- 學位論文
利用Aspen Plus自動化伺服器與模擬退火演算法進行反應蒸餾及挾帶劑增強反應蒸餾之最佳化
Optimization of Reactive Distillation and Entrainer-Enhanced Reactive Distillation using the Aspen Plus Automation Server and a Simulated Annealing Algorithm
摘要
反應蒸餾是一個將分離和反應整合於同一個單元的嶄新程序，相較於傳統設計能大幅降低資本及能源開支。數十年來科學家們已經研究了反應蒸餾的方方面面，而各種確定性或隨機性方法被用於完成其最佳化。有鑒於反應蒸餾複雜的本質，模擬退火法被認為是能克服這種高度非線性程序的合適演算法。儘管它不能百分之百確保全域最佳解，但能夠大幅降低CPU執行時間並且避免落入區域最小值。 此時不少脂化反應的製程已被模擬退火法成功最佳化，然而塔內催化劑負重問題在最佳化過程中缺乏討論。每一板用於容納催化劑的體積可能無法吻合基於水力計算的體積，我們可以通過將此體積差異整合到演算法中來解決這個問題，避免過於繁複的迭代。此外，在模擬過程中也同時考慮壓降的變化，因為它同時影響汽液平衡和反應動力學。反應蒸餾最佳化面臨的另一個問題是如何確保不同設計變數下都能成功獲得收斂結果，因此在本文也提出完整策略。 最後，產出水和醇可能形成最小共沸物作為餾出物，導致反應物損失。一種解決方案是引入夾帶劑以形成新的共沸物。這種三元（或二元）共沸物將在傾析器中分成兩相，並產生含有少量醇類的水相產物，這就是我們所說的夾帶劑增強反應蒸餾(ERD)。其最佳化還未被任何隨機性演算法完成。我們將示範如何使用上述方法來實現夾帶劑增強反應蒸餾的最佳化。結果表明，夾帶劑的使用減少了反應物的損失並相應地給出了較低的年度總成本(TAC)，尤其當產物純度要求提升時，其優勢更為明顯。
並列摘要
Reactive distillation (RD) is an innovative process where separation and reaction are combined into one unit, which reduces the capital cost significantly compared to conventional designs. Researchers have investigated several aspects of RD for decades, and the optimization of RD is carried out by several deterministic and stochastic methods. Considering RD's complicated nature, it is reported that the simulated annealing algorithm is a suitable way to get over these highly nonlinear processes. The algorithm does not guarantee the global solution, but it significantly reduces CPU time and is relatively not vulnerable to being trapped in a local minimum. At the moment, some esterifications have been optimized successfully by the simulated annealing algorithm. However, the issue concerning the loading of catalyst lack discussion. The volume used to hold the catalyst may not match the volume based on the hydraulic calculation. We can solve the problem by integrating the discrepancy of the liquid holdup into the algorithm. Besides, the effect of pressure drop is implemented during the simulations because the change of pressure also affects both vapor-liquid equilibrium and chemical kinetic at the same time. Another problem facing RD optimization is how to ensure the convergence of each simulation with different design variables. A complete strategy to deal with the issues above is accordingly present here. Finally, the produced water and the alcohol may form a minimum azeotrope as a distillate, resulting in the loss of the reactant. One solution is the introduction of an entrainer to form a new azeotrope. This ternary (or binary) azeotrope splits into two phases in a decanter and creates an aqueous product with little alcohol, which is what we call entrainer-enhanced RD (ERD). Its optimization has not yet been carried out by stochastic methods. We demonstrated how to use the procedure mentioned to achieve the optimization of ERD. The result shows that the use of the entrainer decreases the reactant's loss and gives a lower TAC accordingly. Its advantages are apparent, especially when the product requirements are raised.