Sequential Bayesian assurance tests for degradation data

Measuring degradation of items in a reliability context may be all that is feasible. There may simply be neither enough time nor resources to produce a sufficient number of item failures to characterize the underlying time-until-failure distribution. In such contexts, degradation data-based assurance testing can be tuned to strike a compromise between consumer and producer risk when deciding whether to accept or reject a product. A one-stage assurance test counts the number of items in a sample exceeding a fixed degradation threshold at a fixed time and uses this count to make the decision: accept or reject. A general Bayesian framework for extending assurance testing from one-stage to a multi-stage or sequential setting is presented. Our multi-stage assurance tests are shown to compare favorably to their one-stage counterparts by possessing a lower expected time requirement at given sample size and risk constraints. Examples of the methods based on a printhead application are provided and are reproducible with the supplemental R code.