You are on page 1of 20

# C O M P U TE R S &

S TR U C TU R ES
IN C .

R

Software Verification
PROGRAM NAME:
REVISION NO.:

SAP2000
0

EXAMPLE 2-002
SHELL – STRAIGHT BEAM WITH STATIC LOADS
PROBLEM DESCRIPTION
In this example a straight cantilever beam, modeled with shell elements, is
subjected to unit forces at the tip in the three orthogonal directions and unit
moments at the tip about the three orthogonal directions, each in a different load
case. The tip displacements in the direction of the load are compared with hand
calculated results.
It is important to note that this example is an extreme case presented for testing
and verification of the shell element. Shell elements are not in general intended
for use in modeling a beam with a 2 to 1 depth-to-width ratio.
The basic geometry, properties and loading are as described in MacNeal and
Harder 1985. The cantilever beam is 6 inches long, 0.2 inch wide parallel to the
Z direction and 0.1 inch wide parallel to the Y direction. Five different models
are created, each with a different mesh. Models A, B and C use a 6x1 mesh with
rectangular-, trapezoidal- and parallelogram-shaped elements, respectively, as
suggested in MacNeal and Harder 1985. Model D starts with the 6x1 rectangular
mesh and then divides each rectangle into two triangles. Model E starts with the
6x1 rectangular mesh and then divides each rectangle into four triangles. The
meshes used in models D and E are not included in MacNeal and Harder 1985.
Six load cases are created for each model. The six load cases apply a unit axial
force, a unit in-plane force, a unit out-of-plane force, a unit twisting moment, a
unit in-plane moment and a unit out-of-plane moment at the tip of the cantilever,
respectively. The twisting moment is applied as a couple of Y direction forces.
The in-plane moment is applied as a couple of X direction forces. The out-ofplane moment is applied as moments.
The independent solution is derived using elementary beam theory that assumes
no local Poisson’s effect occurs at the support. The beam is modeled in SAP2000
to match this assumption. At the fixed end, joint 1 is restrained in the Ux, Uy, Uz
and Rz degrees of freedom and joint 8 is restrained in the Ux, Uy and Rz degrees
of freedom. Joint 8 is not restrained in the Uz degree of freedom to avoid
imposing the unwanted local Poisson’s effect into the model. Also, when the
beam is loaded with in-plane shear, an in-plane force equal to half the applied tip
load is applied to joint 8 in the opposite direction of the tip load. This special
load at joint 8 is applied to model the reaction without the Poisson’s effect.

EXAMPLE 2-002 - 1

C O M P U TE R S &
S TR U C TU R ES
IN C .

Software Verification

R

SAP2000
0

PROGRAM NAME:
REVISION NO.:

1 - Joint number
1 - Area object number
13
6
5
6

Model A – Rectangular Shaped Elements
9

10

1

11

2

Y1

12

3

2

4

3

4
6 @ 1" = 6"

5

X

8

1.2"

0.8"

9

1.2"

10

1

12

3

2

5

1.2"

14
6

5

4

0.8"

1.1"
13

4

3

1.1"

0.8"

11

2

1

7

1 - Joint number
1 - Area object number

Model B – Trapezoidal Shaped Elements
0.9"

14

6

0.8"

0.2"

8
Z

0.2"

GEOMETRY

7

1.2"

0.9"

1"

0.9"

6"

1.1"

1"

8

1"

9

10

1

11

2

1

1"

2

5

1"

14
6

5

4

1"

13

4

3

0.9"

12

3

6

1"

0.2"

Model C – Parallelogram Shaped Elements

7

1"

1.1"

6"

8
1
1

9
2

3

4

2

10
5
3

11
7
4
6 @ 1" = 6"
6

8

12
9
5

13

14

10 11

0.2"

Model D – Triangular Shaped Elements (2 per Rectangle)
12

6

7

8 3
1
1 2

15

7

9
4

5
2

16
8

6

10 11
9
3 10

17

11 15

18

12 19

12 13

16 17

4 14
12 @ 0.5" = 6"

5

19

13 23
20 21

18

6 22

20

14

0.2"

Model E – Triangular Shaped Elements (4 per Rectangle)
24

7

EXAMPLE 2-002 - 2

000.5 lb at jts 7 and 14 2 Fz = +0.1 in LOADING The following table defines the loading applied to each model. and Fy = +5 lb at jt 14 5 Fx = -5 lb at jt 7.5 lb-in at jt 14 TECHNICAL FEATURES OF SAP2000 TESTED ¾ Membrane analysis using shell elements ¾ Plate bending analysis using shell elements ¾ Effect of shell element aspect ratio ¾ Effect of geometrical distortion of shell element from rectangular ¾ Joint force loading EXAMPLE 2-002 .5 lb at jt 8 3 Fy = +0.5 lb at jts 7 and 14 Fz = -0.: SAP2000 0 PROPERTIES E = 10.3 .846. Load Case Load 1 Fx = +0.3 G = 3.154 lb/in2 Shell section thickness = 0. and Mz = +0.5 lb at jts 7 and 14 4 Fy = -5 lb at jt 7.C O M P U TE R S & S TR U C TU R ES IN C .000 lb/in2 ν = 0. Software Verification R PROGRAM NAME: REVISION NO.5 lb-in at jt 7. and Fx = +5 lb at jt 14 6 Mz = +0.

In addition.Parallelogram D .Parallelogram D .1081 -26% 0.4 .4322 0.Triangle 2 Ux Average of jts 7 and 14 SAP2000 Independent Percent Difference 3E-05 0% 3E-05 0% 3E-05 3E-05 0% 3E-05 0% E – Triangle 4 3E-05 0% A.4314 0% EXAMPLE 2-002 .4321 0% 0.1072 -1% 0.C O M P U TE R S & S TR U C TU R ES IN C .Trapezoid C . the torsional stiffness of the section.4320 0% 0.0066 -94% A.0804 0.Rectangle 0.0032 -97% E – Triangle 4 0. Thin Plate Option Load Case and Type Model and Element Shape Output Parameter A.4322 0% Load case 3 B .0227 -79% B . Software Verification R PROGRAM NAME: REVISION NO.Parallelogram D .Rectangle Load case 1 Axial Extension Load case 2 In-plane shear and bending B .Trapezoid C .Triangle 2 E – Triangle 4 in Uy Average of jts 7 and 14 in 0.Triangle 2 in Uz Average of jts 7 and 14 0.Rectangle 0. J.: SAP2000 0 RESULTS COMPARISON The SAP2000 results are presented separately for the thin plate option and the thick plate option.Trapezoid Out-of-plane shear and bending C . The independent results are hand calculated using the unit load method described on page 244 in Cook and Young 1985. Independent results are also published in MacNeal and Harder 1985.4296 -1% 0. is calculated using item 4 in Table 20 on page 290 in Roark and Young 1975.

03600 0% EXAMPLE 2-002 .395E-04 -85% 7.: SAP2000 0 Thin Plate Option Load Case and Type Load case 4 Twist Load case 5 In-Plane Moment Model and Element Shape Output Parameter SAP2000 A.Parallelogram A.5 .Triangle 2 Average of absolute values at jts 7 and 14 0.Rectangle Load case 6 B .C O M P U TE R S & S TR U C TU R ES IN C . Software Verification R PROGRAM NAME: REVISION NO.00233 -32% 0.Parallelogram C .Rectangle Ux 8.Rectangle Uy 0.550E-04 -94% 0.Trapezoid 1.990E-04 0% B .03600 0% 0.03600 Independent 0.000E-04 0.03600 Percent Difference -32% -20% 0% 0.Triangle 2 Average of absolute values at jts 7 and 14 0.00231 -32% E – Triangle 4 in 0.225E-04 D .03600 0% C .Triangle 2 E – Triangle 4 Rz Average of values at jts 7 and 14 radians 0.Trapezoid 0.00233 D .00341 9.Parallelogram Moment D .00233 -32% B .03600 0% 0.265E-04 -97% E – Triangle 4 in 0.00230 -33% A.Trapezoid Out-of-Plane C .

0032 -97% E – Triangle 4 0.Triangle 2 in Uz Average of jts 7 and 14 0. Software Verification R PROGRAM NAME: REVISION NO.C O M P U TE R S & S TR U C TU R ES IN C .0804 0.Trapezoid C .Trapezoid Out-of-plane shear and bending C .1072 -1% 0.Triangle 2 SAP2000 Independent Percent Difference 3E-05 0% Ux 3E-05 0% Average of jts 7 and 14 3E-05 3E-05 0% 3E-05 0% E – Triangle 4 3E-05 0% A.Triangle 2 E – Triangle 4 in Uy Average of jts 7 and 14 in 0.Rectangle 0.Parallelogram D .4322 0.4321 0% 0.Rectangle 0.Parallelogram D .4328 0% 0.1081 -26% 0.Trapezoid C .0066 -94% A.Parallelogram D .Rectangle Load case 1 Axial Extension Load case 2 In-plane shear and bending B .4307 0% Load case 3 B .4298 -1% EXAMPLE 2-002 .: SAP2000 0 Thick Plate Option Load Case and Type Model and Element Shape Output Parameter A.6 .4321 0% 0.0227 -79% B .

Rectangle Ux 8.C O M P U TE R S & S TR U C TU R ES IN C .550E-04 -94% 0.Rectangle Uy 0.Trapezoid Out-of-Plane C .990E-04 0% B .00224 -34% B .03600 0% EXAMPLE 2-002 . Software Verification R PROGRAM NAME: REVISION NO.Parallelogram Moment D .00458 +34% A.Trapezoid 0.03600 0% 0.265E-04 -97% E – Triangle 4 in 0.Parallelogram C .225E-04 D .395E-04 -85% 7.03600 Independent 0.Parallelogram A.03600 0% C .00466 +37% E – Triangle 4 in 0.Trapezoid 1.00409 +20% 0.Triangle 2 Average of absolute values at jts 7 and 14 0.000E-04 0.00341 9.: SAP2000 0 Thick Plate Option Load Case and Type Load case 4 Twist Load case 5 In-Plane Moment Model and Element Shape Output Parameter SAP2000 A.00240 D .Rectangle Load case 6 B .Triangle 2 E – Triangle 4 Rz Average of values at jts 7 and 14 radians 0.Triangle 2 Average of absolute values at jts 7 and 14 0.03600 Percent Difference -30% -20% 0% 0.7 .03600 0% 0.

Example 2-002e-thick. Example 2-002b-thick. Comparing Models C and F shows that the unacceptable in-plane shear and bending results obtained in Model C are caused by the trapezoidal end elements. showing a difference from the independent results of less than 2%. The parallelogram-shaped elements in Models C and F are identical. Example 2-002d-thin.2" 4 @ 0. Rectangular-shaped elements show acceptable results.9" 1" 1" 1" 1" 19 1" 20 6 6 21 7 7 1" 8 8 1" 1" 4 @ 0. the following discussion illustrates that the trapezoidal elements at each end of Model C actually cause the unacceptable results. Example 2-002e-thin.225“ = 0. Model F 14 15 16 17 18 1 2 3 4 5 1 2 3 4 5 4 @ 0. except that the trapezoidal end elements have been divided into four elements each. parallelogram-shaped elements appear to show unacceptable results. Trapezoidal-shaped elements and triangular-shaped elements show unacceptable results. Example 2-002h-thin DISCUSSION OF IN-PLANE SHEAR AND BENDING (LOAD CASES 2 AND 5) The thin plate option and the thick plate option have essentially the same formulation for in-plane behavior and thus yield essentially the same results for load cases 2 and 5. EXAMPLE 2-002 . Example 2-002c-thin.1" 6" The results for load cases 2 and 5 for both the thin plate and the thick plate option are shown in the following table. Example 2-002g-thin.9" 22 23 24 25 26 9 10 11 12 9 10 11 12 13 4 @ 0. Model F is the same as Model C. Example 2-002c-thick. The in-plane shear and bending results are sensitive to the shape of the element.275“ = 1.8 . The discussion in this section applies to both the thin plate option and the thick plate option. not the parallelogram-shaped interior elements. Example 2-002b-thin. Example 2-002f-thick. Example 2-002d-thick. Software Verification R PROGRAM NAME: REVISION NO. as shown in the following figure.: SAP2000 0 COMPUTER FILES: Example 2-002a-thick. Example 2-002g-thick. The results for Model F are acceptable.C O M P U TE R S & S TR U C TU R ES IN C . Example 2-002f-thin.1" 0.225“ = 0. At first glance.275“ = 1. however. Example 2-002a-thin.

Software Verification R SAP2000 0 PROGRAM NAME: REVISION NO. trapezoidal elements are sensitive to the angle between opposite edges of the trapezoid and to the aspect ratio of the element.9" 8 1.2" 0. The results for Model G are acceptable. as shown in the following figure. however.9% Average of jts 13 and 26 Thick plate Percent Difference in 0. showing a difference from the independent results of less than 2%.C O M P U TE R S & S TR U C TU R ES IN C . Model G 0.1061 -1.3% in Acceptable results can be obtained using trapezoidal-shaped elements.: Model F Results (Parallelogram-shaped elements) Load Case and Type Element Type Load case 2 Thin plate In-plane shear and bending Output Parameter Uz SAP2000 Independent 0.2" 0.8" 9 1 10 2 1.2" 11 3 0. except that each of the six trapezoidal elements in Model B has been further meshed into 10 x 2 elements.1" 14 6 1.9% 8.1061 -1.1081 0.886E-04 -1.9" 6" The results for load cases 2 and 5 for both the thin plate and the thick plate option are shown in the following table. This further meshing reduces the angle between opposite edges of individual trapezoids from 90 to 9 degrees and improves the aspect ratio from approximately 5 to1 to approximately 1 to 1.8" 1.8" 12 4 1.1" 1.2" In-plane moment Average of absolute values at jts 13 and 26 7 0.3% Ux Thin plate Load case 5 Thick plate 9.2" 13 5 0. Model G is similar to Model B. EXAMPLE 2-002 .8" 0.000E-04 8.886E-04 -1.9 .

but not all. The twist results for the thin plate option are relatively insensitive to the shape of the element.000E-04 8. The shell theory used in SAP2000 assumes that the thickness of the shell is very small compared to the EXAMPLE 2-002 . Software Verification R PROGRAM NAME: REVISION NO.9% 0.1071 -0. Both of those items are described in the following text. The value of J calculated is 0. The twist results differ from the independent results by approximately 32%.1081 0.000045776 in4.C O M P U TE R S & S TR U C TU R ES IN C .825E-04 -1.9% in DISCUSSION OF TWIST FOR SHELL WITH THIN PLATE OPTION (LOAD CASE 4) The thin plate option and the thick plate option have different formulations for out-of-plane and twisting behavior and yield different results for load case 4. which is equivalent to 0.9% Ux Load case 5 In-plane moment Thin plate Thick plate Average of absolute values at jts 7 and 14 9. The torsional stiffness.10 . of the discussion in this section is also applicable to the thick plate option.9% 8. The discussion in this section applies specifically to the thin plate option. J.22888bd3. used in the hand calculation is calculated using item 4 in Table 20 on page 290 of Roark and Young 1975.: SAP2000 0 Model G Results (Trapezoidal-shaped elements with 60 x 2 mesh) Load Case and Type Element Type Load case 2 Thin plate In-plane shear and bending Output Parameter Uz SAP2000 0. A minor contributor to the difference is the Rz restraint used at the fixed end of the cantilever. which causes the twisting moment to not be uniform near the fixed end.825E-04 -1. The major contributor to this difference is the difference between the shell theory used by the program and the beam theory used in the calculation of the independent results. Much.1071 Average of jts 7 and 14 Thick plate in Independent Percent Difference -0.

The ratio of the hand calculated J divided by the shell theory J is 0.00341 inch.11 . to vary near the fixed end.00234 inch. This compares well with the computer generated result. as shown in the following M12 twisting moment contour plot. The resulting displacement from the twisting load for Model H is 0.C O M P U TE R S & S TR U C TU R ES IN C . Thus.00234 inch which compares exactly with the hand calculation that has been modified to use the shell theory J of bd3/3.: SAP2000 0 extent of the entire shell mesh. as desired. The Rz restraint used at the fixed end of the beam is needed for the threedimensional computer model to be stable and to resist the moment imposed by the out-of-plane loading in load cases 3 and 6. It does not compare exactly because of the Rz restraint used at the fixed end in the computer model. which should be constant over the length of the beam. the resulting value is 0. The springs are required to keep the three-dimensional model stable. M12 Twisting Moment (per unit length) Contour Plot for Model A-Thin Model H (Example 2-002h-thin) is created for load case 4 only. M12 Twisting Moment (per unit length) Contour Plot for Model H -Thin EXAMPLE 2-002 . Model H uses a rectangular shell element similar to that for Model A.68664.22888bd3 is 0. R Software Verification PROGRAM NAME: REVISION NO. but not exactly.00341 * 0. If this displacement is instead calculated using J = bd3/3. The Rz restraint is removed in Model H and very small rotational Rz springs are provided in its place. This restraint causes the M12 twisting moment resulting from load case 4. The twisting moment is constant over the length of the beam. The following contour plot shows the M12 twisting moment for Model H. the equivalent J for shell theory is bd3/3. The original hand calculated displacement resulting from the twist based on J = 0.68664 = 0.

00341 inches. the equivalent J for shell theory is bd3/3.: SAP2000 0 DISCUSSION OF TWIST FOR SHELL WITH THICK PLATE OPTION (LOAD CASE 4) The thin plate option and the thick plate option have different formulations for out-of-plane and twisting behavior and yield different results for load case 4.68664 = 0.00239 inch.C O M P U TE R S & S TR U C TU R ES IN C .003 inch. This is the theoretical solution that we might expect from the computer. The torsional stiffness. which is equivalent to 0. The ratio of the hand calculated J divided by the shell theory J is 0. used in the hand calculation is calculated using item 4 in Table 20 on page 290 of Roark and Young 1975. if each of the six trapezoidal area objects in Model B is submeshed into a 10 by 2 mesh. The twist results for the thick plate option are smaller than the independent results by as much as 34% and larger than the independent results by as much as 37% depending on the element shape.00341 * 0.22888bd3. Those items are described in the following text.22888bd3 is 0.and parallelogram-shaped elements. For example. It compares fairly well with the results for the rectangular. Even models using equilateral triangles (not included in the example model files) have a twist displacement of approximately 0. which is about 30% different from the theoretical solution that we might expect from the computer. Thus. R Software Verification PROGRAM NAME: REVISION NO. It does not compare well with the results for the trapezoidal. The trapezoidal model can be further meshed to obtain better results.68664.12 .00234 inch. which compares fairly well (approximately 2% difference) with the expected results using shell theory and the results obtained with the rectangular. The shell theory used in SAP2000 assumes that the thickness of the shell is very small compared to the extent of the entire shell mesh. If this displacement is instead calculated using J = bd3/3.000045776 in4. The original hand calculated displacement resulting from the twist based on J = 0.and triangular-shaped elements. The discussion in this section applies specifically to the thick plate option. EXAMPLE 2-002 .and parallelogram-shaped elements (approximately 4% difference). the shell aspect ratio and the variation of the shell shape from a rectangle. J. The triangular models are typically less stiff in twist than the other models. the resulting value is 0. the resulting twist displacement is 0. The contributors to this difference are the difference between the shell theory used by the program and the beam theory used in the calculation of the independent results. The twist results for the thick plate option are sensitive to the shape of the element. The value of J calculated is 0.

For the particular geometry of this example. R Software Verification SAP2000 0 PROGRAM NAME: REVISION NO. 3. J. The SAP2000 results obtained for twist for the thick plate option vary and they differ significantly from the independent results. the twist results are more consistent. When the aspect ratio is approximately one. Triangular-shaped elements are not recommended for use where in-plane shear and/or bending is significant. The SAP2000 results obtained for twist for the thin plate option are quite consistent. Use aspect ratios near one to one. In many types of structures. Trapezoidal-shaped elements should be avoided for use where in-plane shear and/or bending is significant if it is possible to use rectangular-shaped or parallelogram-shaped elements. is given by bd3/3. Minimize the angle between opposite sides of the trapezoid. and because the twist results for the thick shell option are sensitive to shell element aspect ratio and geometric distortions. beam theory calculates J as approximately bd3/4. This difference in J accounts for the difference in results. This occurs because the SAP2000 results are based on shell theory and the hand calculated results are based on beam theory. but they differ significantly from the independent results. 4. particularly those for which the use of shell elements is appropriate. the effect of the twisting behavior is negligible and thus any inaccuracies in the twisting deformations are inconsequential. This occurs because of the difference in J described in the previous paragraph.37. Where the use of trapezoidal elements is necessary. 2.: CONCLUSIONS The SAP2000 results and the independent results have an acceptable and in many cases exact comparison for axial extension and out-of-plane shear and bending for both the thin plate option and the thick plate option.13 . Shell theory assumes the thickness of the shell elements is small compared to the extent of all of the shell elements and thus that the torsional stiffness. The SAP2000 results obtained for in-plane shear and/or bending are acceptable for the rectangular-shaped and parallelogram-shaped elements and poor for other shapes of elements for both the thin plate option and the thick plate option. Review the results carefully to ascertain stress continuity between elements.C O M P U TE R S & S TR U C TU R ES IN C . the following modeling tips are suggested: 1. Always use a mesh that is two or more elements wide. EXAMPLE 2-002 .

R Software Verification PROGRAM NAME: REVISION NO.: SAP2000 0 HAND CALCULATION EXAMPLE 2-002 .14 .C O M P U TE R S & S TR U C TU R ES IN C .

: SAP2000 0 EXAMPLE 2-002 .15 . R Software Verification PROGRAM NAME: REVISION NO.C O M P U TE R S & S TR U C TU R ES IN C .

: SAP2000 0 EXAMPLE 2-002 .C O M P U TE R S & S TR U C TU R ES IN C . R Software Verification PROGRAM NAME: REVISION NO.16 .

: SAP2000 0 EXAMPLE 2-002 . R Software Verification PROGRAM NAME: REVISION NO.17 .C O M P U TE R S & S TR U C TU R ES IN C .

: SAP2000 0 EXAMPLE 2-002 .18 .C O M P U TE R S & S TR U C TU R ES IN C . R Software Verification PROGRAM NAME: REVISION NO.

C O M P U TE R S & S TR U C TU R ES IN C .: SAP2000 0 EXAMPLE 2-002 .19 . R Software Verification PROGRAM NAME: REVISION NO.

C O M P U TE R S & S TR U C TU R ES IN C .20 . R Software Verification PROGRAM NAME: REVISION NO.: SAP2000 0 EXAMPLE 2-002 .