A new processing method for accelerated degradation data based on quantile regression and pseudo-failure lifetime

The key to accelerated degradation testing (ADT) data processing method based on the pseudo-failure lifetime lies in the modelling of the degradation paths. However, the traditional regression method can only illustrate the average changing trend of the degradation, which may lose much information of the actual degradation process. Therefore, for a given accelerated model from engineering practice, we propose a new modelling method for ADT data based on quantile regression (QR). The degradation paths of different quantiles under different stresses can be obtained by QR, which offer a more comprehensive description for ADT data and allow us to use the specific quantile of the degradation to represent the system state. In addition, the proposed method can overcome the problems of the heteroscedasticity and outliers of ADT data. Finally, a double-stresses ADT example of rubber O-rings is given to illustrate the implementation of the proposed method.