I admit I'm not sure how the Knitro solvers in Prime handle integer constraints, but most solvers don't like them. Most solvers head downhill on the error surface, and if the error surface is not smooth they can't handle it. Asking for integer solutions makes it very non-smooth, with many local minima that are singularities at the integer solution points.
Unfortunately my P3.0 can't read your 3.1 file, either, as PTC has decided, that even every subversion should get its own incompatible file format and no one should be able to read newer formats nor be able to save in older file formats.
Sometime (while not often) the symbolic "assume, t=integer" helps as here with Andy's example
But unlike Mathcad 15 and below you can't combine symbolic evaluation with a solve block in Prime. So this may not be an option either.