Decision rule extraction for fault diagnosis by discernibility matrix

This article focuses on algorithm to directly and fast finding out the decision rule from decision tables by using rough set theory. The concept of relative rule is put forward and the method for finding absolute reduction decision rule through combining rules of relative decisional requirement sets is given in this paper. Using this concept, the relative discernibility matrix between decision classes can be created to decide the relative decision requirements which are conjoined to find the decision requirements of all the decision classes. The relative decision requirements are extracted by creating relative discernibility matrix, so that scales of computing can be cut down substantially and saving storage space. Theoretical analysis shows that the proved method is feasible. Last, the proposed method is applied in fault diagnosis of reciprocating machines.