Failure diagnosis system on pneumatic control valves by neural network

A prototype failure diagnosis system is developed using neural network technology for the actuators of pneumatic control valves. The data of 30 failure patterns are experimentally collected using more than ten sensors. A fast Fourier transform (FET) is carried out on the time series of the sensor signals. The data of the magnitude spectrum, phase difference and others are used as the characteristic parameters in the failure diagnosis. Appropriate failure diagnosis information is extracted from the data. Similarities among the failure characteristics are established using fuzzy clustering and statistical analysis. The prototype consists of plural subnetworks and one specific sensor signal, and deals with the magnitude spectra from the sensor signal. The main network makes the final decision according to the output from the sub-networks and other data. The number of network connections can be reduced by approximately 40% without degradation of the recognition capability, in comparison with a conventional system where only one neural network is used