Failure Rates of Non-Homogeneous Parts Populations

摘要

Groups of similar manufactured devices tend to be non-homogeneous due to process variability. Quality control methodology is used to improve the quality of groups of manufactured devices, particularly with respect to reliability by rejecting anomalous devices. In this study, a mathematical model was developed in which the average failure rate observed in a manufactured lot of devices following screening is a function of the failure mechanisms remaining in the non-homogeneous population. Parametric studies of both operating and dormant part failure rates were performed using subgroups having different failure rate characteristics represented by Arrhenius, Eyring, and non-Arrhenius physical failure models. The effect of dormancy on failure rate was explored, and the results were consistent with a previously stated theory which states that the failure rates of high reliability parts remain almost as high when dormant as when operating. (Author)