Abstract: The present system multiple fault diagnosis is generally realized by generating a test sequence with minimal average test cost under fixed initial fault ambiguous groups, and then the faults are isolated through the fault diagnosis tree constructed according to the generated test sequence, so the test sequence and fault diagnosis tree are relatively fixed. When the initial faults ambiguous groups change, the test strategy still keep unchanging. To resolve the problem, an on-line strategy for multiple fault diagnosis is proposed, the iterative formulas to calculate the possibility of ambiguity groups are deduced, and the optimal test is selected by evaluation function based on fault isolation degree and information entropy of unit cost. The faulty unit with highest posterior possibility to be repaired is searched by set covering algorithm. Test case shows that the on-line strategy can isolate multiple faults by flexibly utilizing known information. It has the advantages of simple calculation and small storage space, and is feasible for multiple fault diagnosis of complex system.

Key words: system assessment and feasibility analysis, multiple fault diagnosis, on-line strategy, compact set, fault isolation degree, information entropy, set covering