- 學位論文
應用雙屬性節點分枝決策樹於塑膠射出成型製程績效之研究
The Study of Applying the Decision Tree with Bi-attribute Splitting Criteria on the Process Performance of Plastic Injection Molding
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
射出成型已為現今塑膠產業中重要的生產方式，有大量製造、生產快速、低價成本等優點。而影響塑膠產品的品質問題主要在於機器性能、材料特性、模具設計等所造成不良的缺陷。早期常利用田口(Taguchi)實驗設計方法來探討影響品質的因素，但利用實驗設計需要透過已知或欲設定的製程因素來調整水準，找出較佳參數組合進行生產。在實驗設計水準的設定下，所獲之結果雖然可視為在設定因素及水準下的最佳組合，但每個因素的最佳水準有可能會沒有被考慮到實驗設計中。因此，本研究利用資料挖礦(Data Mining)方法中的決策樹(Decision Tree)分析，迅速且有效率地找出影響塑膠射出成型不良之可能因素及提高其產能的重要因子，並以射出成型製程的真實資料，建立決策樹模型，萃取出分類法則，以診斷影響不良率及產能高低之因素，提供給公司管理人員作為製程改善與生產管理之參考，提早預防產品不良的發生。本研究延伸模型的可能性，提出了透過屬性組合的方法，探討影響生產效率及不良率的因素，以簡化生產製程之管理。
並列摘要
Injection molding has become an important production method in the plastics industry, with advantages such as mass production, high productivity, and low cost. The quality problem of plastic products mainly lies in the defects caused by machine performance, material characteristics, and mold design. In the early days, the Taguchi experimental design method was often adopted to find the factors affecting quality. However, experiments are design and based on known factors and adjust the level through pre setting levels to find the best combination of parameters for production. The best combination of the factors and levels are obtained based on pre setting of the experimental design levels, however, the real optimal levels of factors may not be considered in the experimental design. Therefore, this study aims to use Decision Tree Analysis, one of method in Data Mining, to quickly and efficiently identify possible factors that affect plastic injection molding defects on its productivity. Based on the real data from the injection molding process, the decision tree model is established and the classification rules are extracted to diagnose the factors affecting the defect rates and productivity. The rules can provid to the managers as a reference for process improvement and production management, and to prevent the occurrence of product defects in advance. Extending the possibility of the model, this study proposes the bi-attribute splitting criteria in decision tree analyse to explore the factors affecting productivity and defect rate. The results obtained can be used as the basis for the managers to set up the production parameter to enhance the quality and competitiveness of the company's products.
參考文獻
Behalek L., Seidl M., and Habr J., “Possibility to Detect Degraded Recycled Material in the Moulded Parts from PP,” MM Science Journal, Liberec, Czech Republic, Vol. 2016, Issue September, pp. 989-993 (2016).
Breiman L., Friedman J., Stone C. J., and Olshen R.A., Classification and Regression Trees, Taylor & Francis, pp.1-13 (1984).
Fayyad U., Piatetsky-Shapiro G., and Smyth P., “From Data Mining to Knowledge Discovery in Databases,” AI Magazine, AAAI, Vol. 17, No. 3, pp. 37-54 (1996).
Kozjek, D., Vrabič, R., Kralj, D., Butala, P., and Lavrač, N., “Data mining for fault diagnostics: A case for plastic injection molding,” Procedia CIRP, Vol. 81, pp. 809-814 (2019).
Lorena S. G., Victor S., Elina A. O., Joos V., and Dirk C., “Literature Review of Data Mining Applications in Academic Libraries,” The Journal of Academic Librarianship, ACALIB, Vol. 41, No. 4, pp. 499-510 (2015).