Lifetime prediction of accelerated degradation testing using support vector machine and proportional hazards-proportional odds

A data analysis method for Accelerated Degradation Testing (ADT) is presented in this paper. The degradation paths of each test unit are predicted using support vector machine for regression and the pseudo failure times are obtained as the time points which the predicted degradation paths reach the specified thresholds. Proportional hazards-proportional odds model is then utilized to analyze the pseudo failure times of the test units under each stress level and lifetime and reliability of the product under normal operation conditions is estimated. An accelerated degradation testing system is established and ADT for miniature bulbs under three different stress levels are operated. The method presented in this paper is applied to analyze the miniature bulbs ADT data. Lifetime and reliability of the miniature bulbs under normal operation conditions are estimated finally.