Nonparametric accelerated life testing based on proportional odds model

Accelerated life testing (ALT) is used to obtain failure time data in short duration under high stress levels in order to predict product life and performance under design conditions. The proportional hazards (PH) model, a widely used reliability prediction model, assumes constant ratio between the failure rate at high stress levels and the failure rate at the normal operating conditions. However, this assumption might be violated under some conditions and the prediction of the failure rate at normal conditions becomes inaccurate. We investigate the proportional odds (PO) model, which assumes that the odds ratio under different stress levels is constant, for accelerating life testing. In this research, we propose a nonparametric ALT approach based on the proportional odds model to predict reliability at normal operating conditions. We estimate the parameters of the proposed ALT model using the maximum likelihood estimation method. To verify the new approach, we fit the PO model with simulated failure time datasets and experimental failure data and compare its performance with the PH model. The results show that the new approach based on the PO model is a viable complement to the PH model in estimating reliability of products possessing property of converging hazard rate functions.