Reliability estimate using accelerated degradation data

Reliability estimation models that utilize degradation data can be classified into two groups: physics-based or experimentally-based models and statistics-based models. There are three necessary requirements to estimate reliability using a physics-based model: a subject matter model, which simulates the underlying degradation phenomenon, the distribution of the parameters, and the threshold level at which failure occurs. The application of a developed physics-based model is limited to units that exhibit the same underlying degradation phenomenon. The available statistics-based models are more general than the physics-based models. However, they assume linear degradation paths and/or constant standard deviation for the degradation data. Furthermore, they are developed for a given stress level and reliability can only be estimated at the stress conditions at which the degradation data are collected. This dissertation deals with developing statistics-based models to estimate reliability based on accelerated degradation data in which aforementioned assumptions are relaxed.; We investigate the application of time-dependent parameter distributions in modeling degradation measure, i.e., a random variable that represents ultimate effect of the degradation phenomenon on the device performance. It is shown that reliability functions can be derived using time-dependent parameter distributions that model the degradation measure.; This dissertation introduces degradation hazard function in order to include stress covariates in developing a statistic-based model to estimate reliability based on accelerated degradation data. The model is referred to as Proportional Degradation Hazards Model. It considers nonlinear degradation paths and time-dependent standard deviation for the degradation data. Moreover, unlike the available statistics-based models, it utilizes accelerated degradation data to estimate reliability at normal operating conditions. The only assumption is that the logarithm of the degradation hazard function is a linear function of the stress covariates.; We compare the reliability estimates of an Integrated Logic Family ( ILF) obtained from the proposed model with the reliability estimates obtained from a physics-based model developed for the IFL device. Furthermore, we conduct an accelerated degradation test in the laboratory by subjecting components to multiple stresses and compare the reliability estimates obtained from the proposed model with those obtained from the experimental results.