Optimized Reliability Test Design to Reduce Uncertainties in Reliability Assessment

Abstract

Manufacturers are always asked to deliver products faster while satisfying customer expectations including reliability. Customers cannot wait until they receive their products to assess if the products are reliable. Capability of accurate reliability assessment should yield important information about the likelihood that the manufacturer will provide a reliable product, and therefore add valuable competitive advantage to manufacturer. In the electronics industry, accelerated life testing such as temperature cycling is often used to assess the reliability of products. Careful planning of test conditions such as test duration and sample size is essential, as they directly influence the reliability assessment. Although some industry standards provide guidelines in terms of reliability goals to achieve, uncertainties in reliability assessment associated with test duration and sample size is not clearly defined for most cases. This paper presents sample size and test duration recommendation for knowledge based certification to reduce uncertainties in reliability assessment regardless of stress conditions. Numerical simulations have been conducted in order to determine the appropriate sample size and test duration by generating random time-to-failures from a Weibull distribution. The probability of obtaining accurate reliability assessment is provided according to sample size. Censoring plan is also considered to determine the least test duration. Moreover, temperature cycling tests have been conducted with the minimum sample size identified from this study and various censoring points in order to compare the reliability characteristics. The experimental results were found to be in good agreement with the findings from numerical simulations.