I am modeling the behavior of grains and grain boundaries under impact loading using bulk and cohesive elements in an explicit simulation. My mesh consists of triangular bulk elements (equal size) with zero thickness cohesive elements inserted between bulk elements. The cohesive elements utilize QUADS damage with linear softening. I am interested in the damage dissipation energy associated with the failure of the cohesive elements. As such, I am outputting ELDMD for each element. I’ve run quality checks to see if the ELDMD of a failed element matches the theoretical value and they are orders of magnitude different and don’t seem to follow any obvious scaling pattern. I’m using the default time increment settings, but I’ve reduced the maximum time increment to be much smaller (4.16E-13 vs 1e-14) than the stable time increment and the results don’t differ. It’s also important to note that this simulation has thousands of elements with different traction separation laws. Specifically, each element has the same K and failure displacements, but different nominal stresses.  Thus, their damage initiations and theoretical damage dissipation energies differ. The maximum difference between nominal stress values is about 90% of the maximum. Example: 2,000 MPa and 200 Mpa. 

I made a small model with one cohesive element between two bulk elements and then applied loading. The damage dissipation energy associated with failure matches the theoretical value. However, when I introduced more than one cohesive element into a simple model with different separation laws, the energies start to diverge from the theoretical values. This simple model is four triangular bulk elements (in a square) with four cohesive elements inserted in between. 

Theoretical value: .5 x nominal strength x critical separation x element area 

Note that I am accounting for the different nominal strengths and element areas in my calculations. I assume there is something going on in the solver that I don’t understand. Any help or guidance would be greatly appreciated.