Hierarchically Structured Inductive Learning for Fault Diagnosis

This paper presents a new methodology for fault diagnosis, based on the natural hierarchy of components and sub-components in electrical and/or mechanical systems. In the first section, the advantages of hierarchically decomposing learning tasks are discussed. In the second section, the author’s fault diagnosis system, DE/ IFT, is introduced. The underlying algorithm, the training cycle and the operation of DE/IFT are then discussed. In the third section, the hierarchical methodology for fault diagnosis is presented. In the section following, Hierarchical Condition Description files are introduced and the details of implementing hierarchical fault diagnosis within DE/IFT are explained. Next, an example application is discussed. Results of hierarchical fault diagnosis are presented. These are compared with equivalent results from a non-hierarchical analysis of the same system. Finally, at the end of the paper conclusions are drawn.