Reliability estimation from two types of accelerated testing data based on an artificial neural network supported Wiener process

Abstract

Considering the powerful ability of artificial neural network on data fitting, we have combined it with Wiener process in reliability estimation, while the previously published artificial neural network supported Wiener process based reliability estimation method only can utilize degradation testing data under normal use stress. However, the accelerated degradation tests and accelerated life tests are always performed to shorten the testing time. Hence, based on a general log-linear form acceleration model, we develop the artificial neural network supported Wiener process model into an accelerated testing model to utilize the accelerated testing data and improve the reliability estimating accuracy. Furthermore, the corresponding model training and parameter inferencing approaches are also constructed to apply the proposed model in reliability estimation. A simulation experiment and a case study on stress relaxation dataset are performed to verify the effectiveness of the proposed method. It is concluded that, the proposed method shows superiorities on population evaluation and individual predictions comparing the other methods, considering the accuracies of degradation modeling, life and reliability estimations. As a result, it is more suitable and practical than the previously published approaches.