臺灣博碩士論文加值系統

近年來，得力於資訊科技與人工智慧的發展興盛，工業4.0以及智慧製造已經成為現代製造業的轉型重點，更成為一股勢不可擋的趨勢。
在智慧製造的浪潮下，與塑膠射出A公司合作，配合製造業智慧應用升級輔導計畫(SMB/SMU) ，建立物聯網即時監控系統與製造管理系統，前者連接工廠內所有的塑膠射出機台，可24小時不間斷紀錄並監控機台狀況，後者則針對製程與報工相關資訊加以管制。
本研究在物聯網即時監控系統與製造管理系統的導入下，收集數年來所累計的塑膠射出機台數據，累計805,317筆調機資料，透過資料前處理，配以專家意見，進行製程參數探討與資料分析，建模的部分採用多變量回歸、隨機森林、類神經網路三種經典演算法，並且使用分層抽樣以及K-fold交叉驗證方式確保最佳化模型，為塑膠產業建立智慧化製程參數推薦機制。
研究結果在建立智慧化製程參數推薦機制後，採用隨機森林模型的情況下，關鍵參數冷卻時間可達到76.17%的準確度，保留專業的調機優化智慧，降低調機工程師離職造成的影響，長期來看可持續累積調機經驗的智慧化，達到本研究之目的。

Recently, with the development and prosperity of information technology and artificial intelligence, Industry 4.0 and smart manufacturing have become the focus of transformation of modern manufacturing; and fueled by the epidemic and declining birthrate, it has become an unstoppable trend.
Under the wave of smart manufacturing. It cooperates with Plastic Injection Company A and cooperates with the smart manufacturing application upgrade mentoring program (SMB/SMU) to establish a IoT real-time monitoring system and a manufacturing management system. The former connects All plastic injection machines in the factory can record and monitor the status of the machine 24 hours a day, and the latter controls information related to the manufacturing process and labor reporting.
With the introduction of computer network real-time monitoring system and manufacturing management system, we collected plastic injection data accumulated over the past few years, accumulating 805,317 pieces of machine adjustment information, through data pre-processing and expert opinions, we conduct process parameter discussion and data analysis. The modeling part uses three classic algorithms: linear regression, random forest, and neural network, and uses stratified sampling and K-fold cross-validation methods to ensure model performance and build wisdom for the plastics industry. process parameter recommendation mechanism.
The research results show that after establishing an intelligent process parameter recommendation mechanism and using the random forest model, the key parameter cooling time can achieve 76.17% accuracy, retain professional wisdom in machine tuning and optimization and reducing the impact of the tuning engineer's resignation. The purpose of this study can be achieved through the continuous accumulation of wisdom in machine adjustment experience..

摘要 i
Abstract ii
目次 iii
表目次 v
圖目次 vi
第 1 章 緒論 1
1.1 研究背景與動機 1
1.2 研究目標 2
1.3 研究範圍與限制 2
1.4 研究流程 3
第 2 章 文獻探討 4
2.1 塑膠射出成型簡介 4
2.1.1 射出成型加工過程 4
2.1.2 射出成型關鍵要素 4
2.2 機器學習演算法簡介 5
2.2.1 何謂機器學習 5
2.2.2 多變量回歸 5
2.2.3 隨機森林 6
2.2.4 類神經網路 7
2.3 機器學習在射出成型之相關應用 8
第 3 章 研究步驟與方法 10
3.1 資料蒐集 10
3.2 資料預處理 10
3.2.1 資料過濾/專家法 10
3.2.2 缺失值處理 11
3.2.3 資料標準化 11
3.3 資料分析與探討 11
3.4 資料取樣、建模及評估 12
第 4 章 資料分析與討論 13
4.1 資料分析環境 13
4.2 資料蒐集與彙整 13
4.2.1 製程數據視覺化 14
4.2.2 製程資料預處理 16
4.3 資料切分與實驗 20
4.4 資料採樣與建模 21
4.5 預測結果與效能評估 21
4.5.1 多變量回歸 22
4.5.2 隨機森林 24
4.5.3 類神經網路 26
4.5.4 訓練時間比較 29
4.5.5 預測時間比較 30
第 5 章 結論 31
第 6 章 參考書目 32
第 7 章 附錄 33
7.1 訪談紀錄 33
7.2 測試ChatGPT(v3.5)用於成型條件預測 34
7.3 實驗結果數據 36
7.4 程式碼與原始資料 39

1. Lasi, H., Fettke, P., Kemper, HG. et al, “Industry 4.0,” Bus Inf Syst Eng 6, 239–242 (2014). https://doi.org/10.1007/s12599-014-0334-4 2. Ghobakhloo, Morteza, “Industry 4.0, digitization, and opportunities for sustainability,” Journal of cleaner production 252 (2020): 119869.
3. Sascha, Julian, Oks., Max, Jalowski., Michael, Lechner., Stefan, Mirschberger., Marion, Merklein., Birgit, Vogel-Heuser., Kathrin, M., Möslein., “Cyber-Physical Systems in the Context of Industry 4.0: A Review, Categorization and Outlook,” Information Systems Frontiers, doi: 10.1007/s10796-022-10252-x
4. Raman, Kumar., Sita, Rani., Mohammed, Awadh., “Exploring the Application Sphere of the Internet of Things in Industry 4.0: A Review, Bibliometric and Content Analysis,” Sensors, 2022. doi: 10.3390/s22114276 5. Agrawal, A. R., I. O. Pandelidis, and M. Pecht, “Injection‐molding process control—A review,” Polymer Engineering & Science 27.18 (1987): 1345-1357. 6. R. Ventura and X. Berjaga, “Comparison of multivariate analysis techniques in plastic injection moulding process,” 2015 IEEE 20th Conference on Emerging Technologies & Factory Automation (ETFA), Luxembourg, Luxembourg, 2015, pp. 1-6, doi: 10.1109/ETFA.2015.7301557. 7. J. Posada, C. Toro, I. Barandiaran et al., “Visual Computing as a Key Enabling Technology for Industrie 4.0 and Industrial Internet,” in IEEE Computer Graphics and Applications, vol. 35, no. 2, pp. 26-40, Mar.-Apr. 2015, doi: 10.1109/MCG.2015.45. 8. D. Dujovne, T. Watteyne, X. Vilajosana and P. Thubert, “6TiSCH: deterministic IP-enabled industrial internet (of things),” in IEEE Communications Magazine, vol. 52, no. 12, pp. 36-41, December 2014, doi: 10.1109/MCOM.2014.6979984. 9. Uglov, Arsenii, et al., “Surrogate Modelling for Injection Molding Processes using Machine Learning,” 2021, arXiv:2107.14574 . 10. G. Kim, J. G. Choi, M. Ku, H. Cho and S. Lim, “A Multimodal Deep Learning-Based Fault Detection Model for a Plastic Injection Molding Process,” in IEEE Access, vol. 9, pp. 132455-132467, 2021, doi: 10.1109/ACCESS.2021.3115665. 11. Zhi-Hao, Wang., Fuzhong, Wen., Yi-Ting, Li., Hao-Hsuan, Tsou, “A Novel Sensing Feature Extraction Based on Mold Temperature and Melt Pressure for Plastic Injection Molding Quality Assessment,” in IEEE Sensors Journal, 2023, doi: 10.1109/JSEN.2023.3247597