用無母數回歸的方法推估加速衰變試驗資料的壽命分佈

Abstract

科技進步後，增加產品的使用時間，為了能在短時間內知道產品的壽命分佈，工程師常進行加速衰變試驗，以獲得有關產品壽命分佈的資訊。本文針對加速衰變試驗資料，提出加速衰變隨機模型 ，並假設應力的大小只會影響產品衰變的速度，並不影響產品衰變之路徑。藉由不同應力的加速倍數關係，我們可以得到產品正常使用情況下的壽命分佈情形。文中以發光二極體的亮度衰變資料為例，推估發光二極體在正常使用情況下的壽命分佈。最後並提出模擬分析驗證之。

We propose a nonparametric regression stochastic process model for accelerated degradation data, which consists of groups of degradation curve data. We assume that different stress levels only affect the degradation speed of the product characteristic, but not the degradation trend. A method is developed for estimating the lifetime distribution of the ADT (Accelerated Degradation Test) data. By investigating the relationship between the acceleration factors and the stress levels, we can obtain the lifetime distribution of the product under the normal use condition.We apply the method to data obtained from an accelerated degradation test for an LED (light emitting diode) product.The performance of the method is further checked by a simulated example.