Degradation tests are powerful and useful tools for lifetime assessment of highly reliable
products. In some applications, the degradation measurement process would destroy the physical
characteristic of units when tested at higher than usual stress levels of an accelerating variable such as
temperature, so that only one measurement can be made on each tested unit during the degradation
testing. An accelerated degradation test giving rise to such a degradation data is called an accelerated
destructive degradation test (ADDT). The specification of the size of the total sample, the frequency of
destructive measurements, the number of measurements at each stress level, and other decision variables
are very important to plan and conduct an ADDT efficiently. A wrong choice of these decision variables
may not only result in increasing the experimental cost, but may also yield an imprecise estimate of the
reliability of the product at the use condition. Motivated by a polymer data, this article deals with the
problem of designing an ADDT with a nonlinear model. Under the constraint that the total experimental
cost does not exceed a pre-fixed budget, the optimal test plan is obtained by minimizing the asymptotic
variance of the estimated 100 p th percentile of the product’s lifetime distribution at the use condition. A
sensitivity analysis is also carried out to examine the effects of changes in the decision variables on the
precision of the estimator of the 100 p th percentile. A simulation study further shows that the simulated
values are quite close to the asymptotic values when the sample sizes are large enough.