高可靠度產品壽命預估-以發光二極體為例

Abstract

隨著科技的進步, 各類產品的功能愈來愈強, 構造也日趨複雜, 對於 產品的可靠度必須要有更先進的方法來分析測試.當產品的某品質特性其 衰退的形式與時間有關連性, 則可透過加速衰退試驗進行產品壽命可靠度 的預估. 本研究對高可靠度產品構建一加速衰退試驗程序, 作為實務實驗 者之參考. 此程序具有系統化 流程化與可行性高的特色, 依循著此程 序, 可使實驗的進行更有效率, 並有助於提昇壽命預估的精確度. 另外, 本研究發展以類神經網路以及統計迴歸分析為基礎的方法, 以為壽命預估 的工具. 最後, 以發光二極體為例, 分別利用所發展的方法加以分析. 結 果顯示此二方法在試驗早期就能相當準確地預估產品的壽命, 而類神經網 路方法的預估能力則略優於統計迴歸分析. Rapidly changing technologies, more complicated products, and increasing global competiveness necessiate that manufacturers develop more up-front testing of products. The availability of a quality characteristic for a product whose degradation over time can be relative to reliability would allow us to access the product's life through the accelerated degradation test(ADT). In this study, we develop a procedure for ADTs. The proposed procedure can enhance the efficiency of ADT owing to its systematic and stepwise nature. In addition, artificial neural network-based(ANN-based) and regression-based approachesare also proposed to predict the product's life. A simulated LEDs example demonstrates that the proposed approaches can precisely estimate the producy's life in earlystage of testing.Moreover, the ANN- based approach yields a better performancein life forecasting than the regression-based one.