多階段製造工藝 (MMP) 中缺陷的出現是不可避免的，並且是一個持續存在的問題，會降低產品價值和製造商利潤。 因此，在一定範圍內調整 MMP 參數對於預防缺陷、動態精度和適應製造公差至關重要。 然而，之前的研究主要集中在單個MMP製造階段的優化，例如織物製造中的織造階段。 此外，當前的方法針對單個數字而不是範圍進行優化。 因此，本研究提出了一種稱為 Multistage Parameter Optimization for Rule Generation (MPORG) 的方法來防止 MMP 中出現缺陷。 在所提出的方法中，識別MMP關鍵參數，然後找到每個關鍵參數和缺陷類型的最佳值範圍，然後合併獲得的結果。 本研究的結果為製造工程師在MMP中調整特定產品類型的關鍵參數提供了相關條件（即一組最佳值範圍）。 所提出的方法提供了範圍值優化，在 MMP 參數調整期間為製造工程師提供了靈活的值範圍，這是線性迴歸方法無法提供的。 實驗結果表明，使用真實世界 I-Manufacturing 數據集驗證的 MPORG 方法可以防止單一類型或多類型缺陷的發生，效率約為 89%。 此外，實驗結果表明，在缺陷預防方面，所提出的方法在 MMP 中優於分類迴歸樹 (CART) 和多響應 CART (MR-CART) 方法。;Defects occurring in multistage manufacturing processes (MMPs) are inevitable and a persistent problem that decreases product value and manufacturer profits. As a result, MMP parameter tuning in a range is critical for defect prevention, dynamic precision, and accommodating manufacturing tolerances. Nevertheless, earlier studies primarily concentrated on optimizing single stages MMP, such as the weaving stage in the context of fabric manufacturing. Moreover, current studies focus on optimizing for a single value rather than a range. Therefore, this study introduced an approach named Multistage Parameter Optimization for Rule Generation (MPORG) and aimed at preventing the occurrence of defects in MMPs. In the proposed approach, the MMPs key parameters are identified, following which the optimal value range for each key parameter and defect type is found, and the obtained results are then merged. The results of this study indicate relevant conditions (i.e., a set of the optimal value range) for manufacturing engineers in adjusting the key parameters of a particular product type in an MMP. The proposed approach provides range-value optimization that gives a flexible range of values for manufacturing engineers during the MMPs parameters adjustments, which cannot be provided by the linear regression method. The results from the experiments indicated that the MPORG approach, validated with the real-world I-Manufacturing dataset, can reduce the occurrence of both single-type and multiple-type defects by approximately 89%. Furthermore, the experimental outcomes show that the proposed approach surpasses the performance of both the classification and regression tree (CART) and multiresponse CART (MR-CART) methods in preventing defects within MMP.
