engleski

Optimal Control Of The Batch Rectification Process Using Artificial Neural Network

In pharmaceutical and fine chemicals industry, process of batch rectification is very common and there is a constant interest in energy and material savings trough the optimization of the process control method. From a industrial standpoint most interesting optimal control policy is rectification with the constant distillate composition. This problem represent a optimal control problem that can be very difficult to solve if mixtures that must be separated exhibits highly nonideal behavior, and therefore complex vapor liquid equilibria model must be used. In this work, optimal control method for a batch rectification process is developed based on the artificial neural net. Method is based on the fact that optimal control problem can be transformed in the nonlinear programming problem by discretization of the integral criteria and differential constraints equations. Specially designed neural network is used for solving the nonlinear programming problem. Dynamic augmented Lagrange multiplier method is applied for minimization. Method is tested on the computer simulations and on batch rectification experiments carried out on the laboratory rectification column and performed tests proved that calculated reflux profiles enabled keeping of the desired distillate composition during the rectification of the selected test mixtures.

optimal control; batch rectification process; neural networks

nije evidentirano