Application of knowledge roughnessbased multivariate decision tree in transformer fault diagnosis system
Ran Li · Dianli zidonghua shebei · 2005
The construction of multivariate decision fault tree by rough set theory for transformer is introduced. It considers the correlations among attributes and avoids repeated detecting the fault symptoms during the decision tree generation. With good selfadaptability and selforganization,it is suitable for autoprocessing of large samples. Furthermore,it uses discernibility matrix of rough set to select symptoms,checks multivariate using generalization concept and chooses attributes according to the knowledge roughness. Thus the key problems in building multivariate tree are solved effectively. Compared with entropy,computations of knowledge roughnessbased decision tree are reduced greatly and the tree may be optimum. The result of practical fault examples shows that the proposed method simplifies the decision tree and reduces the redundancy of fault diagnosis information with high efficiency,and easy to be understood.