利用樹脂黏度模型進行LED封裝製程參數設計

LED主要利用環氧樹脂進行封裝，因環氧樹脂具有低成本、成型流程簡單等優點，各家廠商皆致力於提升封裝良率。其中以環氧樹脂交聯現象(凝膠點)容易受到影響，進而影響膠體流動性與產品加工。根據相關文獻討論，產品固化後容易發生氣泡不良現象。有鑒於目前業界對於黏稠度無一函數可循，於新參數設定時往往得經由試誤法尋找最適之製程組合。本研究針對特定環氧樹脂隨溫度與膠配比變化實驗資料建立該黏稠度函數，經驗證後該黏稠度函數可準確地進行預測。於現實生產環境中，材料與製程參數往往存在某種程度之誤差，導致製程品質的不確定性，本研究運用參數設計建議最佳之製程參數組合，以提升黏稠度之穩定性。此方法先運用電腦輔助參數設計方法規劃黏稠度函數實驗，再結合類神經網路與最佳搜尋策略，並以CAFE軟體作為輔助，求取最佳封裝材料製程參數組合。最後驗證智慧型參數設計法與田口方法之間求得最佳參數組合的能力。實驗結果發現，此兩組建議結果較現場實際製程參數穩定，所改善SN比分別達到43.7%與43.8%。本研究最後將參數設計求得之最佳參數組合作為建議廠商生產之參數組合。

關鍵字

樹脂黏度模型；田口方法；電腦輔助參數設計；智慧型參數設計

並列摘要

Epoxy resin is current material for LED packaging process. The curing link (gel time) for epoxy resin is a crucial phenomenon, impact the fluidity and yield of modeling process. A survey of literature reveals that formed voids after the Molding Compound curing. In view of this, electronics manufacturers were not haven viscosity function to determine the process parameters of now product. The viscosity function is recommend in this research could provide prediction accurately. In practice, the phenomenon of process and material variation cause uncertainty of quality. The implement be used in this work include that computer-aided parametric design for experiment of viscosity function, artificial neural networks and location optimizing algorithms (CAFE) for gain the optimum process parametric. At the last, the confirmation experiment is used to proof that parametric recommendatory SN ration increased to 43.7% and 43.8%.