乳化聚合在工業上的操作可分為連續式、半批式和批式操作，其中批式操作多半在實驗室中進行，鮮少應用於工業大量生產，儘管批式操作能有較高的轉化率(約90%)，但不易移除反應熱成為其操作致命傷。丁二烯乳化聚合屬於高放熱反應且丁二烯容易氣化，在升溫期間槽壓會快速上升甚至超過反應器設計壓力造成工安危害，若能掌握反應器壓力變化發展，便可發展其他操作方式來避免超壓危害，此外，目前多數理論模型是以實驗室或試驗場數據進行建模，因此對於預測工業級反應器的預測能力存在許多挑戰。
綜合以上分析，本研究將以聚丁二烯批次乳化聚合作為研究主題，運用Aspen Plus軟體並且根據實廠反應器壓力量測數據，建立批次理論模型，從中探討其複雜的熱力學、動力學機制，提供非線性參數迴歸流程，透過提出不同反應器操作方式探討操作安全和反應時間與操作變數間的關聯性，最後進一步以變動單體入料佔比進行擾動分析。
乳化聚合反應進料有六種成分包含水、起始劑、乳化劑1、乳化劑2以及分子量調整劑，本研究利用Aspen Plus針對丁二烯批次乳化聚合製程建立批次理論模型並且選用POLTNRTL建立熱力學模型，提出新參數迴歸流程(貝葉斯優化結合Nelder-Mead演算法)迴歸動力式參數，以擬合實廠槽壓數據，其擬合誤差MAPE < 2%，最大槽壓誤差< 0.3 kg/cm2。
接續以LG實廠製程操作方式(Case A)為基礎提出額外3種操作方式，分別是區段線性升溫(Case B)、線性升溫(Case C)以及區段壓力控制(Case D)，槽壓模擬結果顯示槽壓開始明顯下降的時機皆在轉化率達40-50%，與文獻實驗相符，由此可透過槽壓數值推算當時轉化率。熱負荷模擬結果顯示，每個操作方式的總反應熱約為升溫總需熱量的10倍，表示有大量多餘的熱需要移除。操作變數靈敏度分析的結果顯示縮短反應時間和製程安全存在trade-off。比較4種操作方式發現，Case D可以避免trade-off問題並且能大幅縮短反應時間從原先的29.03 h縮短至26.38 h，改善幅度達9.13%。以變動單體進料佔比進行製程擾動分析，結果顯示轉化率上升速率、瞬時放熱速率最大值以及反應時間隨單體佔比減少而增加，而反應總放熱量和升溫總需熱量與單體佔比成正比關係。

There are three industrial operations for emulsion polymerization: continuous, semi-batch, and batch operation. Batch operation is predominantly conducted in laboratories and is seldom applied in large-scale industrial production. Although Batch operation can achieve higher conversion rates (90%), it will be challenging to remove reaction heat. Emulsion polymerization of butadiene is a highly exothermic reaction, and butadiene is prone to vaporize. During heating, the reactor pressure increases dramatically, and even exceeds the reactor design pressure. These features render the industrial operation of emulsion polymerization conservative, as the reactor pressure tend to increase dramatically during operation.
There are six components in the reaction system: Water, Butadiene, Initiator, Emulsifier1, Emulsifier2 and Molecular weight regulator. The batch theoretical model for butadiene emulsion polymerization is established using Aspen Plus and the POLYNRTL was selected as the thermodynamic model. A new regression procedure (Bayesian optimization + Nelder-Mead algorithm) is introduced to regress kinetic parameters and fit the reactor pressure data obtained from industrial operation. The fitting error in Mean Absolute Percentage Error (MAPE) is less than 2%, and the maximum reactor pressure error is less than 0.3 kg/cm2. According to the reactor operation of the LG factory process (Case A), three additional procedures were proposed: Piecewise Linear Heating (Case B), Overall Linear Heating (Case C), and Section Pressure Control (Case D). As shown from the simulation result, the reactor pressure begins to drop dramatically when the conversion rate reaches approximately 40–50 %. This is consistent to the previous experimental observation. In addition, the total amount of reaction heat reaches approximately 10 times of heat required for increasing reaction temperature. Hence, successful removal of reaction heat can be critical.
The sensitivity analysis of operation variables reveals a clear trade-off between shortening the reaction time and ensuring process safety. After comparing four operations, Case D represents the best scenario by significantly shorting the reaction time from 29.03 hours to 26.38 hours, which is 9.13% reduction. Finally, the process disturbance is analyzed with monomer feed ratio variations. The results indicate that the rise rate of conversion, maximum instantaneous reaction heat and reaction time will increase with decreasing monomer feed ratio. In contrast, the reaction heat and the total required duty are directly proportional to the monomer ratio.

[1] 徐武軍，石油化學工業-原料製程及市場(第二版)，五南圖書出版股份有限公司，2007年。
[2] Washington, I. D., Dynamic Modelling of Emulsion Polymerization for the Continuous Production of Nitrile Rubber. (Master), Waterloo University, 2008.
[3] Poehlein, G. W., Emulsion polymerisation in continuous reactor systems. British Polymer Journal, 14(4), 153-158, 1982.
[4] Chern, C.-S., Principles and applications of emulsion polymerization: John Wiley & Sons, 2008.
[5] W, B. R.,private communication, 1994.
[6] Meehan, E., Variation of Monomer Pressure with Degree of Conversion in Emulsion Polymerization of Butadiene and of Butadiene--Styrene. Journal of the American Chemical Society, 71(2), 628-633, 1949.
[7] Yeo, Y.-K., Kwon, T.-I., & Lee, K. H., An energy effective PID tuning method for the control of polybutadiene latex reactor based on closed-loop identification. Korean Journal of Chemical Engineering, 21, 935-941, 2004.
[8] Van Krevelen, D. W., & Te Nijenhuis, K., Properties of polymers: their correlation with chemical structure; their numerical estimation and prediction from additive group contributions: Elsevier, 2009.
[9] 西久保忠臣，高分子化學，高立圖書有限公司，2014年。
[10] Lovell, P. A., & Schork, F. J., Fundamentals of emulsion polymerization. Biomacromolecules, 21(11), 4396-4441, 2008.
[11] Herrera‐Ordonez, J., Olayo, R., & Carro, S., The kinetics of emulsion polymerization: some controversial aspects. Journal of Macromolecular Science, Part C: Polymer Reviews, 44(3), 207-229, 2004.
[12] 張世昌，循環類神經網路於連續式化工程序之動態模擬與製程導航系統開發，碩士論文，國立臺灣科技大學，2022年。
[13] Cheng, J. Y., & Mailund, T., Ancestral population genomics using coalescence hidden Markov models and heuristic optimisation algorithms. Computational Biology and Chemistry, 57, 80-92, 2015.
[14] API RP520, Sizing, Selection, and Installation of Pressure-Relieving Devices in Refineries, American Petroleum Institution, 2000.