Blue Room - Session 14

3:30 to 4:30 p.m. Thursday April 23, 2015

Sample Size and Producer's Risk in Reliability Testing

The goal of this presentation is to demonstrate the available methods for sample size determination in reliability testing, appended with new insights on optimizing the probability of good products failing the test (producer's risk) and the probability of bad products passing the test.

Sample size can be affected by several different drivers, such as the purpose of a test, budget, available samples and accelerating failure behavior. In reliability testing, various statistical methods are available for determining the required sample size based on required reliability, allowed number of failures, available test length, failure mechanisms, etc. The actual goal of reliability tests, however, is to prove reliability targets while - at the same time - NOT allowing bad products to pass the reliability test NOR allowing good products to fail the reliability test. Testing more samples might result in a lower probability of good products failing the test (decrease producer's risk), but more testing might not be allowed. Other approaches can then be used to determine the proper sample size for optimal producer's risk. For accelerated testing combined with Weibull techniques, a practical application of these approaches and insights are shown in an actual case.

Key Words: Sample Size, Reliability Testing, Statistical Proof, Producer's Risk, Consumer's Risk, Probability of Failing Test, Weibull, Accelerated Testing, Case Based on Actual Data

Elly Funken

Holland Innovative BV

The Netherlands