Optimal control of the composition of carbon dioxide in the absorption column packed by means of a model of artificial neural network

Abstract

The objective of this research work was perform the optimal control of the composition of carbon dioxide in the absorption column packed of the Laboratory of Unit Operations and Processes of the Faculty of Chemical Engineering by means of a model of artificial neural network. A model of artificial neural network with an input layer, a hidden layer and an output layer was developed, trained with the backpropagation algorithm and validated experimentally for a process of absorption with chemical reaction. In addition, following the same procedure was implemented the model of the neurocontroller (inverse model of the artificial neural network). The experimental work was carried out in an absorption column of pyrex glass of 1.15 meters in height and 8.5 centimeters of internal diameter, packed with glass Raschig rings of 20 millimetres and operates with a range in the gas flow of 12 to 16 L/min and a range in the liquid flow rate of 0.8 to 2.8 L/min. The controlled variable was the concentration of carbon dioxide out of the absorption column in the range 4 to 8 %V and the manipulated variable was the speed of input stream of the sodium hydroxide solution in 0.11 M in a range of 0.8 to 2.8 L/min. To contrast the hypothesis the neurocontroller was tested experimentally establishing different reference values (set points) and the optimal control of the absorption column proved to be feasible by providing adequate control action to achieve the desired reference values.