Theory Of Genetic Algorithm And RBF Neural Network And Its Applications To The Supervisation Of The Atmospheric And Vacuum Towers

This paper aims at the modeling of complicated system and the plant object is the atmospheric and vacuum towers in Kelamayi petrochemical factory. Mainly it discusses the following problems: the SCADA of the atmospheric and vacuum towers; the RBF neural network structure learning approaches involving K-means class learning玦gorithm, orthogonal least squares learning algorithm and genetic learning algorithm; RBF multi-neural network based on generalized information entropy; software instruments of viscosity and flash point using different RBF neural networks and comparations.We use industry configuration software FIX to build the SCADA system of the atmospheric and vacuum towers in Kelamayi petrochemical factory, including the establishment of software instruments of viscosity about No.3 side line of the atmospheric tower and flash point No.2,3,4 side line of the vacuum tower, real-time and history graph, reports.In this paper, the learning algorithm of RBF neural network is the focus. The classic learning algorithm is K-means class algorithm and orthogonal least squares lgorithm.Also we adopt the Genetic algorithm which is a hot point in the area of Intelligent Science, and applied it to the learning of RBF network. To dealing with the features of the different working points when relating plant objects, a RBF multi-neural network based on generalized information entropy is studied and progrmmed .Make use of the simulating object and real plant objects , the ability and accuracy of different RBF neural network is tested and compared. The results show that different algorithm fits for different situation.