Direct Analysis Method – Part I

Solver Prerequisites for Accuracy of Solution

by

Dr. Siriwut Sasibut

(Application Engineer)

and

Kenneth Kutyn
(Jr. Application Engineer)

S-FRAME Software Inc. S-FRAME Software, LLC

#275 - 13500 Maycrest Way #282, 800 Village Walk
Richmond, B.C. Guilford, CT 06437
CANADA USA
V6V 2N8

Phone: 1-604-273-7737 Phone: 1-203-421-4800

Fax: 1-604-273-7731

Please contact us at support@s-frame.com for more information about this publication.

nd
© Copyright by S-FRAME Software Inc. – Version: 5.0 August 2 , 2011
 2

The Direct Analysis Method

Specifications for Structural Steel Buildings, ANSI/AISC 360-10, published by the American Institute
of Steel Construction, outlines the Direct Analysis Method (DM) and specifies its use in the design of steel
structures. DM is defined by AISC as “a design method that captures the effects of residual stresses and
initial out-of-plumbness of frames by reducing stiffness and applying notional loads in a second order
analysis.” DM also accounts for uncertainty in material strength and stiffness as well as stiffness reduction
due to inelasticity. Flexural, shear and axial deformations are considered, and, when necessary, included in
DM.

According to AISC, “recent research has shown that the AISC direct analysis method … is the best
approach to cover all relevant response effects.”

Second Order Analysis

Second order analysis, as required by the direct analysis method, is a means used by engineers to
uncover deformations and forces that would otherwise be overlooked in conventional linear static analysis.
Two common second order effects include P-∆ effects and the less-known P-δ effects.

P-∆ effects, also known as ‘big’ P-delta, refer )to the increase in moment experienced by a fixed
connection at the ith end of a member when the jth end is loaded axially while simultaneously undergoing
lateral translation. Put more simply, one can imagine a cantilever beam with both axial and transverse
loading. Simple static analysis tells us that the fixed end only experiences a moment equal to the magnitude
of the transverse loading multiplied by the length of the member. However, P-∆ analysis reveals that an
additional moment is created due to the application of the axial load (P) no longer lining up with the
support (it is displaced by an amount, ∆). This can be seen in Figure A.

th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011
 3

Figure A: P-∆ Effects Figure B: P-δ Effects

Figure B represents a case where P-δ effects could be considered. Rather than increase the end
forces, P-δ effects (or ‘little’ P-delta) cause an increase in the member internal forces. This is a result of an
axial force acting through lateral translation of a member relative to its own longitudinal axis.

Both P-∆ and P-δ may or may not be negligible in the analysis of a structure. It is the job of a
structural engineer, in consultation with ANSI/AISC 360-10, to decide if they should be included.

S-FRAME

New in S-FRAME version 10.0 is the ability to use the direct analysis method, as outlined in
ANSI/AISC 360-10. Unlike some software, S-FRAME requires only one solution run to perform the direct
analysis method. Additionally, S-FRAME reports the calculated value of B2: the ratio of second order effects
to first order effects. B2 is useful for engineers trying to determine whether consideration of first order
effects alone might be adequate for their analysis.

As mentioned, the use of DM requires a rigorous second order analysis in order to meet the code’s
requirements. ANSI/AISC 360-10 provides two benchmark cases that allow us to “determine whether an

th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011
 4

analysis procedure meets the requirements of rigorous second-order analysis adequate for use in the direct
analysis method.” These cases are quite similar to the P-∆ and P-δ examples presented above. This should
come as no surprise due to the role these second-order effects play in the direct analysis method. If we
model and analyze these cases in S-FRAME, it should be possible to demonstrate that S-FRAME’s
implementation of P-∆ and P-δ analysis is rigorous and adequate. It is worth noting that these are designed
to be first level checks only and that additional checks are available1.

ANSI/AISC 360-10 requires analysis and theoretical values to be within 3% agreement for moments
and 5% for deflections.

Case One

Figure C: Case One

Case One, as seen in Figure C, is designed to reveal errors in calculating P-δ effects in a given analysis
method. As there is no lateral translation of either end, P-∆ effects do not enter consideration. In this case,
the uniformly distributed load causes the member to deform to the right. This deformation results in the
additional internal member forces discussed earlier. As this case investigates moment and deflection at
midspan, we modeled it using two beam elements of equal length. This way, a node was located at our

1
Kehler et al. (2010), Chen and Lui (1987), and McGuire et al. (2000)
th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011
 5

point of interest. Table 1 compares the theoretical deflections and moments to those calculated by S-
FRAME for various values of the axial load, P.

Table 1: Results Comparison for Case One

Axial Load (kips) Theory S-FRAME Difference (%)

Moment at Midspan (kp-in) 235 235.200 0.09
0
Defl. at Midspan (in) 0.202 0.201 0.26
Moment at Midspan (kp-in) 270 269.707 0.11
150
Defl. at Midspan (in) 0.23 0.230 0.08
Moment at Midspan (kp-in) 316 315.610 0.12
300
Defl. at Midspan (in) 0.269 0.268 0.37
Moment at Midspan (kp-in) 380 379.647 0.09
450
Defl. at Midspan (in) 0.322 0.321 0.31

Figures D and E display graphically the results of Case One, where the blue bars represent the
difference between S-FRAME’s computation and the theoretical value. The horizontal red lines indicate the
maximum acceptable difference as stated in ANSI/AISC 360-10. Clearly, the moments and deflections
calculated by S-FRAME are well within the limits of agreement set out by the code. In fact, in this case, S-
FRAME differed from theory by no more that 0.37%,) or less than 1 in 250.

th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011
 6

Case One: Moment at Case One: Deflection at

Midspan Midspan
3.50 6.00
Difference from Theoretical (%)

Difference from Theoretical (%)

3.00 5.00
2.50
4.00
2.00
3.00
1.50
2.00
1.00
0.50 1.00

0.00 0.00
0 150 300 450 0 150 300 450
Axial Load (kips) Axial Load (kips)

Figure D: Difference in Moment (in percent) between Theory Figure E: Difference in Deflection (in percent) between Theory
and S-FRAME and S-FRAME

Case Two

Figure F: Case Two

Case Two, as seen in Figure F, is designed to reveal errors in calculating P-∆ and P-δ effects in a given
analysis method. In this case, the combination of loads applied to the free end of the cantilever causes
additional end forces as well as additional internal member forces. For Case Two, nodes were already
th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011
 7

located at the points of interest, the base and tip, and, as such, it was not necessary to use more than one
beam element. Table 2 compares the theoretical deflections and moments to those calculated by S-FRAME
for various values of the axial load, P.
Table 2: Results Comparison for Case Two

Axial Load (kips) Theory S-FRAME Difference (%)

Moment at Base (kp-in) 336.00 336.000 0.00
0
Defl. at Tip (in) 0.91 0.907 0.01
Moment at Base (kp-in) 470.00 470.141 0.03
100
Defl. at Tip (in) 1.34 1.341 0.07
Moment at Base (kp-in) 601.00 601.809 0.13
150
Defl. at Tip (in) 1.77 1.765 0.26
Moment at Base (kp-in) 856.00 852.705 0.38
200
Defl. at Tip (in) 2.60 2.584 0.63

Figures G and H display graphically the results of Case Two, where the blue bars represent the
difference between S-FRAME’s computation and the theoretical value. The horizontal red lines indicate the
maximum acceptable difference as stated in ANSI/AISC 360-10. Again the moments and deflections
calculated by S-FRAME are well within the limits of agreement set out by the code. In Case Two, S-FRAME
differed from theory by no more than 0.63% or less than
)
1 in 150.

Case Two: Moment at Case Two: Deflection at

Base Tip
3.50 6.00
Difference from Theoretical (%)

Difference from Theoretical (%)

3.00 5.00
2.50
4.00
2.00
3.00
1.50
2.00
1.00
0.50 1.00
0.00 0.00
0 150 300 450 0 150 300 450
Axial Load (kips) Axial Load (kips)

Figure G: Difference in Base Moment (in percent) between Figure H: Difference in Tip Deflection (in percent) between
Theory and S-FRAME Theory and S-FRAME
th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011
 8

Summary

Consistently, S-FRAME’s implementation of second-order effects P-∆ and P-δ, provides results which
differ from theory by well less than one percent. This is well within the 3-5% cut off for acceptable analysis
methods outlined in ANSI/AISC 360-10. Taking into account typical uncertainties associated with material
properties and structure demands, it is safe to conclude that the difference between theoretical values and
S-FRAME’s computations is negligible. Furthermore, this indicates that S-FRAME is a reliable tool for
calculating second-order effects. S-FRAME’s P-∆ and P-δ analysis is accurate and rigorous and more than
sufficient for use alone or as a part of the direct analysis method.

th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011
 9

Appendix 1 – Critical Buckling Loads

S-FRAME’s P-Delta buckling analysis can calculate the load that will cause a structure or member to buckle.
In this appendix, the results of S-FRAME’s buckling analysis are compared to those predicted by theory. The
same two cases as earlier will be used to compare these results.

Case One

The theoretical buckling load for the simply supported beam can be calculated as follows:

Elastic Modulus E = 29000 ksi

Moment of Inertia I = 484 in4
Length L = 28 ft
Effective Length Factor Le = 1

P = E × I × π2 / (L × Le)2 = 1227.056 kips

S-FRAME calculates the buckling load to be 1236.735 kips, a 0.789% difference. Note that, as before, two
analysis members were used for this beam.
)

Case Two

The theoretical buckling load for the cantilever beam can be calculated as follows:

Elastic Modulus E = 29000 ksi

Moment of Inertia I = 484 in4
Length L = 28 ft
Effective Length Factor Le = 2

P = E × I × π2 / (L × Le)2 = 306.764 kips

S-FRAME calculates the buckling load to be 309.184 kips, a 0.789% difference. Note that, as before, one
analysis member was used for this beam.

th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011
 10

Appendix 2 - Buckling of a Frame

It has been shown that S-FRAME’s P-∆ buckling analysis is accurate for calculating the buckling loads of
cantilever and simply supported columns for which there are closed-form analytical solutions available.
However, in practice structural systems are usually more complex than single columns. This appendix
examines the critical buckling loads of a simple 2D frame example by Chen and Lui’s Structural Stability:
Theory and Implementation referenced by ANSI/AISC 360-10: Specifications for Structural Steel Buildings.

The frame will be considered for two support cases: with and without sway allowed. For the sway
prevented case, supports were added at the top of each column which prevented any in-plane lateral
displacement only. Two load cases will be considered for each support case, as shown in Figures A and B,
for a total of four cases. These four cases are outlined in Table 1

Each frame is 10 feet in height and 15 feet in span.

Figure A: Type 1 Loading Figure B: Type 2 Loading

Table 3: Support Conditions and Loading

Sway Loading
Case 1 Permitted Type 1
Case 2 Permitted Type 2
Case 3 Not Permitted Type 1
Case 4 Not Permitted Type 2

th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011
 11

For the entirety of the frame, a W14x48 section was used with the section and material properties as
below. As mentioned, the length of the columns is 10 feet.

Moment of Inertia; Ix = 484 in4;

Elastic Modulus; E = 29000 ksi;

Length; L = 10 ft;

Applying Euler’s formula, we can calculate the critical buckling load, Pe, for a simply supported W14x48
column, 10 feet in length.

Pe = π2 × E × Ix / L2 = 9620.123 kips;

Chen and Lui provide critical buckling loads for the four cases discussed earlier as a fraction of the critical
buckling load for a simply supported column. The critical buckling loads for each case can then be calculated
as follows:

Pcr1 = Pe × 0.6694 = 6439.710 kips

Pcr2 = Pe × 0.661 = 6358.901 kips

Pcr3 = Pe × 2.407 = 23155.635 kips

)
Pcr4 = Pe × 1.314 = 12640.841 kips

Using four analysis members per physical member, the four cases were modeled in S-FRAME and run
through P-∆ buckling analysis. The results showed very good agreement to those predicted by theory. This
can be seen in Table 2.

Table 4: Critical Buckling Load in Kips

Theory S-FRAME Difference in Percent

Case 1 6439.7 6408.5 0.49
Case 2 6358.9 6251.7 1.71
Case 3 23155.6 23217.7 0.27
Case 4 12640.8 12709.7 0.54

It can be concluded that S-FRAME’s P-∆ buckling analysis is very accurate in determining the critical buckling
loads of frames as well as columns. In the cases examined, S-FRAME’s results differed from theory by no
more than two percent.

th
© Copyright by S-FRAME Software Inc. – Version: 4.0 July 13 , 2011