Anda di halaman 1dari 8

Arab J Sci Eng

DOI 10.1007/s13369-014-1232-2

RESEARCH ARTICLE - CIVIL ENGINEERING

Development of 3D Finite Element Code of Incompatible


Displacement Mode for Flexural Analysis
Ahmed Hasan Alwathaf

Received: 15 June 2013 / Accepted: 30 September 2013


King Fahd University of Petroleum and Minerals 2014

Abstract For three-dimensional finite element modeling, it 1 Introduction


is necessary to use a solid element taking into account the
flexural response and also should be simple for computer stor- One of the main causes of the inaccuracies in lower-order
age. One of the elements that can attain the previous require- finite elements is its inability to represent certain simple
ments is the incompatible (non-conforming) element. In this stress gradients and its over-stiffness estimation for element
paper, formulation and algorithm of the developed 3D finite in flexure. In the finite element analysis, the accuracy can be
element code of incompatible displacement mode for flex- improved by using a finer mesh or by using refined (higher
ural analysis are presented using 3D 8-noded isoparametric order) elements. Sometimes, the requirement of finer meshes
incompatible brick element. Several comparisons and a para- over a region of rapidly varying stress leads to bad numer-
metric study have been carried out to verify the incompatible ical condition so that larger or more complicated elements
brick element and the developed code. Accurate results have are mandatory [1,2]. With the introduction of isoparamet-
been obtained by the 3D incompatible element in flexural ric elements, many refined elements with higher-order dis-
analysis of beams in comparison with the standard element. placement field have been developed. Even though these
elements give good results, they require more computer time
Keywords 3D finite element Incompatible displacement and
mode Computer code Algorithm Flexural analysis storage.
Stiffness Furthermore, capability of standard 3D finite element in
flexural modeling of the structures is limited [3]. Figure 1
shows a simple hexahedral (brick) element subjected to pure
bending acting on faces normal to x-axis and in x y plane
(about z-axis). The exact displacement of this type of load-
ing is illustrated in Fig. 1b, whereas the finite element dis-
placement with standard element is shown in Fig. 1c. This
means that the displacement function for standard element,
given in Eq. (1), is not enough because it does not contain
all the quadratic terms in the shape functions. Therefore,
the standard element is not adequate in representing flexural
response.
Other attempts were made to improve the basic accu-
racy of the simpler (first order) elements, without very much
increasing the computer time and storage, by introducing
the incompatible or non-conforming displacement modes
A. H. Alwathaf (B)
[4,5]. Even though the incompatible element reveals slight
Department of Civil Engineering, Faculty of Engineering,
Sanaa University, P.O. Box. 12544, Sanaa, Yemen incompatibility in some cases of loadings [1], in case of
e-mail: aalwathaf@yahoo.com; aalwathaf@suye.ac coarse meshes only, the incompatible element overcomes

123
Arab J Sci Eng

Fig. 1 Brick element under


bending. a Brick element. y
b Exact displacement. c FE
displacement with standard
Natural coordinates
element

x
Mz 2d Mz

z v
2b 2c
(a) u
w

(b) (c)

most shortcomings in the standard element particularly in may be of the following form [7]:
flexural analysis. The addition of incompatible displacement
modes to lower-order displacement-based elements was re- u  8 ui
evaluated and a new method of stress recovery was presented v = Ni vi + [P] {} (2)

in which incompatible modes are introduced [6]. This study w i=1 wi
shows that it is still the most suitable element that can be
where [P] is the additional incompatible displacement modes
used for flexural analysis.
that will be used to correct the displacement error and defined
In this study, the formulation and calculation algorithm
as[7]:
of the 3D incompatible brick element are presented. A com-

puter program (three-dimensional finite element incompati- P1 P2 P3 0 0 0 0 0 0
ble program (3D FEIP)) has been developed in FORTRAN

code to implement the calculation algorithm. Several sub- [P] = 0 0 0 P1 P2 P3 0 0 0 (3)

routines have been written to execute different tasks in the
0 0 0 0 0 0 P1 P2 P3
program. Verification and parametric study have been con-
ducted to validate the code and to study the effect of some and
analytical parameters on flexural analysis of beam.  
{}T = 1 2 3 4 5 6 7 8 9 (4)

where 1 . . ..9 are constants that are functions of dimen-


2 3D Finite Element Formulation sions, curvature, and Poissons ratio of the element. It should
be noted that the i s are nodeless degree of freedom and
2.1 Element Formulation
P1 = (1 2 ) (5a)
The general 8-noded brick element displacement approxima-
P2 = (1 )2
(5b)
tion is
P3 = (1 )2
(5c)
u  8 ui
v = Ni vi (1) As the formulation is achieved in terms of natural coordi-
nates ( , , ) for the isoparametric brick element, we have
w i=1 wi
to establish the relationship between the derivatives of func-
For a general 8-noded brick element, the displacement tions in the natural coordinates and the derivatives in Carte-
approximation that takes into account the flexural behavior sian coordinates (x, y, z). This is derived by evaluating the

123
Arab J Sci Eng

Jacobean, J , given as [7]: where



/ / x
x
u/ x










y

v/ y

/ = J / y
(6) z w/z




{} = = (15)





x y

u/ y + v/ x

/ /z




yz

v/z + w/ y


where zx u/z + w/ x

x/ y/ z/ Ni / x 0 0

0 Ni / y 0
J = x/ y/ z/ (7)

0 0 Ni /z
x/ y/ z/
[Bi ] =
Ni / y Ni / x

(16)
0
Therefore,
0 Ni /z Ni / y

/ x /


Ni /z Ni / x





0
1
/ y = J / (8)





/z /  
{di }T = u i vi wi (17)
where


J11 J12 J13
 
1 [Ba ] = [B1 ] [B2 ] [B8 ] (18)
J = J21 J22 J23 (9)


J31 J32 J33
 
{da }T = u 1 v1 w1 .. u 8 v8 w8 (19)
2.2 StressStrain Relationships

The strain relation can be written in terms of the shape func-    


P
tion derivatives and displacements as: P P2 P3

1
0 0 0 0 0 0
x x x

1
{} = [B] {d} (10) 0 P1
0 0 P2
0 0 P3
0 4


y y y




P1 P2 P3 7
where 0 0 0 0 0 0 2

z z z

= P1 P1 P2 P2 P3 P3 5
y 0 0 0
{d}T x y x y x 8
  3


.. P1 P1 P2 P2 P3 P3

= 0 0 0
u 1 v1 w1 . u 8 v8 w8 . 1 4 7 2 5 8 3 6 9 z y z y z y

6

(11) P1 0 P1 P2
0 P2 P3
0 P3 9
z x z x z x

(20)
 
[B] = [B1 ] [B2 ] . [B8 ] ...  P    P    P   (12) The derivatives of additional shape functions can be written
1 2 3 as follows:
Equation (12) can be written as follows: Pi Pi Pi Pi
= J11 + J12 + J13 (21a)
x

8
   Pi Pi Pi
Pi
{} = [Bi ] {di } + P   (13) = J21 + J22 + J23 (21b)
y
i=1
Pi Pi Pi Pi
or = J31 + J32 + J33 (21c)
z
  
{} = [Ba ] {da } + P   (14) Upon substitution into Eq. (22) in order to obtain [P  ]

123
Arab J Sci Eng


2 J11 0 0 2 J12 0 0 2 J13 0 0
0
2 J21 0 0
2 J22 0 0
2 J23 0

  0 0 2 J 0 0 2 J 0 0 2
J33
P = 31 32 (22)

2 J21 2 J11 0 2 J
2 J22 12 0 2 J
2 J23 13 0
0 2 J
2 J31 0 2 J
2 J32 0 2 J
2 J33
21 22 23

2 J31 0 2 J
2 J11 0 2 J
2 J12 0
2 J13
32 33

or
2.3 Element Stiffness

[K ] = Wi W j Wk f (i , j , k ) |J |i jk (30)
The standard 3D element stiffness matrix is evaluated using
i j k
the following relationship:
 where
[K ] = [B]T [D] [B] dV (23) i, j and k are the sampling points.
V W is the weight of Gauss numerical integration.
The modified element stiffness matrix [K ] reveals the flex-
With the additional mode, the stiffness matrix can be written ural response because it has the effect of incompatible or
as: non-conforming displacement mode, {}. The obtained dis-
  
placement vector, {da }, will be used only to obtain the strain
[Ba ]T   
[K ] =   T [D] [Ba ] P  dV (24) and stress in the element because the displacement mode,
P
V {}, is nodeless displacement vector.
or
    [B ]T [D] [B ] [B ]T [D]  P   
[K ] =  a  a  a
  T   dV 3 ProgramAlgorithm
T
P  [D] [Ba ] P [D] P 
V
Incompatible element stiffness subroutine and different aux-
(25)
iliary subroutines have been developed to include the addi-
The element equilibrium equation can be represented as: tional modes. The calculation algorithm of the element stiff-
     ness is described in the flowchart shown in Fig. 2.
[K aa ]24x24 [K a ]24x9 {da }24x1 {Fa }24x1
= (26)
[K a ]9x24 [K ]9x9 {}9x1 {0}9x1

To remove the nodeless displacements {}, static condensa- 4 Program Code Verification and Parametric Study
tion will be used and Eq. 26 can be written as:
  To verify the developed code of incompatible element, two
[K aa ] [K a ] [K ]1 [K a ] {da } = {Fa } (27) cantilever beams with different loading and different meshes
have been analyzed and the results have been compared.
Therefore, the condensed stiffness matrix (order of 24x24) Moreover, a parametric study has been conducted to investi-
that will be used in the incompatible element is gate the effect of mesh size and type of loading on deflection
     and developed stresses. For each beam, the analysis has been
[K ] = [Ba ]T [D] [Ba ] carried out three times as follows:
V
     T  1
[Ba ]T [D] P  P [D] P  1. Using 3D incompatible element (with incompatible
   mode).
P  [D] [Ba ] dV
T
(28) 2. Using standard 3D element (without incompatible mode).
3. Theoretical solution (closed-form solution).
For the use of Gauss quadrature numerical integration in
order to estimate Eq. , the equation takes the form
The theoretical solution is obtained from the classical beam

theory [8]. Displacement in terms of bending moment in
[K ] = f (i , j , k )dV (29) a beam can be obtained from the second-order differential
V equation as follows, assuming that the beam cross section is

123
Arab J Sci Eng

Fig. 2 Calculation algorithm of


element stiffness with
incompatible modes
Retrieve element geometry and material
properties for the current element

Put initial values as zeros for all arrays

Call a subroutine which sets up the constitutive matrix, [D]

Loops cover all integrated points


Look up sampling position for the current integration point

Call shape function subroutine to find Ni and their derivatives with respect
to x, y and z at integration points G, G and G

Call Jacobean subroutine J and its inverse J-1

Call subroutine to find [Ba] of Eq.18 and [D][Ba] multiplication of Eq. 25


Loop over all elements

Call subroutine to find [P'] of Eq. 22 and [D][P'] multiplication of Eq. 25

Calculate the stiffness matrix entries of Eq. 25

Implement the static condensation procedure to find the condensed stiffness matrix of Eq. 28
[K]

Write stiffness matrix [K] into a file for use in the solution

symmetric about the y-axis, Fig. 3;


y
d2 u(x) Mz (x)
= (31) x
dx 2 E Iz z
Note that the bending moment Mz will in general be a func-
tion of x. Moreover, normal stress in x-direction may be
written in terms of the bending moment as follows: t
1 2 3 N
d2 u(x) Mz (x)
x = y 2
= y (32)
dx E Iz b

Table 1 shows the material, geometrical properties and load- L


ing cases for both beams. Figure 3 illustrates the 3D beam
view, where the geometrical parameters are indicated. Fig. 3 Cantilever Beam I and Beam II

123
Arab J Sci Eng

Table 1 Material and


geometrical properties and Parametric study variables Beam I Beam II
loading cases of studied beams
Material properties Modulus of elasticity, E (N/m2 ) 2.00E+10 2.00E+10
Poisson ratio, 0.15 0
Geometrical properties Length, L (m) 3 2
Width, b (m) 0.2 0.2
Thickness, t (m) 0.2 0.3
Number of elements, N 3, 12, 25, 30 5
Loading Concentrated Load at free end, P (N) 2,000 10,000
Concentrated moment at free end, M (N m) 3,000
Gravity load (self-weight), q (N/m3 ) 2,000
Uniform distributed load, w (N/m2 ) 1,000

0 0
0 10 20 30 40 0 10 20 30 40
-0.001 -0.0005

-0.002 -0.001

Displacement, m
Standard Element Standard Element
Displacement, m

-0.003 Theoretically -0.0015 Theoretically


Incompatible Element -0.002 Incompatible Element
-0.004

-0.005 -0.0025

-0.006 -0.003

-0.007 -0.0035
Number of elements
-0.008
Number of elements
Fig. 5 Displacement versus number of elements due to gravity load,
Fig. 4 Displacement versus number of elements due to concentrated Beam I
load, Beam I
0
-0.0001 0 5 10 15 20 25 30 35

4.1 Beam I -0.0002


Displacement, m

Standard Element
-0.0003
Theoretically
The comparisons were conducted for the displacement at free -0.0004
-0.0005 Incompatible Element
end and normal stress in x-direction near the fixed end versus
the number of elements (mesh size) as shown in Figs. 4, 5, 6, -0.0006

7, 8, and 9 for different loading cases (see Table 1 for loading -0.0007

cases). -0.0008
Number of elements
As shown in Figs. 4, 5, and 6, the displacement results
at free end using incompatible element coincide mostly on Fig. 6 Displacement versus number of elements due to uniform load,
the theoretical values. The standard element reveals over- Beam I
stiffening behavior more than incompatible element for all
loading cases. Moreover, normal stress in x-direction results
ing the analysis were five elements in the longitudinal direc-
is very close to the theoretical solution as shown in Figs. 7,
tion (see Table 1). Figures 10, 11, 12, and 13 show the analysis
8, and 9 especially with finer meshes (small element). These
results of deflection and normal stress for concentrated load
results give an evidence of the accuracy of the 3D incom-
and moment at free end. The concentrated moment on the
patible element in the flexural analysis of beam for different
free end was applied as coupled forces at the free face nodes.
loading cases.
As shown in Figs. 10 and 11, the deflection values in the
longitudinal direction of the beam obtained by the incom-
4.2 Beam II patible element are almost the same as the theoretical values.
The previous observation of over-stiffening behavior of stan-
In this beam, other types of comparison were conducted dard element is also recognized in Beam II. Moreover, the
to investigate the displacement and normal stresses in x- incompatible elements gave normal stresses very close to the
direction along the span. The numbers of elements used dur- theoretical values as shown in Figs. 12 and 13. The calculated

123
Arab J Sci Eng

3.00E+06 0
0 50 100 150 200 250

Vertical Deflection, cm
-0.05
2.50E+06
Normal stress, N/mm 2

-0.1
2.00E+06 -0.15
Standard Element Incompatible Element
-0.2
1.50E+06 Theoretically Standard Element
-0.25
Theoretically
1.00E+06 Incompatible Element -0.3
-0.35
5.00E+05 Distance cm

0.00E+00
Fig. 10 Vertical deflection due to concentrated load, Beam II
0 10 20 30 40
Number of elements
0
Fig. 7 Normal stress versus number of elements due to concentrated
-0.01 0 50 100 150 200 250

Vertical Deflection, cm
load, Beam I
-0.02
-0.03 Incompatible Element
1.80E+06 -0.04 Standard Element
1.60E+06 -0.05 Theoretically
N/mm 2

1.40E+06 -0.06
1.20E+06 -0.07
1.00E+06 -0.08
Normal stress,

Standard Element Distance cm


8.00E+05
6.00E+05 Theoretically
Fig. 11 Vertical deflection due to concentrated moment, Beam II
4.00E+05 Incompatible Element
2.00E+05
0.00E+00 4
N/mm 2

0 5 10 15 20 25 30 35 3.5
Number of elements 3
2.5 Incompatible Element
Normal stress,

Fig. 8 Normal stress versus number of elements due to gravity load, 2 Standard Element
Beam I 1.5 Theoretically
1
4.50E+05 0.5
N/mm 2

4.00E+05 0
0 50 100 150 200 250
3.50E+05
Distance cm
3.00E+05
Normal stress,

2.50E+05 Fig. 12 Normal stress due to concentrated load, Beam II


2.00E+05 Standard Element
1.50E+05 Theoretically
1.00E+05
Incompatible Element
5.00E+04
0.00E+00 0.7
N/mm2

0 10 20 30 40 0.6
Number of elements 0.5
0.4 Incompatible Element
Normal stress,

Fig. 9 Normal stress versus number of elements due to uniform load, 0.3 Standard Element
Beam I 0.2 Theoretically
0.1
0
0 50 100 150 200 250
stresses are located at the Gaussian points, which explains the Distance cm
step distribution shown in Fig. 12 due to the stress gradient.
Fig. 13 Normal stress due to concentrated moment, Beam II

5 Conclusion
accurate simulation for the flexural behavior has been ach-
The incompatible or non-conforming element has improved ieved by the developed 3D FE program code. The developed
the flexural response and gives more accurate results for program code for the 3D incompatible element yields very
deflection and stresses in comparison with the standard ele- accurate results for finer FE meshes for different loading
ment.The analysis results for the studied problems show that cases.

123
Arab J Sci Eng

References 5. Wilson, E.L.; Taylor, R.L.; Doherty, W.P.; Ghaboussi, J.: Incompat-
ible displacement models. In: Fenves et al. (eds.) Numerical and
1. Cook, R.D.; Malkus, D.S.; Plesha, M.E.: Concepts and Applications Computer Method in Structural Mechanics, pp. 4357. Academic
of Finite Element Analysis. Wiley, New York (2002) Press, London (1973)
2. Cook, R.D.: Finite Element Modeling For Stress Analysis. Wiley, 6. Wilson, E.L.; Ibrahimbegovic, A.: Use of incompatible displace-
New York (1995) ment modes for the calculation of element stiffnesses or stresses.
3. Beyabanaki, S.A.R.; Jafari, A.; Biabanaki, S.O.R.; Yeung, M.R.: Finite Elem. Anal. Des. 7, 229241 (1990)
A coupling model Of 3-D discontinuous deformation analysis (3-D 7. Krishnamoorthy, C.S.: Finite Element Analysis, Theory and Pro-
DDA) and finite element method. Arab. J. Sci. Eng. 34(1B), 107119 gramming, 2nd ed. Tata McGraw Hill Comp. Ltd., New Delhi (2001)
(2009) 8. Hinton, E.; Owen, D.R.J.: Finite Element Programming. Academic
4. Taylor, R.L.; Beresford, R.J.; Wilson, E.l.: A non-conforming ele- Press, London (1983)
ment for stress analysis. Int. J. Numer. Method Eng. 10, 12111219
(1976)

123

Anda mungkin juga menyukai