Failure Prediction Through the Theory of Stochastic Excursions of Extreme Vibration Amplitudes

Abstract

This paper attempts to propose a prediction on the failure of machinery resulting from high vibration amplitudes. Since the monitored record is generally in the form of a randomly fluctuating signal, the principles of stochastic excursions are employed in the investigation. Considering the total record to be made up of piecewise stationary signals with normal distributions, an initial estimation on the state of the machine is obtained by determining the first passage probability of the random excursion about a specified safe level. Further information on the incipient failure of the machine is obtained by evaluating the probability of durations of the excursion above the critical level. A possible maintenance program for vibrating machinery is proposed on the basis of two probabilities, and discussed using a particular industrial application.