. (6) after back substitution of nodal displacement functions. The incremental natural deformations expressed in the secant stiffness formulation eliminate rigid body movement, when based on updated Lagrangian formulation. On the other hand, the change of geometry of P- effect due to rigid body movement is then considered by updating the coordinate of the structural system. The incremental secant stiffness [Ks] is a first derivative of the internal strain energy equation.
u1 u L L --------------------2 4L z1 ----- z2 4L y1 ----- y2 --------L 30 30 30 30 U ------- = EA + P + P dk LL u1 + u2 ----4L 4L ----------------------- z1 + ----- z2 ----- y1 + ----- y2 30 30 30 30 L 4 4 2 2 -- z1 + -- z2 -- y1 + -- y2 L L L L + EIz + EIy 2 2 4 4 -- + -- -- + -- - z1 - z2 - y1 - y2 L L L L

= E t

(2)

The incremental internal strain energy U is based on Castiglianos first theorem in terms of stress and strain components as Eq. (3). Therefore, the internal strain energy U can be obtained by integration over the cross section and along the element length as given in Eq. (4). 1 U = E d dAdx = E -- 2 dAdx 2 0 LA0 LA
t

(7)

The tangent stiffness matrix can be obtained by taking a second derivative of the total potential energy in Eq. (6) as, 2 -------------- = [ Kt ] = [ KL ] + [ KG ] dj dk (8)

(3)
2 2

EI d 2 v P dv P dw EA du U = ------ ----- dx + -- ---- dx + -- ------ dx + ------z ------- dx dx dx dx 2 L dx 2 2L 2L 2 L EI d 2 w 2 GJ d 2 + ------y -------- dx + ------ ---- dx dx 2 2 L 2 L dx (4)

The external work done, V, is equal to the applied force multiplied by the corresponding displacement as follows: V = { dx } T { fk } (5)

in which [KL] and [KG] are the 66 linear stiffness matrix and geometric stiffness matrix, which incorporates the member bowing effect, of an element in the local coordinate as stated in Iu and Chan (2004). The tangent stiffness matrix should assemble and transform into global coordinate as written in Eq. (9), because the incremental nodal displacements of a structure can then be obtained on the basis of the tangent stiffness relationship. [ KT ] =
elem n=1

in which u, v, and w are axial deformation and lateral displacements in the direction in y-axis and z-axis, respectively. is
Vol. 12, No. 6 / November 2008 381

[ L ] [ Ke ] [ L ] T = [ L ] ( [ T ] T [ Kt ] [ T ] + [ N ] ) [ L ] T
n=1

elem

(9)

C.K. Iu, W.F. Chen, S.L. Chan, and T.W. Ma

in which [T] is a transformation matrix relating the member forces to element force in local coordinate. [L] is a transformation matrix from local ordinate to global ordinate. And [N] is a stability matrix to allow for the work done of rigid body motion proposed by Meek and Tan (1984) or the geometry change of a structure. In other words, the stability matrix [N] can account for the rigid body movement of the nodal displacement in global structural system, which is neglected in secant stiffness formulation of Eq. (7). Consequently, the incremental displacements evaluated from the tangent stiffness of Eq. (9) include the rigid body movement of P- effect, and the geometry of a structure is then updated by accumulation of both neutral member deformations and rigid body movement.

load increment and unbalanced force. The incremental displacement can be determined by the tangent stiffness in Eq. (12) after the load increment is known. { u }in = [ KT ] 1 { f }in { u }in = { u } n 1 + { u }in (12) (13)

The incremental displacements in global coordinate can be transformed to the member deformation { ue }in by the transformation matrix [L]T for mapping of global to local coordinate. Incremental member resistance in member coordinate { Re } in can then be obtained in local coordinate as, { u e }in = [ L ]T { u }in { Re }in = [ Ks ] { ue }in (14) (15)

3. Nonlinear Solution Procedures


A nonlinear incremental-iterative procedure is used in this study to trace the nonlinear equilibrium path due to the geometric nonlinearities i.e. the P- and P- effects. An incrementaliterative solution procedure is developed based on the NewtonRaphson method for a simple illustration, which is schematically shown in Fig. 1. The equilibrium equation at the solution point is expressed as, { f } [ K s ] { u } = [ K T ] { u } (10)

in which [Ks] is the incremental secant stiffness at { ue }in as given in Eq. (7). After determination of the member resistant forces, the incremental member resistance is accumulated to the total member resistance in global coordinate as, { R }in = { R } n 1 + [ T ] [ L ] { Re }in (16)

in which [T] is the transformation matrix from member coordinate to global coordinate. The unbalanced force { f }in+ 1 at second iteration (i=1) is obtained as, { f }in+ 1 = { f }in { R }in (17)

in which [Ks] is secant stiffness in global coordinate. [KT] is the tangent stiffness matrix as given in Eq. (9). {f} is the applied load. Eq. (10) can then be rewritten in the form of incremental equilibrium equation as, { f } = [ K T ] { u } (11)

in which the {f} is the prescriptive load increment or unbalanced force. {u} is the incremental displacement due to

where { f }in and { R }in are the total applied load and total member resistance at n-th load cycle, respectively. The above incremental procedure is repeated until an equilibrium solution is achieved or divergence is detected. After a nonlinear equilibrium path due to the second-order P- and P- effects is numerically traced by using the above solution procedures, these geometric nonlinear effects can be directly and automatically incorporated in the loading distribution.

4. Design Procedures
Firstly, the flexural buckling on a column of W620 section is studied by the present second-order elastic analysis. The 10 m long steel column is subjected to critical buckling load 350 kN with 1 mm eccentricity. The support conditions of both ends are pinned. In this comparison, the validation of the present secondorder elastic analysis for the P- effect is counter-checked by the well known stability function theory, which was comprehensively discussed (Chen and Lui, 1986). Secondly, a simple portal frame, which stemmed from Vogel (1985), is used as a demonstration of the design procedures by both traditional design (AISC-LRFD, 1993; BS5950, 2000; Eurocode 3, 2005; HKSC, 2005) and second-order elastic analysis. The geometry and member sections of this Vogels portal frame are plotted on Fig. 2. To ensure elastic behaviour of the simple portal frame, the high yield strength S460 of the member is used.
382 KSCE Journal of Civil Engineering

Fig. 1. Incremental-iterative Scheme of Arc-length Method

Direct Second-Order Elastic Analysis for Steel Frame Design

Therefore, this column cannot withstand the axial compression more than 340 kN. Fig. 3 indicates that the linear analysis cannot predict the maximum load allowable on this column. On the contrary, the second-order elastic analysis verges to that total applied load level of 340 kN. Thus, the traditional design necessitates the assumed effective length for the P- effect. The design procedure of compressive buckling resistance of the column by traditional design is as shown below, LE 10 Slenderness ratio, = ---- = --------------- = 148 rx 0.0676 (19)

Compressive strength, pc = 83 N/mm2 (from strut curve b) (20) Compression resistance, Pc = pc A = 83000 3.787 103 (21) = 314.3kN Therefore, the maximum compression load that this column can endure is 314.3 kN, which is less than the critical buckling load as written in Eq. (18) by the traditional design approach. The corresponding load level is 0.924 ultimately. However, no deformation behaviour can be obtained by traditional design. This design specification acts as an indicator to testify whether the member section is vulnerable to flexural buckling. Based on the second-order elastic analysis, no indirect manual calculation (Eqs. (18) to (21)) are present for the stability design of this column. 4.2 Traditional Design BS5950 (Simple portal frame) The procedures of traditional design to incorporate both second-order P- and P- effects generally are summarized as follows: System design level I. Amplifying the first-order loading distribution on an entire structure due to sway effect (P- effect) for subsequent member design.

Fig. 2. Geometry and Member Properties of the Vogels Portal Frame

4.1 Traditional Design BS5950 (Pin-pin column) Fig. 3 shows that the mid-height deflections of the column from both stability function and present second-order elastic analysis reach a good consensus. The mid-height deflection according to first-order analysis is also plotted on Fig. 3 on which the traditional design is usually relied. Based on the linear analysis, the traditional design necessitates the design specifications to account for the second-order P- effect. In this simple case, the compressive buckling load of the column can be theoretically predicted by Euler formula as given,

2EIx 2 2 108 1.72 10 5 Pcr = ------------ = -------------------------------------------------------- = 340kN L2 10 2

(18)

Fig. 3. Load-Deflection of Pin-pin Column Subjected to Axial Load with Eccentricity


Vol. 12, No. 6 / November 2008 383

C.K. Iu, W.F. Chen, S.L. Chan, and T.W. Ma

Member design level II. Calculating the member compression resistance Pc to allow for the member flexural buckling (P- effect). There are two main methods (BS5950-2000; King, 2000) to allow for the P- effect for the regular and typical framed structures. They are sway-check method and amplified moment method, whose design processes are plotted on Fig. 4(a). Also the P- effect is normally considered by buckling resistance of the member. The flow chart of this effective length method is depicted on Fig. 4(b). 4.2.1 Amplified Moments for P- Effect (Simple portal frame) In traditional design, a required load factor (magnification factor) r as shown in Fig. 4(a) is necessary to enlarge the firstorder loading distribution of the frame on Fig. 2 under horizontal

load case in order to include the second-order effect (P- effect) in the loading distribution, i.e. the second-order P- moment. Hence the second-order moment distribution is generally greater than first-order moment distribution. Otherwise, the frame is not vulnerable to P- effect or sway effect. Fig. 5 shows the internal loading distribution on this simple portal frame. To account for the P- effect, the sway-check method is validated to apply for the regular and typical portal frames as shown on Fig. 2, only if the geometrical limits are satisfied. According to the sway-check method, the notional load can be applied to the frame under gravity load case only, which is 0.5% of total designed gravity load. The associated deflection at the top of column is 3.765 mm obtained from the first-order elastic analysis, which satisfies the sway criterion of deflection limit check as indicated in Eq. (22). And the corresponding first-order internal loading distribution is indicated on Fig. 5 by the solid lines with the notional horizontal load.

Fig. 4(a). Member Design Procedures for P- Effect on Portal Frame by using Sway-check Method or Amplified Moment Method
384 KSCE Journal of Civil Engineering

Direct Second-Order Elastic Analysis for Steel Frame Design

Fig. 5. First-order Loading Distribution on the Simple Portal Frame Fig. 4(b). Member Design Procedures for P- Effect on Portal Frame by using Effective Length Method

= 3.765mm < ----------- = ----------- = 5mm

h 1000

5000 1000

(22)

If the deflection i at the top of the column satisfies the sway criterion, the magnification factor r should then be taken as unity under gravity load case as given on Fig. 4(a). Under the horizontal load case for this simple portal frame example on Fig. 2, the elastic critical load factor cr is necessary to evaluate magnification factor r from Eq. (23).
i cr = ------------ = -------------------------- = 6.64 200 i 200 3.765

5000

(23)

Since the critical load factor cr is greater than 5, the portal frame is sway frame and can be designed without using secondorder analysis.

cr 6.64 r = ------------- = ----------------- = 1.177 cr 1 6.64 1

(24)

4.2.2 Member Design for P- Effect (Simple portal frame) According to the traditional design approach, the P- effect of flexural buckling of the two columns in the portal frame should be considered by using compression resistance Pc, which is based on the imaginary member length. In fact, the P- effect induces second-order P moment, which deteriorates the member stiffness, and then enlarges the member deflection. In short, the P- effect can only yield a greater moment distribution resulting from second-order P moment. The member buckling resistance heavily pertains to the estimated effective length (BS5950, 2000; Eurocode 3, 2005; HKSC, 2005) or K factor (AISC-LRFD, 1993), which may be subjective and inaccurate. For instance, the effective length of a column in this simple portal frame is difficult to assess accurately, while end conditions of the column top is heavily dependent of the stiffness contribution from the connected beam. In this case, the sway effect (P effect) is considered by the sway-check method, so the nonsway mode effective length is used for this column. The stiffness distribution factors of both ends are known as given according to BS5950-2000, Ic Lc k1 = --------------------------- = 0.421 and k2=0 (for fixed bottom end) Ic Lc + Ib Lb (25) Effective length, LE=0.575L=2.875 m The use of effective length is to correlate the actual buckling resistance Pc for the ideal elastic buckling load as written in the following equations.

The magnification factor is then 1.177 as given in Eq. (24), which should then multiply with the applied loads under the horizontal load case for this simple portal frame. Under the horizontal load case as shown on Fig. 2, the second-order loading distribution from the first-order analysis is then obtained and represented by the dashed lines on Fig. 5, whose second-order magnitude is shown as the bracketed values. The above traditional design procedures for the P- effect are summarized on Fig. 4(a).
Vol. 12, No. 6 / November 2008

385

C.K. Iu, W.F. Chen, S.L. Chan, and T.W. Ma

LE 2.875 Slenderness ratio, = ---- = ------------ = 22.12 rs 0.13

(26)
2 8 1/2

Cross-section capacity,
68.38 P 2826.4 P Mx M ------- + ------- = ------- + --------- = -------------------------------------------------- + ------------ = 0.503 py A Mcx py A py Zx 4.6 10 5 1.49 102 755.1

E Limiting slenderness, 0 = 0.2 -------- Py


2

1/2

2 10= 0.2 ------------------------- 4.6 10 5

(34) Member buckling resistance, P mx Mx 2826.4 0.44 80.52 ------- + ------------ = --------------- + ---------------------------- = 0.474 p c A py Zx 6621.6 755.1 Member buckling resistance, P mx Mx 2826.4 0.44 68.38 ------- + ------------ = --------------- + ---------------------------- = 0.497 p c A py Zx 6177.5 755.1 (35)

= 13.1

(27)
2 2 8

E 2 10Elastic Euler buckling stress, pE = -------- = ------------------------2 22.12 2


= 4.034 106 kN m 2

(28)

(36)

Ayrton and Perry (1886) and Robertson (1925) presented the formulas of the member imperfection for member buckling resistance. This is called Perry factor as given in Eq. (29). When the Perry factor is zero, it is equivalent to the ideal Euler buckling resistance in which a is a parameter to adjust the amplitude of member imperfection. a ( 0 ) 3.5 ( 22.12 13.1 ) = -------------------- = ---------------------------------------- = 0.032 1000 1000 py + ( + 1 )pE 4.6 10 5 + 1.032 4.034 106 = ------------------------------- = ---------------------------------------------------------------------2 2 = 2.31 10 kN m
6 2

(29)

(30)

The compressive strength of buckling resistance with member imperfection is given,


pE p y 4.034 106 4.6 105 pc = ------------------------------------ = ------------------------------------------------------------------------------------------------------------------------- + (2 pE py )1 / 2 2.31 106 + [ (2.31 106 )2 4.034 106 4.6 105 ]1 / 2 = 4.44 105 kN m2

(31)

The member buckling load is therefore obtained as given, Pc = pc A = 1.49 10 2 4.44 105 = 6621.6kN > 2826.4kN (32) From Eq. (32), the columns in the simple portal frame are adequate to withstand the axial load without flexural buckling. The above design procedures are given in the flowchart of Fig. 4(b). The structural adequacies of interaction effect on material yielding and instability are also necessary to check as given in Eqs. (33) to (36), respectively. Alternatively, if both P- and P- effects are incorporated into the buckling resistance of column, the sway mode effective length (LE=1.175L) is used for the column to determine its buckling resistance according to Eqs. (25) to (32) similarly. The member buckling load is therefore obtained as given, Pc = pc A = 1.49 10 2 4.14 105 = 6177.546kN > 2826.4kN (33) After evaluating amplified moment due to the P- effect and buckling resistance of the column with either P- effect or both effects, the interaction effect under axial and bending action can then be checked as follows:

From Eqs. (34) to (36), the columns in the portal frame are adequate to resist external load. It should be noted that the firstorder moment should be used in interaction effect against crosssection capacity as Eq. (34), when no geometric nonlinear effects, such as P- and P- effects, are taken into consideration. In respect of Eq. (35), the first term of axial load contains the P- effect only by non-sway mode compressive strength pc, whereas, second term of bending effect includes P- effect when using second-order moment 80.52kNm from sway-check method. In Eq. (36), first term of axial load includes both P- and P- effects, so first-order moment distribution is used. The interaction factors from Eqs. (35) and (36) are closely consistent. In summary, according to the traditional design, P- effect is considered by using the amplified moment distribution. And the P- and P- effects are included by means of the member compression resistance Pc in relation to its effective length in non-sway and sway mode, respectively. In actual structural behaviour subjected to the geometric nonlinearities, both of these second-order P- and P- effects increase the moment distribution on a structure due to both second-order P and P moments, respectively. 4.3 Second-order Elastic Analysis (Simple portal frame) The second-order loading distribution is plotted on Fig. 6. No design specification for these second-order effects is thus allowed in the second-order elastic analysis. For example, there is no fictitious horizontal load and no subjective effective length in its design methodology. Therefore, the present system design method integrates the system analysis and member design for the second-order effects. According to second-order elastic analysis, the structural adequacy of interaction effects on capacity check only is expressed in Eq. (37). The design procedures from Eqs. (22) to (32) for assessing the second-order effects are therefore omitted by using second-order elastic analysis. The second-order effects are included in the moment distribution Mx in Eq. (37) instead of axial compression resistance Pc as given in Eq. (35) or (36). Cross-section capacity, P Mx M 2832.3 - 79.61 --- + ------- = ------- + --------- = -------------------------------------------------- + ------------ = 0.52 - Pp y Mcx py A py Zx 4.6 10 5 1.49 102 755.1 (37)
KSCE Journal of Civil Engineering

386

Direct Second-Order Elastic Analysis for Steel Frame Design

Fig. 7. Geometry of the Shallow Dome Skylight Structure Fig. 6. Second-order Loading Distribution on the Simple Portal Frame

In conclusion, the interaction factor in Eq. (37) by secondorder elastic analysis is 0.52, which is slightly greater than those by traditional design (member buckling resistance check as Eq. (36) or (36)). There is only at most 9% difference between traditional design and second-order elastic analysis. It implies that the design by second-order elastic analysis is slightly less conservative in order to provide a somewhat lesser safety margin.

which are the same 660350 mm built-up section. The system analysis and design (Integrated method) of this complicated skylight was carried out by nonlinear analysis NIDA (2007). 5.1 Design Example Traditional Design Approach In the procedures of traditional design methods (AISC-LRFD, 1993; BS5950, 2000; Eurocode 3, 2005), the linear elastic analysis is conducted to determine the first-order loading distribution on the structure. The first-order elastic loading distribution of two critical members in this skylight structure is shown in
Table 1. Load distributions from linear analysis (1.35DL+1.5LL +WL+0.9TL) Member Arch beam Bracing Interaction factor -1.0186 -0.6910 Node 1 2 1 2 Mx (kNm) My (kNm) 562.0 -493.0 -5.1 -121.0 -41.1 -113.0 27.8 17.7 P (kN) -1720 -3620

5. Design Example (Skylight Structure)


A practical design of a large-scale complicated structure was carried out by elastic design. If the structural inadequacy of a member is found as determined by using Eq. (37), the strengthening work of the member section is compulsory. This practical design example was a steel skylight frame in Macau. This steel frame is 11.8 m height and 62 m length and has a shallow arch dome shape with a 57 m span, which is rested by nine pinned supports as depicted on Fig. 7. The cladding is mounted on the longitudinal members with rectangular hollow section, which are supported by main arch beams with built-up I-section. Their sections are also displayed on Fig. 7. The main arch beams also carry the loadings of lightings and utilities, and transfer all these loadings to the supports. The bracings on the verges made of built-up I-section play as stability for whole skylight structure. Fig. 7 shows that there are two critical members by the bold lines, which attempt to be checked by both traditional design and second-order elastic analysis for comparative purpose. One is the bracing and the other is the end segment of the main arch beam,
Vol. 12, No. 6 / November 2008

Table 2. Load distributions from advanced analysis (1.35DL+1.5LL +WL+0.9TL) Member Arch beam Bracing Interaction factor -1.2405 -1.1105 Node 1 2 1 2 Mx (kNm) My (kNm) 700.0 -605.0 -50.3 -156.0 -45.5 -131 -26.2 63.9 P (kN) 700.0 -3640

387

C.K. Iu, W.F. Chen, S.L. Chan, and T.W. Ma

Table 1. Interaction factor of main arch beam is only slightly greater than unit, because no second-order effect is considered in the loading distribution in Table 1. It is remarked that when the section capacity factor is negative on Table 2, it implies that the member is generally critical to member buckling due to compression. For P- effect, the skylight structure does not meet the geometrical requirements of sway-check method, so amplified moment method may be used for this shallow dome structure. The critical load factor cr of the first buckling mode evaluated from NIDA (2007) is 0.269. It indicates that, because of its large clear span, this skylight structure is so flexible that the traditional design is not suitable to allow for the P- effect on this structure. For P- effect, the structural adequacies of both main arch beam and bracing may be carried out on the basis of the design specifications. However, there is no design specifications (AISCLRFD, 1993; BS5950, 2000; Eurocode 3, 2005; HKSC, 2005) suitable for main arch beam owing to its peculiar geometry. Because of its end conditions, the effective length of bracing is also difficult to assess in which the approximation and subjective judgment is involved. Further, the stiffness contribution of connected members to the bracing is unknown on the basis of traditional design, which heavily influences the buckling load Pc of bracing. In addition, the buckling load Pc of the bracing is only meaningful, if the sway effect of the whole structure is known. After incorporating the second-order P- effect, the interaction factor of bracing increases from 0.691 to 1.032, which is obtained by the standard design procedure of BS5950-2000 and neglects P- effect. In conclusion the traditional design method may not be suitable to the uncommon and complicated structure. Further, the procedure of traditional design is tedious for a large-scale complicated structure. 5.2 Design Example Second-order Elastic Analysis The loading distributions on two critical members are indicated on Table 2 under a critical load combination by using secondorder elastic analysis, which numerically and directly evaluates both second-order P- and P- effects. By compared with Tables 1 and 2, the moment distribution from second-order elastic analysis is greater than those from linear analysis, because the present analysis takes the second-order moments due to P- and P- effects into account. On the other hand, the axial forces in Tables 1 and 2 on the members are similar. Since the second-order effects are considered in the loading distribution based on present analysis, the structural capacity adequacy of the member should be checked for its interaction effects by using the following criterion similar to Eq. (38). Mx My P Cross-section capacity, --------- + ------- + ------- 1 py Ag Mcx Mcy (38)

whole. Therefore, for some complicated cases, the effective length and magnification factor of the member are difficult or impossible to obtain according to traditional design.

6. Conclusions
According to the traditional design method, the empirical design parameters, such as magnification factor and effective length, are imperative to simulate the second-order P- and P- effects indirectly, respectively. In addition, the restrained conditions, member length, geometry and loading of every member may not be same, so these empirical design parameters may be applied for all members differently. As a result, the design procedure or manual calculation of traditional design is separately implemented for member-by-member stability check inevitably. On the other hand, a second-order elastic analysis can numerically and directly calculate the loading distributions and deformations with inclusion of both P- and P- effects on a structural system. Consequently, no design parameter is needed for the member. It implies that there is no member-by-member stability design procedure. All the onerous and cumbersome manual calculations for second-order effects are implemented directly in the numerical process of second-order elastic analysis, which is mentioned in Sections 2 & 3. It makes a full utilization of the computing technology in the present age. As a result, the structural design based on second-order elastic analysis is computer-friendly and computer-based approach. Also this system design approach is convenient, versatile and adaptable for different structural forms, and even for a complicated large-scale structure. It leads to an efficient and competitive design process and non-labour-intensive, and reduce the cost of the design subsequently, even the complicated large-scale structure.

Acknowledgements
The authors would like to give a gratitude to the financial supports from the University of Hawaii at Manoa to accomplish this research work of system analysis for elastic steel framed structures.

References
AISC-LRFD (1993). Load and resistance factor design, manual of steel construction, Vol. 1 and 2, 2nd edition, American Institute of Steel Construction, AISC, Chicago, III. Ayrton, W.E. and Perry, J. (1886). On struts. The Engineer, Vol. 62, pp. 464. Bathe, K.J. (1982). Finite element procedures in engineering analysis, Prentice-Hall, Englewood Cliffs, NJ. Batterman, R.H. and Johnston, B.G. (1967). Behaviour and maximum strength of metal columns. Journal of the Structural Division, ASCE, Vol. 93, No. 2, pp. 205-231. Bjorhovde, R., Galambos, T.V., and Ravindra, M.K. (1978). LRFD criteria for steel beam-columns. Journal of the Structural Division, ASCE, Vol. 108, No. 9, pp. 1371-1387.
KSCE Journal of Civil Engineering

in which Mcx and Mcy are the section capacity of the member. The cross-section capacity check for every member in an entire structure can be carried out directly in the system analysis as a

388

Direct Second-Order Elastic Analysis for Steel Frame Design

Bleich, F. (1952). Buckling strength of metal structures, New York, McGraw Hill, Chaps. 6-7. Bridge, R.Q. and Fraser, D.J. (1987). Improved G-factor method for evaluating effective lengths of columns. Journal of Structural Engineering, ASCE, Vol. 113, No. 6, pp. 1341-1356. BS449 (1969). Specification for the Use of Structural Steel in Building Part 2. BS5950 (2000). Part 1, Structural Use of Steelwork in Building: Code of Practice for Design Rolled and welded sections. Chan, S.L. and Kitipornchai, S. (1987). Geometric nonlinear analysis of asymmetric thin-walled beam-columns. Engineering Structures, Vol. 9, pp. 243-254. Chan, S.L. and Zhou, Z.H. (1998). On the development of a robust element for second-order nonlinear integrated design and analysis (NIDA). Journal of Constructional Steel Research, Vol. 47, No. 12, pp. 169-190. Chen, W.F. and Lui, E.M. (1986). Structural stability Theory and implementation, Elsevier, New York. Chen, W.F. and Kim, S.E. (1997). LRFD steel design using advanced analysis, CRC Press, New York. El-Zanaty, M.H. and Murray, D.W. (1983). Nonlinear finite element analysis of steel frames. Journal of Structural Engineering, ASCE, Vol. 109, pp. 353-368. Euler, L. (1759). Sur la force de colonnes, Memoires de lacademie de, Berlin. Eurocode 3 (2005). Part 1-1, Design of Steel Structures: General rules and rules for buildings. European Convention for Constructional Steelwork (ECCS) (1984). Ultimate limit state calculation of sway frames with rigid joints. Publication No. 33, Technical Committee 8 Structural Stability Technical Working Group 8.2, Brussels, Belgium. European Convention for Constructional Steelwork (ECCS) (1991). Essentials of Eurocode 3 design manual for steel structures in building, Publication No. 65, ECCS Advisory Committee 5, Brussels, Belgium. Goto, Y. and Chen, W.F. (1987). Second-order elastic analysis for frame design. Journal of Structural Engineering, ASCE, Vol. 113, No. 7, pp. 1501-1519. Hancock, G.J., Papangelis, J.P., and Clarke, M.J. (1995). PRFSA users manual, Centre for Advanced Structural Engineering, University of Sydney. Harrison, H.B. (1973). Computer methods in structural analysis, Prentice-Hall, Englewood Cliffsm, New Jersey. HKSC (2005) Code of Practice for the Structural Use of Steel, Hong Kong. Ioannidis, G.I. and Raftoyiannis, I.G. (2005). A simplified nonlinear stability analysis of an imperfect rectangular two-bar frame. Computational Mechanics, Vol. 35, pp. 127-133. Iu, C.K. and Chan, S.L. (2004). A simulation-based large deflection and inelastic analysis of steel frame under fire. Journal of Constructional Steel Research, Vol. 60, pp. 1495-1524.

Kanchanalai, T. (1977). The design and behaviour of beam-columns in unbraced steel frames, AISI Project No. 189, Report No. 2, Civil Engineering/ Structures Research Laboratory, University of Texas at Austin, Austin. Kim, S.E. and Chen, W.F. (1996a). Practical advanced analysis for braced steel frame design. Journal of Structural Engineering, ASCE, Vol. 122, No. 11, pp. 1266-1274. Kim, S.E. and Chen, W.F. (1996b). Practical advanced analysis for unbraced steel frame design. Journal of Structural Engineering, ASCE, Vol. 122, No. 11, pp. 1259-1265. King, C. (2000). In-plane Stability of Portal Frames to BS5950-1, The Steel Construction Institute. Meek, J.L. and Tan, H.S. (1984). Geometrically nonlinear analysis of space frames by an incremental iterative technique. Computers Methods in Applied Mechanic and Engineering, Vol. 47, pp. 261282. Merchant, W. (1954). The failure load of rigid joined frameworks as influenced by stability. The Structural Engineer, Vol. 32, pp. 185190. NIDA (2007). Non-linear Integrated Design and Analysis (NAF-NIDA) version 7. Oda, H. and Usami, T. (1997). Stability design of steel plane frames by second-order elastic analysis. Engineering Structures, Vol. 19, No. 8, pp. 617-627. Robertson, A. (1925). The strength of struts. ICE Selected Engineering Paper, Vol. 28. Timoshenko, S.P. and Gere, J.M. (1961). Theory of elastic stability, McGraw-Hill, New York. Vogel, U. (1985). Calibrating frames. STAHLBAU, Vol. 1, pp. 295301. Wood, R.H. (1974a). Effective lengths of columns in multi-storey buildings Part I. The Structural Engineer, Vol. 52, No. 7, pp. 235244. Wood, R.H. (1974b). Effective lengths of columns in multi-storey buildings Part II. The Structural Engineer, Vol. 52, No. 8, pp. 295302. Wood, R.H. (1974c). Effective lengths of columns in multi-storey buildings Part III. The Structural Engineer, Vol. 52, No. 9, pp. 341346. Wood, B.R., Beaulieu, D., and Adams, P.F. (1976a). Column design by P-delta method. Journal of the Structural Division, ASCE, Vol. 102, No. 2, pp. 411-427. Wood, B.R., Beaulieu, D., and Adams, P.F. (1976b). Further aspects of design by P-delta method. Journal of the Structural Division, ASCE, Vol. 102, No. 3, pp. 487-500. Ziemian, R.D., McGuire, W., and Dierlein, G.G. (1992a). Inelastic limit states design part I: Planar frame studies. Journal of Structural Engineering, ASCE, Vol. 118, No. 9, pp. 2532-2549. Ziemian, R.D., McGuire, W., and Dierlein, G.G. (1992b). Inelastic limit states design part II: Three-dimensional frame study. Journal of Structural Engineering, ASCE, Vol. 118, No. 9, pp. 2550-2568.

Vol. 12, No. 6 / November 2008

389

Das könnte Ihnen auch gefallen