Neural network application to diagnostics of pneumatic servo-motor actuated control valve

This paper presents an approach for development of a diagnostic system on the base of ANFIS networks for creation of reference models of pneumatic servo-motor actuated control valve in the mode without faults. Detection, localization and identification of faults is carried out on the basis of the analysis of extreme values of residuals. The idea of this approach is illustrated on a practical example of the diagnostic task solution for benchmark model of pneumatic servo-motor actuated control valve developed by the European university project DAMADIC'S in MATLAB/SIMULINK.