Modeling of Chemical Plant´s Rectifying Towers Using Artificial Neural Networks

An artificial neural network based modeling method of a chemical plant´s rectifying towers is presented in this paper. There are many approaches on chemical plant modeling. Some of them use neural networks to model some part of chemical plants or processes. This paper also tries to model a component of chemical plants. Standard multilayer perceptron (MLP) and back-propagation (BP) learning algorithm are used in this study. Using actual data obtained from real operation of rectifying towers MLP is trained at first and then tested for real data not used for training. Experimental results for two O₂ production increase cases, 3000nm³/h and 5000nm³/h, NN based modeling shows that the model mimics well actual rectifying towers. In the experiments, 22 inputs are selected as inputs and 5 outputs are selected as outs to model rectifying towers.