Fault diagnosis of rotating machinery based on SVD, FCM and RST

Because there is no criterion available, discretization of numerous data about objects of information systems may be the biggest obstacle to performing inductive learning from instances. Based on a Singular value decomposition of matrix and Fuzzy C-means clustering, a discrete approach of continuous attribute values in a decision table, with continuous condition attribute values and discrete decision attribute values, has been proposed. Rough sets theory has been employed for diagnostic rule acquisition of rotating machinery with consideration of conflicting objects of decision table. A weak matching mode of objects with diagnostic rules has been proposed, with diagnostic conclusion and its belief degree obtained by comparing new objects with the standard diagnostic rules. An example at the end of this paper shows that the acquired rules have good merits of the ability of generalization and extension, and to some extent, improves classification level.