Neural Network Control Of The Distillation Column Operation

Based on the theory of Parallel Distributed Processing, the neural network control system has agreat capability of doing unsupervised learning and adapting to the system automatically. It hasrealized a predictive control over the nonlinear systems and the systems that have obvious lag.Apparently, it is superior to the traditional PID control process. Due to the obvious advantages it has,it is more and more widely used in the areas such as industrial production, agriculture, petroleumexploration, intelligent controlling and scientific experiments etc. The NN control method has beendeveloping rapidly in the field of industrial control.Based on the study on the application of NN in the automatic control in the refining column, thethesis analyzes the feedstock, the temperature changes on the sensitive points, reflux, overhead,bottoms and determines the major parameters that influence the effect of the refining column. Thethesis also studies the method of gaining sample data applied to train the NN. The thesis adopts theBP network based on its reliability. The model of BP network has been established;the structure ofthe network in the refining column, the inputting parameters and the objectivity has been defined. Thespan of learning, the times of training and the accuracy of training have been determined. The neuralnetwork has been trained based on the statistics gathered in the plant. Then more statistics collected atsite has been input into the trained neural network as simulated data. A comparison is made betweenthe simulated outlet and the practical outlet.The result indicates that the trained neural network can predicate the quality of the productsuccessfully whether it is inside the operation range or outside of it. Furthermore, it can make timelyadjustment on the parameters that can influence the quality of products. The stability of the systemhas been improved and the rates of wastes and energy consumption have been decreased. At the sametime, the introduction of NN into the refining column controlling system has alleviated the intentionof work and achieved good social effect. Thus, the prospect of NN in the automatic control area ofpetrochemical industry is prosperous.