Quasi-best linear unbiased estimate for non-parallel constant stress accelerated storage life test data

A quasi-best linear unbiased estimation method for non-parallel accelerated storage life test data is proposed in this paper. As the stress imposed on the products is at a low level in storage state, the physical process of life depletion (corrosion, aging, stress relaxation, etc.) is extremely slow. So it is difficult to obtain sufficient information about the storage failure time of the products in a reasonably short period of time. Therefore, accelerated life test is widely used to estimate the storage lifetime of products. Accelerated storage life test is a method to estimate product lifetime in which products are placed under the test stress level significantly higher than storage environment stress level. Through accelerating the physical process of storage life depletion which the products need to experience, all or part of the products failure information is obtained, and then the service life under normal storage stress level can be derived by acceleration equations. However, the traditional estimation methods which are applied to life-determination test data of new products generally fail to analysis accelerated storage test data from non-parallel storage products which are stored for different periods of time before test. This problem is widespread in life-extension test. If the different storage time is ignored, estimation results will deviate from the real values. However, when maximum likelihood estimate method is used to analysis non-parallel accelerated storage life test data, there exist relatively large errors in estimation results. For this reason, a new method - quasi-best linear unbiased estimation method for non-parallel accelerated storage life test data is established. Firstly, the initial values of the parameters to be estimated are selected to calculate the acceleration coefficient. And then through acceleration coefficient, the storage time is converted to the equivalent time under test stress. Next the equivalent time is combined with the failure time observed in non-parallel storage test. Finally, the point estimates of the parameters and the confidence interval estimate of the reliability index can be obtained through iterating best linear unbiased estimation method. At the end of the paper, the rationality and feasibility of the proposed method in this paper are illustrated through simulation.