Machine performance monitoring and fault classification using an exponentially weighted moving average scheme

Performance monitoring is crucial in maintaining normal machine operating conditions for the continued production of high quality parts. The objective of the following research is to develop an effective performance monitoring technique to assess the condition of rotating machinery through vibration signature analysis. An autoregressive model is utilized to characterize normal vibration signals, the modified covariance method calculates the deviation of the current condition from a normal condition, and an exponentially weighted moving average statistic measures the current machine condition by signalling either a normal or an out-of-control condition. The following paper discusses the statistical techniques employed in performance monitoring. Preliminary studies show that these techniques provide accurate performance monitoring of the operating condition of rotating machinery using vibration signals.