A Method for Degradation Modeling and Prediction Based on Inverse Gaussian Process Supported by Artificial Neural Network

Abstract

This paper proposed a method based on inverse Gaussian process supported by the artificial neural network for degradation model and predict the lifetime. To overcome the uncertainly of the degradation path, we trained the artificial neural network to get the path of degradation. It is no longer necessary to assume the initial degradation when establishing the degradation model. The artificial neural network is trained by the run-to-failure degradation data. And the minus log-likelihood is used as the loss function. Considering the differences of individual, assumed that the parameters of IG process are obey Gamma distribution. The Gamma distribution parameters assessment by the method of moment estimation based on the degradation path trained by the artificial neural network. And predicted the lifetime by real-time degradation dataset. The method proposed is verified by the actual degradation dataset. The actual example results show that the degradation model based on inverse Gaussian process supported by the artificial neural network can represent the process of the degradation and predict the service life, though no prior knowledge about the degradation path.