Abaqus query on UMAT: NaN issues

Hello, I have a written a UMAT for a hyperfoam type material in Abaqus. To validate it I am comparing my UMAT results with that from a *HYPERFOAM material definition. I am facing a particular problem:

Under a displacement load on the boundary the UMAT results are comparable to that of the *HYPERFOAM results. I compare displacement U, nodal forces TF, deformation gradient DG and total strain energy and work done. For total strain energy and work done the first few significant digits match and the difference between the two results increase slightly with increase in increment, but still the accuracy is acceptable and hence I conclude that my UMAT is probably working correctly.

Next I apply a different set of displacement BC (of lower magnitudes) on the same set of boundary nodes where I had prescribed BC in the above case. I observe that the solution from UMAT gives me NaN values for all fields from increment 3 onwards, whereas with *HYPERFOAM case no such problem occurs.

On further investigation I find that in the UMAT Abaqus passes in NaN values for DFGRD1 and STRESS during the initial iterations (while searching for an equilibrium solution) in increment 3. I also verify that in the previous 2 increments there is no NaN or 1/0 (inifinity) type values returned from UMAT.

Can anyone please suggest what can be a plausible reason for this problem ?

Thank you for your time and advice.

--

Sourish