| Abstrakt: |
The dilemma for any model-based controller is its process model's development effort and accuracy. Thus, selecting an appropriate modeling technique for the model-based controller is noteworthy to achieve the best control performance. In this work, a block-oriented modeling technique, namely the neural wiener (NW) model, is developed for the Ethylene Glycol (EG) plug flow hydrogenation reactor. The NW model consists of a linear block using state space (SS) model and a nonlinear element using a static neural network model. Presently, no prior research has been carried out previously using the NW model for EG production. The EG hydrogenation reactor is modeled using Aspen Plus for steady state and Aspen Dynamic for dynamic simulation. For modeling purposes, the selected process inputs are jacket flow rate and hydrogen feed flow rate. Meanwhile, the EG output mole fraction and product temperature are chosen for the process outputs. The performance of the developed NW model is evaluated using a validation dataset, and a state space (SS) model is used as a comparison. Based on the model validation results, the NW model has achieved R2 (coefficient of determination) of 0.97 compared to the SS model with R2 of 0.78 for Output 1(EG mole fraction). For Output 2 (Product temperature), the NW model has produced R2 of 0.94 in comparison with the SS model with R2 of 0.69. Thus, the NW model has outperformed the SS model in modeling the EG hydrogenation reactor. [ABSTRACT FROM AUTHOR] |
