臺灣博碩士論文加值系統

高分子加工中，射出成型製程具備大量生產、高重複精度與低成本等生產優勢。然而，傳統製程中，機台生產前需經諸多前置試模流程以找尋最佳參數配置；同時，製程中無法對每模次成品進行品質控管成了其無法忽視的問題。綜上所述，此兩痛點在缺乏數位化與結構性調整策略下，將於時間與人力上消耗大量成本，進而成為產業生產中亟需克服之痛點。
本研究為克服上述之痛點，並延伸改善模具交付客戶後因場域環境與人為操作等變動所導致重新進行前置實驗所消耗的成本，提出一種以多層感知器模型結合製程參數與品質指標之品質預測技術，此方法不僅預測成品品質，更能節省量測所需之大量時間與人力成本，提升整體生產效率與效益。此外，本研究以實驗與模擬兩種主要方案，討論相同模具與相異機台下之生產品質差異，並從中進行射出件之品質轉移預測研究，並以此探討與分析模具於相異機台下，仍維持預期表現並產生符合品質要求之成效。
本研究具體分為三大部份，第一，先於實驗調動機台參數，分別為熔膠溫度、保壓壓力及冷卻時間，並以過往實驗所得出最佳參數配置為中心，以一定區間進行隨機抽樣，藉此還原且放大實際射出過程中的環境變動造成的品質不穩定影響，進而透過壓力感測技術與品質指標化技術，擷取模內壓力並進行特徵萃取與資料降維；第二，建立實際機台生產參數、品質指標與成品品質之三者製程關鍵資訊中，以多層感知器模型搭建起兩者間之關係並進行數據預測，達到資料擬合與學習效果；第三，建立模擬環境並模擬射出過程，提取感測點壓力曲線並以品質指標化技術與遷移式品質擬合技術，使模擬品質指標與實驗品質指標兩者擬合，進而直接預測重量與幾何品質。
經由實驗驗證，本方法可良好預測原機台生產之重量品質，同時亦能透過模擬數據提取之品質指標達到異機成品品質預測之成效。其多層感知器模型預測的RMSE值皆小於0.1，顯示其預測準確度具可信任性，並且擬合後之品質指標預測重量及幾何品質結果皆與模擬品質平均誤差<1%，顯示本研究所提出之研究架構具備可行性。

In polymer processing, injection molding has the advantages of high production volume, high repeatability, and low cost. However, traditional production requires various pre-molding processes to find the optimal parameter combination. Also, the lack of quality control for each molded product during the process becomes a significant issue. These two pain points, without digitalization and structured adjustment strategies, result in significant time and labor costs and become urgent challenges in industrial production.
To overcome the above-mentioned challenges and address the cost consumed by conducting pre-experiments due to environmental and operational changes after delivering molds to customers, this study proposes a quality prediction technique that combines a multi-layer perceptron model with process parameters and quality indicators. This method not only predicts the quality of finished products but also saves a significant amount of time and labor costs required for measurements, thereby enhancing overall production efficiency and effectiveness. Furthermore, this study discusses the differences in production quality between molds produced under different but similar machines through two main approaches: experimentation and simulation. It also investigates quality transfer prediction to ensure that molds maintain expected performance and produce results that meet quality requirements under different machines.
This study consists of three main parts. First, experimental adjustments are made to machine parameters, including melt temperature, holding pressure, and cooling time, centered around the optimal parameter configuration obtained from previous experiments. Random sampling is performed within a specific positive and negative range to restore and amplify the impact of environmental variations during the actual injection process. Pressure sensing technology and quality indicator techniques are then utilized to extract in-mold pressure and perform feature extraction and data dimensionality reduction. Second, a multi-layer perceptron model is established to predict data by mutually establishing relationships between actual machine production parameters, quality indicators, and finished product quality. This achieves data fitting and learning effects. Third, a simulation environment is created to simulate the injection molding process. Pressure curves at sensing points are extracted, and quality indicator techniques and transferable quality fitting techniques are employed to achieve fitting between simulated quality indicators and experimental quality indicators. This enables direct prediction of weight and geometric quality.
Experimental validation shows that this method can accurately predict the weight quality of original machine production. Moreover, through the extraction of simulation data, the effectiveness of predicting different machine product quality is achieved. The root means square error (RMSE) values of the multi-layer perceptron model predictions are all less than 0.1, indicating their reliability in prediction accuracy. Additionally, the average errors between the fitted quality indicator predictions for weight and geometric quality and the simulated quality are both less than 1%, demonstrating the feasibility of the research framework proposed in this study.

摘要 i
ABSTRACT ii
誌謝 iv
目錄 v
表目錄 viii
圖目錄 x
符號目錄 xii
第一章 緒論 1
1-1 前言 1
1-2 傳統試模技術 2
1-3 科學化成型方法 3
1-4 人工智慧在射出成型的應用 5
1-5 研究動機與目的 6
1-6 論文架構 7
第二章 文獻回顧 9
2-1 傳統射出成型 10
2-2 工業4.0 12
2-3 感測技術與大數據處理技術 14
2-4 智慧製造與彈性製造系統 16
2-5 人工智慧與AI射出成型 19
第三章 研究方法 22
3-1 建立移機品質預測架構 22
3-1-1 單機製程關鍵資訊獲取 23
3-1-2 單機品質預測 23
3-1-3 機械觀點品質預測 24
3-1-4 跨機品質預測 24
3-2 建立數據庫與資料處理流程 25
3-3 品質指標化 25
3-4 相關性分析 28
3-5 多層感知器模型 29
3-6 田口實驗設計法 30
3-7 多層感知器超參數優化方法 31
3-7-1 田口方法L12(21×35)直交表 32
3-7-2 訊噪比分析 33
3-8 平均絕對百分比誤差 34
3-9 國際標準公差 34
第四章 實驗設計與實驗設備 36
4-1 實驗設備 39
4-1-1 射出成型機與模溫機 39
4-1-2 模具設計與實驗材料 40
4-1-3 感測器與感測設備 43
4-1-4 量測設備 44
4-1-5 軟體設備 45
4-2 選定實驗製程參數 46
第五章 結果與討論 48
5-1 製程參數對壓力曲線與品質指標之數值關係分析 48
5-1-1 A機台數值轉換關係之說明 48
5-1-2 B機台數值轉換關係之說明 50
5-2 單機製程參數對成品品質之相關性分析 52
5-2-1 A機台製程參數對成品品質之相關性分析 52
5-2-2 B機台製程參數對成品品質之相關性分析 53
5-3 單機品質指標對成品品質之相關性分析 54
5-3-1 A機台品質指標對成品品質之相關性分析 54
5-3-2 B機台品質指標對成品品質之相關性分析 57
5-4 MLP模型之預測效果 59
5-4-1 品質指標與成品品質關係之MLPIQ模型預測效果 59
5-4-2 製程參數與品質指標關係之MLPPI模型預測效果 63
5-4-3 預測品質指標(Î)與成品品質關係之MLPÎQ模型預測效果 65
5-4-4 製程參數與成品品質關係之MLPPQ模型預測效果 66
5-4-5 跨機品質指標間的關係之MLPIBIA模型預測效果 67
5-5 單機與機對機品質預測方法之分析與討論 70
5-5-1 A機台品質QA 對B機台品質QB 71
5-5-2 Model A1品質QA vs 實際品質QA 72
5-5-3 Model A2品質QA vs 實際品質QA 73
5-5-4 單機品質預測方法之結果與討論 74
5-5-5 Model A1B品質 QB 對A機台品質QA與B機台品質QB 75
第六章 結論與未來展望 80
6-1 結論 80
6-2 未來展望 82
參考文獻 84

[1]Rosti, Injection molding: complete guide to scientific molding, (2023). https://www.plasticcomponents.com/scientific-molding-comprehensive-guide (accessed June 13, 2023).
[2]FimmTech, Scientific molding – the 6-step study, (2018). https://fimmtech.com/knowledgebase-2/scientific-molding-the-6-step-study/ (accessed June 22, 2023).
[3]A. Kaplan, M. Haenlein, Siri, siri, in my hand: who’s the fairest in the land? on the interpretations, illustrations, and implications of artificial intelligence, Bus. Horiz. 62 (2019) 15–25. https://doi.org/10.1016/j.bushor.2018.08.004.
[4]Y. Lockner, C. Hopmann, W. Zhao, Transfer learning with artificial neural networks between injection molding processes and different polymer materials, J. Manuf. Process. 73 (2022) 395–408. https://doi.org/10.1016/j.jmapro.2021.11.014.
[5]A. Mourya, A. Nanda, K. Parashar, Sushant, R. Kumar, An explanatory study on defects in plastic molding parts caused by machine parameters in injection molding process, Mater. Today Proc. 78 (2023) 656–661. https://doi.org/10.1016/j.matpr.2022.12.070.
[6]C.-J. Tzeng, Y.-K. Yang, Y.-H. Lin, C.-H. Tsai, A study of optimization of injection molding process parameters for SGF and PTFE reinforced PC composites using neural network and response surface methodology, Int. J. Adv. Manuf. Technol. 63 (2012) 691–704. https://doi.org/10.1007/s00170-012-3933-6.
[7]P. Zhao, Z. Dong, J. Zhang, Y. Zhang, M. Cao, Z. Zhu, H. Zhou, J. Fu, Optimization of injection-molding process parameters for weight control: converting optimization problem to classification problem, Adv. Polym. Technol. 2020 (2020) 1–9. https://doi.org/10.1155/2020/7654249.
[8]S. Mukras, Experimental-based optimization of injection molding process parameters for short product cycle time, Adv. Polym. Technol. 2020 (2020) 1–15. https://doi.org/10.1155/2020/1309209.
[9]T. Ginghtong, N. Nakpathomkun, C. Pechyen, Effect of injection parameters on mechanical and physical properties of super ultra-thin wall propylene packaging by Taguchi method, Results Phys. 9 (2018) 987–995. https://doi.org/10.1016/j.rinp.2018.04.001.
[10]TPRMA, 德國工業4.0政府計畫推動趨勢與智慧製造亮點案例, (2017). https://www.tprma.org.tw/industry-news/203-德國工業4-0政府計畫推動趨勢與智慧製造亮點案例.html (accessed June 14, 2023).
[11]J. Bilbao-Ubillos, V. Camino-Beldarrain, G. Intxaurburu-Clemente, E. Velasco-Balmaseda, Industry 4.0 and potential for reshoring: A typology of technology profiles of manufacturing firms, Comput. Ind. 148 (2023) 103904. https://doi.org/10.1016/j.compind.2023.103904.
[12]行政院經濟建設委員會, 歐美「再工業化」策略及 對台灣產業政策意涵, (2013). https://ws.ndc.gov.tw/001/administrator/10/relfile/5636/27188/f8ec4bf5-71a6-467b-a8d4-faac2022e449.pdf (accessed June 27, 2023).
[13]T. Ageyeva, S. Horváth, J.G. Kovács, In-mold sensors for injection molding: on the way to industry 4.0, Sensors. 19 (2019) 3551. https://doi.org/10.3390/s19163551.
[14]Y. Li, J.C. Chen, W.M. Ali, Process optimization and in-mold sensing enabled dimensional prediction for high precision injection molding, Int. J. Interact. Des. Manuf. IJIDeM. 16 (2022) 997–1013. https://doi.org/10.1007/s12008-021-00800-1.
[15]S.-C. Nian, Y.-C. Fang, M.-S. Huang, In-mold and machine sensing and feature extraction for optimized IC-tray manufacturing, Polymers. 11 (2019) 1348. https://doi.org/10.3390/polym11081348.
[16]P. Zheng, H. wang, Z. Sang, R.Y. Zhong, Y. Liu, C. Liu, K. Mubarok, S. Yu, X. Xu, Smart manufacturing systems for Industry 4.0: Conceptual framework, scenarios, and future perspectives, Front. Mech. Eng. 13 (2018) 137–150. https://doi.org/10.1007/s11465-018-0499-5.
[17]J.-C. Fan-Jiang, C.-W. Su, G.-Y. Liou, S.-J. Hwang, H.-H. Lee, H.-S. Peng, H.-Y. Chu, Study of an online monitoring adaptive system for an injection molding process based on a nozzle pressure curve, Polymers. 13 (2021) 555. https://doi.org/10.3390/polym13040555.
[18]I.R. Schiffers, G.P. Holzinger, G. Huster, Adaptive process control for stabilizing the production process in injection moulding machines, in: Proc. 10th Int. Fluid Power Conf., Dresden, Germany, 2016: pp. 341–352. https://nbn-resolving.org/urn:nbn:de:bsz:14-qucosa-200201 (accessed June 28, 2023).
[19]Y.-S. Chen, K.-T. Wu, M.-H. Tsai, S.-J. Hwang, H.-H. Lee, H.-S. Peng, H.-Y. Chu, Adaptive process control of the changeover point for injection molding process, J. Low Freq. Noise Vib. Act. Control. 40 (2021) 383–394. https://doi.org/10.1177/1461348419875057.
[20]S. Farahani, V. Khade, S. Basu, S. Pilla, A data-driven predictive maintenance framework for injection molding process, J. Manuf. Process. 80 (2022) 887–897. https://doi.org/10.1016/j.jmapro.2022.06.013.
[21]S. Kumar, H.S. Park, C.M. Lee, Data-driven smart control of injection molding process, CIRP J. Manuf. Sci. Technol. 31 (2020) 439–449. https://doi.org/10.1016/j.cirpj.2020.07.006.
[22]ENGEL, Compensate for fluctuations in plastic viscosity, (2023). https://www.engelglobal.com/en/us/digital-solutions/digital-solutions-injection-molding-manufacturing-process/optimize-plastic-viscosity (accessed June 27, 2023).
[23]K.-C. Ke, M.-S. Huang, Quality classification of injection-molded components by using quality indices, grading, and machine learning, Polymers. 13 (2021) 353. https://doi.org/10.3390/polym13030353.
[24]K.-C. Ke, M.-S. Huang, Quality prediction for injection molding by using a multilayer perceptron neural network, Polymers. 12 (2020) 1812. https://doi.org/10.3390/polym12081812.
[25]陳銘德, ISO國際版公差分析之2─標準公差值解析, (2020). http://digorlon.com/home/post/1201 (accessed July 15, 2023).
[26]SMART Molding, 射出成品精度與成型參數及模穴公差之關係, (2020). https://www.smartmolding.com/20-09c05/ (accessed July 15, 2023).
[27]B.-W. Wang, M.-S. Huang, S.-C. Nian, 射出成型之虛實整合技術, 2021. https://hdl.handle.net/11296/5q3fta (accessed July 12, 2023).