步進應力加速衰退試驗曁高可靠度產品壽命之硏究—以發光二極體爲例

摘要

隨著產品品質的不斷提升，消費者對各部份組成零件的要求亦隨之日趨嚴格。同時，新產品的開發由研究、測試至上市的時間也曰益縮短。其中有關步進應力加速衰退實驗終止時間的決定時機與方式，因涉及到實驗分析結果的準確性，與其有時間及成本上之考量，對實驗之進行者而言，一直是件很困難也急欲解決的事。針對此一問題，在本研究中嘗試以發光二極體(LEDs)爲例，首先應用統計的迴歸理論來判斷適當的應力調整時機與實驗終止時間點。再使用類神經網路(neural network)的技術，利用用其模擬人類大腦神經架構的原理，描述實驗應力與產品亮度間之複雜關係，並對產品壽命進行預測。根據分析結果顯示，本研究所提出之方法，可有效的判斷出適當的應力調整時間。實驗之時間成本可減少約25%（833個小時），但此時產品壽命之預測差異約爲2.5%。

Abstract

Rapidly changing technologies, more complicated products, and increasing global competitiveness necessitate that manufacturers develop more sophisticate and powerful testing of products. The step-stress accelerated degradation test is one of these effective experiments. In the test, the determination of the shifting time for controllable factor is always a critical problem for the experiment conductors because it will affect the analysis results. In this study, we developed a termination rule for step-stress experiment based on traditional regression theory. In addition, an artificial neural network (ANN) is utilized to predict the product's life. To demonstrate the efficiency of the proposed rule and the constructed model, product's life forecasting is performed on a real LEDs data. The results reveal that the proposed approach can successfully identify the termination time for the experiment and the product's life in the normal situation can be estimated precisely.

關鍵字:

• 衰減性資料
• 步進應力加速衰退實驗
• 實驗終止規則
• 迴歸理論
• 類神經網路

Keywords:

• degradation data
• step-stress accelerated degradation experiment
• termination rule
• regression
• artificial neural network