Ways to improve the analysis of multi-level accelerated life testing

Abstract

Multi-level accelerated life tests (ALTs), MEOST (multiple extreme overstress testing) and traditional step-stress approaches (hereafter all called simply step-stress) have a number of worthwhile applications for the analysis of the projected reliability of components and systems. These test methods are considered for many common situations of accelerated life testing (ALT). Some of the main purposes of ALT are to measure projected system life and to identify product weaknesses in order to improve reliability. These methods may apply well when only a small number of systems are available, when extremely long, specialized test equipment is required, when limited environmental chamber capability and/or test fixtures are involved and, lastly, when very expensive support equipment represents a serious test limit. The accelerated testing methods mentioned have had some limited applications in the past. The limits have been sometimes due to poorly described degradation or accumulative fatigue and the subsequent difficulty with the analysis of the failure data themselves. A tight series of ground rules is proposed in this paper that may help solve, mitigate or prevent this possibility. This paper also presents methods to improve the analysis of these multi-level stress tests. While the methods shown are approximate, they easily lend themselves to analysis on a computer by standard software packages or through manual techniques. Additionally, the ground rules presented suggest that wider limits may often be taken for running this type of accelerated testing than has been suggested in the past. These ground rules aid the analysis by ensuring uniformity of results. As an aid to test set-up, neither the step intervals need be of the same length nor the stress steps uniform. A recent paper by van Dorp et al. (IEEE Trans. Reliab., REL-45, 491–497 (1996)) provides a theoretical basis for some of these suggestions. Another paper by Bhote (Trans. Am. Soc. Met., 237–243 (1985)) identifies many reasons to perform this type of reliability test and credits Dorian Shainin with the development of many of the early MEOST techniques. Clearly, step-stress approaches and applications have not been exhausted for measuring life and identifying design or manufacturing flaws. Copyright © 1998 John Wiley & Sons, Ltd.