| Summary: | 碩士 === 國立清華大學 === 工業工程與工程管理學系 === 90 === Accelerated Degradation Test (ADT) is widely used to assess lifetimes of highly reliable products. How to design an ADT under the restrictions of the finite experiment cost and to make the test have the better performance is the question for the experimenters. The problem is to find the optimal test plans for the sample size, inspection frequency, and the termination time such that the variance of the estimated 100p-th percentile is minimized subject to the constraint that the total experimental cost doesn't exceed a pre-determined budget.
In this talk, the (general) random degradation model is used
to describe the accelerated degradation path that the error term follows a zero-drift Wiener process. In the case of the LED's accelerated degradation model, the estimated lifetime (100p-th percentile) in normal use condition and its variance are figured out. They are used in designing an ADT giving the accelerated stress level is known, the ratio of the sample size in each stress level is fixed and the observation interval and observation times are the same in every stress level. Finally, Compare the performances between the optimal design of ADT and that of Degradation Test. ADT is benefit on curtail the experiment time, but it needs more test sample items and has larger lifetime variance. In the other word, ADT can shorten experiment time but spends more. This is helpful to inference the lifetime of the high-reliable products.
|
|---|
