Aside from posting your model to actually identify the problem, possible solutions could be:
1) Initialize your blocks (Integration, Unitdelays, whatever is used in the calculation)
2) Make sure no division by zero occurs
3) Make the solver fixed step (in model properties - solver options) instead of variable step, use a small step size (for example: 1e-3 )
4) Add a "Rate Transition" block before/after the block
Mind that all of these are just arbitrary solutions, and you should upload the actual model and workspace (if needed) to identify the problem.
