光阻噴管配置問題之探討

Abstract

本研究探討光阻噴管在機台上的配置問題，在本季產品需求已知及上季噴管配置給訂的情境下，考慮機台產能限制與光阻相關性限制，以總成本最小化為目標，建構光阻噴管在機台上的配置整數規畫模式。總成本考慮光阻噴管的閒置成本、保養成本與沖洗成本三種，利用LINGO求解最佳各機台光阻噴管配置。由於在求解時間與機台數及光阻數成指數成長，故本研究另行提出一啟發式演算法，先構建一些法則，以縮小收尋範圍，縮短求解時間。隨之，設計575個案例驗證本論文的啟發式演算法之正確性，經求證後可知此啟發式演算法確實能有效的降低求解時間，且98.6%可求出最佳解，僅1.4%(8個案例)與最佳解有稍許誤差，其誤差皆小於4%。經由台灣某晶圓代工廠提供的一接近實例中驗證，本論文的模式可應用於業界。

This study addresses the photoresist nozzle allocation problem. Given the product demands of a planning period, and the photoresist nozzle allocation of the previous period, considered the machine related constraints and the relationship among the types of photo resist , we formulate an integer programming model for minimizing the total cost to determine the optimal photoresist nozzle allocation. The items in the total cost function are dummy cost of photoresist nozzle, OK82 cost, and purge cost. Then we use LINGO to solve the IP model. The computational timeis exponential growth with the numbers of photo resist and machine. .For reducing the computational time we propose a heuristic method with reducing the searching space to solve the IP model. We also verify the accuracy of the proposed heuristic method. Approximate 98.6% examples the optimal solution can be found. In 1.4% examples the deviations of total cost from those in the optimal solution are not over 4%. Finally a near real case example is given to illustrate the application of the IP model.