Sistema de controle inteligente baseado em redes neurais artificiais aplicado ao processo de destilação extrativa.

Abstract

Extractive distillation processes are widely used in chemical industries for the separation of non-ideal mixtures, for example, in the dehydration of ethanol due to its great industrial interest because of its diverse applications, requiring that production adjust to different market demands, mainly about composition. In addition, strict product quality specifications and strict environmental regulations require that the control system has a high degree of performance. In addition, strict product quality specifications and strict environmental regulations require that the control system has high performance. Due to the difficulty of measuring composition continuously, the control in a distillation column is still a challenge, and a widely used alternative is the installation of temperature sensors to infer the concentration. However, for non-ideal multicomponent mixtures, with complex thermodynamic behavior, temperature is a weak indicator of composition. The main objective of this work is to develop alternatives for intelligent control, based on Artificial Neural Networks, capable of making the composition of the products in an extractive distillation process move towards new specifications and that, regardless of the disturbances, keep the output in the set point established. Control loops include conventional, advanced and model-based control, without the use of expensive, highmaintenance composition analyzers. The models were based on the use of Artificial Neural Networks (ANN), developed in the MATLAB® software, being necessary to use as much data as possible in order to build models that cover a wide range of operational conditions of the process, which were obtained with the help of Aspen Plus ™. An analysis in Aspen Plus Dynamics ™ showed that the intelligent control through the set points modification gifts controllers in the original instrumentation is able to cause disturbances in power do not affect the quality of the final product or through a simple command operator, the control system is able to use mathematical logic to modify the composition of the product to achieve the desired specification depending on production planning. In view of these characteristics, intelligent control, in relation to conventional control, presented better performance and flexibility for the proposed problem, with low oscillation and fast responses.