Non-invasive diagnostics of motor-operated valves

This paper is concerned with the development of data analysis methods to be used in a model-based approach to the online monitoring and diagnosis of motor-operated valves (MOVs). The technique to be utilized will include the extensive integration of mechanical and electrical measurements and signal processing. A torque estimator is developed and tested to obtain electric torque of the induction motors which are attached to the MOV system. Transfer functions between the actuator housing vibration and gear meshing force are measured and inverse filters are designed to recover the source waveforms. To monitor the operating condition, amplitudes and frequencies of the recovered signal are examined. Various frequency demodulation techniques are studied to find the most robust method. An adaptive linear phase bandpass filter is developed to improve signal-to-noise ratio and to track the important frequency components for diagnostic purposes during the operation. Finally, stand-alone valve experiments are carried out to validate the developed estimation scheme. Faults are introduced into the experiment set-up by placing obstructions in the path of the closing valve. The results of this study will be directly applied in the monitoring of MOVs and the methods developed can be applied to other diagnostic system as well. An additional benefit of this study will be a reduction in reliability problems associated with torque switch failures in existing MOVs.