電子產品最佳預燒條件、預燒時間與成本之決定―以交換式整流器為例

Abstract

預燒(Burn-in)是一種用來篩選早衰期零件的有效方法，過去都是以失效分配曲線模式、成本模式或是環境應力模式(Environmental stress model)進行預燒研究，對不同特性產品而言，這些研究模式皆無法有效提供預燒條件、預燒時間與預燒成本之決定。本研究利用零件應力分析法(Part stress analysis)與穩健式設計方法(Robust design)獲得產品之最佳預燒條件。由於在實務製造上，一新產品的產生往往是由舊有產品延伸而來，此舊產品可稱為基產品(Base product)。利用新產品與舊產品之相關性，本研究提出以類神經網路方法決定最佳預燒時間與預燒成本。此外，由於靜態預燒作業往往需較長之作業時間，而過去研究中指出應力加速方法(Accelerated stress)可產生與靜態預燒方法同等值之效能，因此，可利用應力加速方法將用來產生動態的預燒時間與成本，此可提供管理者最好之預燒決策。最後，本研究利用所提出之方法在交換式整流器進行實證研究。

Burn-in is an effective means of screening out the infant mortality components of a system. Most related studies investigated burn-in by curve fitting for a failure model, cost for an optimal burn-in time model and environmental stress model. However, those investigations did not provide an effective method to determine the optimal burn-in condition, burn-in time and burn-in cost for a practical operation. As a result, this study presents an effective procedure that uses part stress analysis and robust design techniques to determine the optimal burn-in operational environment. Meanwhile, in the actual manufacturing process, a new electronic product is always extended from an old product, called the base product. By adopting the relationship between new product and base product, this study also presents a neural network-based approach to determine the optimal burn-in time and cost without any assumptions. Moreover, burn-in consumes much time and energy during production. To solve this problem, stress burn-in can be applied to eliminate failures before appearing in the field. A case study of the production of a switch mode rectifier will be performed and compared with the traditional approaches, to examine the proposed procedures’ effectiveness.