You are on page 1of 15

286

displacement-based beam-column elements and the second-order mixed beam-column


element. All of these elements are capable of undergoing large rotations and large
displacements. The only difference is the formulation utilized in the corotational frame.
This section represents several numerical examples that verify the developed
formulations. A full Newton-Raphson iterative solution procedure with an arc-length
iterative strategy and an automatic arc-length incrementation strategy is utilized in the
following numerical examples unless otherwise specified. Convergence of the finite
element solution is based on the Euclidean norm of the out-of-balance force vector as
well as the incremental displacement vector.

7.7.1

Elastic Benchmark Problems

7.7.1.1 Forty-Five Degree Cantilever Bend


This example involves a three-dimensional response and has been used by
number of authors (Crisfield 1990, Nukala 1997, Cardona and Geradin 1988, Bathe and
Bolourchi 1979, Simo and Vu-Quoc 1988). Figure 7.8 shows the initial geometry. The
section of the cantilever bend is of unit cross-section and E is 107. It is modeled using
eight straight elements as compared to eight elements with initial imperfections
(Crissfield 1990, Nukala 1997, Simo and Vu-Quoc 1988). Table 7.1 compares the
present solutions (with geometrically linear and nonlinear displacement-based elements
and with mixed elements) with those given by a number of authors for the tip-geometry
at loads 300,450 and 600. Table 7.2 gives the incremental/iterative performance of the
solution with the displacement-based finite elements in which the load is applied in eight

287

0
10

P
X

45 o

Figure 7.8 Initial geometry of forty-five degree cantilever bend

Table 7.1 Tip coordinates (x,y,z) for 45 degree bend (initially, 29.29, 0.0, 70.71)

Load Level

Model

Present Solution
(Geometrically
Linear,
Displacement-based element)
Present solution
(Geometrically
Nonlinear
Displacement-based
Element)
Present Solution
(Mixed Element)
Bathe and Bolourchi

300

450

600

22.14, 40.50, 58.56

18.39, 48.75, 51.98

15.57, 53.68, 46.89

22.26, 40.14, 58.81

18.51, 48.46, 52.27

15.68, 53.44, 47.18

22.16, 40.42, 58.56

18.43, 48.65, 52.00

15.62, 53.56, 46.93

22.5, 39.5, 59.20

15.9, 53.4, 47.2

Simo and Vu-Quoc

22.33, 40.08, 58.84

18.62, 48.39, 52.32

15.79, 53.37, 47.23

Cardona and Geradin

22.14, 40.35, 58.64

18.38, 48.59, 52.11

15.55, 53.50, 47.04

Crisfield

22.16, 40.53, 58.53

18.43, 48.79, 51.93

15.61, 53.71, 46.84

Nukala

22.28, 40.09, 58.81

18.58, 48.37, 52.30

15.78, 53.33, 47.23

288

Table 7.2 Incremental/Iterative performance at P=450


Iteration Number
(mixed solution)

Iteration Number
(Displacementbased solution)
4

Increment Number

increments up to 450. Furthermore, Table 7.3 gives the residual force norm of this
solution at the second increment and it is observed that a very fast convergence rate is
achieved.

Table 7.3. Convergence rate for the second increment in Table 7.2

Number of iteration

Residual Force Norm


(Mixed Solution)

Residual Force Norm


(Displacement-based
solution)
3457.8

25.4731

26.4665

6.15823

7.56097

0.0026223

0.06633

0.00100574

0.01046

3.22409E-9

2.769E-8

3617.38

289

7.7.1.2 Lateral Buckling of a Cantilever Right-Angled Frame under End-Load


This example involves a full three-dimensional response of a right-angled frame
shown in Fig. 7.9. A load P is applied in z-direction at the element tip and a small
perturbation load (Pr = 0.0002P) is given in x-direction. A thin narrow rectangular crosssection with width and depth of 0.6 and 30, respectively, is used. The same problem is
used by number authors (Crisfield 1990, Nukala 1997, Simo and Vu-Quoc 1988). The
computed response between the applied load and the lateral (x-direction) tip-deflection is
shown in Fig. 7.10. The right-angle frame is solved by geometrically nonlinear
displacement-based and mixed elements and only two elements per each leg are utilized
for both solutions. Solution with one element per member for each leg is also performed
and it predicts the correct limit load but it shows convergence difficulties for post-limit
range. An arc-length iterative solution procedure is used in the present solutions whereas
Simo and Vu-Quoc (1988) used a displacement control iterative solution method. It
should be mentioned that both references used five elements per each

30

0.6

240

X
Y

240

P
Pr

Figure 7.9 Initial geometry for the right-angled frame

290

1.5

Applied Load (lb)

1.25

0.75

0.5
Crisfield (1990)
Simo&Vu-Quoc (1988)

0.25
Displacement-based solution
Mixed Element Solution

0
0

20

40

60

Lateral Tip Displacement (in.)

Figure 7.10 Load/tip x-displacement for right-angled frame

leg. The buckling load is found to be 1.10 and it is reported 1.086 in Nukala (1997) and
1.09 in Simo and Vu-Quoc. It should be mentioned that the converged solution is similar
to one obtained by Crisfield (1990).

7.7.1.3 Lateral Torsional Buckling of an I-Beam


The lateral buckling of a W10x100 beam is studied in this example. The beam is
loaded in pure bending about its major axis and a small perturbation moment is applied at
one end about its minor bending axis (Fig. 7.11). The same problem is also solved with

291

introduced initial imperfections only such that a sinusoidal function in z-direction with an
amplitude of 0.00001L is defined (see Appendix 2). The beam is simply supported at its
ends and lateral movement and twisting are restrained at its ends. The theoretical critical
load obtained by considering the pre-buckling deformation effects for the warping free
conditions is

M cr =

2 E Cw

E I y G J 1 +
G J L2

L
I
GJ 2 E Cw
1 y 1

1 +
I z
EI z
G J L2

It should be mentioned that the above equation also gives the critical moment for the case
of pre-buckling deformations are ignored if the value at the denominator of the equation
is not accounted for. Thus, for W10x100, the analytical buckling moments with and
without pre-buckling deformations are 15118 k-in. and 12303 k-in., respectively. Applied
moment versus mid-span out-of-plane deflection for this example is shown in Fig. 7.12.

E = 29000 ksi
v = 0.3
W 10x100
L=240 in.

0.01M

Figure 7.11 Lateral buckling of an I-beam

292

25000

Applied Moment (k-in.)

20000

15000
Critical Load with pre-buckling deformations

Critical Load without pre-buckling deformations

10000

Converged Solution
Displacement-based solution (2 elements)
Displacement-based solution (4 elements)

5000

Mixed solution (2 elements)


Mixed Solution (4 elements)
Displacement-based solution (4 elements, with
initial imperfections)

0
0

Central out-of-plane displacement (in.)

Figure 7.12 Lateral buckling of an I-beam (W10x100): applied moment versus central
out-of plane deflection

7.7.1.4 Nonuniform Torsion of a Cantilever Beam


The shape functions derived for the angle of twist are given is Section 7.4. It
should be mentioned that they are the exact shape functions for a prismatic, elastic and
geometrically linear beam element. One can also use cubic Hermitian functions instead of
these hyperbolic functions (see Eq. 7.62) but the accuracy of this solution is not good as

293

compared to the exact solution that is obtained with hyperbolic functions. In this
example, a cantilever beam with L = 136 in. subjected to a nonuniform torque is studied
for an elastic case and the responses of the beam element based on the hyperbolic and the
cubic functions are compared. Three different section sizes are chosen: W24x68,
W14x68 and W8x67. The beam is torsionally fixed at one end and free at the other end.
and a torque (T) is applied at free end. The analysis results are shown in Fig. 7.13. It is
inferred from the figure that the solution with cubic functions yield comparable results
for the cases with W24x68 and W14x68. However, its accuracy for the section W8x67 is
not good. In the current study, the hyperbolic functions are utilized both for elastic and
inelastic cases even though these shape functions are not correct shape functions for
inelastic case. However, it is checked that the convergence of these shape functions is
guaranteed at the limit when the length of the element is decreased.

1.2
W8x67

Tsv/T

0.8

W14x68

0.6

W24x68

0.4
0.2

Solution with exact shape functions


Solution with cubic approximations

0
0

0.25

0.5

0.75

x/L
Figure 7.13 One element solution for nonuniform torsion problem

294

7.7.2

Inelastic Benchmark Problems

7.7.2.1 Farwell and Galambos Experimental Test


The analysis results are compared to the experimental test results obtained by
Farwell and Galambos (1969) for the non-uniform torsion of an I-beam in which a
concentrated torque is applied at the center of the specimen. The problem definition is
given in Fig. 7.14. The beam is assumed to be initially straight, and to have zero residual
stresses. The ends of the beam are torsionally simply supported, i.e., twist rotations are
not allowed and the cross-sections are allowed to warp freely. The section is divided into
4x16 and 8x16 fibers at the web and at the flanges, respectively. Two and four
displacement-based and mixed elements are used and the analysis results are compared to
the experimental data and to the converged solution in Fig. 7.15. It is shown that the
displacement-based solution with two elements is stiffer than corresponding mixed
solution.
5.94 in.

T
0.313 in.
6 in.
L=76 in.
0.481 in.
E = 30924 ksi
Est = 900 ksi

y = 41.3 ksi

Figure 7.14 Farwell and Galamboss Test No. 5, initial geometry and cross-section
dimensions

295

200

Applied Torque (k-in.)

150

100

Farwell and Galambos test data (1969)

50

Converged Solution
Displacement-based solution (2 elements)
Displacement-based solution (4 elements)
Mixed Solution (2 elements)
Mixed Solution (4 elements)

0
0

0.4

0.8

1.2

1.6

Mid-span angle of twist (rad.)

Figure 7.15 Analysis results for inelastic non-uniform torque versus twist relation

7.7.2.2 Lateral Torsional Buckling of an I-Beam


The same problem in Section 7.7.1.3 is resolved for inelastic case. The Youngs
modulus and the yield stress are 29000 ksi and 50 ksi, respectively and a Poissons ratio
of 0.3 is used. A stiffness of 217.5 ksi is chosen after yielding. The element is modeled
with four displacements-based and mixed elements with 4x16 and 8x16 fibers at the web
and the flange, respectively and the analysis results are compared to Abaqus (1999)
results in which 16 B33 elements are utilized. It should be noted that the chosen beam
element in Abaqus (1999) utilizes one surface plasticity model that does not account for

296

shear stresses. Figure 7.16 portrays the analysis results graphically and it is shown that
the results with four elements for displacement-based and mixed solutions are similar and
for the both cases, the analysis is completed within 16 load increments with an average of
3 iterations per increment. Four-element solution predicts the buckling load of 6018 k. It
is found 6050k with ABAQUS but the post-buckling behavior is different as observed in
the figure due to the constitutive model.

7000

Applied Moment (k-in.)

6000

5000

4000

3000

2000

ABAQUS
Displacment-based solution (4 elements)

1000

Mixed solution (4 elements)

0
0

Mid-span out-of-plane displacement (in.)

Figure 7.16 The analysis results of lateral buckling of an I-beam for inelastic case

297

7.7.2.3 Aalbergs Tests


In this section the proposed solution is compared to experimental test results
obtained by Aalberg (1995). The test setup, loads and support conditions are shown in
Fig. 7.17. The beam-columns are assumed to be initially straight and have zero residual
stresses. The Youngs modulus and the yield stress are 210000 N/mm2 (30460 ksi) and
279 N/mm2 (40.47 ksi) respectively and a poison ratio of 0.3 is used. A stiffness of 4000
N/mm2 (580.14 ksi) is utilized after yielding. It should be mentioned that the yield stress
used in this study was reported in the reference as the average yield stresses measured in
the beam-column flanges whereas the measured average web yield stress was around 290
N/mm2 (42.06 ksi). A full length of 2030 mm is used in all calculations and 4x16 and
8x16 fibers for the web and the flange, respectively are utilized after any yielding at a
section is detected.

2090 mm

140.9 mm

Z,w

No

To
u=0
v=0
w=0

To

7.3 mm
11.4 mm

Y,v

z=0
u=0

u=0
v=0

HEB 140

Figure 7.17 The problem definition of Aalbergs Test (1995)

140.45mm

2030 mm

X,u

298

Uniform torsion test: the load deflection curve is shown in Fig. 7.18 and analytical
results are obtained with one displacement and one mixed element. Both results are
similar and they are compared to the test results. The difference between the analysis
results and the test results may be attributed to section dimension measures that lead
to different values of St Venant constant, the section warping constant and fillet areas.
In addition, the yield value and strain hardening stiffness utilized in the analysis are
rough estimations obtained from the experimental results. Nevertheless, the analysis
results yield good comparable values.

Bending and torsion combination: The analysis solution is compared to the test results
for the case H-2-MT in the reference. Four displacement-based and two mixed
elements are used in the analysis. The applied loads and displacements are
normalized with respect to their yield values as:

N=

M=

T=

w
wy

v
vy

N
Ny

w=

M
My

v=

T
Ty

in which
N y = 1198 kN

w y = 2.85 mm

M y = 58.9 kNm v y = 7.3 mm


Ty = 2.65 kN

y = 4.95 o

Normalized load versus normalized displacements are plotted in Fig. 7.19. It is


observed that both mixed and displacement-based solutions overpredict test data.

299

This can be related to chosen linear strain hardening modulus (4000 N/mm2),
predicted from the reference. The same problem is solved with hardening stiffness of
1500 N/mm2 (217.5ksi) and better results are obtained.
Axial force and torsion: The analysis solution is compared to the test results for the
case H-9-NT in the reference. The test was loaded sequentially in such a way that
axial force was first applied and then kept constant while torsional load was applied at
the ends. Figure 7.20 compares analysis results and the experimental data. Two
displacement-based and mixed elements are used to model the beam-column.

6
T

5
Torsional Moment (kNm)

T
HEB 140

Experimental (Aalberg, 1995)


1 Displacement-based element

1 Mixed element
0
0

0.25

0.5

0.75

1.25

1.5

Twist Rotation (rad.)

Figure 7.18 Uniform torsion test on HEB 140 beam

300

1.5

Normalized Moment

E s t = 4000 N /mm 2 ( 580 . 14 ksi)

1.0

E st = 1500 N/mm 2 ( 217 .5 ksi)


T

0.5

HEB 140

Experimental Data (Aalberg, 1995)


Converged Solution
Mixed Solution (2 elements)
Displacement-based solution (4 elements)
Displacement-based solution (4 elements)
0.0
0

Normalized displacement (v)

Figure 7.19 Bending and torsion combination

Normalized Applied Torque

N = 0.50
1.5

1
HEB 140

Experimental Data (Aalberg, 1995)


0.5

Converged Solution
Displacement-Based Solution (2 elements)
Mixed Solution (2 elements)

0
0

Normalized Rotation at the end

Figure 7.20 Axial force and torsion combination

10

You might also like