銲接製程穩健設計最佳化之研究

Abstract

在自動化銲接的製程領域中，影響銲接品質的參數頗多。在銲接實務上，製程參數一般根據過去的經驗，或是參考文獻資料及設備供應商建議的數據來決定。對於特定的銲接系統或環境條件，此方式難以確保可得到最佳化的銲接品質。一般業界使用田口方法解決上述問題，然而田口方法在實務應用上存在一些缺失。由於應用田口方法於類神經網路設計，可得到許多網路設計的效益，因此，本論文提出一結合田口方法與類神經網路的方法，以改善參數設計最佳化的銲接問題。此方法包括二階段， 階段一利用田口方法針對銲接製程執行初始最佳化的實驗，以建立後續訓練類神經網路的資料庫。階段二應用類神經網路來搜尋最佳的參數組合，並採用Levenberg-Marquardt 倒傳遞演算法。本論文利用三個銲接的實務案例，包括氣體鎢極電弧銲、脈衝式Nd:YAG 雷射微接合及汽車電阻點銲等製程，來說明所提方法的有效性。實驗結果顯示本論文所提的方法優於傳統應用田口方法；氣體鎢極電弧銲接平均可提昇11.96%的銲道深寬比，脈衝式Nd:YAG 雷射微接合可降低3.37%的不良品率，電阻點銲平均可提昇7.26%的拉剪強度值；由此實務操作及結果，可說明所提方法具備高度的可行性。

Many parameters affect the automatic welding quality. In practice, the desired welding parameters are usually determined based on experience or handbook values. It does not insure that the selected welding parameters result in optimal or near optimal welding quality characteristics for that particular welding system and environmental conditions. To solve such problems, engineers conventionally apply the Taguchi method. However, the Taguchi method has some limitations in practice. Many benefits can arise from using the Taguchi method for neural network design. A proposed approach that combine the Taguchi method and a neural network to determine optimal welding conditions for improving the effectiveness of the optimization of parameter design is presented. The proposed approach includes two phases. Phase 1 executes initial optimization via Taguchi method to construct a database for the neural network. Phase 2 applies a neural network with the Levenberg-Marquardt back-propagation (LMBP) algorithm to search for the optimal parameter combination. Three examples involving the gas tungsten arc (GTA) welding, the pulsed Nd:YAG laser micro-weld process, and the resistance spot welding (RSW) process in automotive industry demonstrate the effectiveness of the proposed approach. The experimental results show that the proposed procedures excel the Taguchi method in this dissertation. It has demonstrated the practicability of the proposed procedures.