A reliability estimation method based on signal feature extraction and artificial neural network supported Wiener process with random effects

Abstract

Accurately predicting degradation and reliability is the key to effective condition-based machinery maintenance. This paper presents a reliability estimation method based on an artificial neural network (ANN)-supported Wiener process model with random effects. A number of time-domain features, frequency-domain features, and intrinsic energy features are considered to reflect the health condition of products. These features can be selected by the PROMETHEE II method considering multi-sample and multi-measure conditions. An ANN-supported health index function is defined to describe the relationship between the health condition and the selected signal features. Based on the health index function, an ANN-supported Wiener process is used to model machine degradation. The corresponding health index function training and process parameter inference approaches are presented. An actual testing dataset of a machine bearing is used to demonstrate the proposed method. The proposed method provides accurate service life and reliability predictions.