A New Lifetime Estimation Method Using the Combined Data of the Accelerated Test Data and the Inspection Data.

使用全新的电气绝缘材料的加速寿命试验长期以来一直用于估计绝缘材料的寿命。虽然在加速寿命试验中对绝缘材料施加了各种类型的应力，但由于意外的劣化因素，绝缘材料的破裂往往比预计的寿命更快。然后使用工作绝缘进行升压击穿电压试验，以评估绝缘的实际剩余寿命。然而，这两种测试方法之间的关系还没有得到严格的理论解释。将两者结合起来的一种方法可能是Nelson提出的累积暴露模型，或者通常用于短时间加速寿命试验的Sedyakin模型。在研究累积暴露模型时，我们发现了模型参数估计的各种困难；因此，这一事实促使我们首先研究用升压法估计击穿电压。当绝缘被击穿后不能使用时，采用升压法来估计冲击击穿电压。我们分析了升压法中潜在击穿概率分布估计的可靠性，当(1)观察到的击穿电压本身是可用的(2)它是不可用的。与后一种情况相比，前一种情况有许多优点，如：(1)估计的置信区间变小；(2)可以稳定地获得估计。我们总结了假定击穿电压的潜在概率分布为正态、威布尔和甘贝尔型的三种情况的结果。简单明了地描述了各种分布模型的最优测试方法。在假定基础概率分布为某些可靠性分布模型的情况下，推荐了新的升压检验方法。我们首先建议使用基本概率分布的参数。其次，如果观测到的击穿电压本身比击穿和非击穿的二值信息更有利于采用新的升压方法。与传统的升压方法相比，采用新的升压方法可以大大减少试样的数量，以获得相同的估计信度。升压试验获得50%击穿电压的最优策略是将升压距离设置为大于底层分布标准差的一半。少

Accelerated life tests using the brand-new electrical insulation have long been used to estimate the lifetime of the insulation. Although a variety of types of stresses are imposed on the insulation in the accelerated life tests, the insulation is often broken more quickly than the estimated lifetime due to unexpected deterioration factors. The step-up breakdown voltage tests using the working insulation are then performed to assess the actual remaining lifetime of the insulation. However, theoretical explanations of the relation between these two test methods have not been made rigorously. One way to combine these two may be the cumulative exposure model proposed by Nelson, or Sedyakin model which is usually applied to short time accelerated life testing.In studying the cumulative exposure model, we have found a variety of difficulties in parameter estimation for the model; thereby this fact leads us to study first the breakdown voltage estimation using the step-up method. The step-up … More method is used to estimate the impulse breakdown voltages when the electrical insulation is not usable after it is broken. We analyses the reliability of the estimates of the underlying breakdown probability distribution in the step-up method, when (1)the observed breakdown voltage itself is available and (2)it is not available. The former case has many advantages compared to the latter case such that (1)the confidence intervals of the estimates become smaller and (2)the estimates can be obtained in stable. We summarize the results of the three cases that the underlying probability distribution for the breakdown voltage is assumed to be normal, Weibull, and gumbel types. The optimal test method is simply and clearly described for various distribution models. The new step-up test method is recommended when the underlying probability distribution is assumed to be some reliability distribution models. We first recommend the use of the parameters of the underlying probability distribution. Second, it is advantageous to use the new step-up method if the observed breakdown voltage itself rather than the two-valued information of breakdown and non-breakdown is available. Using the new step-up method, the number of test specimens can be substantially reduced comparing to that in the conventional step-up method to obtain the same estimation reliability. The optimal strategy of the step-up test to obtain the 50% breakdown voltage is to set the step-up distance to larger than the half of the standard deviation of the underlying distribution. Less