Burn-in with Mixed Populations
Abstract
Any electronic component may be either perfect or defective. The lifetime distributions of both types are assumed known. We assume that perfect items never fail. Before being used, each production lot is tested to eliminate some of its defectives, i.e., the lot is subjected to burn-in. Here, the purpose of burn-in is to ensure with a given confidence level that an item chosen randomly from the test survivors has a given probability of operating properly for a given time period. Three procedures are considered. Small sample theory is investigated for various assumptions about the information available concerning the number of defectives by using both analytic techniques and simulation. Large sample theory is studied, as well. This study shows that the first two procedures are sensitive to the number of defective items assumed and the performance of the third procedure is not. Keywords: Burn-in; Sequential stopping; Reliability; Mixed populations.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1987
- Accession Number
- ADA194651
Entities
People
- Saul Blumenthal
- Un-quei W. Pan
Organizations
- Ohio State University