ABSTRACT

Predictive monitoring aims to identify developing problems in equipment before an actual breakdown occurs. The goal is to reduce unnecessary maintenance while ensuring that equipment does not cause unexpected work stoppages.

We are developing an automatic predictive monitoring system to detect imminent motor failures which is based on a neural network autoassociator used as a novelty or anomaly detector. The autoassociator is trained to reconstruct spectra obtained from the healthy motor so that abnormal spectra are perceived as anomalies or novel events.

In laboratory tests, we have demonstrated that the trained autoassociator has a small reconstruction error on measurements recorded from healthy motors but a larger error on those recorded from a motor with a fault. We have designed and built a motor monitoring system using an autoassociator for anomaly detection and are in the process of installing several test systems to monitor motors in several industrial and commercial plants.

ABOUT THE SPEAKER

Thomas Petsche is a Project Manager at Siemens Corporate Research, Inc. where he has worked on the development of an on-line motor monitoring system. He received a PhD in Electrical Engineering from Princeton University in 1988, and an ME and BE in Electrical Engineering from Cooper Union in 1983 and 1984, respectively. His research interests include neural networks and learning systems. He is a co-editor of volumes 2 and 3 of "Computational Learing Theory and Natural Learning Systems" (MIT Press) and an associate editor of the IEEE Transaction on Neural Networks.