Hybrid Method Based on the Inverse of the Belonging Joint Probability and the PSD for Fault Diagnosis

The hybrid method presented in this paper, is based on the inverse of the belonging joint probability (IBJP) and the power spectral density (PSD) to detect new anomalous in a random signal representing normal state behavior of a given machine. We first compute the power spectral density (PSD) of the normal state signal and then represent this PSD signal by an adequate Gaussian white noise (GWN) model. The new changes or anomalous in the machine behavior are, usually, represented by new picks in the PSD curve. Since the latter is usually random, these picks may be, however, smeared out and thus, they will not be well detected. To overcome this problem, we propose, in this work a technique based on the belonging joint probability inverse (IBJP) combined with the PSD to detect these faults. We have, furthermore, performed a simulation to clarify this hybrid technique efficiency. Key-Words: Faults diagnosis; inverse of the belonging probability; power spectral density; Gaussian stationary white noise model.