Reliability growth testing, what is the real final result? Accelerated test methodology for reliability growth

Abstract

This paper discusses the error in the understanding and representation of the traditional reliability growth model where the failure rates of different failure modes are treated in the same manner. The total initial failure rate of an item, consisting of the constant and changing components, is modeled mathematically as changing, but during the test, the random failures (a part of the overall failure rate) are disregarded. This introduces inevitable overestimation of the reliability growth and the reliability improvement. Only a smaller portion of the total tested item (design errors' where the failure rate is only a fraction of the total item failure rate) can be improved Improving a much smaller rate to a desired reliability goal requires considerably and often cost prohibitive test time To perform correct calculations for realistic test results and with reasonable resources, accelerated testing based on the Physics of Failure is recommended to use. This method is capable of providing a reasonable test duration which would be technically correct and economically feasible. This tutorial presents the methodology for the design of the accelerated reliability growth tests and their data analysis.