Failure diagnosis for pneumatic control valves by neural network

This paper deals with the failure recognition of pneumatic control valves using a neural network system. Pneumatic control valves are usually the final controlling element of process control equipment at power systems. As they are non-liner and involve a dead time, the signs of failure are complex. For the advance of reliability and safety at power systems, a diagnosis support technology for predictive maintenance of expert-level by simple operations seems worthwhile. The authors have tried to develop a new failure diagnosis method that classifies the response waveform of valve components using a neural network when the valve is activated in a step manner.