Optimal Burn-in for Percentile Residual Life
- 2001-04-19 (Thu.), 14:30 PM
- Recreation Hall, 2F, Institute of Statistical Science
- Prof. Way Kuo
- Texas A&M University U.S.A.
Abstract
Burn-in is an engineering method for testing a population of products under electrical or thermal conditions. The objective of burn-in is to detect and to remove weak products from their population, and thus burn-in becomes meaningful when a population is not homogeneous. One problem with burn-in is determining the best time to stop testing, called the optimal burn-in period. Strictly speaking, burn-in should be continued until most of the weak products have been detected, leaving only strong products for field delivery. Traditionally, the optimal burn-in period is constructed by maximizing conditional reliability subject to the cost functions. However, many products are deployed in the field for as long as they function, without a specified mission time. In such cases, the percentile residual life can be considered. But, the main difficulty with percentile residual life as a performance measure is the infeasibility of finding its close form in terms of the burn-in period. In this talk, we will present an overview on burn-in study and problems that are being examined in practice. We will also show you the relation between maximizing reliability and maximizing the percentile residual life.