Research Of Reheating Furnace Modeling Based On Genetic Algorithm And Artificial Neural Network

The reheating furnace is a complex industrial object.The traditional method for its modeling is often inaccurate,and the effect is not satisfactory.With the vigorous development of intelligent technology in recent years,the method of modelling by means of neural network,especially back propagation(BP)neural network,is becoming increasingly widespread in industrial process.As a kind of evolutionary computation,genetic algorithm(GA)can overcome the disadvantage that BP neural network can easily fall into local minimum because of its global searching ability.Therefore,it is necessary to study the combination of BP neural network and genetic algorithm(GA).A solution of reheating furnace modeling based on historical data with neural network is introduced in this dissertation.Firstly,the structure of neural network is determined and the number of the model parameters in BP neural network is obtained.Then the genetic algorithm is used to select,cross and mutate the survived individuals repeatedly to get the best individual which contains the weights and thresholds of the BP neural network,and its decoding is taken as the initial value of the parameters of the BP neural network.Finally,the neural network model of the reheating furnace is obtained by the training of BP algorithm.The main contents and contributions of this dissertation are as follows:(1)The validity of neural network model of the furnace temperature in term of characteristics of its input and output is studied,the problem that the structure of neural network is difficult to determine can be solved by comparing the chosen effectiveness index.(2)The genetic algorithm is suggested to be applied to the reheating furnace modeling in this dissertation.By the advantage of its global search ability in large scale,the weight parameters of the BP neural network model are optimized to avoid falling into local minimum and to accelerate its convergence speed..the model acquired using this method has stronger learning ability and better prediction accuracy.(3)In order to accelerate the convergence speed of genetic algorithm and improve its efficiency,some parameters can be redesigned to be adaptive.In this dissertation,a genetic algorithm with adaptive mutation probability is designed and implemented in the neural network model of the reheating furnace to accelerate the convergence speed of model parameters.(4)The experiments are conducted to acquire the model of reheating furnace with its real dataset generated in a steel rolling mill of Tangshan,verifying the feasibility and effectiveness of the proposed method by comparison.It is supported in the results that neural network can be used to build the model of reheating furnace effectively and efficiently.