Research on the Blast Furnace Charge Position Tracking Based on Machine Learning Regression Model

Abstract

In this paper, the detection data of the mechanical probe and radar are analyzed. Combined with blast furnace position system, a real-time prediction method of blast furnace charge position based on the data fusion of mechanical probe and radar detection data is proposed. A piecewise linear regression model is established, combining the periodic characteristics of the charge position of the blast furnace, the prediction regression curve of the position of the blast furnace is obtained. Then, the regression curve of the blast furnace charge position and the current working condition parameters are expressed as input. Taking the regression statistics as the weight adjustment parameter, the training input sample is constructed, and using the cross validation k nearest neighbor algorithm (KNN), the position of eight point radar is predicted in real time, and the predictive weighting coefficient is obtained.