The 2000 T.R.

Higgins Lecture: A Practical Look at Frame Analysis, Stability, and Leaning Columns

Author

Summary

Louis F. Geschwindner

ouis F. Geschwindner is professor of architectural engineering at Pennsylvania State University and is a Registered Professional Engineer. He received his bachelor's degree in building science from Rensselaer Polytechnic Institute and both his master of science in architectural engineering and his Ph.D. in civil engineering from Pennsylvania State University. He has been a faculty member at Pennsylvania State for over 30 years, teaching and conducting research in building structures and he is also in charge of the undergraduate program in architectural engineering. Professor Geschwindner is chairman of the Committee on Design of Steel Building Structures of the American Society of Civil Engineers (ASCE). He is also a member of the Committee on Metals, the LRFD Committee and vice-chair of the Tension Membrane Structures Standards Committee, all of ASCE. Dr. Geschwindner is a member of the AISC Committee on Specifications as well as its Task Committee 10 - Stability. His memberships also include the Masonry Society and the American Society for Engineering Education. Teaching advanced level structural courses in the five-year architectural engineering program at Pennsylvania State University, has been the primary thrust of Dr. Geschwindner's academic career. He has participated as a lecturer in the American Institute of Steel Construction's lecture series and has developed a short course on LRFD for practicing professionals. He has received numerous awards for outstanding teaching, including the University-wide AMOCO Foundation Outstanding Teaching Award and the AT&T Foundation Award for Excellence from ASEE.

he analysis and design of unbraced moment frames is a fairly regular activity in structural engineering practice yet it can be a complex structural engineering problem. Numerous analysis methodologies are available and the many commercial software packages used in practice provide a variety of approaches to the problem. Questions arise as to whether a first-order or secondorder analysis is appropriate, should an elastic or inelastic analysis be carried out, what moment magnifiers should be used when axial load and moment act together, and how should column capacity be determined, with effective length factors or some other approach. Stability of a column, although often expressed as a function of the individual column, is actually a function of all of the members in the story. Thus, column design is a story problem, not an individual column problem. When unbraced moment frames support pinended columns, additional problems arise. These pin-ended columns do not participate in the lateral resistance of the structure, but instead rely on the unbraced frame for their lateral stability. Thus, the frame must be designed to accommodate the loads that are applied to it and provide sufficient stiffness to support any of these "leaning" columns. Frame analysis may be approached by a variety of paths but linear elastic analysis is perhaps the most common, as well as the least complete. A second order inelastic analysis, while perhaps the most comprehensive, is also the most complex. Whichever analysis method is chosen, the design approach must be compatible. Numerous approaches have been presented in the literature to address the design of frames both with and without leaning columns. Although a direct stability analysis

1-1
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

may be performed, the most common approaches still appear to be those that utilize some form of simplification. This paper will briefly review first- and secondorder elastic and inelastic analysis approaches and a few of the commonly used computer software packages for unbraced moment frames. This will be followed with a discussion of the impact of lean-

ing columns on these analyses. Once analysis approaches have been identified, the design process, where the inclusion of column effective length comes into play, will be addressed. The use of effective length to predict column capacity will be discussed and its use to account for the inclusion of leaning columns will be addressed. Effective length

calculations will be reviewed with particular attention to the approaches presented by Yura, Lim & McNamara, LeMessurier, and the equations found in the AISC LRFD Commentary. The results from these approaches will be compared to those of a stability analysis for simple frames that have been found in the literature.

1-2
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

A Practical Look at Frame Analysis, Stability, and Leaning Columns

By Louis F. Geschwindner
Introduction

The analysis and design of unbraced moment frames is a fairly regular activity in structural engineering practice yet it can be a complex structural engineering problem. Numerous analysis methodologies are available and the many commercial software packages used in practice provide a variety of approaches to the problem. Questions arise as to whether a first-order or second-order analysis is appropriate, should an elastic or inelastic analysis be carried out, what moment magnifiers should be used when axial load and moment act together, and how should column capacity be determined, with effective length factors or some other approach. Stability of a column, although often expressed as a function of the individual column, is actually a function of all of the members in the story. Thus, column design is a story problem, not an individual column problem. When unbraced moment frames support pin-ended columns, additional problems arise. These pin-ended columns do not participate in the lateral resistance of the structure, but instead, rely on the unbraced frame for their lateral stability. Thus, the frame must be designed to accommodate the loads that are applied to it despite the reduction in overall frame stability that results from these "leaning" columns. Frame analysis may be approached by a variety of paths. Linear elastic analysis is perhaps the most common, although the least complete. A second order inelastic analysis, while perhaps the most comprehensive, is also the most complex. And there are many approaches between these. Whichever analysis method is chosen, the design approach must be compatible. Numerous approaches have been presented in the literature to address the design of frames both with and without leaning columns. Although a direct stability analysis may be performed, the most common approaches still appear to be those that utilize some form of simplification. This paper will briefly review first- and secondorder elastic and inelastic analysis approaches and a few of the commonly used computer software packages for unbraced moment frames. This will be followed with a discussion of the impact of leaning columns on these analyses. Once analysis approaches have been identified, the design process, where the inclusion of column effective length comes into play, will be addressed. The use of effective length to predict column capacity will be discussed and its use to account for the inclusion of leaning columns will be addressed. Effective length calculations will be reviewed with particular attention to the approaches presented by Yura, Lim & McNamara, LeMessurier, and the equations found in the AISC LRFD Commentary. The results from these approaches will be compared to those of a stability analysis for simple frames that have been found in the literature.
Analysis The state of the art of structural analysis encompasses a wide range of possible approaches to the determination of member response to structural loading. Each new

Geschwindner
1-3
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

approach adds or subtracts some aspect of frame or member behavior in an attempt to properly model the true behavior of the structure. Before looking at the requirements of the current LRFD specification, it will be helpful to categorize these analysis approaches and discuss their characteristics. Figure 1 shows a comparison between the loaddeflection curves of a series of analysis approaches.

First-order elastic analysis (1): The first and most common approach to structural analysis is the linear elastic analysis. In this case, deformations are assumed to be small so that the equations of equilibrium may be written about the undeformed configuration. In this case, superposition is valid, any inelastic behavior of the material is ignored and the load-deflection curve is linear. This is the approach used in the development of the common analysis tools of the profession, such as slope-deflection, moment distribution and the stiffness method found in most commercial computer software. In a first-order analysis, the deformations are determined and then used in turn to determine forces. Once the forces are determined, the analysis is complete.

Figure 1. Load-deflection history

Linear buckling analysis (2): An analysis carried out for the determination of the elastic buckling load will result in the determination of a single critical buckling load for a frame. The critical buckling load may be determined through an eigenvalue solution or through a number of iterative schemes based on equilibrium equations written about the deformed configuration. This is the type of analysis that yields the critical buckling load of the single column and is the basis upon which the effective length factor is founded. It can be seen in Figure 1 that the results of this analysis do not provide a load-displacement curve but rather the single value of load at which the structure fails.

Second-order elastic analysis (2): When the equations of equilibrium are written about the deformed configuration of the structure and the deflections corresponding to a given

Geschwindner
1-4
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

set of loads are determined, the resulting analysis is a second-order analysis. This is the analysis generally referred to as the P-delta analysis. Two components of these secondorder effects are included in the analysis. When the influence of axial force and member curvature is included, it is said that the effects are included and when the sidesway effects are included it is said that the effects are included. It is seen in Figure 1 that the load-deflection history obtained through this analysis approach the value obtained from the eigenvalue solution. This analysis usually requires an iterative solution so it is a bit more complex than the first-order analysis. Because of the problems inherent with iterative solutions, many researchers have proposed one-step approximations to the second-order analysis.

First-order plastic-mechanism analysis (3): As load is increased on a structure, it is assumed that defined locations within the structure will reach their plastic capacity. When that happens, the particular location continues to resist that plastic moment but undergoes unrestrained deformation. These sections are called plastic hinges. Once a sufficient number of plastic hinges have formed so that the structure will collapse, it is said that a mechanism has formed. No additional load can be placed on the structure. This limit can be seen in Figure 1. First-order elastic-plastic analysis (4): If the approach to determination of the collapse mechanism tracks the development of individual hinges, more information is obtained from this analysis than from the mechanism analysis. It is clear however, that if the same assumptions as to hinge length and geometry are maintained, the limit of the elasticplastic analysis will be the mechanism analysis as seen in Figure 1.

Second-order inelastic analysis (5): This approach to analysis combines the same principles of second-order analysis discussed previously with the plastic hinge analysis just discussed. It is obvious that this approach to analysis is much more complex. It does, however, yield a more complete and accurate picture of the behavior of the structure, depending on the completeness of the model used. This type of analysis is what has come to be known as "advanced analysis." The load-deformation curve for a second-order inelastic analysis is shown in Figure 1.
In summary, it can be seen that as more complex behavior is taken into account in the analysis, the predicted load level is reduced. Thus, designers must exercise extreme caution when using an analysis approach with which they are unfamiliar. Design The approach taken for member design must be consistent with the approach chosen for analysis. There are currently three approaches to the design of steel structures acceptable under US building codes as they incorporate AISC Specifications (6,7). Currently, the best tool for steel design is the Load and Resistance Factor Design Specification (LRFD). However, the Plastic Design (PD) approach is also permitted and the Allowable Stress Design Specification (ASD) is still used.

Geschwindner 1-5
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

The LRFD specification stipulates, in Section C1, that "Second order effects shall be considered in the design of frames." The comparable statement in the ASD specification states, in Section A5.3, that "Selection of the method of analysis is the prerogative of the responsible engineer." And in Section C1 that "frames...shall be designed to provide the needed deformation capacity and to assure overall frame stability." The normal analysis method is one of the linear elastic approaches. The satisfaction of the deformation capacity and the assurance of stability are left to the engineer. In order to use the Plastic Design Specification for structural design, a first-order mechanism analysis must be carried out. Additionally, other restrictions found within the specification must be satisfied. Thus, regardless of the specification used, the engineer is required to address the second order effects. This may be done using a first-order analysis and a code provided correction for second order effects or it may be addressed through direct use of a secondorder analysis. Impact of Second-order Effects Two different second-order effects will impact on the design of a single column. The first, illustrated in Fig. 2a for a column in which the ends are prevented from displacing, is the result of the deflection along the length of the column. It can be seen that the moments along the column will be increased due to the column deflection, by an amount This increase in moment due to member deflection is referred to as the member effect.

Figure 2. Influence of Second-Order Effects The column in Fig. 2b is part of a structure that is permitted to sway laterally an amount As a result, the moment required in the end of the column to maintain equilibrium in the displaced configuration is given as This member moment is referred to as the structure effect, since the lateral displacement of the column ends is a function of the properties of all of the members of the structure participating in sway resistance.
Geschwindner 1-6
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

The deflections, and shown in Fig. 2 are second-order deflections, resulting from the applied loads plus the second-order forces. Although it requires a second-order analysis to determine the magnitude of the forces, both of these effects appear to be straightforward for the individual column of Fig. 2. However, when columns are combined to form frames, the interaction of all of the members of the frame significantly increases the complexity of the problem. The addition of columns which do not participate in lateral resistance but which do carry gravity load brings further complexity to the problem. Computer Approaches Although the influence of second order effects was easily observed in the single column of Figure 2, the analysis of an unbraced frame of any realistic proportions could be expected to require the use of some computer software. The particular software being used will dictate the methods of analysis available to the engineer. Based on their showing in a recent AISC survey (8), five commercial software packages were reviewed to determine their approach to a second order analysis. They are STAAD-III, GTSTRUDL, RISA-2D, ETABS, and RAM FRAME. All of these programs indicate, in their literature that they include some form of second-order analysis.

STAAD-III (9) adopts a simplified method to perform the P-delta analysis. The users manual provides the following steps as a guide for using its P-delta analysis: 1. Perform a first order analysis to obtain deflections and member forces based on external loads, 2. Create an additional load vector by combining the previously calculated member forces and deflections with the external loads, 3. Use this new load vector and the original stiffness matrix to calculate a revised set of deflections, 4. Determine member forces and reactions from these revised deflections. Although this is not an exact approach, the program authors indicate their belief that their approach has been justified by the literature.
GTSTRUDL (10) uses a direct stiffness formulation about the displaced geometry of the structure. A secant stiffness approach is used where the structure stiffness matrix is modified in each step of the analysis until the resulting change in displacements or forces from one step to the next is within the bounds specified or the maximum number of specified cycles is reached. In RISA-2D (11), the P-delta effects are "accurately approximated." Their approach is to solve the first-order problem and determine the member forces. From these member forces and displacements, a member shear is determined as These shears are added to the original loads and the first-order problem is resolved. The new displacements are compared to the previously calculated displacements. If the difference is small enough, the analysis is complete. If a more accurate solution is needed, additional cycles are performed until the required level of accuracy is reached.

ETABS uses a geometric stiffness correction in which the effects are "exactly represented"(12). The approach is to use the load P determined from the story mass to

Geschwindner

1-7
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

modify the stiffness matrix, which is then used throughout the analysis, both static and dynamic. This method has been presented in the literature by the program authors (13).

RAM FRAME (14) has also adopted the geometric stiffness method. However, they indicate that certain modifications have been made in the implementation of the method within the specific formulation of their program.
It appears that commercial software for second-order analysis of steel frames is readily available to the design engineer. Thus, it can be expected that the use of a second-order analysis will become a more realistic approach for normal design. Predicting the Critical Buckling Load

When an analysis tool is available to determine the critical buckling load of a frame, there is no need to worry about "predicting" that load through some other means. Thus, it might be said that if all structural analysis were carried out as a second order buckling analysis, there would be no need to spend time discussing the correct approach to determining a K-factor to use in design. It seems that ever since the K-factor was introduced into the 1961 AISC Specification, it has generated extensive discussion and misunderstanding (15). To understand the debate over the K-factor, one must understand what the K-factor is intended to accomplish. It will be helpful to consider the critical buckling load of a particular column, determined by one of the second-order analysis programs, as It will also be helpful to remember that the critical buckling load of the perfect column, as derived by Euler, is given as
(1) Since the column in a steel frame is not likely to be a perfect column, but rather a real column with multiple imperfections, its critical buckling capacity can be said to be somewhat less than the Euler column, thus
(2)

If that reduction factor is defined as

it is seen that
(3)

Thus, the K-factor is simply a mathematical adjustment to the perfect column equation to try to predict the capacity of an actual column. Every method or equation that is proposed is simply trying to predict accurately the actual column capacity as a function of the perfect column.

Geschwindner 1-8
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

Perhaps the most commonly used approach to the determination of K-factors is the nomograph found in the commentary to the LRFD and ASD Specifications (6,7). The equation upon which the nomograph is based is given here as Eq. 4 (3).
(4)

with

and the A and B subscripts referring to the ends of the column under consideration.
The assumptions used in the development of the nomograph are detailed in the Commentary to the Specification. One of these important assumptions is "all columns in a story buckle simultaneously." Although this assumption was essential in the derivation of this useful equation, it is also one that is regularly violated in practical structures. This assumption is critical since it eliminates the possibility that any column in an unbraced frame might contribute to the lateral sway resistance of any other column. A reasoned analysis of the behavior of columns in actual structures would seem to indicate that strong columns, ones whose load is not near their capacity, should be able to help restrain weaker columns. Thus, other approaches to determining the K-factor should be considered.

Figure 3. Symmetric Portal Frame

Second-Order Analysis and Leaning Columns

Without leaning columns: Two simple frames will be discussed so that the impact of secondorder analysis and leaning columns may be understood. The symmetric frame shown in Fig. 3a is subjected to a symmetrically placed gravity load. A first order analysis yields the

Geschwindner 1-9
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

forces shown. Note that there are no column moments and thus, there will be no axial force and moment interaction. When a lateral load is added as shown in Fig 3b, forces and moments as shown result from a first-order analysis. In this case, all moment is due to the lateral load and these moments must be amplified to account for the second-order effects. For both cases, axial capacity may be determined using the K-factor from the nomograph. A similar frame, with two different columns, is shown in Fig. 4a, loaded with a nonsymmetric gravity load. For an LRFD-based design and a first-order analysis, the sway and non-sway moments must be separated. To do this, the first step is to restrain sidesway and perform a first-order analysis. A fictitious restraining force, AJR, results as shown in Fig. 4b. Next, the fictitious force is removed by applying CJF = -AJR, removing the restraint, and carrying out another first-order analysis. These results are shown in Fig. 4c. The capacity of the columns must then be checked according to the code specified interaction equations. Regardless of method, LRFD or ASD, the column buckling capacity for the first term of the interaction equation can be determined from the nomograph and the column moments will be amplified to account for second-order effects using the in-plane K-factor in the calculation of the amplification factors. Normally, the sway effects of the non-symmetric load will be quite small. If a lateral load is added, as shown in Fig. 4d, a similar approach can be used; however, the sway moments will now be significant.

Figure 4. Nonsymmetric Portal Frame

When the frame of Fig. 3a is subjected to a second-order analysis, there will be no change from the results for a first-order analysis. Thus, the design approach should be the same; K-factors from the nomograph and no moments to be amplified. If a second-order analysis is performed for the frame of Fig. 3b, the results will be different from the first-

Geschwindner
1-10
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

order results. The buckling capacity may still be determined through the nomograph Kfactor; however, the moments will not need to be amplified to account for sway since that was already accomplished through the second-order analysis.
When the frames of Fig. 4 are subjected to a second-order analysis, different moments result, which account for sway. The buckling capacity may be determined from the Kfactors of the nomograph and moments may be used directly, with no need for amplification.

With Leaning Columns: When an unbraced frame is called upon to provide lateral restraint for leaning columns, the first- and second-order analysis may or may not be sufficient, depending on the simplifications used. In addition, the nomograph approach (Eq. 4) to determining buckling load may not be a good indicator of column capacity. When a leaning column is added to the frame of Fig. 3a, the resulting structure is as shown in Fig. 5. A first-order analysis will yield the same member forces for the unbraced frame as had been determined for the structure in Fig. 3a. Thus, it appears that the leaning column has no impact on the original structure. If the structure is subjected to a second-order analysis, again, no change will be noted.

Figure 5. Symmetric Frame with Leaning Column

Figure 6. Nonsymmetric Frame with Leaning Column

If a leaning column is added to the frame of Fig. 4, as shown in Fig. 6, and both the gravity and lateral loads shown are applied, a first-order analysis will again repeat the results from the frame of Fig. 4. If a second-order analysis is performed, the results will be different from those previously determined. They will account for the amplification of moment due to sidesway of the structure and both loads P and Q; but, as for the frame of Fig. 5, no account will be taken of the reduced buckling capacity due to the presence of the leaning column.
If the buckling load for a frame member is to be determined through an approach other than a complete buckling analysis, a model that will reasonably predict the capacity of a frame including these leaning columns is needed. Numerous approaches intended to account for the effect of leaning columns and the sharing of lateral resistance have been presented in the
Geschwindner
1-11
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

literature. These approaches offer a wide range of mathematical complexity and practical usefulness. Four approaches that have been presented in the literature for including the leaning column in the determination of column capacity will be discussed along with some simplified equations that are included in the Commentary of the LRFD Specification. As always, the designer is called upon to decide on the appropriate approach to use in a particular design situation.

Effective Length
Modified Nomograph Equation (16): The derivation of Eq. 4 is available in numerous references, including (3). Following the same procedures and assumptions, with the addition of the leaning column, as shown in Figure 7, a new equation may be developed.

Figure 7. Restraining and Leaning Columns

Viewing the structure in its displaced equilibrium configuration, the leaning column and the restraining column are separated as shown in Fig. 7b and c. The load Q on the leaning column CD must be balanced by the horizontal force, at D, for equilibrium of the leaning column. This force must then be applied as a load at B on the restraining column AB. Equations of equilibrium at the joints of column AB and the sway equilibrium equation can be written for the structure in the displaced configuration. Member end moment equations are then written using the slope deflection method, incorporating the stability functions (17) necessary to account for the influence of axial load on column AB. Combining these equations and setting the determinate of the coefficients equal to zero will yield the following buckling condition equation.

Geschwindner 1-12
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

If the leaning column load is zero, Eq. 5 reduces to Eq. 4. Since neither of these equations can be solved explicitly, an iterative approach may be used or, in the case of the frame without leaning columns, the nomograph already discussed may be used.
The Yura Approach (18): This is perhaps the easiest approach to develop since it relies on a straightforward interpretation of the physical problem. For the unbraced frame shown in Fig. 8, equilibrium will be established for the structure in the undeflected configuration and again in the deflected configuration. The first-order, undeflected equilibrium configuration forces are shown in Fig. 8a. If the frame is permitted to displace an amount A, equilibrium in this displaced configuration will be as shown in Fig. 8b. In order for column EF to be in equilibrium, a lateral force, as shown at F is required. This force must be equilibrated by an equal and opposite force shown at B. Thus, when column AB buckles, it buckles with a moment of at its base. It is observed that this is the same moment that would result if the individual column AB were to buckle under the axial load of (P + Q). The assumption that the buckling load is (P + Q) is only slightly conservative for the individual column AB, since the deflected shape due to an axial load and a lateral load differ only slightly. In order to insure sufficient lateral resistance to buckling for column EF, column AB must be designed to carry a fictitious load (P + Q).

Figure 8. Equilibrium forces for Yura derivation

In order to compare this approach to others presented in the literature, it is helpful to convert
it to an effective length approach. If column AB is to be designed to carry the load P but have the capacity (P + Q), a modified effective length factor will be required. is defined

as the effective length factor that would be determined from the nomograph or Eq. 4, which does not account for the leaning column. In this case is defined as the effective length factor that will account for the leaning column. Thus, based on the buckling load being (P + Q)
(6)

Geschwindner
1-13
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

If the column is to be designed to carry the actual applied load, P, with the leaning column accounted for through then
(7)

Solving equations 6 and 7 for their corresponding K's and taking the ratio

yields
(8)

which may be solved for

as
(9)

If column AB from Fig. 8a were designed to carry the load P using the effective length factor it would provide sufficient lateral restraint to permit column EF to be designed to carry the load Q using K = 1.0. For frames with more than one leaning column and more than one restraining column, and will replace P and Q. It should also be noted that this approach maintains the assumption that all restraining columns in a story buckle in a sidesway mode simultaneously. Lim & McNamara Approach (19): Another approach that will account for the leaning column was proposed by Lim and McNamara for columns of unbraced tube buildings. Their development is also based on the assumption that all columns in the restraining frame buckle in a sidesway mode simultaneously; however, they developed the sway buckling equation through the use of stability functions and an eigenvalue solution. The resulting effective length factor, accounting for leaning columns is given in their paper as
(10)

where and are as defined earlier, and is the eigenvalue solution for a frame without leaning columns and is the eigenvalue solution for a frame with leaning columns. The authors suggest that for normal column end conditions, should provide a K-factor on the conservative side by at most 2%. Substituting for n and using the Lim & McNamara approach gives the same K-factor as the modified Yura approach where

(11)

Geschwindner
1-14
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

Thus, for the story buckling approach, a single multiplier for each story will be sufficient to modify the individual nomograph K-factors to account for leaning columns.
LeMessurier Approach (20): In his landmark paper, LeMessurier presented a more complex, yet still very practical approach for frames with and without leaning columns. The basic equations were developed for a single cantilever column and then extended to the general frame. Where the previous approach determined a constant value for a story by which the nomograph value of was modified, this approach determines a constant value for a story which then multiplies the individual column moment of inertia divided by the column load, , for each column, i. Thus, the contribution of each column to the lateral resistance is accounted for individually. The effective length factor for each column that participates in resisting sidesway buckling, Eq. 46c from the original paper, expressed in the notation of this paper, is given by
(12)

where (13)

(14)

= effective length of column i, accounting for leaning columns.

= 0 for leaning columns.
= load on restraining column, i. = moment of inertia for column, i,

= load on the restraining columns in a story.

= load on the leaning columns in a story = sum of for each column in the story.

= sum of

for each column participating in lateral sway resistance

Commentary Equations (6): Although use of Eq. 12 is not particularly complex, the second edition of the Commentary to the LRFD Specification presented two modified LeMessurier equations that were thought to be of value to the practicing engineer. Although the third

Geschwindner
1-15
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

edition of the commentary will be significantly simplified in the relevant section, it is anticipated that there will still be two simplified LeMessurier equations presented. One is based on the story buckling model while the other is based on a story stiffness model. For the story buckling model, it is assumed that there is no reduction in column stiffness due to the presence of axial load. This is accomplished by taking for all columns, which leads to Substitution of these values into Eq. 12 yields:
(15)

Which reduces to
(16)

Equation 16 can be recast into the form of the commentary equation, which is currently proposed to be
(17)

For a structure in which only one column can be considered to provide lateral stability, the summation in Eq. 16 are unnecessary and the equation reduces to
(18)

which is the same as the equation that resulted from the modified Yura and Lim & McNamara approaches, Eqs. 9 and 11 respectively.

For the story stiffness model, stiffness reduction due to axial load is included as though all columns were cantilevers with a buckled shape in the form of a half sine curve as shown in Fig. 2b, thus Since the leaning columns have no lateral stability of their own, = 0.0 for all leaning columns. The equation given in this paper as Eq. 12 is just one form of the effective length factor equations given by LeMessurier. Another form that uses, as a measure of lateral stiffness, the ratio of lateral displacement of a story to the lateral load, is also available through the same derivation (19). Equation 46d from the original paper, in the notation of this paper, is given as
(19)

where the total lateral load supported by the level under consideration, is the corresponding lateral displacement of the level and is the total load on the
Geschwindner

1-16
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

given story. In order to account for on the leaning columns, the load on these leaning columns must be subtracted from the total load on the story so that Making this substitution and factoring out yields

(20)

This equation is somewhat simplified in the second edition commentary as

(21)

If the stiffness reduction due to axial load is applied to all columns, that is the leaning columns are not excluded, then would be applied to the total load on the story and the separation taken to arrive at Eq. 20 would not be necessary. Thus, Eq. 19 would become
(22)

This equation, recast in the form of the third edition commentary equation is
(23)

These simplifications may not really be necessary since, in the original form, the equations presented by LeMessurier are not much more complex. Examples The following examples will show how these approaches may be used to evaluate columns in unbraced frames.

Example 1: The unbraced frame with leaning columns as shown in Fig. 9 is to be checked for strength and stability. This is the frame introduced by Higgins (21) and modified with a load factor of 1.43 to permit a check by LRFD. The frame is braced out of the plane.

Geschwindner 1-17
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

Yura Approach: Since sway will likely control the design of column AB, a W10x39 will be investigated. Due to symmetry, one half of the load on the leaning columns will be assigned to a single column, AB.

Figure 9. Symmetric Frame for Example 1

Note that to account for the pin ended beam, the length is modified by the factor 2.

For strength in the y-axis, thus, use LRFD E2-2. kips > 57 kips thus, the strength is adequate.

For stability about the x-axis, therefore from LRFD E2-2, = 228.0 kips > 207 kips, thus the column will be sufficient to provide stability for the remaining leaning columns. Lim & McNamara Approach: Again, a W10x39 will be considered for column AB. Using the ratio of As before, so As already shown, the column will be adequate for strength. Now, checking for stability, and Thus, applied load which shows that the column is also adequate for stability.

Geschwindner
1-18
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

LeMessurier Approach: Since the W10x39 column was shown to be adequate, it will again be checked. Using the values obtained above, for column AB,

Thus, with restraint.

from above, the W10x39 will be sufficient to provide lateral

Modified Nomograph Equation: An iterative solution of Eq. 5 with Q and P as given above yields As can be seen, this value compares quite well with the values already obtained and the column will be adequate.

Commentary Equations: For the first simplified equation, Eq. 16, as would be expected from the derivation shown above, since there is only one restraining column.
The use of the second simplified equation, which is based on Eq. 19, requires an analysis of the structure to determine the ratio of lateral displacement to load. Since the analysis carried out for the previous approaches assumed that the pin connection at the base was not a true pin but one which resulted in this must be included in the analysis. With this provision accounted for, an arbitrary lateral load of 5 kips results in a deflection in. Using Eq. 20 yields

Using the equation from the second edition commentary, Eq 21, results in

Geschwindner
1-19
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

The proposed equation for the third edition, in the form of Eq. 22, yields

Using any of these K values, the column will prove to be adequate to provide the required lateral resistance. The results obtained from Eqs. 20 and 21 appear to be quite similar to those determined from the other approaches presented. The result from Eq. 22 appears to be somewhat distant from the other results, although the column will still prove to be adequate.
Example 2: The frame shown in Fig. 10 was used by Cheong-Siat-Moy (22) to show that Yura's approach would not work and by de Buen (23) to present his new approach. The truss is assumed to provide sufficient rotational restraint at the top to permit that end of column AB to be treated as a fixed end while the bottom of the column is pinned, thus from the nomograph, Recognizing that Yura's approach will yield a larger column than required and that LeMessurier's approach will give the correct results, as shown by the two previously cited references, the Lim & McNamara approach will be compared to the LeMessurier approach. For this example, the column will be taken as a W12x136 and 50 ksi.

Figure 10. Frame for Example 2

Lim & McNamara: With one half of the leaning column load resisted by the single restraining column, n = 990/330 = 3.0. Thus,

Using thus from LRFD E2-3, = 8.48 ksi, thus Since this approach is essentially the same as the Yura approach, this result is not entirely unexpected. LeMessurier: For a fixed-pinned column, and so that Eq. 12 yields:

Geschwindner

1-20
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

as shown in Ref. 22 and 23.

Using thus thus from LRFD E2-3,

Modified Nomograph Equation: Using G values consistent with a true fixed end and a true This pinned end, along with an iterative solution of Eq. 5 results in result is similar to that previously obtained from the LeMessurier approach.
Commentary Equations: As was shown for example 1, where only one column provides lateral support, the results from Lim & McNamara and the simplified equation, Eq. 16, are identical, The LeMessurier analysis, Eq. 12, assumed that the columns were fixed at the upper end and had true pins at the lower end. This results in which is the assumption used to develop the second simplified equation, Eq. 20. With the lateral displacement calculated for a cantilever beam with a 5 kip load, in. This yields from Eq. 20 and from Eq. 22. The results from Eq. 20 are identical to the previously calculated value using the complete LeMessurier approach and the results from Eq. 22 are the same as that from the Lim & McNamara approach, as would be expected from the derivations of the simplified equations.

Example 3: An interesting problem originally proposed by Zweig (24) is shown in Fig. 11. A portion of a large, unbraced one story industrial building with deep roof trusses is shown. The trusses, which frame in each direction, exhibit an infinitely large stiffness when compared to the columns. In order to equalize sway restraint in each direction, alternate columns have their strong axes turned 90. The nomograph approach with and = 10 yields Without considering leaning columns and using the W12x65 proposed by Zweig would be slightly undersized for the required By this approach, each column in the building would be the same, designed to carry 234 kips about its weak axis. However, if the strong axis column is used to brace the weak axis column, there should be some savings available. Each of the approaches previously discussed will be used to check a W12x53 column to determine whether this smaller column would be adequate. Throughout this example, the columns will be taken as pairs, one strong axis and one weak axis column.
Yura Approach: Using the capacity of the W12x53 for sway buckling about the x-axis, column 1, and the y-axis, column 2, will be added. The combined capacity must be equal to or greater than the total load 2(234) = 468 kips. For
The

Geschwindner
1-21
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

Figure 11. Frame for Example 3

combined capacity is 432 + 127 = 559 kips which is greater than the 468 kips combined load. Thus, the W12x53 will be acceptable for all of the columns.
Lim & McNamara Approach: If column 1 is taken as the restraining column and column 2 as the leaning column, n = 234/234 = 1.0. Thus, for column 2, K = 1.0 and KL = 1.0(20) = 20 ft. For the W12x53 buckling about the y-axis, For column 1, and For buckling about the x-axis, 280 kips > 234 kips. Thus, the W12x53 is adequate for both columns in each direction. LeMessurier Approach: With and for both columns 1 and 2, and

Thus, from Eq. 12

which yields

Thus, for column 2, thus and For column 1, which is the same as for column 2 so This shows that the W12x53 column is adequate for both carrying the load and providing lateral restraint. Modified Nomograph Equation: In this case, with end stiffness and load ratio as for the other approaches, Eq. 5 yields for the restraining column. The leaning column would use K=1.0 as with the Lim & McNamara approach. Again, the W12x53 will be acceptable.

and

Commentary Equations: For the first simplified equation, Eq. 16, columns. Thus,

for both

Geschwindner

1-22
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

As was the case in example 1, the application of second simplified equation requires the analysis of the structure with a connection stiffness at the base that will result in With a 5 kip lateral load, the resulting lateral deflection, Thus, from Eq. 20,

while Eq. 21 yields

and Eq. 22 yields

As can be seen from these K values, the W12x53 will work satisfactorily for all columns in the structure when checked through these simplified equations. The significance of this example, as it was when originally published, is to show that the sharing of lateral stiffness can be used to the advantage of the structure.
Example 4: The frame shown in Fig. 12, introduced by Geschwindner (25), will be used to compare the simplified methods for determination of effective length factors with a true buckling analysis. The frame is supported in such a way that in plane behavior will be critical. The columns AB and CD as well as the beam BC are W12x136. The other members are of such a size that their individual characteristics will not control. The results of a GTSTRUDL buckling analysis for equal loads on columns AB and CD yields Using Eq. 3 this is equivalent to K=2.232. The nomograph equation yields K=2.166 or

Figure 12. Frame for Example 4 with Leaning Columns When equal loads are also applied to columns EF, GH, and JK, GTSTRUDL yields 528.7 kips or K=3.414. The loading on the structure shows two equal loads on the restraining columns and three equal loads on the leaning columns. This gives P = 2 and Q = 3. From the modified nomograph equation, K=3.289 and Eq. 11 yields K = 3.425 and The LeMessurier equation, using G = 100,000 to represent the pin end, yields K = 3.295 and
Geschwindner
1-23
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

No matter what approach is taken to account for the leaning columns, it is clear that they have a significant impact on the stability of the structure. It is also evident that a second-order elastic analysis will yield the same forces for members AB, CD, and BC, whether there are loads on the leaning columns or not. Thus, more than a second-order elastic analysis is needed for the complete design of the structure. Example 5: Factored loads are now applied to the frame of example 4, as shown in Fig. 13. First- and second-order elastic analyses are performed and, along with the results from example 4, a check on column CD, with is carried out.

Using the results from the first order analysis, and and from the LeMessurier analysis including the leaning columns, K=3.295 and From this, so LRFD Eq. H1-1a is used. Since the column moment is from a first order analysis, it must be amplified. This will be accomplished using the second suggested equation for Since there are no moments in the non-sway analysis, The results of the first-order analysis give a lateral deflection due to the 20 kip load of 1.738 in. Using these values, so that The interaction equation becomes

Since this is less than 1.0, the column will be adequate.

If the results of the second-order analysis are used, and ft. Again, the effect of the leaning columns will be included from the LeMessurier analysis so that, and LRFD Eq. H1-1a is used again. Since the column moment results from a second-order analysis, there is no need to amplify it prior to using the interaction equation, thus

Geschwindner
1-24
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

Again, the column is seen to be adequate. It is interesting to note that there is an increase in the column axial load due to the second order effects that is not included in the simplified code approach to second-order analysis and that the second-order moments obtained from the two approaches are quite similar.
Example 6: An interesting structure was presented by Baker (26) to demonstrate the problems associated with effective length determination when the assumptions of the nomograph are violated. The frame shown in Figure 14 represents one of two frames

Figure 14. Example 6 Frame with lateral service load participating in providing lateral resistance for the building. This frame carries a gravity load of and provides lateral stability to columns carrying an additional = 1875 kips, both being one half of the total load for the building. The lateral deflection of the frame due to is with The total gravity load is and

The results presented by Baker were in the form of column capacities, However, a review of his equations shows that his solution actually uses Eq. 21. Table 1 shows the Kfactors for each of the five columns in this frame as determined through nomograph, the four simplified equations, and an elastic buckling analysis. It can be seen from Table 1 that the use of the nomograph, Eq. 4, does not predict effective length values that would subsequently produce accurate elastic buckling values for the columns of this frame. This is due to the fact that this frame significantly violates the assumptions used to develop the nomograph. If this structure had been designed with those values, the columns would have had an expected capacity significantly larger than their true capacity.
The goal in determining an appropriate K-factor is to predict the results of a linear buckling analysis. The results of a linear buckling analysis from GTSTRUDL, converted to effective length, are also given in Table 1. The values determined from Eqs. 16, 20, and 21 appear to be fairly consistent although slightly below the elastic buckling values. The

Geschwindner
1-25
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

results of Eq. 22 are significantly higher than the elastic buckling values, although they will yield a conservative solution.

Table 1. Summary of Effective Length Calculations for Example 6

Conclusions

Structural systems which combine unbraced frames with simple gravity columns provide the structural engineer with a system requiring special attention. In addition, unbraced frames that do not meet the restrictive assumptions permitting use of the nomograph also pose interesting problems. It is clear that leaning columns result from the structural framing arrangement, not from the use of a particular design philosophy. It is also clear that a specific accounting must be made for the leaning column, even when a second-order analysis has been carried out. In addition, the design philosophy and the analysis approach must be compatible.
This paper presented a brief discussion of the full range of approaches that might be used to carry out a structural analysis. It is well understood that the critical buckling load of a frame can be determined through an eigenvalue analysis. It is this load that the K-factor is attempting to predict. Although the K-factor has been a controversial topic from its initial introduction, it remains a useful tool to measure column capacity. Perhaps the most troubling aspect associated with the use of the K-factor has been the assumptions included in the most common predictor equations. Four approaches from the literature for determination of the K-factor were presented, along with several simplified equations derived from those procedures. It was shown that through an iterative solution of Eq. 5, a more accurate value of could be obtained than that from the nomograph, Eq. 4, when leaning columns are present. In this case, the leaning column loads are accounted for; however, the other limitations of the nomograph solution are still present.

The equations proposed by LeMessurier are generally recognized as the most accurate of those presented. There are two approaches to the use of the LeMessurier equations. One requires the determination of which may be accomplished through the nomograph, as is
Geschwindner 1-26
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

normally done, or by an iterative solution of equation, Eq. 4. The other approach uses the lateral stiffness of the frame, as measured by its lateral deflection due to a lateral load. Either of these approaches will provide a practical solution to determining column capacity. Thus, it is not unrealistic to use the LeMessurier equations for effective length factors in normal engineering practice.
The commentary to the LRFD Specification provides simplified equations, based on the LeMessurier equations, which are felt to be useful to the designer. The examples presented here allow for a comparison of results between several of these equations. It appears that there is some wide variation in results, depending on the choice of approximation. The assumptions used to develop these simplified equations are presented so the engineer will be in a better position to decide which should be used in a particular situation.

It was also shown that the use of a second-order analysis does not automatically account for the presence of leaning columns, but does meet the need for determination of moments for use in the interaction equations. Thus, both the second-order effects and buckling capacity must be determined for the analysis and design to be complete. It is simply a matter of deciding which approach is to be used.
Although the LeMessurier approach is not overly complicated to use, designers wishing to use an even simpler approach may find that the Lim & McNamara equation for provides a sufficiently accurate way to account for leaning columns, particularly in preliminary stages of design. In addition, the LeMessurier equation, which is based on the lateral deflection of the frame, provides a straightforward approach to the actual calculations, as suggested by Baker. Although simplified equations are presented in the Commentary to the 1993 LRFD Specification, and will be included in the 1999 version, there is really no need to use them. Once the elastic-buckling load of the frame has been determined and the appropriate amount attributed to the individual columns, design by any approved method may proceed.

References

1. West, Harry H., Analysis of Structures. John Wiley and Sons, New York, 1989.
2. Galambos, T. V., Structural Members and Frames. Prentice-Hall, Inc., Englewood Cliffs, NJ, 1968. 3. Disque, R. O., Applied Plastic Design in Steel. Van Nostrand Reinhold Co. New York, 1971.
4. Chen, W. F., Goto, Y., and Liew, J. Y. R., Stability Design of Semi-Rigid Frames. John Wiley and Sons, New York, 1996.

5. Chen, W. F. and Toma, S., Advanced Analysis of Steel Frames, CRC Press, Boca Raton, 1994.

Geschwindner 1-27
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

6. Load and Resistance Factor Design Specification for Steel Buildings. 2nd edition, AISC, Chicago, IL, 1994.

7. Specification for Structural Steel Buildings - Allowable Stress Design and Plastic Design. AISC, Chicago, Ill., 1989. 8. AISC, "1998 Structural Engineering Software Survey," Modern Steel Construction. AISC, Jan. 1998, pp 62-63. 9. STAAD-III Structural Analysis and Design Program User's Manual. Revision 16, Research Engineers, Inc., Marlton, N.J., 1992 10. GTSTRUDL User's Manual. Georgia Tech Research Corp., Atlanta, GA,

11. RISA-2D User's Guide, RISA Technologies, Lake Forest, CA.

12. ETABS User's Manual, Computers and Structures Inc., Berkeley, CA. 1988. 13. Wilson, E.L. and Habibullah, A., "Static and Dynamic Analysis of Multi-story Buildings including P-Delta Effects," Earthquake Spectra Journal, EERI, Vol. 3, No.2, 1987, pp.289-298.

14. The RAM Structural System, RAM Frame, Version 6, RAM International, Carlsbad, CA, 1998.
15. Higgins, T. R., "Effective Column Length - Tier Buildings", Engineering Journal, AISC Vol. 1, No.4, 1964, pp. 12-15.

16. Geschwindner, Louis F., "A Practical Approach to the Leaning Column", Engineering Journal, AISC Vol. 31, No. 4, 1994, pp. 141-149.
17. Chen, W.F. and Lui, E.M.. Stability Design of Steel Frames, CRC Press, Boca Raton, FL., 1991.
18. Yura, J. A., "The Effective Length of Columns in Unbraced Frames," Engineering Journal, AISC, Vol. 8, No. 2, 1971, pp. 37-42.

19. Lim, L. C. and McNamara, R. J., "Stability of Novel Building System," Structural Design of Tall Steel Buildings. Vol. II-16, Proceedings, ASCE-IABSE International Conference on the Planning and Design of Tall Buildings, Bethlehem, Pa. 1972, pp. 499-524.
20. LeMessurier, Wm. J., "A Practical Method of Second Order Analysis," Engineering Journal, AISC, Vol. 14, No. 2, 1977, pp. 49-67.

Geschwindner
1-28
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.

21. Higgins, T. R., "Column Stability under Elastic Support", Engineering Journal, AISC, Vol. 2, No. 2, 1965, pp. 46-49. 22. Cheong-Siat-Moy, F., "Column Design in Gravity-Loaded Frame," Journal of Structural Engineering, ASCE, Vol. 117, No. 5, May 1991, pp.1448-1461.

23. de Buen, Oscar, "Column Design in Steel Frames Under Gravity Loads," Journal of Structural Engineering, ASCE, Vol. 118, No. 10, October, 1992, pp. 2928-2935.
24. Zweig, A., Discussion of "Column Stability Under Elastic Support," by T. R. Higgins, Engineering Journal, AISC, Vol. 2, No. 3, 1965, pp. 105-106.

25. Geschwindner, L.F., "Practical Design of Unbraced Moment Frames with Leaning Columns," Proceedings of Structures Congress 13, Boston, Mass., April 2-5, 1995, ASCE, pp. 527-542. 26. Baker, W. F., "Practical Problems in Stability of Steel Structures," Proceedings of the National Steel Construction Conference, Chicago, IL, AISC 1997, pp. 2.1-2.24.

Geschwindner 1-29
2003 by American Institute of Steel Construction, Inc. All rights reserved. This publication or any part thereof must not be reproduced in any form without permission of the publisher.