Multiple-fault Diagnosis of Finite State Machine Based on Dynamic Clustering

To solve the problem of too many enumerations in multiple-fault diagnosis, a method based on dynamic clustering analysis was proposed. First, the failed test cases were clustered according to whether they have the same conflict set of initial symptom, and the suspicious degree of the conflict set of initial symptom and its transitions was calculated with this method. Then, the possible fault combinations of transitions were enumerated according to the suspicious degrees. In this process, the failed test cases needed clustering again. Finally, the test set was used to verify the possibility of faults and to generate the fault diagnosis set. Results show that this method can effectively reduce the number of faults enumeration and improve the diagnosis efficiency.