Recent advances and prospects in condition monitoring

Abstract

The underlying reason for pursuing Condition Monitoring activities is to enhance the overall performance of plant either by accurately predicting the end of effective life or by diagnosis and location of incipient faults. The achievement of both these objectives requires the use of a combination of measurement, modelling and statistical techniques, but the balance of these three strands varies with the plant under study. Practitioners of these different disciplines sometimes develop rivalries, but in reality all three are essential. The fundamental problem involved is the determination of the physical process within a machine which cannot be measured directly and so must be inferred from external measurements, the most common of which is vibration. This inference from measurement to diagnosis requires some form of model, whether physics or statistics based. The paper describes some work on the monitoring of large turbines and in particular, the ways in which measurements and models can be combined to enhance insight into a machine’s operation. Similarly, our understanding of complex data can be enhanced by the use of Artificial Neural Networks and these can be used to enhance understanding. Combining various approaches leads to some new possibilities which are briefly outlined.