On-line condition monitoring of railway equipment using neural networks

The work reported here is concerned with presenting the results of the condition monitoring research work carried out at the University of Birmingham, UK. The paper begins with outlining the motivation for the research in terms of improved reliability of equipment and quality of service to metro passengers as well as the need for cost-effective and efficient maintenance management. An overview of the requirements for condition monitoring systems is presented and their role in a number of maintenance management strategies are discussed. A summary of the comprehensive survey of various condition monitoring and fault diagnosis techniques in railways and other industries is outlined. A generic early-warning system is proposed for mechanical equipment currently in operation in many railways. The concept has so far been validated on the train-stops during both laboratory and field trials and is currently being extended to other equipment. Two case studies, a train-stop and a DC circuit breaker, are described in this paper.