Fuzzy model-based condition monitoring of a dry vacuum pump via time and frequency analysis of the exhaust pressure signal

Abstract

A fuzzy model-based diagnostic scheme is designed to monitor dry vacuum pump performance and detect two fault conditions, mechanical inefficiency and exhaust system blockage. The diagnostic scheme is based on time and frequency analysis of the exhaust pressure signal. Power ratios of certain frequency components in the signal spectrum can be used to predict the gas load, motor current and hence mechanical efficiency. Changes in the periodic features of the signal, symptomatic of fault conditions can be detected using a fuzzy reference model. A fuzzy rule base is used to analyse outputs of the reference model and the load estimator and produce a diagnosis of the pump condition. Experimental results show that the motor current estimation had a root mean squared error of 0.11 A (∼5 per cent). Two fault symptoms, a 29 per cent obstruction of the exhaust silencer and an 8 per cent increase in current with respect to gas load, were simulated on the pump test bed and successfully diagnosed.