The use of dividing-wall columns applied to extractive distillation has been assessed regarding many factors, from design to the best control strategies. In this study, it is proposed an intelligent control system, based on artificial neural networks, to reject disturbances in an extractive distillation dividing-wall column used for anhydrous ethanol production with ethylene glycol as solvent. Disturbances to the azeotropic mixture feed stream (flow rate and composition) are applied in order to compare the performance of the conventional control system to that of the intelligent control system; the latter accounting for the self-adjusting split ratio between the vapor flow rates directed to each side of the dividing-wall. The results showed the capability of the intelligent control system to reject all sources of disturbances applied to the process, maintaining low offsets to product specifications with respect to their nominal values (99.5 %). Moreover, the intelligent control system provided a decrease in reboiler steam consumption when compared to the conventional control system, indicating to be a better strategy, both dynamic and energy-wise. The strategy of changing controller setpoints by using an artificial neural network is easy to implement, does not require investment to acquire new instrumentation, and minimizes human intervention in the process.

从设计到最佳控制策略，已经对许多因素对分隔壁塔用于萃取蒸馏的使用进行了评估。在这项研究中，提出了一种基于人工神经网络的智能控制系统，以排除在以乙二醇为溶剂的无水乙醇生产中使用的萃取蒸馏分隔壁塔中的干扰。为了将常规控制系统的性能与智能控制系统的性能进行比较，需要对共沸混合物的进料流（流速和组成）进行干扰。后者考虑了导向分隔壁每一侧的蒸汽流量之间的自调节分配比。结果表明，智能控制系统具有抑制过程中所有干扰源的能力，相对于其标称值（99.5％）保持较低的产品规格偏差。此外，与传统控制系统相比，智能控制系统减少了再沸器的蒸汽消耗，这表明在动态和能量方面都是更好的策略。通过使用人工神经网络来更改控制器设定点的策略易于实施，不需要投资来购买新的仪器，并且将过程中的人工干预降至最低。与传统控制系统相比，该智能控制系统减少了再沸器的蒸汽消耗，这表明在动态和能源方面都是更好的策略。通过使用人工神经网络来更改控制器设定点的策略易于实施，不需要投资来购买新的仪器，并且将过程中的人工干预降至最低。与传统控制系统相比，该智能控制系统减少了再沸器的蒸汽消耗，这表明在动态和能源方面都是更好的策略。通过使用人工神经网络来更改控制器设定点的策略易于实施，不需要投资来购买新的仪器，并且将过程中的人工干预降至最低。