在化學製程中，觸媒會隨著操作時間逐漸失活，因此需要對反應器的溫度
和壓力進行調整以維持產量，然而如果沒有對觸媒失活的定量預測和剩餘使用
壽命的估算，這些調整是經驗性且次優的。在本研究中，我們使用ASPEN Plus
模擬了一個酯化反應製程。新鮮原料和回收進料加壓進入恆溫觸媒固定床反應
器。反應器出料加壓後進入蒸餾塔將乙酸正丁酯與其他成分分離。乙酸正丁酯
從蒸餾塔底部排出。未反應的乙酸、正丁醇與水則是從蒸餾塔頂部離開該製程。
最近，ASPEN Plus™ 發布了一個AI 混合模型，其可以訓練一個神經網路模
型，使用模擬器中的操作自由度（獨立變量）預測未知係數（神經網路輸出），
從而使模擬結果與測量的傳感器變量（依賴變量）吻合。在本研究中，我們調
查了AI 混合模型預測觸媒失活的能力。
另一方面，使用神經網路形式的代理編碼器，將所有獨立變量和依賴變量
作為輸入，而將潛在變量（指數前因子𝑘0）作為神經網路輸出。這個代理編碼
器通過在由獨立變量和𝑘0 空間組成的多維空間中生成大量模擬數據來進行訓練。
通過這種方式，可以有效地預測觸媒衰減路徑。

In a chemical process, the catalyst gradually deactivates over operational time, prompting adjustments in reactor temperature and pressure to maintain yield. Without a quantitative prediction of catalyst deactivation and an estimate of remaining useful life, these adjustments are empirical and suboptimal. In this study, an esterification process was simulated using ASPEN Plus. The fresh feed and recycled stream are pumped into an isothermal catalytic fixed-bed reactor. The reactor effluent is pumped into a distillation column to separate BuAc from the other components. BuAc leaves the bottom of the distillation tower. The unreacted HAc and BuOH are separated with water at the top of the distillation tower.
Recently, ASPEN Plus™ released an AI Hybrid model that allows us to train a neural network model to predict unknown coefficients (NN output) in the simulator using operating degrees of freedom in the simulator (independent variables) so that simulation results match measured sensor variables (dependent variables). In this work, the ability of the AI Hybrid model to predict catalyst deactivation is
investigated.
On the other hand, a surrogate encoder in the form of a neural network is built with all independent and dependent variables as input and the latent parameter, the pre-exponential factor 𝑘0, as output. This surrogate encoder is trained by generating numerous Aspen simulation results within the multidimensional space defined by independent variables and the k0. In this way, the catalyst decay path can be properly modeled.

1. McCulloch, W. S., Pitts, W. (1943). A logical calculus of the ideas immanent in
nervous activity. The bulletin of mathematical biophysics, 5, 115-133.
2. Haykin, S. (2009). Neural networks and learning machines, 3/E. Pearson Education India.
3. Nair, V., & Hinton, G. E. (2010). Rectified linear units improve restricted boltzmann machines. In Proceedings of the 27th international conference on machine learning (ICML-10) (pp. 807-814).
4. Han, J., & Moraga, C. (1995, June). The influence of the sigmoid function parameters on the speed of backpropagation learning. In International workshop on artificial neural networks (pp. 195-201). Berlin, Heidelberg: Springer Berlin Heidelberg.
5. Kingma, D. P., & Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.
6. Cybenko, G. (1989). Approximation by superpositions of a sigmoidal function. Mathematics of control, signals and systems, 2(4), 303-314.
7. Arpornwichanop, A., Koomsup, K., Assabumrungrat, S. (2008). Hybrid reactive distillation systems for n-butyl acetate production from dilute acetic acid. Journal of industrial and engineering chemistry, 14(6), 796-803.
8. Fang, D., Wen, Z., Lu, M., Li, A., Ma, Y., Tao, Y., Jin, M. (2020). Metabolic and process engineering of Clostridium beijerinckii for butyl acetate production in one step. Journal of Agricultural and Food Chemistry, 68(35), 9475-9487.
9. Shen, Y., Zhao, F., Qiu, X., Zhang, H., Yao, D., Wang, S., Gao, J. (2020). Economic, thermodynamic, and environmental analysis and comparison of the synthesis process of butyl acetate. Industrial & Engineering Chemistry Research, 59(50), 21869-21881.
10. Al-Rabiah, A. A., Alqahtani, A. E., Al Darwish, R. K., Bin Naqyah, A. S. (2022). Novel process for butyl acetate production via membrane reactor: A comparative study with the conventional and reactive distillation processes. Processes, 10(9),1801.
11. Duan, G. Study on catalytic reactive distillation process for synthesis of butyl acetate. Master’s Thesis, Hebei University of Technology, 2009.
12. Kunin, R., Meitzner, E. A., Oline, J. A., Fisher, S. A., & Frisch, N. (1962). Characterization of amberlyst 15. macroreticular sulfonic acid cation exchange resin. Industrial & Engineering Chemistry Product Research and Development, 1(2), 140-144.
13. Klambauer, G., Unterthiner, T., Mayr, A., & Hochreiter, S. (2017). Self-normalizing neural networks. Advances in neural information processing systems,30.
14. Lyche, S. (2018). On deep learning and neural networks.
15. Polyak, B. T. (1964). Some methods of speeding up the convergence of iteration methods. Ussr computational mathematics and mathematical physics, 4(5), 1-17.
16. Huber, P. J. (1992). Robust estimation of a location parameter. In Breakthroughs in statistics: Methodology and distribution (pp. 492-518). New York, NY: Springer New York.
17. McKay, M. D., Beckman, R. J., & Conover, W. J. (2000). A comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics, 42(1), 55-61.
18. Stein, Michael. "Large sample properties of simulations using Latin hypercube sampling." Technometrics 29.2 (1987): 143-151.
19. Kingma, D. P., & Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.
20. Draper, N. R., Smith, H. (1998). Applied regression analysis (Vol. 326). John Wiley & Sons.