Time limit in a programatically programmed parameter estimation

I have defined an estimation object to identify the best parameter set for a simulink model to fit a desired transient behavior. However, since the optimization problem is highly non-linear, it is far from being convex and many local minima appears. As a result of this, when a pattern search algorithm is selected the simulink model to be identified may enter in regions of the parameter space where some model assumptions are invalid, the model behavior become erratic and it stalls. When this situation arises, the only solution is to restart Matlab and therefore no solution can be found to the problem.

In this connection, is there a way to fix a time limit (as it exist for Genetic Algorithms) for the optimization algorithm in order to avoid Matlab crashes?