A Simulation-Based Test Generation Scheme Using Genetic Algorithms

Abstract

This paper discusses a Genetic Algorithm-based method of generating test vectors for detecting faults in combinational circuits. The GA-based approach com-bines the merits of two techniques that have been used previously for generating test vectors- the directed search approach and the random test method. We em-ploy a variant of the traditional GA, the Adaptive G A ( A G A), to improve the e f i c a c y of the genetic search. Two cost functions that are used for assessing the qual-ity of the vectors are discussed. The performance of the AGA-based test generation approach has been evaluated using ISCAS-85 bench-mark circuits, In our approach, the number of vectors that need to be simulated for detecting all detectable faults is significantly smaller than that required for a random test method. Even when optimized input dis-tributions are used to generate the random test vec-tors, the A G A sustains its superior performance over the random test method. 1