Genetic algorithm based approach for segmented testing

Segmented testing, in which a set of test patterns are partitioned into several segments, has been shown to be applicable for on-line testing as it can shorten the mean time to fault detection. One problem that exists for segmented testing is how to partition the set of tests so that the detection latency can be minimized. In this paper, we first propose a method to compute a lower bound of detection latency. Then we present a genetic algorithm (GA) based procedure to partition a given test set into several test segments aiming to reduce the detection latency. Experimental results on ISCAS´89 benchmark circuits demonstrate that the proposed approach can effectively reduce detection latency.