Providing software solutions since 1976
Quality Improvement Papers
New Methods for Modeling Reliability Using Degradation Data
José G. Ramírez and Gordon Johnston, SUGI Proceedings, 2001
Abstract
Engineers and researchers use life tests in a wide variety of industrial applications to determine how well critical components and materials will perform under different operating conditions. By estimating the failure time distribution, decisions can be made that influence the design, improvement, and storage of products so they meet customer requirements. Manufacturer warranty decisions may also be influenced by knowledge of product reliability performance.
Traditional life tests are often not the most efficient way to obtain reliability information because few, if any, actual failures are observed. On the other hand, it is often possible to obtain pseudofailure data using degradation measurements. Consequently, there is growing interest in studying the degradation of product performance or material properties over time.
In this paper we describe two methods for the analysis of such data. The first uses the SAS/QC® RELIABILITY procedure with pseudo-failure times to predict a given percentile of the failure time distribution, and the second uses the SAS/STAT® NLMIXED procedure to fit concave degradation models, and some SAS® macros to estimate the failure distribution. These methods are illustrated with a degradation study on the strength of an electronic material, which was subjected to four different temperatures and measured over time. We compare the two approaches, and show the advantages of using concave degradation models in this setting.
