I am running a Latin Hyprecube Sampling routine for statistical reliability analysis. I randomdly sample from 9 probability distributions of parameters that from inputs to a crack growth integral equation. When I take a larger number of samples you get unlucky and at least one of combinations of parameter inputs to the integral results in the integral failing to converge for that combination and this results in an error. Is there a way of including a diagnostic line in the program loop that will flag that a combination does not converge (eg like a "time-out" in convergence iterations) but then record this combination as not converging and then continue on with the loop?
The attached sheet usually works for several hundred samples but when say n > 300 it often flags an error (sometimes you may need a large number of samples before you get unlucky eg. n = 10,000) which I have been able to isolate through trial and error to prove there is a particular combination that does not converge.
Any assistance would be appreciated.
Cheers
Ross
