You are on page 1of 19

Evaluation of the Effective Length Method and the Direct Analysis Method for the

Design of Steel Columns in Frame Structures.

Author
Justin Prajzner
University of Maryland
College Park, MD 20742
Fall 2006

Advisor
Ricardo A. Medina, Ph.D.
Assistant Professor
Department of Civil and Environmental Engineering
University of Maryland
College Park, MD 20742

1
1. Introduction

Current AISC (American Institute of Steel Construction) design specifications [1]


provide engineers with different methods to design columns that form part of steel frame
structures. Traditionally, the Effective Length Method (ELM) has been used in the design
of steel columns. This method relies on the use of effective length factors, K, that account
for the contribution of boundary conditions to the axial load carrying capacity of a steel
column. Because the ELM is based on several assumptions on geometry, boundary
condition, and material properties of columns, it may not always be appropriate for the
design of steel columns, especially for frame configurations in which the conditions of a
given column are not consistent with these assumptions. In an attempt to provide a design
methodology that would more accurately capture the main factors that affect column
behavior, the Direct Analysis Method (DAM) was introduced into the steel design
specifications in 2005 as an alternative to the more traditional ELM. As stated by Andrea
Suroveck [2], a major advantage of the DAM is its ability to account for member and
construction imperfections within a frame which creates additional stresses and reduces
the load-carrying capacity of the structure. In this context, imperfections have to do with
member camber, out-of-plumbness of the frame, and bearing elevations. To capture the
effects from imperfections, the DAM relies on the use of notional loads that are applied
to the structure in the horizontal direction. These notional loads are meant to mimic the
influence of second-order effects in the structural response, which could be viewed as
effects that reduce the effective lateral stiffness of a frame. The DAM is based on
calculations that assume an effective length factor, K, equal to one, which significantly
reduces the number of calculations as compared to those required to design a column
based on the ELM.

Both the ELM and the DAM use column interaction equations to estimate the capacity of
individual steel columns. However, because these two methods are different, they will
not necessarily produce the same column sizes for a given structural configuration. If
substantial differences in size are produced by the two methods, the adequacy of each
method to provide structural members with sufficient capacity to resist the imposed
demands becomes a concern. In order to assess the adequacy of both the ELM to the
DAM, this study deals with the evaluation of case studies that correspond to steel
columns that form part of various structural configurations. These case studies include a
portal frame, a leaning column frame, a multi-story structure, and a multi-bay frame. If
any of these frames produce significant differences in the member sizes when the two
design approaches are implemented, an additional analysis method will be needed for
comparison. This additional method should provide a reasonable representation of real
frame behavior, which includes the incorporation of material and geometric nonlinearities
in the analysis. In this study, the basic analysis approach used to quantify frame behavior
and assess the adequacy of design methods is a 2nd order plastic analysis. As it is shown
in subsequent sections of this paper, in most cases, both the ELM and the DAM provide
consistent results that adequately capture the most important characteristics that control
the behavior of a steel frame. In addition, an evaluation of cases and conditions in which
each one of these methods may not be appropriate is also conducted.

2
2. Overview of Design Methods

2a. Effective Length Method (ELM)

The ELM accounts for the influence of the total frame on the behavior of an individual
column. In order to do this, alignment charts are used to determine the effective length
factor of a member. However, these alignment charts are based on several assumptions as
stated in the AISC specifications. These assumptions are presented next:

(1) Behavior is purely elastic.


(2) All members have constant cross section.
(3) All joints are rigid.
(4) For braced frames, rotations at opposite end of beams are equal in magnitude,
producing single curvature.
(5) For unbraced frames, rotations at opposite ends of the restraining beams are
equal in magnitude, producing reverse-curvature bending.
(6) The stiffness parameters of all columns are equal.
(7) Joint restraint is distributed to the column above and below the joint in
proportion to I/L of the two columns.
(8) All columns in a frame buckle simultaneously.
(9) No significant axial compression force exists in the girders.

As expected, in many cases, the frame to be analyzed does not comply with all the
assumptions listed above, yet designers still use this method to design frame structures.
A significant flaw with this method is that for many frame structures to ‘fail’, several
members within the structure need to fail first. Therefore, the load-carrying capacity for
the entire frame is overestimated. In many situations, this causes the ELM to be too
conservative, which tends to counterbalance the effect of inconsistencies between frame
behavior and the assumptions that form the basis for this method. Another drawback of
this method is its inability to clearly account for the effects of imperfections, which
include the out-of-plumbness of the frame.

2b. Direct Analysis Method (DAM)

The DAM was introduced into the AISC Specification in an attempt to account for the
drawbacks present with the ELM. The DAM uses a 2nd order elastic analysis (EA) of a
frame with notional loads and a reduced stiffness to design members. The net effect of
incorporating notional loads and reduced stiffness properties is to approximately account
for member inelastic behavior, imperfections, and load re-distribution effects within the
frame structure. In this method, the 2nd order elastic analysis used to estimate demands is
conducted with the following loads and reduced stiffness properties:

(1) A notional load, Ni = 0.002Yi, is applied horizontally to the structure at each level
where Yi is the factored gravity load from that level. This 0.002 factor was
calibrated using plastic zone solutions that included the effects of slenderness, yield
strength, and end-moment restraints [3].

3
(2)The reduced stiffness of members that contribute to the stability of the frame is
given by EI* = 0.8τbEI, where τb = 1 for Pr/Py <= 0.5 or 4[Pr/Py (1- Pr/Py)] for Pr/Py >
0.5. Pr is the required axial compressive strength for the column in question and Py =
AFy.
(3)The reduced axial strength of members that contribute to the stability of the frame
is given by EA* = 0.8EA.

A major advantage of the DAM is that it simplifies the calculations by letting K be a


constant equal to 1, which eliminates the need for alignment charts. In addition, the
implementation of notional loads and stiffness-reduction factors provide a better
representation of the distribution of demands due to the member-system interdependence
and inelastic behavior of the frame.

3. Analysis Methodology

3a. 2nd Order Analyses

As stated before, results from 2nd order analyses are utilized for each case study to assess
the adequacy of the design method under consideration. In order to conduct second-order
analyses, the structural analysis program Mastan2 [4] was used. Mastan2 was selected
for its ability to capture the additional demands imposed by 2nd order effects. Two types
of second-order analyses were conducted in this study: a 2nd order elastic and a 2nd order
inelastic (plastic) analysis. For each type of 2nd order analyses, a solver based on
predictor-corrector calculations was utilized. Frame capacities for the plastic analyses
were determined once the first column had plasticized due to stresses developed from
both bending and compression. These numerical models used a concentrated plasticity
approach as opposed to a distributed plasticity one. Calculations conducted by the author
with frame structures similar to those used in this study where concentrated plasticity was
used provided frame capacities that were reasonable close to those obtained assuming a
distributed plasticity model. Each member within the frame was divided into 10 smaller
elements so that 2nd order effects could be calculated more accurately. Out-of-plane
buckling, lateral-torsional buckling, and local buckling were ignored.

3b. Frame Configurations

Each frame structure has a tributary width, i.e., beam spans of 35’with the same loading.
A uniformed load is applied across each horizontal beam. The basic design loads in all
analyses corresponds to a dead load equal to 80 psf and a live load of 40 psf. All
connections are either fully restrained or pinned. All frames are analyzed with four
methods:

(1) 2nd order plastic analysis with unfactored loads (PA).


(2) 2nd order plastic analysis with factored loads and a
reduced stiffness properties 0.9EI and 0.9EA (FPA).
(3) 2nd order elastic analysis with factored loads (ELM).
(4) 2nd order elastic analysis with factored loads including

4
notional loads (DAM).

The FPA is based on the work of Maleck [3]. The purpose of providing stiffness
reduction factors is to account for the effect of load factors on the calculated demand-to-
capacity ratios. An out-of-plumbness of L/500 [1] was utilized in all cases to account for
the effect of geometric imperfections. In concept, the PA can be considered the most
accurate model and the FPA a reasonable representation of ‘true’ behavior. To validate
the adequacy of the stiffness-reduction factors in the FPA, the PA along with the ELM
and DAM will be used.

3c. Validation of FPA

In order to use FPA on different frames as a basis for design, the validation of the
stiffness-reduction factors of the FPA is conducted by analyzing a simple portal frame as
shown in Fig. 1.

.
Figure 1 Portal Frame

Table 1 and 2 show the ratio of total load that caused failure under the ‘Failure’ column
and the reciprocal of that ratio under ‘Pu/Pf’ column for various beams and column sizes.
The loading was applied to the structure in incremental steps of 1% of the total applied
load to the structure. After 100% of the factored load or unfactored load, the program
continues to apply a larger percent until the structure fails. The failure is written in terms
of the ratio of (% Load that causes failure / 100% Load, Pf / Pu). A clear relationship
between the two methods is obvious if the failure ratios of the FPA to PA are compared.
This is shown in Table 3.

5
Table 1
PA
Column Beam Failure Pu/Pf
W33x118 W18x175 4.28 0.233645
W27x84 W24x104 2.84 0.352113
W18x76 W14x90 1.93 0.518135
W18x65 W14x90 1.54 0.649351

Table 2
FPA
Column Beam Failure Pu/Pf
W33x118 W18x175 2.89 0.346021
W27x84 W24x104 1.92 0.520833
W18x76 W14x90 1.32 0.757576
W18x65 W14x90 1.04 0.961538

Table 3
FPA Failure/ PA Failure
0.675234
0.676056
0.683938
0.675325

The ratios obtained in Table 3 are very similar to one another with a mean ratio of 0.678.
The reciprocal of the mean ratio is 1.476, which can be thought of as a “Safety Factor”.
This “Safety Factor” is used as a basis to evaluate the reliability obtained using code-
compliant design methods.

Table 4
ELM
Pu Mu φPn φMn
Columns Beams (kips) (kip-in) (kips) (kip-in) Ratio
W33x118 W18x175 98.07 5955 1432.5 18675 0.353106
W27x84 W24x104 98.11 5348 1010 10980 0.535637
W18x76 W14x90 98.27 5769 873 7335 0.842786
W18x65 W14x90 98.31 5549 744 6650 0.900505

Table 5
DAM
Pu Mu φPn φMn
Columns Beams (kips) (kip-in) (kips) (kip-in) Ratio
W33x118 W18x175 98.15 5971 1445.3 18675 0.353687
W27x84 W24x104 98.21 5366 1023 10980 0.536708
W18x76 W14x90 98.41 5791 919.9 7335 0.842992
W18x65 W14x90 98.46 5572 763.9 6650 0.902340

6
Table 4 and 5 show the results for the ELM and the DAM. The term ‘Ratio’ in the last
column is calculated from the design beam-column interaction equations. In concept, this
‘Ratio’ should provide a reasonable representation for the safety factor in the design of
each member. In this case, both the ELM and the DAM produced similar results for this
case. For a correlation to be made between the FPA to the ELM and the DAM, the ratio
Pu/Pf is compared to the ‘Ratio’ obtained using the ELM and DAM (Fig. 2). It can be
observed that the ELM and the DAM correlate well with the FPA. As a result of this
comparison, two important observations can be made:

(1) The ELM and DAM when used to analyze a simple frame produce similar
results.
(2) The FPA can be used as a basis to compare the accuracy of the ELM and the
DAM to estimate the reliability of a design.

With the FPA as a good basis for design, it can be used to compare the ELM and the
DAM when designing a frame structure that may be susceptible to increased stress due to
the presence of imperfections. This is the subject of the following section.

Ratio Comparison for Simple Frame

1
0.9

0.8
0.7
0.6
Applied
0.5
Load/Capacity Elastic
0.4
Notional Load Analysis
0.3 Plastic .9E +.9Fy
0.2
0.1

0
W33x118 W27x84 W18x76 W18x65
Columns

Figure 2 Portal Frame Capacities

4. Case Studies

4a. Case 1: Frame with a Leaning Column

In this case, a column leaning with a slope of 2.25, rigidly attached to a beam, W27x84,
is designed as shown in Fig. 3. The beam is supported at the other end by a pin/pin
connected column, W10x33. The length of the beam is 27’ while the height of beam at
both ends is 18’. This frame geometry was chosen because the slope of the leaning
column may affect the ability for the two design methods to adequately predict the
capacity its capacity.

7
Figure 3 Leaning Column Portal Frame

Table 6
FPA
Columns Failure Pu/Pf
W10x54 0.80 1.2500
W10x60 0.88 1.1299
W10x68 0.99 1.0101
W10x77 1.11 0.9009
W10x88 1.25 0.8000
W10x100 1.39 0.7194

Table 7
ELM
ΦPn Pu ΦMn Mu
Columns g k λ (kips) (kips) (kip-in) (kip-in) Ratio
W10x54 0.1459 1.18 0.842265 528.3 59.23 2997 3698 1.289953
W10x60 0.1642 1.18 0.838428 590.1 59.48 3357 3729 1.161209
W10x68 0.1897 1.19 0.836011 671.7 59.78 3839 3780 1.029256
W10x77 0.2191 1.20 0.833649 760.3 60.09 4392 3843 0.914517
W10x88 0.2572 1.20 0.824467 876.9 60.43 5085 3931 0.807514
W10x100 0.3001 1.21 0.820494 998.1 60.79 5850 4032 0.719683

Table 8
DAM
ΦPn Pu ΦMn Mu
Columns λ (kips) (kips) (kip-in) (kip-in) Ratio
W10x54 0.71378 574.45 58.58 2997 3797 1.317921
W10x60 0.71053 641.14 58.87 3357 3822 1.184427
W10x68 0.70253 732.02 59.22 3839 3866 1.047613
W10x77 0.69470 830.98 59.57 4392 3924 0.929286
W10x88 0.68705 956.54 59.95 5085 4007 0.819341
W10x100 0.67809 1091.3 60.37 5850 4105 0.729367

8
Leaning Column Design

1.4
1.2
1
0.8
Ratio
0.6
0.4 Factored Plastic Analysis
0.2 Alignment Chart Analysis
0 W W W W W W
Direct Anlysis
1 1 1 1 1 1
0 0 0 0 0 0
x x x x x x
5 6 6 7 8 1
4 0 8 7 8 0
0

Columns

Figure 4 Leaning Column Capacities

Table 7 and 8 show the results for both the ELM and the DAM. Each of the analyses
predicted similar results with the W10x77 as the column of choice. In Figure 4, both the
ELM and the DAM remained slightly conservative as compared to the results for the FPA
in Table 6. It is clear in this case that each of these methods correlate well and are
adequate design methods.

4b. Case 2: Multi-Story Structure

The general configuration of the multi-story, moment-resisting frame structure is


presented in Fig. 5. The beams correspond to W27x84 sections and have a span of 35 ft.
The story heights are each 15 ft. This frame is utilized to evaluate the contribution of out-
of-plumbness along several story heights on the design of steel columns.

TC

MC

LC

Figure 5 Multi-Story Frame

9
Table 9
FPA
Columns Failure Pu/Pf
W12x58 1.10 0.9124
W10x54 1.06 0.9433
W10x49 1.02 0.9832
W10x45 0.97 1.0309
W10x39 0.90 1.1135
W10x33 0.81 1.2391
W12x26 0.68 1.4705
W12x22 0.61 1.6528

Table 10
ELM
Mu ΦPn ΦMn
Columns Section Pu (kips) (kip-in) (kips) (kip-in) Ratio Worst Ratio
W12x58 TC 98.23 3376 640.6 4320 0.858152 0.85815179
MC 196.1 2569 634.0 4320 0.837907
BC 294.1 1850 631.3 4320 0.846523
W10x54 TC 98.22 2679 573.1 3330 0.890196 0.93199024
MC 196.1 2173 566.6 3330 0.926146
BC 294.1 1512 556.6 3330 0.93199
W10x49 TC 98.22 2517 526.0 3020 0.926809 0.99614155
MC 196.1 2071 521.6 3020 0.985524
BC 294.1 1429 511.0 3020 0.996142
W10x45 TC 98.22 2383 484.8 2745 0.974265 1.06475425
MC 196.1 1983 482.0 2745 1.048984
BC 294.1 1359 470.8 2745 1.064754
W10x39 TC 98.22 2141 416.5 2340 1.049118 1.19249193
MC 196.1 1819 409.1 2340 1.170323
BC 294.1 1229 405.3 2340 1.192492
W10x33 TC 98.22 1874 350.5 1940 1.138877 1.34974864
MC 196.1 1627 348.4 1940 1.308334
BC 294.1 1081 344.2 1940 1.349749
W12x26 TC 98.22 2108 294.3 1860 1.341148 1.57649869
MC 196.1 1796 293.2 1860 1.52713
BC 294.1 1212 294.9 1860 1.576499
W12x22 TC 98.22 1759 246.6 1465 1.46557 1.79474335
MC 196.1 1541 248.1 1465 1.725409
BC 294.1 1016 249.6 1465 1.794743

10
Table 11
DAM
Mu ΦPn ΦMn
Columns Section Pu (kips) (kip-in) (kips) (kip-in) Ratio Worst Ratio
W12x58 TC 98.39 3397 663.6 4320 0.860476 0.86047611
MC 196.5 2605 663.6 4320 0.83212
BC 294.9 1894 663.6 4320 0.834106
W10x54 TC 98.37 2697 593.2 3330 0.892825 0.92117747
MC 196.5 2210 593.2 3330 0.921177
BC 294.9 1558 593.2 3330 0.913017
W10x49 TC 98.37 2535 540.0 3020 0.930487 0.98463944
MC 196.5 2109 540.0 3020 0.984639
BC 294.9 1476 540.0 3020 0.980548
W10x45 TC 98.36 2400 497.9 2745 0.97472 1.04930083
MC 196.6 2021 497.9 2745 1.049301
BC 295.0 1406 497.9 2745 1.047781
W10x39 TC 98.35 2158 429.2 2340 1.0489 1.17279534
MC 196.6 1857 429.2 2340 1.163475
BC 295.0 1278 429.2 2340 1.172795
W10x33 TC 98.33 1891 360.6 1940 1.139122 1.33675222
MC 196.6 1665 360.6 1940 1.308089
BC 295.0 1132 360.6 1940 1.336752
W12x26 TC 98.34 2125 297.5 1860 1.346086 1.59374718
MC 196.6 1834 297.5 1860 1.537304
BC 295.0 1260 297.5 1860 1.593747
W12x22 TC 98.32 1775 249.6 1465 1.470892 1.82929338
MC 196.6 1579 249.6 1465 1.745719
BC 295.0 1067 249.6 1465 1.829293

The FPA results seen in Table 9 predict that a W10x49 would be the last adequate
column for design. The ELM and DAM both conclude that the same column is also the
most efficient design as seen in Table 10 and 11. When comparing results from all four
analysis methods, the ELM and the DAM produced slightly conservative results when
compared to the FPA as shown in Fig. 6. Once again, the results obtained using the ELM
and the DAM are consistent with those obtained using the FPA. This observation is also
consistent with those made in 3c and 4a. The FPA has given consistent results for three
different frames and proven its effectiveness to judge the accuracy of the ELM and the
DAM.

11
Multi-Story Ratio Comparison

2
1.8
1.6
1.4
1.2 Factored Plastic
Ratio 1 Analysis
0.8
0.6 Effective Length
0.4 Method
0.2 Direct Analysis
0 W
1
W
1
W
1
W
1
W
1
W
1
W
1
W
1
Method
2 0 0 0 0 0 2 2
x x x x x x x x
5 5 4 4 3 3 2 2
8 4 9 5 9 3 6 2

Columns
Figure 6 Multi-Story Capacities

It is important to note that this frame, though small and not overly complex, became time
consuming to design when using the ELM. Due to the geometry of the frame, various
column sizes caused different columns to fail first for each analysis. As a result, the
calculation of every effective length factor for each column was needed; this resulted in 3
additional calculations for each change in column size when compared to the DAM,
which required only the calculation of the notional load - one additional calculation for
the entire design. With similar results from each method, the fact that the DAM required
less calculations makes this method a more efficient design alternative.

4c. Case 3: Multi-Bay Structure

A multi-bay structure was generated by expanding the simple frame utilized in section 3c
into 11 bays with additional pin/pin columns at every 35 ft (see Fig. 7). These new
columns are the same size as the two moment restrained columns in the center bay. Each
of the beams connecting the pin/pin columns are W27x102 simply-supported beams. The
beams which are rigidly attached to the two moment columns are W27x84. This 11-bay
frame was first developed by Maleck and was chosen because of its susceptibility to an
out-of-plumbness, which creates addition stresses in the central columns.

12
Figure 7 11-Bay Frame

Table 12
FPA
Columns Failure Pu/Pf
W10x54 1.306 0.76569
W10x49 1.243 0.80450
W10x45 1.143 0.87489
W10x39 0.980 1.02040
W10x33 0.817 1.22399
W10x30 0.797 1.25470
W12x26 0.859 1.16414
W12x22 0.690 1.44927

Table 13
PA
Columns Failure Pu/Pf
W10x54 1.934 0.517063
W10x49 1.840 0.543478
W10x45 1.694 0.590319
W10x39 1.453 0.688231
W10x33 1.210 0.826446
W10x30 1.180 0.847458
W12x26 1.273 0.785546
W12x22 1.022 0.978474

For FPA and PA, various sized columns were used with the failure ratios and Pu/Pf
calculated seen in Tables 12 and 13. The comparison of failure ratios can be made
between this frame and the frame in 3d as shown in Table 14.

13
Table 14
Ratio of Factored to
Unfactored
Portal 11-Bay
Frame Frame
0.675234 0.675284
0.676056 0.675543
0.683938 0.674734
0.675325 0.674467
0.675207
0.675424
0.674784
0.675147

The mean of the values in Table 14 for the 11-bay frame is 0.675. This nearly matches
the mean value corresponding to the portal frame (0.678). This implies that a similar
‘safety factor’ would result in both cases. Therefore, by using the factored plastic
analysis as a basis, the last predicted safe column is a W10x45.

Table 15
ELM
Pu Mu φPn φMn
Columns (kips) (kip-in) (kips) (kip-in) Ratio
W10x54 216.2 466.4 523.6 2997 0.551242
W10x49 216.2 422.7 476.1 2718 0.592345
W10x45 216.1 387.9 438.2 2471 0.632692
W10x39 216.0 329.7 376.7 2106 0.712559
W10x33 216.0 270.8 315.0 1746 0.823579
W10x30 215.9 268.9 293.5 1647 0.880731
W12x26 216.0 320.5 272.3 1674 0.963427
W12x22 215.9 245.6 226.2 1465 1.103483

Table 16
DAM
Pu Mu φPn φMn
Columns (kips) (kip-in) (kips) (kip-in) Ratio
W10x54 214.7 1126 561.6 2997 0.716264
W10x49 214.2 1257 511.0 2718 0.830265
W10x45 213.5 1501 470.8 2471 0.993436
W10X39 207.1 4067 405.3 2106 2.227556

Based on the information presented in Tables 15 and 16, the ELM predicts a safe column
design with a W10x26, while the DAM predicts a safe column design with a W10x45.
Comparing the two design methods for a W10x45, the difference in ratios is 0.993 to
0.633. The difference between the two columns size and properties is significant. These

14
sizeable differences in ratios can be seen for each one of the columns in the
aforementioned tables.

11-Bay Frame Ratio Comparison

2.5

R
a 1.5
t Unfactored Plastic Analysis
i 1
o Factored Plastic Analysis

0.5 Direct Analysis Method

Effective Length Method


0
W W W W W W W W
1 1 1 1 1 1 1 1
0 0 0 0 0 0 2 2
x x x x x x x x
5 4 4 3 3 3 2 2
4 9 5 9 3 0 6 2

Columns

Figure 8 11-Bay Frame Capacities

True Failure vs. Alignment Charts

1.2

R 0.8
a
t 0.6
i
o 0.4 Unfactored Plastic Analysis
Effective Length Method
0.2

0
W W W W W W W W
1 1 1 1 1 1 1 1
0 0 0 0 0 0 2 2
x x x x x x x x
5 4 4 3 3 3 2 2
4 9 5 9 3 0 6 2
Columns

Figure 9 11-Bay Frame PA and ELM Capacity Comparison

Figure 8 shows a comparison between all four analysis methods in this case. Using the
FPA as the benchmark to estimate design capacity, the two design methods can be
evaluated. For the DAM, the capacity ratios are similar up until the plastic design predicts
an unsafe member (failure). It is at this point that the DAM predicts a ratio of 2.23. This

15
does not match with the plastic analysis which is only slightly over failure. As a result,
the DAM produces a member that is more conservative than required. On the other hand,
the ELM predicts that the W12x26 is the most efficient member, but when compared to
the FPA, it is too liberal of a design. The FPA gives a capacity ratio of 1.164 which is
much greater than 1. Furthermore, when compared to the PA, as seen in Fig. 9, the ELM
closely follows the PA capacity ratio, and at one point is even greater than the ‘true’
failure ratio. One could argue that because the ELM correlates well with the results from
the PA, the ELM is the better design approach. But as was shown earlier in each section
3d, an appropriate “Safety Factor” of 1.476 must be applied. When comparing the ratios
for the ELM to the PA for this frame, a safety factor of 1.078 is calculated. This is not an
adequate safety factor for the design of beam columns.

4d. Case 4: 13-Bay Structure

Due to the inadequacy in properly designing columns with a reasonable safety factor, the
11-bay frame is adapted to see if a worse capacity ratio for the PA could occur. In order
to do this, a bay is added to each side of the structure with another leaning column. The
ELM and PA were evaluated (see Table 17).

Table 17
13 Bay to 11 Bay Comparison
Mu Capacity
W12x26 Columns Pu (kips) (kip-in) Ratio
13-Bay
Effective Length 216 320.5 0.963
True Plastic 0.873
Factored Plastic 1.316
11-Bay
Effective Length 216 320.5 0.963
True Plastic 0.786
Factored Plastic 1.168

In both cases, the ELM predicts the same capacity. It is seen though that the 13-bay
frame has a decrease in its PA available capacity. This occurs due to the added stress on
the moment-restrained columns caused by the out-of-plumbness of the frame. It is
evident that if additional frames are added to the structure, the ELM would provide
member sizes that would fail before the entire unfactored loads are applied. Though it is
not shown, in this case, the DAM produces a conservative column size since the added
bays would add an additional notional load. It now becomes apparent that the ELM is not
always an effective or adequate design approach, particularly when second-order effects
become significant.

16
5. Discussion & Conclusion

Various design methods are currently available to design steel columns that form part of
frame structures. The ELM is the traditional method used for analysis by designers.
Restrictions and assumptions inherent in the ELM do not make it appropriate for all
frame geometries. However, it is not clear when this design method will provide
unreasonable member sizes. Therefore, in this study, 2nd order plastic analyses of various
frame configurations were conducted to evaluate the ability of the ELM to provide an
adequate level of safety in the design of steel columns. It was concluded that the FPA,
which is based on factor loads and reduced member stiffnessess was appropriate to
evaluate the adequacy of design methods.

By using the developed FPA as the “true” behavior for the structure, the ELM’s accuracy
could be quantified. In several case studies with varying structural configurations, the
ELM provided results which reasonably matched the FPA. Such structural systems
included a multi-story frame and a leaning column frame. As expected, the ELM
provided accurate results for the multi-story structure since this configuration is similar to
that of the portal frame, which is consistent with the assumptions that form the basis of
the ELM. Moreover, the leaning column frame also produced accurate design results
when using the ELM, even though this structural configuration was not consistent with
the assumptions behind the development of this method. The implication is that the
conservativism present in the evaluation of empirical effective length factors may offset
the effects produced by the lack of consistency of a given structural frame configuration
with the assumptions present in the ELM.

The DAM, which has been newly introduced into the Specification, was developed so
stresses created by inaccuracies in construction and imperfections in structural members
would be accounted for when designing a frame. A major advantage of this method is its
simplicity, i.e., a relatively small number of calculations are required when designing for
compression members. In order to test the accuracy of the DAM and evaluate whether or
not this method correlates well with the ELM, the DAM was used on the same design
examples. Similar to the ELM, the DAM also provided results consistent with those
obtained using the FPA. The DAM in each example provided equally reliable results as
the ELM, which based on the results from this study, raises the question of why a newer
method would be needed if it provides results consistent with those obtained using the
traditional ELM. Based on the results from this study, the DAM has several advantages
with respect to the ELM: (a) it is simpler to implement, (b) it is more accurate when
second-order effects are significant.

The simplicity of the DAM was demonstrated with the design of the multi-story structure.
In this case, for each column capacity check, nine additional equations needed to be
calculated when using the ELM as compared to the DAM. The conclusion is that the
DAM has the advantage of being easier and faster to implement. The case study with the
11-bay frame demonstrated that the ELM, though adequate for most models, provided an
unreliable design, while the DAM produced adequately designed members. For a
designer, it is imperative that a member designed by any method has a capacity at least

17
equal to the expected loading conditions the frame may encounter during its service life.
Such considerations have been adopted into the AISC, and can be seen as Safety Factors,
Load Factors, and Resistance Factors. Without being able to neither predict every type of
load, including the magnitude of that load, nor predict every member imperfection and
construction flaw, the method of analysis must produce conservative results. Therefore,
if a specific method of analysis provides design that either is not "safe enough," or for
that matter would fail under just the service loads, this method must be restricted and
boundaries must be specified for its implementation. The ELM produced ‘unsafe’
designs for the case of the multi-bay structure analyzed in the third cases study and
should be restricted in such cases. In this case, second-order effects due to the
combination of loading and imperfections became significant and the ELM was unable to
appropriately account for them. These second-order effects could be quantified by using a
“sway” factor. The sway factor should be related to the total area loaded that contributes
to the capacity of the moment-restrained column along with other variables such as
geometry of the frame, intensity of the load, and the loading pattern. All of these factors
and the way they affect the “sway” factor are not completely understood at this time, and
additional research is needed to have a better understanding of the role of second-order
effects on current design approaches and quantify appropriate “sway” factors.

In concept, the “sway” factor for each column may be reduced by adding additional
moment-restraining columns in the structure. Stresses which develop due to the out-of-
plumbness of the structure are now distributed between all the moment-restraining
columns. In addition to distributing the stresses, the added columns would stiffen the
structure, which then would reduce the stresses developed from 2nd order effects.

The DAM captures the 2nd order effects due to out-of-plumbness, but at times can
significantly overestimate these effects. As seen in the 11-Bay frame example, when the
design capacity ratio is near 1.0, the design becomes overly conservative with a ratio of
2.23. The DAM could be improved in situations where the sway factor is substantial
with a reduction in the magnitude of the specified notional load. This reduction factor,
possibly developed from a quantification of the sway factor, would make the DAM a
more accurate method when designing framing structures.

The results from this study demonstrated that the FPA can be used to evaluate and
quantify the capacity of individual steel frames. In addition, in general, the DAM when
compared to the ELM, was just as accurate or better. The DAM’s calculations were also
easier and fewer. The DAM, though not perfect, is a step toward more accurate analysis
approaches which will allow the designer to have more confidence in his or her design. It
would be valuable if a clear definition of the sway factor could be derived so that the
DAM could be more accurate in all frames. It is with further research that new and old
analysis methods can be tested and modified so that the design of structures can become
more accurate and reliable.

18
6. References

1. AISC (2005), Specification for Structural Steel Buildings, American Institute of


Steel Construction, Inc., Chicago, IL, March 9, 2005.

2. Suroveck,S.E. and Zienmian, R.D. (2005), “The Direct Analysis Method:


Bridging the Gap from Linear Elastic Analysis to Advanced Analysis in Steel
Frame Design”, Structures Congress, New York, April 2005.

3. Maleck, A.E., “Alternative Approaches for Elastic Analysis and Design of Steel
Frames”, Journal of Structural Engineering, ASCE, August 2004. pp 1186-1205.

4. Zienmian, R.D. and McGuire, W., Mastan2 v2.0, John Wiley and Sons Inc. 2002.

19

You might also like