How Few Tests Can Demonstrate the Operational Reliability of Products

Highly reliable industrial products impose insisted demands on manufacturers to carry out, before starting mass production, reliability demonstration tests for their new products in order to verify how they meet the reliability targets. However, due to cost and time constraints, a small number of the new products is available for testing and the reliability demonstration is not that easy, as conventional statistics cannot be directly applied. In this paper, four probabilistic approaches are discussed and compared in case of reliability demonstration with few test data. These approaches are the confidence interval, the test of hypothesis, the Bayesian approach and the compound uncertainties. The latter is proposed in this work as a robust tool for demonstration tests with small sample size.