The Residual Life Prediction Based On Degradation Stochastic Process

The performance degradation data of product often contain abundant reliability information about the failure mechanism of the products,which leads to a link between the degradation and failure of product.Therefore,it provides a new path for product reliability assessment.The prediction of residual life based on performance degradation is a hot research area in reliability research.The degradation model are proposed for the degradation situation of different product based on the Wiener process in this paper.The corresponding remaining life distribution are concluded for the product.The specific contents are as follows:A new random-effect Wiener degradation model is proposed for the product with high degradation rate,in which both the drift parameter and the diffusion parameter are random variables and they have linear relationship.This model can express the unstable degradation path of the product and take into account the differences between individuals.The residual life distribution of the product and the analytical expression of the reliability function are obtained based on this model.Then the estimation of the model parameters are obtained using the EM algorithm based on the common degenerate data.Finally,an illustrative example of titanium alloy fatigue crack data is conducted and compared with the existing model.The result shows the validity and accuracy of the model.Aim at the product degradation data containing stress information,an Wiener accelerated degradation model based on the random-effect is proposed,in which the drift parameter is a random variable and follows the normal distribution.The two-step maximum likelihood estimation methods are used to obtain the estimation of the unknown parameters and of the model based on the performance degradation data under stress state.Finally,the proposed model is compared with the existing model through the performance degradation data of the olefin insulation cables.The results showed that the proposed model had a higher degree of fit and the estimation results are more accurate.