ABSTRACT
To collect the information about the lifetime distribution of a product, a standard life testing method at normal working conditions is impractical when the product has a substantially long lifespan. Accelerated life testing solves this problem by subjecting the test units at higher stress levels for quicker and more failure data. Due to constrained resources in practice, several decision variables such as the allocation proportions and stress durations must be determined carefully at the design stage in order to run an accelerated life test efficiently. These decision variables directly affect the experimental cost as well as the estimate precision of the parameters of interest. This article investigates these optimal decision variables based on several well-known optimality criteria under the constraint that the total experimental cost does not exceed a pre-specified budget. A general scale family of distributions is considered for the underlying lifetimes to accommodate different lifetime models at different stress levels for flexible modeling. The constant-stress and step-stress accelerated life tests are then studied in detail with linearly decreasing stress durations as the stress level progresses. Under the identical budget constraint, the efficiencies of these two stress loading schemes are compared using two case studies.
Additional information
Notes on contributors
David Han
David Han is an Associate Professor at the Department of Management Science & Statistics in the University of Texas at San Antonio, TX. He received two Honours B.Sc. degrees, one in Biochemistry and the other in Computer Science & Statistics, both from McMaster University in Canada. Continuing his studies at McMaster, he received the M.Sc. and Ph.D. in Statistics. His research interests include the operations research and statistical inference for accelerated life testing in reliability and survival analysis, analyses of censored data, optimal censoring plans, competing risks analyses, and statistical quality control.