臺灣博碩士論文加值系統

在現今化工製程中，經常使用軟儀表模型來預測製程中關鍵的品質變量，以解決其量測成本與量測時間過長等問題。藉由製程中較容易取得的操作變量與感測變量透過神經網路模型，獲得與品質變量之間的關聯性。本研究目標為建立粗製間苯二甲酸製程軟儀表與人工智慧操作導航系統，但由於取樣時間間隔較長與製程穩定控制的需求，模型需要具備長時間預測的能力。配合品質變量取樣時間間隔不同，提出增益一致性遮罩式序列對序列式模型，建立具備物理解釋性的高頻率預測模型，確保能夠同時滿足即時預測與控制的需求。模型的建立使用移動式窗的方法作為輸入，讓實際測量數據能夠在各個預測時間點進行訓練，同時在損失函數中引入遮罩變數來遮蔽未取樣時間點的數據對於模型訓練的不合理性，並導入增益一致性損失函數讓模型具有正確的物理解釋性。本研究最終提出以主成分分析的方法結合模型預測控制概念作為人工智慧導航的架構。主成分分析提供當前製程操作範圍內最佳的操作條件;模型則藉由調整解碼器時間點的操作變量，輸入目標方程式與限制條件進行預測與規劃出最佳的操作策略。實際測試結果顯示，依照最佳操作條件調整與導航建議的配合下，能夠有效降低單位投料量(間二甲苯)下一氧化碳與二氧化碳總合濃度所佔的比例。當品質變量 3-羧基苯甲醛超過限制條件時亦能提供合理的操作建議進行穩定製程的控制。

Nowadays, soft-sensors are often used to predict quality variables in chemical engineering process to reduce the measurement cost and time. There are several
variables that can be detected in real time, which are manipulated variables and sensor variables. By using neural network can construct the correlation between them. The objective of this research is to establish a soft-sensor and an artificial intelligent navigation on a Crude Isophthalic Acid process. However, due to the long sampling interval of quality variables and demand of stabilizing the process, the model needs to have the capability of long-term prediction. According to the dislocation of quality variable data, this research proposed to use sequence-to-sequence model combine with mask and gain consistency loss functions to establish a model with multi-steps, high-frequency prediction and physical interpretability. The model was built by using the moving window method as the input, so that the actual measurement data could be trained at each time step. The mask method introduced mask variables in the loss function to avoid the irrationality of the non-sampling data for model training, while the gain consistency loss function forced the model to be physically interpretable. This
research ended up using principle component analysis method combined with the
concept of model predictive control as the architecture of artificial intelligent navigation. Principle component analysis can provide the optimal operation conditions within current process and the model can also provide suitable control strategies by adjusting manipulated variables on decoder. The results showed that the adjustment of the process followed by the suggestion can reduce the ratio of the summation of CO and CO2 in per unit Meta Xylene (MX) effectively. In addition, while quality variable 3-carboxybenzaldehyde (3CBA) exceed the constraint, this system can also provide reasonable suggestion to stabilize the process.

摘要 i
Abstract iii
致謝 v
目錄 vi
表目錄 viii
圖目錄 ix
第一章 緒論 1
1.1 研究動機 1
1.2 研究目的 2
1.3 論文架構 2
第二章 文獻回顧 3
2.1 軟儀表 3
2.2 人工神經網路 4
2.3 遞迴神經網路 5
2.4 序列對序列式模型(Sequence to Sequence, StS) 5
2.5 人工智慧操作導航 6
第三章 CIA氧化塔製程介紹 8
第四章 研究方法 10
4.1 模型數據介紹 10
4.2 數據前處理 12
4.2.1 數據標準化 12
4.2.2 移動視窗 12
4.3 序列對序列式模型結構 14
4.4 遮罩序列對序列式模型結構(Mask Sequence to Sequence, MStS) 15
4.5 增益一致性損失函數 17
4.5.1 增益一致性 17
4.5.2 增益一致性損失函數計算 20
4.6 增益一致性遮罩式序列對序列式模型(Gain-Informed Mask Sequence-to-Sequence, GI-MStS) 22
4.7 主成分分析 23
4.8 製程導航系統架構 24
4.9 最適化方法 25
4.9.1 最適化決策 25
4.9.2 差分進化演算法 26
4.10 目標函數與製程限制式 28
第五章 結果與討論 32
5.1 預測能力分析 32
5.1.1 遮罩式序列對序列式模型 32
5.1.2 增益一致性遮罩式序列對序列式模型 35
5.2 PCA 歷史操作分析與最佳操作建議 46
5.3 製程導航系統實際操作結果 49
第六章 結論 55
參考文獻 56

1. J.Mohd Ali, M.A.Hussain, M.O.Tade, J.Zhang, Artificial Intelligence techniques applied as estimator in chemical process systems - A literature survey, Expert Systems with Applications, 42, 5915-5931, 2015
2. Kadlec, P., Gabrys, B., and Strandt, S., "Data-driven soft sensors in the process industry, " Computers & chemical engineering, vol. 33, no. 4, pp. 795-814, 2009.
3. Acock, A. C., "Working with missing values," Journal of Marriage and family, vol. 67, no. 4, pp. 1012-1028, 2005.
4. Shao, W. and Tian, X., "Adaptive soft sensor for quality prediction of chemical processes based on selective ensemble of local partial least squares models, " Chemical Engineering Research and Design, vol. 95, pp. 113-132, 2015.
5. Khatibisepehr, S., Huang, B., and Khare, S., "Design of inferential sensors in the process industry: A review of Bayesian methods, " Journal of Process Control, vol. 23, no. 10, pp. 1575-1596, 2013.
6. Shardt, Y. A., Statistics for chemical and process engineers. Springer, 2015.
7. Lin, B., Recke, B., Knudsen, J. K., and Jørgensen, S. B., "A systematic approach for soft sensor development, " Computers & chemical engineering, vol. 31, no. 5- 6, pp. 419-425, 2007.
8. Facco, P., Doplicher, F., Bezzo, F., and Barolo, M., "Moving average PLS soft sensor for online product quality estimation in an industrial batch polymerization 72 process, " Journal of Process Control, vol. 19, no. 3, pp. 520-529, 2009.
9. Fujiwara, K., Kano, M., Hasebe, S., and Takinami, A., "Soft‐sensor development using correlation‐based just‐in‐time modeling, " AIChE Journal, vol. 55, no. 7, pp. 1754-1765, 2009.
10. da Silva Bispo, V. D., Scheid, C. M., Calçada, L. A., and da Cruz Meleiro, L. A., "Development of an ANN-based soft-sensor to estimate the apparent viscosity of water-based drilling fluids, " Journal of Petroleum Science and Engineering, vol. 150, pp. 69-73, 2017.
11. Liu, Y., Yang, C., Gao, Z., and Yao, Y., "Ensemble deep kernel learning with application to quality prediction in industrial polymerization processes, " Chemometrics and Intelligent Laboratory Systems, vol. 174, pp. 15-21, 2018.
12. Qin, S. J., "Neural networks for intelligent sensors and control—Practical issues and some solutions, " in Neural systems for control: Elsevier, 1997, pp. 213-234.
13. Fortuna, L., Graziani, S., and Xibilia, M. G., "Soft sensors for product quality monitoring in debutanizer distillation columns, " Control Engineering Practice, vol. 13, no. 4, pp. 499-508, 2005.
14. Rallo, R., Ferre-Gine, J., Arenas, A., and Giralt, F., "Neural virtual sensor for the inferential prediction of product quality from process variables, " Computers & Chemical Engineering, vol. 26, no. 12, pp. 1735-1754, 2002.
15. I.Sutskever, O.Vinyals, Q.V.Le, Sequence to sequence learning with neural networks. In Advances in neural information processing systems, 3104-3112, 2014.
16. D.Bahdanau, K.Cho, Y.Bengio, Neural machine translation by jointly learning to align and translate, arXiv preprint arXiv:1409.0473, 2014.
17. Cho, K., Van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., and Bengio, Y., "Learning phrase representations using RNN encoder-decoder for statistical machine translation, " arXiv preprint arXiv:1406.1078, 2014.
18. A.Kannan, K.Kurach, S.Ravi, T.Kaufmann, A.Tomkins, B.Miklos, ... and V.Ramavajjala, Smart reply: Automated response suggestion for email. In Proceedings of the 22nd ACM SIGKDD, International Conference on Knowledge Discovery and Data Mining, 955-964, 2016.
19. Chou, C.-H., Wu, H., Kang, J.-L., Wong, D. S.-H., Yao, Y., Chuang, Y.-C., Jang, S.-S., and Ou, J. D.-Y., "Physically consistent soft-sensor development using sequence-to-sequence neural networks," IEEE Transactions on Industrial Informatics, vol. 16, no. 4, pp. 2829-2838, 2019.
20. Bengio, Y., Simard, P., and Frasconi, P., "Learning long-term dependencies with gradient descent is difficult, " IEEE transactions on neural networks, vol. 5, no. 2, pp. 157-166, 1994.
21. Dey, R. and Salem, F. M., "Gate-variants of gated recurrent unit (GRU) neural networks," in 2017 IEEE 60th international midwest symposium on circuits and systems (MWSCAS), 2017: IEEE, pp. 1597-1600.
22. Lee, J.H., Model predictive control: Review of the three decades of development. International Journal of Control, Automation and Systems, 9(3): p. 415-424, 2011.
23. Rachael, J.A., Rault, A., Testud, J. L., & Papon, J., Model Predictive heuristic control: application to an Industrial Process. Automatica, 14(5): p. 413- 428, 1978.
24. W.AI-Gherwi, H.Budman, A.Elkamel, A robust distributed model predictive control based on a dual-mode approach, Computers and Chemical Engineering, 130-138, 2013.
25. M.Morari, J.H.Lee, Model predictive control: past, present and future, Computers and Chemical Engineering, 667-682, 1999.
26. Lu, Z., Pu, H., Wang, F., Hu, Z., and Wang, L., "The expressive power of neural networks: A view from the width," in Proceedings of the 31st International Conference on Neural Information Processing Systems, 2017, pp. 6232-6240.
27. Hsiao, Y.-D., Kang, J.-L., and Wong, D. S.-H., "Development of Robust and Physically Interpretable Soft Sensor for Industrial Distillation Column Using Transfer Learning with Small Datasets," Processes, vol. 9, no. 4, p. 667, 2021.
28. S.Wold, K.Esbensen, P.Geladi, "Principal component analysis, Chemometrics and intelligent laboratory systems, 2, 37-52, 1987.
29. Mohamed, A.W., Sabry, H. Z., & Abd-Elaziz, T, "Real parameter optimization by an effective differential evolution algorithm. Egyptian Informatics Journal, 2013. 14(1): p. 37-53.