Hi, Unless, I misunderstood the question, with 3D, I believe your curve is correct since for example for linear elastic behavior, the Poisson effect (thickness) cannot provide stress increase in the same way as in the load direction. It depends on both the damage model and load (if it is stress or maxstress or homogeneous damage used).

Also, looking at your input file, you are doing dynamic explicit using combination of different element types. I suggest for testing purposes of damage law to do a simpler analysis which will highlight only the damage behavior and not some other coupled effects. You should then be able to extract theoretically the damage behavior.

A linear traction-separation cohesive law does not follow a linear trend in the SDEG variable. You can analitically extract the shape of the SDEG curve and use the correct expression for the tabular input of the cohesive law. I have experience on that and you can input any cohesive law you imagine, the only constraint is that SDEG monotonically increases. By the way I recommend that you use a very simple single element model to post process it easily and avoid any other effects as Daniel  suggested.

Thank you for your responses. Miguel's suggestion of the non-linear relationship between the damage variable and stress was the problem.