Reliability Evaluation and Prediction Method with Small Samples

Abstract

How to accurately evaluate and predict the degradation state of the components with small samples is a critical and practical problem. To address the problems of unknown degradation state of components, difficulty in obtaining relevant environmental data and small sample size in the field of reliability prediction, a reliability evaluation and prediction method based on Cox model and 1D CNN-BiLSTM model is proposed in this paper. Taking the historical fault data of six components of a typical load-haul-dump (LHD) machine as an example, a reliability evaluation method based on Cox model with small sample size is applied by comparing the reliability evaluation models such as logistic regression (LR) model, support vector machine (SVM) model and back propagation neural network (BPNN) model in a comprehensive manner. On this basis, a reliability prediction method based on one-dimensional convolutional neural network-bi-directional long and short-term memory network (1D CNN-BiLSTM) is applied with the objective of minimizing the prediction error. The applicability as well as the effectiveness of the proposed model is verified by comparing typical time series prediction models such as the autoregressive integrated moving average (ARIMA) model and multiple linear regression (MLR). The experimental results show that the proposed model is valuable for the development of reliability plans and for the implementation of reliability maintenance activities.