Professional Documents
Culture Documents
19933-Article Text-25252-1-10-20170410
19933-Article Text-25252-1-10-20170410
19933-Article Text-25252-1-10-20170410
54-62
ISSN 1410-9530 print / ISSN 1979-570X online
Abstract: A major difficulty in formulating a finite element for shear-deformable beams, plates,
and shells is the shear locking phenomenon. A recently proposed general technique to overcome
this difficulty is the discrete shear gap (DSG) technique. In this study, the DSG technique was
applied to the linear, quadratic, and cubic Timoshenko beam elements. With this technique, the
displacement-based shear strain field was replaced with a substitute shear strain field obtained
from the derivative of the interpolated shear gap. A series of numerical tests were conducted to
assess the elements performance. The results showed that the DSG technique works perfectly to
eliminate the shear locking. The resulting deflection, rotation, bending moment, and shear force
distributions were very accurate and converged optimally to the corresponding analytical
solutions. Thus the beam elements with the DSG technique are better alternatives than those
with the classical selective-reduced integration.
Keywords: Discrete shear gap; selective-reduced integration; shear locking; Timoshenko beam
element.
54
Wong, F.T. et al../ Study of the Discrete Shear Gap Technique / CED, Vol. 19, No. 1, March 2017, pp. 54–62
Although the DSG technique was firstly introduced In this equation, δw and δβ are a virtual deflection
in the context of Timoshenko beam [11], subsequent and a virtual cross section rotation, respectively. The
applications of the technique were focused on the term „admissible‟ means that the δw and δβ satisfy
plate and shell problems. In the context of the the requirements of C0 continuity and homogenous
Kriging-based finite element method [16], the displacement boundary conditions. The constant G is
application of the DSG technique has been studied the shear modulus given as
by Wong and Sulistio [17]. It was found that the E
DSG is effective to eliminate shear locking for a G (1b)
2(1 )
Kriging-based element with cubic polynomial basis,
but not effective for those with quadratic and linear and k is a shear correction factor that is dependent
bases. In the context of the standard FEM, however, upon the cross-section geometry and Poisson ratio
there has been no formulation and numerical test of [18]. The comma signifies the first derivative with
the DSG technique on higher-order Timoshenko respect to the variable next to it (i.e. x).
beam elements. It is thus the aim of this paper to
study the application of DSG technique in the The bending moment M and shear force Q along the
standard linear, quadratic, and cubic Timoshenko beam are given as
beam elements. The application of the DSG is accom- M EI , x (2)
plished by replacing the shear strain-displacement Q kGA ( w, x ) (3)
matrix with the one derived based on the DSG
concept. A series of numerical tests is carried out to Finite Element Discretization
evaluate the performance of the DSG beam ele-
ments, in particular to evaluate the effectiveness of Let the beam is subdivided into Nel elements and
the DSG technique at eliminating shear locking. The consider a typical element having n nodes. The
comparison is made with beam elements using the deflection and rotation along the element are
selective-reduced as well as full integrations. approximated as follows:
w i 1 N i wi N w d
n
Timoshenko Beam Elements (4)
i 1 Ni i N d
n
(5)
Governing Equations
in which [Nw] is the shape function matrix for the
A beam model of length L with the cross sectional deflection, i.e.
area A and moment of inertia I is considered. The N w N1 0 N 2 0 N n 0 (6)
material properties are a constant modulus elasticity
[Nβ] is the shape function matrix for the rotation, i.e.
E and a constant Poisson‟s ratio ν. The beam is
subjected to a static transversal distributed load N 0 N1 0 N 2 0 Nn (7)
q=q(x). The displacement of any point in the beam and {d} is the unknown nodal displacement vector,
can be expressed in terms of two independent field i.e.
variables, i.e. the vertical displacement of the neutral
d w1 1 w2 2 wn n
T
axis, or, the deflection w=w(x) and the cross section (8)
rotation β= β(x). The coordinate system and sign
convention for positive w and β are depicted in The number of element node n depends on the
Figure 1. polynomial degree employed to approximate w and
β, that is, n=2 for a linear element, n=3 for a
The weak form of the governing equation for static quadratic element, and n=4 for a cubic element, as
deformation of the Timoshenko beam model is given shown in Figure 2.
as:
L L L Using the isoparametric mapping concept, the
0
, x EI , x dx ( w, x ) kGA ( w, x ) dx q wdx
0 0 coordinate x is expressed in terms of natural coor-
for all admissible δw and δβ (1a) dinate [1, 1] in the same manner as the
approximation for the field variables, i.e.
x i 1 N i xi
n
(9)
where xi, i=1,…,n are the nodal coordinates. The
expression for the shape functions are listed in Table
1 in terms of ξ. The Jacobian of the mapping is
defined as
Figure 1. Coordinate System and Positive Directions for
the Deflection and Rotation
J x, (10)
55
Wong, F.T. et al../ Study of the Discrete Shear Gap Technique / CED, Vol. 19, No. 1, March 2017, pp. 54–62
Figure 2. One-dimensional Linear, Quadratic, and Cubic It is worth mentioning here that while the SRI tech-
Elements nique is effective to eliminate shear locking, the
resulting shear forces are correct only at the inte-
Table 1. Shape Functions for a 1D Element of Two to Four gration sampling points. The shear forces evaluated
Nodes (taken from [10], p. 343) with the Numbering Labels directly at the element nodes are in great error. The
as Shown in Figure 2. shear force distribution fluctuates heavily [20].
If the
If the element has nodes 3 Table 2. Number of Integration Sampling Points Required
element
and 4 for Full and Selective-reduced Integrations.
has node 3
N1= 1
2 (1 ) 12 (1 2 ) 161 (9 3 2 9 1) Element Technique Bending Shear Force
Stiffness Stiffness
N2= 1
2 (1 ) 12 (1 ) 161 (9 9 1)
2 3 2
Linear Full 1 2 2
N3= (1 2 ) 161 (27 3 7 2 27 7) SRI 1 1 2
Quadratic Full 2 3 3
N4= 161 (27 3 9 2 27 9)
SRI 2 2 3
Cubic Full 3 5 4
Substituting Equations (4) and (5) into Equation SRI 3 3 4
(1) and carrying out the standard finite element Note: Full: full integration;
formulation procedure yield the element system of SRI: selective-reduced integration
equations
k d f (11) Discrete Shear Gap Technique
56
Wong, F.T. et al../ Study of the Discrete Shear Gap Technique / CED, Vol. 19, No. 1, March 2017, pp. 54–62
w 2 w2 w1 i 1 ( N i Jd ) i
2 1
Linear
1
w 2 w2 w1 i 1 ( N i Jd ) i
3 1
Quadratic
1
w 3 w3 w1 i 1 ( N i Jd ) i
0
Figure 3. Shear Gap Interpreted as the Difference bet- 3
57
Wong, F.T. et al../ Study of the Discrete Shear Gap Technique / CED, Vol. 19, No. 1, March 2017, pp. 54–62
two different meshes, that is, a mesh with four The results of the deflection and rotation at the free
elements of equal length and a mesh with four end, and the bending moment and shear force at the
elements of unequal length shown in Figure 4 [17]. fixed end were observed and compared to the
The material, geometrical, and loading parameters corresponding exact solutions. For the sake of bre-
were taken as follows: L = 10 m, b = 2 m, E = 2000 vity, only the results for the extremely thin beam
kN/m2, ν = 0.3, and M = 1 kN-m. Two different using the irregular mesh were presented here in
thickness was considered, that is, h = 2 m (L/h = 5), Table 4.
which represents a thick beam, and h=0.001 m
(L/h=10000), which represents an extremely thin It is seen that even for the most critical case (i.e. very
beam. Certainly, the beam with L/h=10000 is outside thin beam and irregular mesh), the DSG beam
the range of practical length-to-thickness ratio. The elements are not only free from shear locking but
purpose to include it here is merely to test the also are able to reproduce exact results. In contrast,
tendency of the beam elements to shear locking. the results obtained using the linear beam element
with full integration are erroneous because of shear
M locking. While the SRI M technique can eliminate the
locking, as addressed by Prathap [20], the shear
forces calculated directly at the fixed-end node using
0.1L 0.1L
L/4 L/4 L/4 L/4 Equation (3) are in great error, even worse that the
0.6L 0.2L
shear force obtained with full integration. The
overall results showed that that all DSG beam
(a) Equal-length four element model elementsfour
(b) Unequal-length passed this
element pure bending test for both the
model
thick and extremely thin beams, with the regular
M M and irregular meshes. The DSG technique is better
than the SRI because the former can give accurate
shear force along the beam while the later can only
0.1L 0.1L give accurate shear force at the integration sampling
4
0.6L 0.2L points.
Table 4. Analysis Results for the Cantilever Beam of L/h=10000 Modeled using Four Elements of Unequal
Length
(a) Normalized Deflection and Rotation
w/wL exact β/βL exact
Element
DSG SRI Full DSG SRI Full
Linear 0.9999999 0.9999999 0.0000008 0.9999999 0.9999999 0.0000008
Quadratic 0.9999993 0.9999994 0.9999999 0.9999995 0.9999996 0.9999999
Cubic 1.0000003 1.0000012 1.0000006 1.0000002 1.0000008 1.0000004
Full: full integration; SRI: selective-reduced integration; SRI: selective-reduced integration,
58
Wong, F.T. et al../ Study of the Discrete Shear Gap Technique / CED, Vol. 19, No. 1, March 2017, pp. 54–62
qL4 qL2 q0 L4 5
w( L2 )exact (31) w( L)exact (1 ) (32a)
384 EI 8GkA 30EI 12
2
1 h (32b)
q (12 11 )
5 L
1 2 3 4 5 6 7 8
h
M (0)exact 6 q0 L2 , Q(0) exact 12 q0 L
1 (33c, d)
L b
The beam were modeled using 1, 2, 4, and 8 elements.
(a) Finite element model of the beam (b) Beam cross section
The analysis results, normalized to the corres-
Figure 5. Fixed-fixed Beam Subjected to Uniform Load ponding exact values, were presented in Table 6.
and its Finite Element Model
59
Wong, F.T. et al../ Study of the Discrete Shear Gap Technique / CED, Vol. 19, No. 1, March 2017, pp. 54–62
Table 6. Normalized Free-end Deflections, Fixed-end Bending Moments, and Fixed-end Shear Forces of the
Cantilever Beam for Different Number of Elements, Nel, Obtained using Different Beam Elements
Figure 7. Deflection Profile Obtained using Four Linear, Figure 9. Bending Moment Profile Obtained using Four
Quadratic, and Cubic DSG Beam Elements Linear, Quadratic, and Cubic DSG Beam Elements
Figure 8. Rotation Profile Obtained using Four Linear, Figure 10. Shear Force Profile Obtained using Four
Quadratic, and Cubic DSG Beam Elements Linear, Quadratic, and Cubic DSG Beam Elements
60
Wong, F.T. et al../ Study of the Discrete Shear Gap Technique / CED, Vol. 19, No. 1, March 2017, pp. 54–62
61
Wong, F.T. et al../ Study of the Discrete Shear Gap Technique / CED, Vol. 19, No. 1, March 2017, pp. 54–62
Mechanics and the 6th Asia-Pacific Congress on 20. Prathap, G. The Finite Element Method in
Computational Mechanics, Seoul, Korea, July Structural Mechanics, Springer, 1993, retrieved
24-29, 2016, pp. 187-206. from http://www.springer.com/gp/book/9780792-
18. Cowper, G.R., The Shear Coefficient in Timo- 324928 [Accessed: 25-Feb-2016].
shenko‟s Beam Theory, Journal of Applied 21. Chapman, S.J., Matlab Programming for
Mechanics, ASME, 33(2), 1966, pp. 335–340. Engineers, Second edition, Brooks/Cole, Pacific
19. Cook, R.D., Malkus, D.S., Plesha, M.E., and Grove, California, 2002.
22. Friedman, Z. and Kosmatka, J.B., An Improved
Witt, R.J., Concepts and Applications of Finite
Two-node Timoshenko Beam Finite Element,
Element Analysis, Fourth edition, John Wiley
Computers and Structures, Elsevier, 47(3), 1993,
and Sons, New York, 2002. pp. 473–481.
62