Research On The Prediction Model Of Converter Steel-making Eendpoint Temperature

It is one extraordinary complicated high temperature, polyphase physical chemical process to steelmaking. By reasons of enchancement factor and operation factor interference in steelmaking proocess, for quite some time, the problem of process control and End-Hitting is always the difficulty in steelmaking. Therefore, it is necessary to establish the steelmaking endpoint temperature prediction model.This article will be based upon practicle steelmaking real production data, incorporate with the steelmaking technical process and the thermal balance, analyses the effects for each factor towards end point temperature in steelmaking, ascertained the controlled variables for the end point temperature,divides the operation data set under different conditions,and establishes the steelmaking end point temperature’s prediction model which is base on the Least Square Support Vector Machine algorithm.then chooses the suitable modeling kernel function type according to the prediction results,and analyses the influences of regularization parameter and kernel parameter on the model’s learning ability and generalization ability.To abtain the best regularization parameter and kernel parameter for building the endpoint temperature predictive model of converter steelmaking,researcher tries to optimize the two parameters through the cross validation technology, particle swarm optimization algorithm and traverse optimization method.The endpoint temperature predictive model of converter steelmaking based on the Least Square Support Vector Machine algorithm has inaccurate prediction results to the test samples,which contain same outliers.researcher sloves the unavoidable phenomenon by calculating the weigth of the model based on the Least Square Support Vector Machine algorithm.Every input variable’s data includes some outliers,we can calculate their membership degree belongs to its reasonable data region.,and work out the weigthes of models that based on Least Square Support Vector Machine algorithm and the mechanism of converter steelmaking.The error values of temperature between the model prediction and the actual measured in±15℃for more than80%. The results show that the predicted of liquid steel temperature and measured values with good agreement.