Fault prognosis based on fault reconstruction: Application to a mechatronic system

Abstract

The fault prognosis method developed in this work has a horizontal structure, and aims the estimate the RUL by the reconstruction of the fault trend after detecting the degradation beginning. The diagnosis part is realized using a Principal Component Analysis (PCA), the fault reconstruction is done using the fault direction matrix, and the RUL is estimated using an Auto-Regressive Recurrent Radial Based Function (ARRRBF) neural network. The developed method is implemented on a mechatronic system dedicated to the prognosis, which offers the possibility of introducing gradual and controlled degradations.