engleski

Generic Neural Model Control of Batch Distillation Process

The batch distillation process is common in production of pharmaceuticals, fine chemical products and similar industries. Purpose of this work is to develop a method for distillate composition control during the batch distillation process by use of generic model control method applied on previously developed dynamic neural model. As a neural model of the process, externally recurrent continuous dynamic neural net (also called state space neural net) was chosen. Net was trained with experimental data obtained from sets of batch distillation runs performed on laboratory column at several different constant reflux ratios with acetone water mixture. Using the generic model control concept, state space neural net is used for calculation of appropriate control action during the process, namely reflux ratio. Developed method was tested on the same laboratory column used for collection of experimental data. Results showed good performance on task of constant composition control.

batch distillation; neural net; generic model control

nije evidentirano