表面黏著技術(Surface mount technology)中以迴焊(Reflow)方式將錫膏融熔焊接電子元件及電路板。為使錫膏形成優良銲點(Solder Joint)產品於量產前,製程工程師通常以試誤法(Trail and Error)調整迴焊爐設定參數,即反覆改變加熱區設定溫度或輸送帶速度等,直到的量測之迴焊溫度曲線能符合溫度曲線規範,因此整備時間相當冗長連帶影響產能,故此乃迴焊製程中所須解決之問題。本文以高階伺服器產品之電路板為研究對象,依據迴焊溫度曲線推算流程,並蒐集影響迴焊之因子資料(電路板總熱容量、加熱區設定溫度及輸送帶速度),利用逐步迴歸分析方法及類神經網路建立修正原設定溫度。研究發現,逐步迴歸分析方法之修正函數於特定加熱區(Z2、Z3及Z10)存在顯著之預測能力;由類神經網路建立之修正模型,經修正後之建議設定溫度後與實際設定溫度較為接近。研究最後以高階伺服器產品之電路板,利用上述方法求得之各設定溫度進行實際迴焊驗證,發現由類神經網路修正之設定溫度所得之量測值均符合迴焊各階段之規範,且位於規範中間值,故利用此流程可達到降低測溫次數及減少測溫時間之目的,同時確保迴焊製程之穩定性。
The temperature profiles on the critical locations of the board during reflow soldering are important to result in the desired solder joint quality. This research establishes a procedure to predict the temperature profile based on given information about the PCB designs and components loading. First, critical factors such as the conveyor speed and temperature settings of adjacent zones that may influence the heating process are identified and investigated. A regression model and artificial neural network are constructed to more accurately predict the temperature profile. Results of this study will help improve the efficiency of temperature setting process in the pilot run stage.