Near-optimal Test Sequencing Algorithms For Sequential Fault Diagnosisl

In [1], optimal AN DIOR, graph search algorithms were developed to analyze the testability of a systeln design and to determine an optimal sequence of tests for the trouble-shooting of hierarchical systems. However, NP-hardness of the test sequencing problem makes the computation of optimal testseq{lence impractical for even moderate-sized problems. Hence, there is a need for near-optimal test sequencing algorithms that provide a trade-off between optimality and computational complexity. Consequently, we have developed three classes of nearoptimal algorithms: multi-step information heuristic, hybrid breadth-depth search, and rninimax heuristic. These algorithms compute fault isolation strategies with significantly lower computational requirements than the optimal A0" algorithm and have enabled us to solve testability analysis problems with as many as 50,000 failure sources and 45,000 test points. In ac\dition, our test-sequencing algorithms handle precedence constraints on tests, overlapping setup operations for tests, rectification/replacement of modules, and fault isolation to any desired level of the system hierarchy. The above suite of test-sequencing algorithms were extended to handle the case of imperfect tests, where in missed detections and false alarms add another dimension of uncertainty to the testing process .