A demonstration of perceived practical benefits of assessing product reliability from degradation data

Thetraditionalfailuretimeanalysismethodhasbeenproventobelesseffectiveinassessingreliability of highly reliable products. It is a prerequisite that failures are recorded or observed when one considersthetraditionalcensoredfailuretimeanalysisapproach. Asfordegradationdata,withfeworno failures, it is possible to obtain useful and significant reliability information, on condition that there is a strong correlation between the underlying degradation process and the product’s failure. This research study seeks to demonstrate the advantages of using degradation data via the data on GaAs (Gallium Arsenide) lasers adopted from Meeker and Escobar (1998). In particular, to illustrate that with degradation data, meaningful conclusions can be reached much earlier. It does this by proposing the general degradation path model and infers the time to failure distribution from the registered degradationdata. Thelognormaldistributionisdeemedtobetheappropriatefailuretimemodelfrom experience with the lasers and this information is taken into account in a simulation procedure. The laser is regarded to have failed the first time a 10% increase in current is required to maintain the constant light output. Lasers whose degradation paths have not reached the failure threshold are extrapolatedtofailure(generatingpseudo-failuretimes)orto5000hours,thedesiredyearsofoperation forthelaser. Parameterestimationforthebestrankedlifetimedistributionisdoneviathemedianrank regression method. Important pieces of reliability information are derived from the inferred failure time distribution. Shorter and longer simulated test data were found to be comparable with respect to most of the reliability metrics and their confidence bounds. This demonstrates the advantages of using degradation data when assessing product reliability
A Masters Research Report submitted to the Faculty of Science, University of the Witwatersrand, in partial fulfillment of the requirements for the degree of Master of Science May, 2019