Вы находитесь на странице: 1из 13

Engineering Structures 22 (2000) 155167 www.elsevier.

com/locate/engstruct

Stability of steel frames: the cases for simple elastic and rigorous inelastic analysis/design procedures
Donald W. White
a

a,*

, Jerome F. Hajjar

School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0355, USA b Department of Civil Engineering, University of Minnesota, Minneapolis, MN 55455-0220, USA Received 20 August 1997; accepted 9 January 1998

Abstract In this paper, two stability design approaches that are alternatives to traditional buckling-solution based procedures are outlined and their merits are discussed. The rst approach, which is commonly referred to as a notional load type procedure, is effectively a modied rst plastic hinge technique. This approach provides a simple elastic analysis based solution for the stability design of steel frames. The second approach, termed an advanced analysis design procedure, involves the use of a rigorous second-order inelastic analysis as part of the design assessment. Several examples are provided to illustrate the implications of these alternative procedures relative to the use of traditional buckling solution based design equations. 1999 Elsevier Science Ltd. All rights reserved.
Keywords: Advanced analysis; Effective length; Inelastic redistribution; Modied rst-plastic hinge design; Notional load; Stability design; Steel structures

1. Introduction Steel building structures are often designed using planar unbraced moment frames for the lateral-load resisting systems along with a signicant number of gravity columns throughout the structure. Fullyrestrained or partially-restrained connections may be employed within the lateral-load resisting frames, whereas oor girders are connected into the gravity columns using only simple shear connections. The gravity columns are assumed not to contribute to the structures lateral resistance, and the lateral-load resisting systems must be designed to insure that they provide adequate sidesway stability to these leaning columns. In many cases, braced bents or structural walls might be employed in one direction within the oor plan, and correspondingly, all the columns are assumed to be braced at the oor levels in that direction. However, it is common that the structure may be unbraced (sideswayuninhibited) in the orthogonal direction. Furthermore, it

* Corresponding author. Tel.: 001-404-894-5839; Fax: 001-404894-0211

is common in current practice that sidesway stiffness and/or out-of-plane strength may control the design of the beamcolumns in the lateral-load resisting sway frames, rather than in-plane strength [1]. As a result, engineers often initiate their design of these frames by proportioning the girders and beamcolumns to satisfy total sidesway and/or interstory drift limits at service loads [2,3]. Approximate optimization techniques, e.g., Ref. [4], are useful for obtaining well balanced, adequately stiff framing systems, which must then be checked for strength using the AISC Specications [5,6] or other standards. Within the vast literature on stability design of frames, much has been written about determining the in-plane strength of various types of critically loaded frames in which the beamcolumns are assumed to be fullybraced in the out-of-plane direction. While this work is important, typically much less attention has been devoted to accurate and simple design calculation of outof-plane strengths. Furthermore, established design procedures for checking in-plane strength often are tedious and prone to error. For instance, the alignment chart based procedures [5,6], which are used widely by engineers within the US, often require numerous adjustments.

0141-0296/00/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 1 - 0 2 9 6 ( 9 8 ) 0 0 1 0 5 - 9

156

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

These include modications to account for sidesway inection points not being located at the mid-span of the girders, the fact that the stronger framed columns tend to stabilize the weaker ones, and various other attributes of the stability behavior. Mistakes in adjusting alignment chart K factors to account for the physical characteristics of a frame often are on the unconservative side [7]. If simpler, less tedious procedures can be identied that give comparable accuracy in their assessment (particularly of the in-plane structural response), such approaches deserve careful consideration. Again, it is important to keep in mind that the in-plane strength of sway frames often does not control the design in the rst place. Baker [8] has summarized one such approach that is based on determination of elastic sidesway buckling loads using rst-order elastic drift calculations. Bakers approach is one of several methods outlined in the AISC LRFD Commentary [6], and it is a subset of a more general elastic/inelastic procedure detailed by White and Hajjar [9]. This approach, which White and Hajjar term the Pe(RL) method, offers major advantages in simplicity and accuracy over traditional procedures for evaluating the stability of unbraced frames. However, both the Pe(RL) approach and the more general elastic/inelastic version of this approach require that the designer must check for extreme cases of very light, very heavilyloaded individual columns, to insure against signicant loss of accuracy. Checks for these extreme cases are also required by the alignment-chart based procedures, and equations that implement these checks are provided in the AISC LRFD Commentary [6] and are discussed in Ref. [9]. Furthermore, the Pe(RL) and alignment chart based methods can falsely penalize lightly-loaded columns that are actually not providing a signicant contribution to the story sidesway stability. The engineer must use judgement to override the design assessment in these cases [9]. In Ref. [10], the merits and limitations of designing based on second-order elastic analysis with a rst-plastic hinge check are discussed in detail. In this paper, the authors would like to explain the motivation and logic behind a simple modication of the rst-hinge concept, to alleviate the limitations of that approach. This modied rst-hinge approach involves: (1) replacing the section full-plastication strength associated with traditional elastic-plastic hinge analysis with member strength interaction equations in which the axial strength is based on the actual member unsupported length; and (2) application of notional horizontal loads at each oor level of a building structure. The strength checks are based only on individual member properties, and therefore, they are more easily implemented than buckling-based checks. Based on its ties to a traditional rst-hinge procedure, this approach eliminates the need to assess the inuence of neighboring framing on member strength in calcu-

lations that are separate from the analysis itself (in contrast, the buckling-solution based methods require separate calculations that assess the inuence of adjacent framing on the member strength). It is assumed that the system strength is predicted adequately as the load level associated with the member strength being reached in the most critically loaded member. Several examples are provided here to highlight key advantages and limitations of the resulting method. This approach has been referred to in the recent literature [7,912] as an AISC LRFD-based notional load design procedure (by virtue of its use of notional lateral loads). However, the authors suggest that the term modied rst hinge (MFH) is more descriptive of the characteristics of the procedure. Therefore, the term modied rst hinge (MFH) will be used predominantly in the following discussions. It should be emphasized that the MFH approach discussed in this paper is exactly the same as the AISC LRFD based notional load procedure discussed in Refs. [7,9,10,12,11]. The main limitation of the MFH and Pe(RL) approaches is that they ignore the inelastic reserve strength that may exist within a structure. In certain cases, it can be benecial to consider at least some of the inelastic reserve capacity in order to achieve design exibility and/or economy. Therefore, the authors also discuss the potential use and implications of advanced three-dimensional second-order inelastic analysis procedures within a design context. In advanced analysis/design, a secondorder inelastic analysis is conducted for all the relevant load combinations. The analysis formulation must be sufciently rened to account comprehensively for all key behavioral characteristics implicit within the specication member capacity equations. A structure that is then proportioned to support the design loads, without detection of a structural failure in the analysis, implicitly satises the intent of the specication beamcolumn limit states provisions. Separate checking of the corresponding limit states equations, e.g., the beamcolumn interaction formulas, is not necessary. This can allow the engineer to focus on the design of a structure for stiffness, followed by a simple automatic check of the structure for strength. Also, an advanced analysis/design approach can provide signicantly more information about the idealized structural performance at service and factored load levels than elastic analysis/design procedures. This additional information can be valuable in many situations, and can be considered with little additional design effort if software is available to help the engineer rapidly synthesize and interpret the analysis results.

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

157

2. Elastic design approach based on modication of the rst-plastic hinge concept (notional load approach) 2.1. In-plane strength of frames composed of compact sections in compression and strong-axis bending It is well known that, with the exception of small nonredundant benchmark frames, elastic-plastic hinge procedures1 generally provide a sufcient design estimate of the strength of frames composed of compact-section members, fully-braced out-of-plane, which fail in an inplane sidesway mode [10,1315]. The AISC LRFD beamcolumn interaction curve in the limit of zero member length, shown in Fig. 1, can be used as a good approximation of the cross-section full plastication strength for frames of this type in which the members are subjected to strong-axis bending and axial force [13]. This curve may be factored by c 0.85 on the vertical axis and b 0.9 on the horizontal axis, as shown in the gure, to account for uncertainties in the resistances [14]. If the engineer wishes to perform an elastic analysis and design with these procedures, the analysis can be stopped at the onset of the rst plastic hinge, and this point can be accepted as the strength of the structural system. As noted above, the elastic-plastic hinge procedures generally work quite well in predicting the sidesway system strength of frames composed of fully-braced compact section members. This is due in part to the fact that beamcolumns in sidesway moment frames are often

Fig. 1. Modication of section full-plastication strengths to account for reduced member strengths.

stiff enough, due to service drift considerations, such that their maximum second-order moments are located at the member ends. However, if certain members are light enough and are subjected to large axial forces, the maximum second-order moments in these members can occur within the member length. These types of members can potentially fail in a non-sway mode that often is not predicted with sufcient accuracy by elastic-plastic hinge models. The problem lies in the fact that elastic-plastic hinge models do not account for the detrimental effects of distributed plasticity within an imperfect beamcolumn due to residual stresses and applied loadings. Therefore, for a rst-hinge based design procedure to be able to assess the in-plane strength of any compact section frame, this approach must be modied in some way to capture potential non-sway member failure modes. For strong-axis in-plane bending, the potential non-sway failure of light, heavily-loaded frame members, can be captured conservatively by using the column strength based on the actual member unsupported length, Pn(L), in the AISC LRFD beamcolumn interaction formula rather than the cross-section yield load, Py. This reduction in the strength along the vertical axis in Fig. 1 accounts for the detrimental effects of both initial geometric imperfections (i.e., member out-of-straightness) and distributed yielding along the member length [12]. Overall out-of-plumbness within the erection tolerances permitted by the AISC Code of Standard Practice [6] generally can have a non-negligible effect on the inplane strength of frames that fail in a sidesway mode. In the context of a rigorous second-order inelastic analysis that accounts for distributed yielding, and for frames of up to about nine stories in height, it is sufcient to specify the maximum AISC out-of-plumb tolerance on a single shipping piece ( o L/500) for all the columns within the analysis model to account for these effects [10,16].2 However, in the context of elastic-plastic hinge analysis, this uniform out-of-plumbness needs to be increased to account approximately for possible amplication of the second-order sidesway forces within the system due to distributed plasticity effects. Liew et al. [13] studied the in-plane behavior of a large number of sensitive benchmark frames and concluded that an elastic-plastic hinge analysis with an exaggerated out-ofplumbness of o L/200 is sufcient to account for the combined effects of distributed plasticity and a real outof-plumbness of o L/500 on the sidesway strength of steel frames composed of compact section members. Similar conclusions have been reached in earlier European work [17]. Generally, a frame would need to be

1 In an elastic-plastic hinge analysis approach, the member crosssection is assumed to be completely elastic until its full plastic capacity is reached. At this stage, the cross-section response is assumed to become perfectly plastic (i.e., no strain hardening). The inelastic behavior is approximated by zero-length plastic hinges, and the members are modeled as fully elastic elements between the hinge locations.

2 For taller frames, specication of a uniform o L/500 over the entire building height would tend to be conservative, since this would signicantly exceed the envelope within which column line working points must fall, specied in Ref. [6].

158

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

analyzed for this out-of-plumbness in both the positive and negative horizontal directions within its plane. Clarke and Bridge [11] have shown that the effects of the above uniform o L/200 on the second-order moments can be duplicated by applying lateral loads in the direction of the out-of-plumbness at each of the oor levels of Hnl Q/200, where Q is the total factored gravity load stabilized by the lateral-load resisting frame at a given oor. The forces Hnl must be added to any factored lateral loadings that the structure is being designed for. The resulting procedure is referred to as an AISC LRFD-based notional load approach in Ref. [7]. In Ref. [7], it is developed by calibrating to inelastic buckling-solution based column strengths for sensitive gravity loaded structures. However, this method can be viewed more intuitively as a simple modication of the rst-plastic hinge approach to account approximately for potential non-sway member failures (by using Pn(L) instead of Py in the beamcolumn rst-hinge strength checks) and increased sidesway deections and forces due to distributed yielding at factored load levels (by giving the frame an additional horizontal kicker force at each of the oor levels of Hnl Q/200). The term modied rst-hinge (MFH) is more descriptive of the characteristics of this approach than the commonly used notional load term. 2.2. In-plane strength of frames composed of compact sections in weak-axis bending and axial compression For W shapes, the normalized full-plastication strength is signicantly more convex for weak-axis inplane bending and axial loading than for strong-axis cases. This is illustrated in Fig. 1 by comparison of the light-grey curve, which is a typical idealization of the weak-axis full plastication strength of a wide-ange section (factored by c 0.85 and b 0.9), to the solid black AISC LRFD beamcolumn interaction curve for L 0. Nevertheless, it is well documented that the maximum second-order forces in beamcolumns loaded in weak-axis bending often cannot reach the cross-section full-plastication strength prior to failure of the member [7]. This is due to the fact that partial yielding at the ange tips dramatically reduces the weak-axis exural stiffness of W shapes. Fortunately, it turns out that the interaction curve based on Pn(L) in Fig. 1 can be used quite successfully to account for the maximum resistance of such a member, either in a sway- or a nonsway mode of failure [12]. That is, the member strength shown in Fig. 1 can be used effectively for both strongand weak-axis bending checks. Also, this parallels the current AISC LRFD buckling-solution based approach, which species the same interaction curve for both strong- and weak-axis cases. A similar approach was suggested by Liew et al. [13] to capture the sidesway strength of frames composed of beamcolumns in weak-

axis bending, but Liews approach used Py for the axial capacity. Use of Pn(L) for this capacity captures conservatively any potential non-sway failure modes in light, heavily loaded members. 2.3. Out-of-plane strength of compact beamcolumns subjected to in-plane strong-axis bending Generally, a beamcolumn member loaded by strongaxis bending and axial compression within the plane of the frame may fail by out-of-plane lateral-torsional buckling. The rst-plastic hinge approach can be modied to accommodate this type of situation by calculating Pn(L) in Fig. 1 as the smaller of the column strengths associated with the unsupported lengths about the strong and weak axes. Furthermore, due to lateral-torsional buckling, the bending strength in the absence of axial loading, Mn, can in general be smaller than the plastic moment strength Mp, although for many building columns in reverse curvature bending, Mn is equal to Mp in AISC LRFD. It is important to note, however, that for typical column-type wide-ange sections (i.e., d bf) in multistory frames (i.e., sections subjected to reversed-curvature bending), rigorous second-order inelastic analyses generally show that an out-of-plane failure does not govern the strength unless the axial force is close to the column axial capacity in the weak-axis direction [16]. Consequently, the AS4100 Standard [18] takes a different approach that involves a separate out-of-plane strength check for these types of cases. The combination of the AS4100 out-of-plane strength equation with a separate in-plane strength interaction curve gives a more accurate match with advanced analysis and experimental results than the above traditional single-equation approach, although with some increase in the complexity of the beamcolumn strength assessment [16,19]. This combination of out-of-plane and in-plane interaction curves is illustrated in one of the subsequent examples. The reader is referred to Ref. [12] for more detailed discussions. 2.4. Strength of frames composed of compact sections loaded in compression and biaxial bending For general three-dimensional framing systems in which certain beamcolumns may be loaded spatially, the strength interaction effects can be handled conservatively by additively combining the moment terms of the member strength curve shown in Fig. 1 for both strong and weak-axis bending. It is well known that the actual strength of biaxially-loaded beamcolumn members is somewhat larger than the resulting predicted capacities [10]. However, this additional strength is not recognized for sway frames in current AISC LRFD practice. The resulting modied rst-hinge procedure is similar to the current AISC LRFD design approach, except the con-

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

159

trolling column strength about the strong or weak axis of the section is based on the actual unsupported length(s), and the second-order inelastic forces within the structure at the factored load state are approximated by including notional horizontal loads within the analysis. Detailed three-dimensional studies with this approach have not been performed to date. However, it is expected that the use of an exaggerated out-of-plumbness in two orthogonal directions of a building structure, or the equivalent simultaneous application of notional horizontal loads, should produce acceptable solutions. 2.5. Strength of frames containing non-compact members The above approach can be extended to address frames containing members of non-compact cross-section by modifying the terms Pn(L) and Mn to account for the effects of local buckling on the strength. Again, this parallels the current buckling-solution based procedures specied in the AISC LRFD Specication [6].

3. Design based on advanced analysis Recent research on second-order inelastic analysis of steel frames, e.g., Refs. [20,21], has resulted in efcient nite elements that are capable of capturing member three-dimensional inelastic stability behavior quite accurately with only one or a few elements per member. This research has been motivated in large part by discussions by Springeld [22], McGuire [23] and others, who have pointed out that until advanced analysis can capture torsional-exural stability behavior more reliably, it will have no signicant impact. The key means behind these research advancements include: (1) interpolation of stress-resultant forces along the member length as part of the basis of the element formulation, (2) expression of the gradual yielding behavior at a cross-section in terms of a cross-section stress-resultant constitutive model, and (3) efcient integration of the cross-section inelastic behavior along the element length. By basing the formulation on interpolation of forces, distributed plasticity effects can be modeled accurately using only a few elements compared to displacement-based approaches, which have difculty in matching sharp variations in inelastic curvatures along the length of a member. It can be argued that a distributed plasticity approach is essential to be able to capture beamcolumn inelastic torsional-exural stability effects with sufcient accuracy [16]. By use of a cross-section stress-resultant plasticity model, the element calculations become nearly as efcient as those of elastic-plastic hinge models, while providing the rigor of directly accounting for the distributed plasticity effects. Continued research is still needed for tuning and

renement of these techniques, but the results to date are promising. If the distributed plasticity behavior can be modeled rigorously and directly without signicant computational expense, it would seem that direct distributed plasticity modeling would be preferable over simpler representations of the inelasticity. Each of the approaches cited above are based on Vlasov kinematics, i.e., plane sections remain plane and normal to the axis of the member with the exception of warping deformations normal to the cross-section associated with nonuniform torsion, and the shape of the cross-section is assumed to remain undistorted. Therefore, capturing of local buckling and/or cross-section distortion effects is precluded. This implies that inelastic deformation limits must be checked, but it is not expected to restrict the application to hot-rolled steel construction by any signicant degree. White and Nukala [16] have attempted to synthesize much of the recent technical discussions that have taken place in the literature and among several task groups of the Structural Stability Research Council regarding advanced analysis and design. They outline in detail one possible way in which the above distributed-plasticity types of models can be applied within a design context. Based on studies with distributed plasticity models that have been conducted to date, e.g., Refs. [11,21], they recommend that general advanced analysis/design solutions for frames composed of rolled wide-ange shapes can be obtained by basing the analyses on: (1) the Galambos and Ketter [24] self-equilibrating residual stress pattern, which involves a linearly varying residual stress in the anges from 0.3Fy at the ange tips to a constant residual tension in the web, (2) a sinusoidal member out-of-straightness of o L/1000, specied in a direction that produces P moments that are additive with member primary moments, and (3) a uniform outof-plumbness of o L/500 applied to the structure in mutually orthogonal directions. White and Nukala [16] have shown that for certain simple cases, these imperfections produce nominal beamcolumn strengths that are practically identical with current AISC LRFD equations. Specication of the above uniform out-of-plumbness within an analysis model is reasonably straightforward. However, specication of an out-of-straightness within all the beamcolumns of a large structure generally would be quite cumbersome. Fortunately, out-ofstraightness effects should be negligible in most of the beamcolumn members of practical sway frames. White and Nukala [16] recommend that member out-ofstraightness effects can be neglected for steel wideange members whenever Pu/Pe 1/7, where Pu is the maximum factored axial load in the beamcolumn and 2 EI/L2. This limit is somewhat more restrictive Pe than a similar limit of Pu/Pe 1/4 for which Eurocode 3 [25] species that out-of-straightness effects do not need to be considered in an analysis.

160

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

Based on the fact that nominal column strengths can be predicted with greater precision by the above advanced analysis models, compared to the use of a single column strength curve, White and Nukala [16] suggest that a uniform resistance factor of 0.9 may be appropriate for determination of design resistances with these methods. For simple column benchmark problems, the resulting strong-axis column design strengths (including ) are slightly more liberal than in current LRFD practice, whereas because of the more signicant reductions in exural stiffness with yielding at the ange tips, the resulting weak-axis column design strengths are slightly more conservative. The differences between the advanced analysis based strong- and weak-axis column strengths are similar to the differences between these strengths predicted by multiple column curve formulas in Eurocode 3 [25]. White and Nukala [16] recommend that, within the context of an advanced analysis, the above uniform 0.9 can be moved to the load side of the limit states checks such that the analyses can be conducted with nominal attributes for all quantities. The structure is then required to support 1/0.9 of the factored load combinations.

Fig. 2.

Unsymmetric two-story two-bay frame.

4. Illustrative examples The implications of the suggested elastic analysis/design and advanced analysis/design procedures can be illustrated more clearly by considering a few simple design case studies. In the following subsections, the application of these procedures to a two-story two-bay frame that has been utilized in a number of other related studies [7,9] is presented rst. This is followed by portal frame and single member examples, which serve to highlight how the different methods capture key attributes of the frame behavior that can be difcult to understand in large and/or complex problems. A common characteristic of all these examples is that substantial distributed plasticity is encountered in at least certain portions of the structure. This characteristic helps to verify the applicability of the elastic design approaches as well as to demonstrate the benets of advanced analysis/design. 4.1. Unsymmetric two-story two-bay frame The two-story two-bay rigidly-connected frame shown in Fig. 2 is a modied version of one investigated by Ziemian et al. [14]. Some of the calculations for this frame are demonstrated in Ref. [7], and detailed design solutions for ve different elastic design procedures including the Pe(RL) and notional load methods are presented in Ref. [9]. In this frame, the girders are assumed to remain elastic throughout all the loadings. This facilitates the understanding of the different procedures by eliminating the issue of inelastic redistri-

bution and reserve strength associated with plastic hinging in these members, i.e., the example focuses on proper capturing of beamcolumn stability effects. Also, the frame is subjected to large gravity loads, to accentuate these effects. The elastic and advanced analysis procedures described in the previous sections are applied to this frame, with the exception that the ECCS [17] residual stress pattern, in which the residual stresses vary linearly between 0.5Fy in the web and in the anges, is specied for the advanced analysis. Strain hardening is neglected in the advanced analysis except that a small tangent modulus equal to 700 MPa is assumed within the yielded regions. A sinusoidal out-of-straightness of L/1000 is modeled in all the columns, although the recommendations by White and Nukala [16] suggest that out-of-straightness does not need to be considered for any of the members of this frame. These imperfections are specied such that the associated P effects are additive with the maximum primary bending moments. All the frame members are compact and are assumed to be fully braced in the out-of-plane direction to simplify the discussions. Fig. 3 shows the moments at the tops of the bottom story columns versus the fraction of the total factored gravity load, , obtained from two separate advanced analyses of the structure. The dashed curves correspond to an initial out-of-plumbness of the frame of o L/500 to the right, and the solid curves correspond to the same initial out-of-plumbness to the left. It can be seen that the direction of the out-of-plumbness has a signicant effect on the maximum strength for this structure. The frame is able to sustain 1.50 times the factored design load if it is out-of-plumb to the right, while collapse occurs for the idealized structure at 1.29 times the factored load level for an out-of-plumbness to the left. If a uniform of 0.9 is applied to these results, a design

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

161

Fig. 3. Moments at the tops of the bottom story columns versus the load parameter .

Fig. 4. Axial force versus moment interaction at the tops of the bottom story colums frame out-of-plumb to the right.

capacity of 0.9(1.29) 1.16 times the factored load level is obtained. The axial force-bending moment load paths (P/Py versus Mtop/Mp) at the tops of the bottom story columns, obtained from both the modied rst-hinge and the advanced analyses, are shown in Figs. 4 and 5. These gures correspond to an initial out-of-plumbness to the right and to the left, respectively. The advanced analysis

load paths are shown by the solid black lines whereas the load paths for the elastic analysis/design procedure are shown by the solid light-grey lines. In addition to the load paths, the cross-section initial yield and full plastication strengths, and the modied member strengths (based on the adjustment of the axial strength from Py to Pn(L)) are shown. All the results in these gures are plotted in terms of nominal values (i.e., no factors included). The actual points on the elastic analy0.85 along sis load paths are therefore divided by c the vertical axis and b 0.9 along the horizontal axis to obtain a valid comparison with the nominal strength curves and advanced analysis results. In both Figs. 4 and 5, it can be observed that the modied rst-hinge approach tends to overestimate the portion of the moments due to sidesway in the advanced analysis as long as the frame is predominantly elastic. The sidesway contributions to the moments at the tops of columns c2 and c3 are signicantly overpredicted at the factored design load level, while the sidesway contribution to the moment at the top of column c1 is slightly overpredicted3 (the force states associated with the design load level are indicated by the open circles in the gures). However, as the columns begin to yield substantially, and the bottom story of the frame begins to fail in sidesway, the advanced analysis moments change rapidly. As a result, the rst-hinge analysis moment in column c2 at the factored load level is essentially equal to this columns advanced analysis moment at the collapse of the story, both for an out-of-plumbness to the left and to the right. This is signicant since this column provides the largest contribution to the sidesway stiffness and strength of the bottom story. The beamcolumn capacity check is controlled for c2 by the rst-hinge analysis forces associated with an out-of-plumbness to the left. The controlling factored-load moments in columns c1 and c3 are also predicted conservatively by the elastic procedure, with out-of-plumbness to the left contributing to the controlling force for c1 and out-ofplumbness to the right being the worst case for c3. The implication of the notional load calibration in Ref. [7] is that if the factored loads correspond to collapse, the member forces within the structural system at this state will be predicted adequately. More detailed discussions of this behavior are provided in Ref. [9]. It is useful to assess how the modied rst-hinge and the advanced analysis results compare to the more conventional, but quite useful, Pe(RL) approach for the two-

Fig. 5. Axial force versus moment interaction at the tops of the bottom story colums frame out-of-plumb to the left.

3 In the advanced analysis results shown in Figs. 4 and 5, the column moments are predominantly due to the gravity loading as long as the frame is mostly elastic. The rapid changes in the moments as the maximum strength is approached are due to the sidesway deformations associated with the failure of the structure.

162

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

story two-bay frame. Fig. 6 shows the load paths and beamcolumn interaction curves for the buckling-solution based Pe(RL) approach in the format of Figs. 4 and 5. The load paths in Fig. 6 correspond to the secondorder elastic moments calculated under the assumption that the frame geometry is initially perfect (no notional loads). There is no hint of any nonlinearity in these load paths. The effects of distributed plasticity and initial geometric imperfections on the in-plane strength are handled implicitly by the calculation of column strengths, Pn, associated with an inelastic sidesway failure under pure axial loading (zero moments). These axial strengths are the intercepts of the beamcolumn interaction curves with the vertical axis in Fig. 6. Two interaction curves are shown for both columns c1 and c3 in the gure. The more restrictive lower interaction curves correspond to an estimated uniform Pu/ cPn in the bottom story of the Fig. 2 frame [8,9]. The less restrictive upper curves correspond to direct calculation of the Pns based on the elastic story buckling load (or elastic effective length) in each of the columns. It should be noted that the ratio Pu/Pe(RL), where Pe(RL) is the force in a given column at elastic story sidesway buckling, is always the same for all the columns of a story [7]. Also, the Pu/ cPn obtained using a converged inelastic buckling analysis will be the same for each of the columns. This must be the case since the column loads are assumed to be increased proportionally until a story-buckling failure occurs, and the Pn values are directly proportional to the inelastic buckling loads in a converged inelastic buckling solution [7,9]. However, the Pu/ cPn values determined by substituting the individual story-based column elastic buckling loads (or the elastic effective lengths) into the AISC LRFD column strength formulas are different in general for all of the columns. Due to the fact that columns c1 and c3 tend to be destabilized by leaning effects from column c2 as this column becomes heavily plastied under pure vertical load (zero moment), the inelastic-buckling based Pu/ cPn values for c1 and c3 are larger than the Pu/ cPn

values obtained using the individual column Pe(RL) values (i.e., the inelastic-buckling based Pns for these columns are smaller than the strengths obtained directly from the elastic story buckling loads). This situation can happen whenever there are disparate Pu/Py values among the columns of a story. Baker [8] and White and Hajjar [9] explain that the Pu/ cPn for the most highly stressed column (c2 in this example) is usually an acceptable conservative estimate of the uniform inelastic-buckling based Pu/ cPn.4 Based on the above behavior, it can be concluded that the more restrictive interaction curves in Fig. 6 should be used to assess the adequacy of columns c1 and c3 for strength. However, this leads to large values of the beamcolumn interaction equation for columns c1 and c3 relative to the values predicted using the individual column elastic story-buckling loads in the calculation of Pu/ cPn. Furthermore, the controlling interaction checks for columns c1 and c3 are even smaller in the modied rst-hinge procedure. For column c2, which has the highest Pu/Py, the interaction values from the different procedures are reasonably comparable. The numerical results of the beamcolumn interaction checks shown in Figs. 46 are summarized in Fig. 7. The results based on converged inelastic story-buckling calculations [9] are also shown to illustrate the effect of this renement. As noted previously, the advanced analysis solution indicates that collapse would occur at 1.16 of the factored load level. If the inverse of the maximum interaction value is taken as an indicator of the ratio of the frame capacity to the factored load level, the elastic Pe(RL) approach predicts that this ratio is 0.87 1/1.155 based on an estimated uniform Pu/ cPn (1.155 is the maximum of the second row of values in Fig. 7) and 0.99

Fig. 7.

Summary of interaction equation values.

Fig. 6. Comparison of second-order elastic force paths at the tops of the bottom story columns to the corresponding AISC LRFD beamcolumn strength interaction curves obtained using the Pe(RL) approach.

4 In the application of the Pe(RL) approach, the individual column elastic buckling loads used in determining the column axial capacities Pn are restricted to a maximum of 1.7PL, where PL is the individual columns contribution to the elastic story sidesway stiffness. The elastic buckling load for column c1 is controlled by this limit. Although in many practical situations, it is clear that the 1.7PL limit does not control by simple inspection, in frames with disparate values of Pu/Py among the columns within a story (such as in the example frame) this limit must be checked explicitly for the lighter and/or more heavily loaded columns. This check increases the complexity of the Pe(RL) procedure. The reader is referred to Ref. [9] for a detailed step-by-step outline of the Pe(RL) calculations. For the other examples which follow in this paper, the 1.7PL limit does not control for any of the columns when the elastic Pe(RL) approach is employed.

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

163

1/1.013 based on the calculation of Pn using the individual column elastic story-buckling loads (from the rst row of values in Fig. 7). The modied rst-hinge procedure predicts that the frame will fail at 1.02 1/0.985 times the factored load level, and the converged inelastic-buckling based solution predicts a capacity ratio of 0.90 1/1.110. The modied rst-hinge approach and the Pe(RL) approach based on the individual elastic story-buckling loads each suggest that columns c1 and c3 might be reduced in size. This assessment is correct for this frame. Although the AISC beamcolumn interaction equations based on a converged inelastic buckling solution are generally expected to give the most accurate description of the strengths for non-redundant benchmark frames, as well as for the most highly-stressed columns in general redundant systems, the modied rst-hinge procedure (i.e., the AISC LRFD-based notional load approach) generally provides a better assessment of the strength interaction between axial force and moment in the lesscritical beamcolumn members of redundant frames [9]. In this example, the interaction values based on the individual column elastic story-buckling loads, Pe(RL), tend to follow the same trend as the modied rst-hinge interaction values for the less critically loaded columns. However, there is no formal rationale behind these results (i.e., the elastic buckling-based solution is based on the relative elastic stiffnesses of the columns). The reason that the converged inelastic buckling-based solutions do not characterize the strength of the less-critical beamcolumn members as well as that of the more critical ones is as follows. The discrepancies between the inelastic stiffnesses of columns with different Pu/Py values tend to be much greater at the higher load levels associated with buckling under pure axial compression compared to the smaller axial load levels often associated with the design loadings. In the converged inelastic buckling solution, the more lightly-loaded columns are penalized by leaning effects from the more heavily plastied columns, based on the state at inelastic buckling of the story, even though the structure may not ever be loaded by pure gravity load. The interpolation between cPn and bMn by the AISC LRFD beamcolumn interaction curve, which represents the advanced analysis behavior for non-redundant benchmark frames quite well, is not as representative of the behavior for redundant structures with disparate levels of Pu/Py in the lateral-load resisting columns. Because of these issues, the MFH procedure should be the preferred method for the assessment of this frame. The reader is referred to Refs. [7,9,12] for a more comprehensive assessment of the different elastic design procedures. Several simple examples are discussed in the following sections that highlight the validity of each of the elastic design procedures and the merits of advanced analysis/design for capturing benecial contributions to the overall strength.

4.2. Inelastic redistribution of lateral stiffness and strength between column members In the above two-story two-bay frame example, the behavior is complicated by inelastic redistribution of the lateral stiffness between the columns. Fig. 8 is a simpler example that illustrates some of the implications of inelastic stiffness redistribution between column members. The frame is composed of two cantilever columns tied together at their tops by a pinned-end strut and loaded by equal axial forces. These members have approximately the same elastic moment of inertia, but the area of the column on the left is only one-half that of the column on the right. Therefore, the left-hand column tends to lean more and more on the right-hand column as it becomes plastied by the applied loadings. Several strength interaction curves for the entire frame, relating the ratio of the total axial and shear loadings to the sum of the axial and shear capacities of the two columns (i.e., P/ Py versus H/ Hp), are shown in the gure. The thick solid curve shows the results from 0.9. The advanced analysis, factored by a uniform other three curves are based on elastic analysis/design approaches, and correspond to the P/ Py and H/ Hp applied to the frame when the strength of the most critically loaded member is achieved. These curves are all factored on the horizontal axis by b 0.9 and on the 0.85. In all of the elastic vertical axis by c approaches, the smaller left-hand column is the most critical in the interaction equation checks. Therefore, P in Fig. 8 is equal to two times the axial force in the lefthand column when its beamcolumn interaction curve by the different procedures is breached. It can be observed from the gure that the modied rst-hinge and

Fig. 8. System strength of portal frame with inelastic redistribution of lateral stiffness and strength between column members.

164

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

the Pe(RL) procedures (where the Pn values are calculated directly from the individual column elastic story-buckling loads) predict essentially the same frame capacity for all levels of H (i.e., they predict essentially the same strength of the most critically-loaded member). Furthermore, this capacity is a conservative prediction of the strength computed by the advanced analysis. This conservatism is due to the neglect of the inelastic reserve strength within the system. The third elastic designbased strength, shown by the dotted curve in the gure, is obtained using converged inelastic buckling loads. This approach predicts essentially the same design capacity as the advanced analysis solutions for the pure vertical loading case. That is, the inelastic redistribution of stiffnesses between the columns is accounted for. However, as H/ Hp is increased, the inelastic buckling-based solution approaches that of the other two elastic design checks. This is due to the fact that inelastic redistribution of the sidesway moments between the columns is not accounted for. Advanced analysis consistently accounts for all the inelastic redistribution effects. It is informative to also consider the interaction equation results for the non-critical right-hand column in this simple frame. These curves are plotted in Fig. 9 in terms of the P/ Py and H/ Hp on the entire frame when the different strength curves are reached by the load paths for this member. This allows us to address the appropriateness of the low interaction equation values that were obtained for the outside columns in the individual column Pe(RL) and modied rst hinge approaches for the two-story two-bay frame of the previous example (see Fig. 7). Since the member strength check for the more critical left-hand column does not enter at all into the strength assessments in Fig. 9, the strength interaction curves in terms of P/ Py and H/ Hp can exceed a value of one where they intersect the plot axes. The axial load term P is equal to two

Fig. 9. Elastic design checks for the non-critical right-hand column of the portal frame.

times the axial force in the right-hand column when this members beamcolumn interaction check is breached by the different procedures, and the shear term H is equal to the sum of the total shear force applied to the frame when the corresponding beamcolumn interaction check is breached. In other words, when member-bymember checks are used to assess the adequacy of a structure, the fact that another members interaction equation may have been violated is not considered. It is assumed that the system capacity is equal to the load level at which the strength interaction curve for the most critically loaded member is breached. As observed for columns c1 and c3 of the two-bay two-story frame, the Pe(RL) and modied rst-hinge predictions are more liberal than the inelastic buckling-solution based assessment for the non-critical member. For the Pe(RL) procedure, there is no rational explanation for the validity of the more liberal interaction equation values based on the individual column elastic buckling loads (since this approach is based only on the relative elastic stiffnesses of the columns). However, for the modied rst-hinge procedure (i.e., the AISC LRFDbased notional load approach), the large strength of the right-hand column compared to that of the left indicates that there is substantial inelastic member strength still available in the right-hand column when the predicted capacity of the frame is reached. As a nal note, a key attribute of the modied rst hinge procedure can be observed by supposing that the frame in Fig. 8 is loaded by pure vertical load up to its predicted capacity by this procedure (i.e., P/ Py 0.47). The MFH interaction equation for the left-hand column will then equal 1.0, whereas the interaction equation for the right-hand column will be substantially less than one. The designer might conclude that the frame should be able to support substantially larger load based on the small interaction value of the right column. However, if the size of the right-hand column is reduced without changing the left-hand column, the left-hand column will be found to violate its interaction equation check. This indicates that the MFH interaction equations are self-checking. Generally, if the interaction equation checks are satised for the most critically loaded members, the predicted frame capacity is always conservative. The discrepancies between the interaction values for the different columns is an indicator of the fact that the frame is designed inefciently for this particular loading, with the column on the left being loaded severely relative to the column on the right. This information would also be available from an advanced analysis by inspecting the relative amounts of plasticity in each of the members. Conversely, the inelastic buckling solution based interaction checks do not provide this information. They predict equal values for Pu/ cPn, or P/ Py at the strength of the frame, in each of the columns. For the most-critically loaded members, the MFH

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

165

procedure tends to give member strength assessments that have roughly about the same maximum conservatism as design calculations based on elastic story-buckling loads [7]. Given the simplicity of the modied rst-plastic hinge procedure, compared to buckling-solution based procedures, this approach should be particularly benecial for cases in which in-plane stiffness or out-of-plane strength of the members within an unbraced moment frame controls the design. As noted by White and Hajjar [9], this approach is very amenable to preliminary analysis and design. If a more rigorous check on the system strength is desired, an advanced analysis can be conducted. 4.3. Out-of-plane strength of columns in unbraced frames Fig. 10 shows example results from a three-dimensional advanced analysis as well as the two modied rst-hinge approaches outlined earlier in the paper for checking the combined in-plane and out-of-plane strength of beamcolumns subjected to strong-axis bending in the plane of a frame (see Section 2.3). Procedure 1 is a simpler and more traditional single-equation approach whereas Procedure 2 involves the combination of separate in-plane and out-of-plane strength equations with the beamcolumn capacity given by the minimum of the two equations. The in-plane advanced analysis strength also is shown in the gure for comparison. The shape of the single interaction equation of the rst of the two MFH procedures is representative of the shape of the in-plane advanced analysis strength curve. However, it is not representative of the strength interaction computed by the three-dimensional advanced analysis. It is apparent that for the W200 46 column considered in

Fig. 10. Strength of a sidesway beam-column member braced at its top and bottom in the out-of-plane direction.

this study, the advanced analysis strength is not controlled by an out-of-plane failure mode unless the axial load is nearly equal to that corresponding to a exuralbuckling type failure in the out-of-plane direction. To generalize upon these results, it appears that the strengths of beam-type wide ange sections with high rx/ry values and negligible out-of-plane end restraint tend to be approximated reasonably well by the AISC LRFD based curves. However, the results shown in the gure are quite representative of the behavior of column-type wide ange sections (d bf) subjected to reversed-curvature bending. This observation is signicant, since it indicates that many of the present design standards are often conservative in basing the strength of beamcolumns in multistory frames on out-of-plane lateral-torsional buckling predictions. The true failure mode is essentially in-plane for lower axial load levels, but becomes predominantly out-of-plane as the axial load level approaches the out-of-plane exural buckling strength. It should be emphasized that the three-dimensional advanced analysis strengths shown in Fig. 10 are based on unrestrained warping and unrestrained out-ofplane rotation at the ends of the column. In practice, some restraint often exists at the member ends. This would tend to make the out-of-plane failure mode even less likely. In the second of the modied rst-hinge approaches shown in the gure, the nonlinear interaction equation for the out-of-plane strength from AS4100 [18] is employed along with the MFH in-plane strength equation, but with the basic strength terms (Pn(L), Mn, etc.) substituted from the AISC LRFD provisions. It can be seen that the correlation between the three-dimensional advanced analysis strengths and the strengths predicted by these combined, but separate, in-plane and out-ofplane strength checks is quite good. It should be noted that the AISC LRFD axial strengths for this problem (factored by c 0.85) are 0.75Py and 0.66Py, respectively, for in-plane and out-of-plane failure under pure vertical loading, if the in-plane capacity is based on an in-plane elastic sidesway buckling calculation (i.e., based on an elastic sidesway-uninhibited effective length). The corresponding factored LRFD inplane and out-of-plane beamcolumn interaction curves, not shown in Fig. 10, are very similar in shape to the MFH in-plane strength and the MFH out-of-plane strength by procedure 1, which predict pure vertical load capacities of 0.71Py and 0.61Py, respectively. This illustrates the fact that the MFH procedure can be conservative relative to conventional elastic buckling-based design checks in certain cases. It is even more conservative relative to the AISC LRFD in-plane strength based on a converged in-plane inelastic sidesway buckling calculation (i.e., based on a converged inelastic sideswayuninhibited effective length). The factored AISC LRFD in-plane vertical load capacity for this problem based on a converged inelastic buckling analysis is 0.77Py. This

166

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

example represents what the authors feel is approximately the maximum conservatism for practical sway frames. If the MFH procedure 2 is used for the out-ofplane check, the MFH approach predicts slightly more capacity than the AISC LRFD strength equations for all combinations of P and H in this problem. The AISC LRFD equations imply that the strength is controlled by out-of-plane lateral-torsional buckling, whereas the advanced analysis solutions show that this is not the case except at high axial load levels.

5. Conclusions In current US practice, tedious and error-prone calculations are required to compute accurate nomographbased effective length factors to assess the in-plane strength of unbraced steel frames, although the in-plane sidesway stiffness or the out-of-plane strength checks may often control the design. The Pe(RL) story bucklingbased procedure offers a highly efcient method of assessing the in-plane strength, by virtue of the fact that it is based on the elastic story lateral stiffness, a quantity that is readily obtained in conjunction with preliminary stiffness-based design. This paper has outlined the benets of two additional alternative procedures: (1) the modied rst-hinge (MFH) approachalso referred to as an AISC LRFD-based notional load approachwhich provides an elastic analysis based assessment of the inplane, out-of-plane and/or spatial strength of framing systems, and (2) a rigorous second-order inelastic analysis (i.e., advanced analysis). The modied rst-plastic hinge procedure involves the application of a notional horizontal load coupled with an assessment of beam column strength based on the actual member unsupported lengths about the strong and weak axes of bending. As shown in the examples, this approach generally provides a more liberal strength assessment than buckling-based solution procedures for the less-critically loaded beamcolumns within a structural system. It tends to provide member strength assessments that have roughly the same maximum conservatism as design calculations based on elastic story-buckling loads for the most critical members (i.e., the beamcolumns with the highest Pu/Py). This approach generates additional sidesway moments in the beams and connections associated with its approximation of the inuence of distributed plasticity on the forces in the structural system at factored load levels. In situations involving disparate column Pu/Py values, the MFH procedure is simpler to apply than the Pe(RL) approach (in these types of cases, the 1.7PL limits on the applicability of the Pe(RL) equations usually need to be checked). Furthermore, it has been illustrated that the MFH procedure is more representative of the physical behavior in these situations. It has been shown that for beamcolumns loaded by in-

plane strong-axis bending, separate MFH equations can be applied for the in-plane and out-of-plane member strengths to more accurately predict the true behavior, at the expense of some additional complexity. Furthermore, if this approach is employed, it has been shown that the in-plane strength often governs rather than the out-of-plane strength for typical column-type wideange sections in multi-story frame structures (i.e., single equation based interaction checks tend to be conservative for some cases of lateral-torsional beamcolumn failure, since the true interaction curve associated with out-of-plane lateral-torsional failure is not necessarily the same shape as the in-plane failure curve). Illustrative example design calculations are outlined for this and for other buckling-solution based design procedures in Ref. [9]. For frames in which the engineer wants to account explicitly for at least some inelastic reserve capacity, advanced analysis, in turn, captures the interaction between framing members due to progressive inelastic redistribution of stiffness and strength. In addition, this procedure provides an efcient and exible analysis/design methodology by alleviating the need to compute explicit individual beamcolumn interaction checks, while providing signicantly more information about the idealized structural performance at service and factored load levels than elastic analysis/design procedures.

References
[1] LeMessurier WJ. Simplied K factors for stiffness controlled designs. In: Proceedings of ASCE 13th Structures Congress 95. New York: ASCE, 1995: 1797812. [2] Gaylord CN, Watabe M, editors. Structural design of tall steel buildings, vol. SB. New York: Council on Tall Buildings and Urban Habitat and American Society of Civil Engineers, 1979. [3] Ellingwood B. Structural serviceability review and standard implementation. In: Proceedings of ASCE 14th Structures Congress 96. New York: ASCE, 1996: 43643. [4] Chan C-M, Sherbourne AN, Grierson DE. Stiffness optimization technique for 3D tall steel building frameworks under multiple lateral loadings. Engrg Struc 1994;16(8):5706. [5] American Institute of Steel Construction. Manual of steel construction: allowable stress design, 9th ed. Chicago, IL: AISC, Inc., 1989. [6] American Institute of Steel Construction. Manual of steel construction: load and resistance factor design, 2nd ed. Chicago, IL: AISC, Inc., 1994. [7] ASCE Committee on LRFD. Effective length and notional load approaches for assessing frame stability: implications for American steel design. New York: ASCE, 1997. [8] Baker WF. Practical problems in stability of steel structures. In: Proceedings, 1997 National Steel Construction Conference, AISC, Chicago, IL, 1997. [9] White DW, Hajjar JF. Accuracy and simplicity of alternative procedures for stability design of steel frames. J Constr Stl Res 1997;42(3):20961. [10] SSRC. In: Galambos TV, editor. Guide to stability design criteria

D.W. White, J.F. Hajjar / Engineering Structures 22 (2000) 155167

167

[11]

[12]

[13]

[14]

[15]

[16]

[17]

for metal structures, 5th ed. New York: John Wiley and Sons, 1998. Clarke MJ, Bridge RQ. The notional load approach for the design of frames. Research Report No. R718, School of Civil and Mining Engineering, University of Sydney, 1995. White DW, Clarke MJ. Design of beamcolumns in steel frames. Part II: comparison of current standards. J Struc Engrg, ASCE 1997;123(12):156575. Liew JYR, White DW, Chen W-F. Notional load plastic hinge method for frame design. J Struc Engrg ASCE 1994;120(5):143454. Ziemian RD, McGuire W, Deierlein GG. Inelastic limit states design: part Iplanar frame studies. J Struc Engrg, ASCE 1992;118(9):253249. Ziemian RD, Miller AR. Inelastic analysis and design: frames with members in minor-axis bending. J Struc Engrg, ASCE 1997;123(2):1516. White DW, Nukala PKVV. Recent advances in methods for inelastic frame analysis: implications for design and a look toward the future. In: Proceedings, 1997 National Steel Construction Conference, AISC, Chicago, IL, 1997. ECCS. Ultimate limit state calculation of sway frames with rigid joints. Pub. No. 33, ECCS Technical Committee 8Structural Stability, Technical Working Group 8.2System, 1984.

[18] SAA. AS4100-1990, steel structures. Standards Association of Australia, Australian Institute of Steel Construction, Sydney, Australia, 1990. [19] Cuk PE, Rogers DF, Trahair NS. Inelastic buckling of continuous steel beamcolumns. J Constr Stl Res 1986;6(1):2153. [20] Attalla, M. Inelastic torsional-exural behavior and the threedimensional analysis of steel frames. Doctoral dissertation, School of Civil and Environmental Engineering, Cornell University, 1995. [21] Nukala PNVV. Three-dimensional second-order inelastic analysis of steel frames. Doctoral dissertation, School of Civil Engineering, Purdue University, West Lafayette, IN, 1997. [22] Springeld J. Limits on second order elastic analysis. In: Proceedings, 1991 Annual Technical Session, SSRC, 1991: 8999. [23] McGuire W. Inelastic analysis and design in steel, a critique. In: Sanayei M, editor. Restructuring: America and beyond, Proceedings of Structures Congress XIII. New York: ASCE, 1995: 182932. [24] Galambos T, Ketter R. Columns under combined bending and thrust. Journal of the Engineering Mechanics Division, ASCE 1959;85(EM2):130. [25] CEN. ENV 1993-1-1 Eurocode 3: design of steel structures, Part 1.1general rules and rules for buildings. Brussels: European Committee for Standardization, 1992.

Вам также может понравиться