The Search of Optimal Operation Parameters of a Belt Furnace for Microelectronic Packaging Process Using Artificial Neural Networks

以類神經網路法尋找半導體陶瓷構裝基板在覆帶燒爐的製作過程中之最佳操作參數

廖朝光(Leo C.-K. Liau) ； 楊重光(Thomas C.-K. Yang) ； 黃獻平(Shien-Ping Huang) ； 張志成(Chih-Chen Chang)

《Journal of the Chinese Institute of Chemical Engineers》 34卷2期 (2003/03) Pp. 219-227

https://doi.org/10.6967/JCICE.200303.0219

摘要

本文以計算流體力學（CFD）技術，研究多層陶瓷模組（MCM）元件在覆帶式燒爐之製作過程中，能量耗損與陶瓷體的輸送關係。透過覆帶系統內每一格點所計算出的溫度分佈、佐以材料熱膨脹係數與其它熱物理性質，如此半導體陶瓷構裝基板之暫態熱張力便可求得。在CFD的計算模擬中，將同時考慮不同覆帶速度與通氣速度對成形之陶瓷體所引發的熱張力。結果顯示MCM元件所引發的熱張力相當易受熱爐內加熱歷史、流場與熱場的影響。接著，利用統計實驗中反應曲面法與類神經網路法，以陶瓷體內產生最小張力為目標，進行這個加熱程序的最佳操作條件的尋求。這些成效證實，多層陶瓷元件的製作可達到高程序良率與優良的產品可信度之地步。

並列摘要

Energy dissipation and the transport of ceramic packaging substrates in a belt furnace for the manufacture of multi chip module (MCM) devices were investigated based on computational fluid dynamics (CFD). Using the results of temperature profiles and knowledge of thermal expansion coefficients and thermophysical properties, the transient thermal stresses of the substrates were calculated through structure analysis at each grid point along the belt. Various levels of the belt speed and purging-gas flow rate were considered in the CFD simulation to evaluate their influence on stress formation inside the solid-state substrates. Results show that the induced thermal stress in MCM substrates is highly sensitive to their heating histories as well as the flow fields that develop inside the furnace. Subsequently, a search for optimal operating conditions for the firing process was conducted, with the goal of minimizing stress generation inside the glass-ceramic substrates, using the surface response method and an artificial neural network (ANN) approach. These efforts verified that high process-yields and excellent product-reliability for ceramic packaging devices could be achieved.

並列關鍵字

Stress analysis ； Process optimization ； Computational fluid dynamics ； Artificial neural networks