Signal Processing Methods for Gearbox Fault Detection

摘要

Methods of accounting for load variation in vibration signals from helicoptertransmission systems are presented. These methods are based on autoregressive moving-average (ARMA) models, and several ARMA parameter estimation schemes are presented. Simulations of load variation are carried out, and a prediction error filter, based on the ARMA models, is used to generate a residual signal. Fault indices extracted from the residential signal are used to indicate the presence or absence of a fault. The results of the simulations suggest that this method of fault detection is able to detect both general and local fault conditions.