Diagnostics of a coolant system via neural networks

The application of a neural network (NN) to diagnose faults in a machine tool coolant system is described. The measured variable in the system is the pump outlet pressure; the transient response of this as the flow valve is closed is used as a pattern for fault recognition. A two-stage diagnostic system using a back propagation NN at each stage is described and this is trained by using data from the coolant system under healthy (i.e. unfaulty) and faulty conditions. The faults are simulated on the real coolant system. Novel (i.e. previously unmet) faults are defined by maximum values of a 'deviation' which is used to allocate faults. The diagnostic system is shown to be capable of first deciding whether the system is healthy or faulty; if faulty, it then decides whether one of the three common faults or a novel fault is occurring. Having made the decision that one of the common faults is occurring, it is then capable of deciding, from four different levels, the approximate severity level of the fault. Of 345 tests on the coolant system the diagnostic system allocated the fault 99 per cent correctly and the severity level 96 per cent correctly.