Multicriteria optimization of Ethylene Dichloride (EDC) Dehydration Process with Use of Neural Networks

The article raises the issue of the determination of optimum values of technological parameters, necessary for the development of an effective control system of a technological process, on the basis of the experimental data collected when carrying out such a technological process for receiving a ready-made product of a required quality-1.2 dichloroethane with the moisture content of no more than 0.001%. For the solution of the problem, the following tasks were set: to find the optimum values of the technological process parameters and to develop the operating program for the programmable logical controller which will provide for a necessary quality of a product of the technological process. As a criterion function for finding the optimum parameters of the moisture content of the dehydrated dichloroethane, the temperature and a pressure cubed column are selected as changeable parameters. Multicriteria optimization is performed with the use of a combination of neural network and the methods of minimization on the correct simplex and minimization of the Nelder-Mid on correct triangle and on basic vectors where neural network is applied to the description of a criterion function. As a result, after the calculation the values of technological parameters are received in three ways, and those values at which the quality of the received product corresponds to required quality are chosen. The correctness of the received values is checked with the use of the map of lines of the criterion function level. Also, with the use of the calculated parameters of a technological process, the operating program for the programmable logical controller, thanks to which the values of technological parameters are achieved and maintained so that the obtained ready-made product of the studied technological process - dehydrated dichloroethane- had the quality necessary for its further use in production, was developed.