Benabdellah Yagoubi
Multi-fault detection, random process, inverse of the belonging individual probability, PSD, Gaussian white noise
This work addresses the problem of detecting multiple faults in a random process. This is a vibrant research domain in control that has been approached by many researchers using various methods such as stochastic techniques, neural network-based method and the power spectral density (PSD). The latter is frequently used particularly in electrical engineering to detect faults. However, as most of the PSD curves are random in practice, the peaks representing eventual faults may be hidden or smeared in the PSD process and hence their detection becomes a difficult task. To overcome this problem by making these peaks more visible in the PSD representation, we suggest in this paper a novel approach, based on the inverse of the belonging individual probability (IBIP) and the PSD hybrid method. We show that this IBIP–PSD hybrid technique is extremely effective in detecting multiple faults that may coexist in the system. The application of this technique to multi- faults detection is still in its infancy, but we believe that it has a great ability to enhance this area, thereby improving the reliability, safety and efficiency techniques of supervision and monitoring.
Important Links:
Go Back
