Professional Documents
Culture Documents
INTRODUCTION
All physical problems can be
expressed mathematically in terms of ordinary or partial differential equations. All differential equations cannot be solved analytically. Many DEs do not possess closed form solutions. Finite Element Method is an effective numerical method for solving differential equations. The differential equation in structural mechanics is the Differential equation of equilibrium along with displacement and force boundary conditions.
the physical system in FEM is by using nodes and elements. The process of creating nodes and elements is said as discretization or meshing. The number of degrees of freedom in each node depends on the element type. As the FEM model has finite DOF as against infinite DOF in actual structure, FEM solutions are approximate and FEM model is stiffer
problems which cannot be solved analytically, there are some situations for which convergence is poor in FEM.
The aim of our project is to bring out such an error
called locking and how it is being overcome by the commercial FEA codes.
The FEA softwares used are ANSYS and ABAQUS.
LOCKING :
Locking means the effect of a reduced rate of
convergence in dependence of a critical parameter. eg.,slenderness of the plate in case of transverse shear locking. Because of locking , wrong displacements , false stresses, spurious natural frequencies are encountered. Locking may be
Contd
Shear locking occurs on use of fully integrated linear
elements that are subjected to pure bending. The formulation of these elements promotes shear strains that do not really exist. Volumetric locking occurs in fully integrated elements when the material behavior is nearly or fully incompressible( poissons ratio approaches or equals 0.5) Membrane locking occurs only in Shell and Curved beam elements.
Shear locking:
Timoshenko beam theory works for all l/d ratio
whereas Bernoullie beam theory works for only large values of l/d ratio. In thick beams , shear deformation becomes predominant and Bernoullis theory wont give accurate result. Shear locking occurs in elements which account for shear deformation ,when these elements are used in modeling bending predominant beams .
Modelling
A
Cantilever beam subjected to loading at the free end is analysed with the following properties : Youngs Modulus = 200 GPa Poissons Ratio = 0.3
Ansys Element SOLID 45 and
MESHING :
The beam is meshed such that there are 4 elements in height , 10 elements along the length and 1 element along the width.
L/d ratio = 25
L/d ratio = 5
L/D ratio 25 is 0.312 cm The free end deflection obtained from SOLID45 is 0.308 cm whereas from SOLID185 element the tip deflection is found to be 0.1514 cm. Thus it is apparent that the results from SOLID185 cannot be relied upon. The difference between the results of Solid45 and Solid185 is due to shear locking and it is not an approximation error.
ANALYTICAL DISCUSSION :
Consider Stiffness matrix, K , for the beam element and the expression
is given by
oThe bending stiffness can be exactly evaluated by one point gauss quadrature, on the other hand the shear stiffness matrix (shear strain energy) contains second order terms and two point gauss quadrature is required to exactly integrate it.
term is given as
given as
and
Contd..
Solving we get, W =
and
=
In the limiting condition i.e., tending to thin beam limit
W=l
and expression for remains unchanged. Thus the beam deformation is solely due to bending as given by the equation and this is correct.
Contd..
Solving for , we get =
W =
=
It can be observed that these expressions lead to erroneous results as they contain only the co-efficient corresponding to shear deformation.
overstiffness in bending problems. This formulation includes shear strains in bending which do not physically exist, called parasitic shear.
L/D ratio
ANALYTICAL SOLUTION
SOLID 45
% Error
1. 2. 3. 4. 5.
25 20 15 10 5
incorporated
gives exact solution irrespective of the L/d ratios. SOLID185 has reduced integration option which when enabled produces closer results.
ABAQUS MODELLING
The 8 noded brick element in abaqus is the continuum
problem.
Cantilever beam is modeled for various L/D ratios &
the results from both ansys and abaqus packages are tabulated.
L/d Ratio = 25
L/d Ratio = 5
DEFLECTION RESULTS:
L/D ratio 25 20 15 10 5
ANALYTICAL SOLUTION
C3D8(mm C3D8R ) (mm) 0.9189 0.6312 0.362 0.1454 0.0235 3.32 1.701 0.7183 0.2137 0.02752
ABAQUS SOLUTION
Abaqus has got the following element technologies to
overcome locking: Reduced integration(C3D8R) incompatible modes(C3D8I) C3D8 element has default setting as reduced integration. Enabling incompatible modes option, the answer obtained is still more accurate.
INCOMPATIBLE MODES:
8-node brick elements are frequently used with
incompatible modes to improve the bending behavior in case of pure displacement problems. These elements add additional degrees of freedom to improve the accuracy of the element. The internal nodes are condensed out of the element equations, so the final element matrices have their original DOF s but the element retains its accuracy.
two integration points in each co-ordinate system. The reduced integration elements use fewer integration points and will decrease shear locking because some terms in the gauss integration are eliminated.
r = Ndof / Ncon
given by
Contd.,
It can be seen that the constrain ratio r gets smaller as
issue.
MEMBRANE LOCKING:
Occurs only in curved beam and shell elements.
In curved elements , both flexural & membrane
inextensional bending becomes predominant i.e. Membrane strain becomes vanishingly small when compared to bending strain.
Theoretical Proof
u, s w / R u, s / R w, ss
Where is the membrane strain and is the bending strain
The circumferential and radial displacements
are given by
u a0 a1
w b0 b1 b2 2 b3 3
The strain field interpolations
6b a1 2b2 2 ) ( 23 ) Rl l l
b a1 b0 b2 b 3b b ) ( 1 3 ) 2 (1 3 2 ) 3 (3 5 3 ) l R 3R R 5R 3R 5R
INEXTENSIONAL BENDING
If a thin and deep arch such that L/T>>1 and R/H is small is modeled by a curved shell element , the physical response is inextensional bending such that the membrane strains tends to vanish.
b a1 b 0 2 0 l R 3R b1 3b3 0 5
b2 0
b3 0
Each in turn implies the conditions
Contd.,
A shell was modeled using ansys elements SHELL 93 &
SHELL 63 having the following properties: youngs modulus = 30 Gpa Poissons ratio = 0 shell thickness = 3 cm radius of curvature = 300 cm length of the shell = 600 cm pressure applied = 6250 N/m2
Shell Elements :
ANSYS MODEL :
2.
120
-1.809
-1.7769
-1.4312
19.4
3.
80
-2.4012
-2.23
-2.0979
5.9
4.
40
-1.385
-1.3582
-1.3509
0.0053
Interpretation :
As the angle becomes larger (as the shell becomes deeper),
the difference between the results of edge deflection obtained from shell63 and shell93 is greater.
But Shell63 results are almost closer to the StaadPro
results. This shows that for deeps shells Shell63 should be used instead of Shell93.
Shell93 has spurious membrane strains introduced in it as
the shell becomes deep and Shell63 donot encounter such an issue.
REMEDY :
Reduced integration techniques are not always
successful in overcoming locking behavior and result in overstiffness. Shell63 is a linear element whereas Shell93 is a curved element with midside nodes. Faceted modeling(planar elements)is currently the most reliable and successful method. Curved modeling(curved elements)has only limited success.
CONCLUSION
In Ansys , Element technologies like Reduced Integration
and Extra Shape Function has been adopted to overcome Shear locking.
Similarly in Abaqus , Reduced Integration and Incompatible
selection of appropriate elements i.e. Linear elements instead of curved elements in deep arches and shells.
References :
Finite Element Analysis by C.S.Krishnamoorthy , IIT , Madras. Strength of Materials- Elementary theory and problems by
Stephen Timoshenko Eric Quili Sun, Shear locking and hourglassing in MSC Nastran, Abaqus and Ansys www.cmmacs.ernet.in/cmmacs/pdf/ www.imechanica.org www.mscsoftware.com/events/vpd2006/na/presentations/.../27.p df www.eng-tips.org www.springerlink.com www.openpdf.com www.sciencedirect.com
THANK YOU