Why in the case of plane strain results from FEM code in Matlab and Abaqus don't match?

[wikiiri]

Why in the case of plane strain results from FEM code in Matlab and Abaqus don't match whereas the plane stress results do?
I am trying to simulate a simple pure shear problem while considering a viscous fluid with my own FEM code and I compare it with Abaqus. I use 4 node bilinear quadrilateral element in my FEM code in Matlab. I apply pure shear boundary conditions. I consider Plane stress and Plane strain cases.
My results match perfectly (Displacements, reaction forces and shear stress) with Abaqus in Plane stress case (I use CPS4 element with full integration). Whereas in the case of Plane strain my results don't match (I use CPE4 (A 4-node bilinear plane strain quadrilateral) element with full integration).
Plain strain results only match in case I consider just one element for the whole problem but as soon as I increase the number of elements I get a difference in values.
I checked my stiffness matrix for both cases and they also have some values which are the same and some not. Stiffness matrix. Any suggestions to get rid of this problem?
Thanks

 

[FEA way]

Well, it's hard (if not impossible) to say without more details. It may be caused by error made in Abaqus (seems unlikely though) or glitch in your code. You may have wrong element formulation or some mistake in matrix assembly proces (since it only happens when there are multiple elements considered). Make sure that your analysis in Abaqus is correct, maybe check it with another software and then you can treat it as a reference solution for your code. It would also be good to have an analytical solution for comparison.
Think whether the differences are big or small enough to be neglected. Remember that Abaqus features highly optimized code developed and upgraded through many years by a large group of specialists. There's nothing wrong in fact that your Matlab results differ from these obtained in Abaqus.

One more thing - make sure that you solve exactly the same problem. In case of your code you know exactly what it does but in Abaqus there might be some automatic options that change the definitione of your problem.

 

[wikiiri]

Dear FEA way, thanks for your reply. I have check my matrix assembly and element formulation and they look ok!
I am able to fix my code to some extent but have still some issue. What I mean is that when I consider one material (e.g a fluid) for my whole domain (9 elements), my plain strain results match with Abaqus. Now, when I change the properties of the middle element to a fibre reinforced material(in this element now I will have different stiffness in the axial and transverse direction), then I again have a difference in my plane strain calculations. Can that be due to volume locking? I am not sure.

 

[FEA way]

Volumetric locking occurs in case of almost incompressible materials when second order full integration elements are used (reduced integration elements may suffer from hourglassing).
Abaqus has special elements called hybrid elements used for almost incompressible materials that eliminate the problem of volumetric locking. If your code doesn’t have such elements then indeed volumetric locking may be an issue.