Spatial Feature Learning Based Gaussian Process Regression for Blast Furnace Raceway Temperature Prediction

The tuyere raceway refers to the area where the coke is burned and rotated under the action of high speed hot air in the blast furnace ironmaking process. It is the main place for carbon gasification in the hearth reaction. The adiabatic combustion temperature of the raceway is an important index for judging the combustion state of the furnace. It affects the isostatic depression of the burden at the upper part of the blast furnace, dripping of molten slag iron and the gas distribution in the hearth. And it plays a vital role in ensuring the continuous and stable smooth smelting of the whole blast furnace. Therefore, in order to study the combustion state in the tuyere raceway, a prediction model of the combustion temperature in the tuyere raceway based on image data and physical variables is proposed. First, we use the spatial geometric relationship of the image and physical variable information to learn their spatial feature matrix. Second applying the multi-kernel learning algorithm to combine them, using the combined heterogeneous spatial feature matrix as the covariance matrix of Gaussian process regression (GPR), and regressing to obtain the temperature prediction value. Finally, the model was verified by taking the data of a 2580m ³ blast furnace as an example. Experimental results show that the model has high prediction accuracy, can quickly and accurately predict the combustion temperature of the tuyere raceway, and provides a new method for predicting the temperature of the blast furnace raceway.