Real-time fault diagnosis — Using occupancy grids and neural network techniques

Abstract

This paper presents a methodology for real-time fault diagnosis of manufacturing systems using occupancy grids and neural network techniques. Themain advantages ofthe system over other existing methods are its ability to capture imprecise and time dependent information, ability to accommodate nonlinear relationships, ability to learn and acquire knowledge automatically. A case study related to real-time milling machine fault diagnosis is discussed. The paper also discusses the problems with the proposed method and the future research directions.KeywordsNeural networksOccupancy gridsInformation measureBack-propagation algorithm