Abstract: A method of detecting and isolating critical faults is proposed, which is based on the problem that the critical fault is easy to be missed due to the small probability fault in the traditional test optimization method. First, from the perspective of system security, the importance of critical faults is analyzed. Then, based on the fault test dependency matrix, an optimal test selection model taking critical fault detection rate and critical fault isolation rate into account is proposed. The optimization target is minimization of cost of selected tests. The constraints of the model involve fault detection rate and fault isolation rate of system. The binary particle swarm optimization algorithm based on centroid, whose inertia weight is adjusted based on population diversity, is used to solve the optimization model. The simulation results show that the optimization method of considering critical faults can effectively eliminate the serious threat to the safety of the equipment due to the missing of critical faults.

Key words: design for testability, test selection, critical fault, centroid, binary particle swarm optimization