Modeling and planning accelerated life testing based on mean residual life

Accelerated life testing (ALT) is a widely used approach for reliability demonstration and prediction of components or systems at normal operating conditions using data obtained at accelerated condition. Extensive research in the development of ALT models has been focused on Accelerated Failure Time (AFT) models, Proportional Hazards (PH) models, and some extensions of these two types of models. Mean residual life function provides a more descriptive measure of the aging process than these two types of models; however, it has not been used in modeling and planning ALT.;In this dissertation, first we introduce the proportional mean residual life (PMRL) distribution family, and then study the different reliability aging properties under PMRL model, including failure rate, mean residual life, new better (worse) than used, new better (worse) than used in expectation, new better (worse) than used of aged, harmonic new better (worse) than used in expectation, and increasing (decreasing) uncertainty of residual life. The second problem incorporates the PMRL concept into the ALT field. We extend the PMRL to a more general framework with a covariate vector Z, and develop a new general nonparametric model for ALT.;The third problem deals with competing risk in ALT. We demonstrate the use of the proposed model in modeling the failure times obtained from accelerated life testing with multiple failure modes, the exponential and linear functions are used to present the baseline mean residual life function. The fourth problem deals with optimum ALT plans based on PMRL model. We design the first analytically based accelerated life testing plans that utilize the proposed PMRL model for reliability prediction. We also adopt the most widely used cumulative exposure model to derive the cumulative density function of the failure time for a test unit experiencing step-stress loading, and present the optimum simple step-stress test. These ALT plans need not assume the time to failure distribution and life stress relationship. It can also be applied in situations with more than more than a single accelerating stress without any difficulties, which addresses cases where there is a high degree of uncertainty about the model and the stresses.;Furthermore, in order to validate the proposed ALT model and optimum ALT plans, we conduct an accelerated life testing on the Light Emitting Diodes (LEDs) in the laboratory. The test results indicate that the proposed model is valid and accurate. It is shown that it is a excellent alternative to the accelerated failure time (AFT) models and the proportional hazards (PH) models.