Hybrid Soft-Sensing Model of Tuyere Raceway Temperature Based on Improved RBF Network

Blast furnace is an important unit in iron-steel making. The temperature in tuyere raceway of blast furnace has been a primary concern due to that it plays a crucial role in meeting the production requirements of high product quality and low energy consumption. In this paper, an improved Radial Basis Function (RBF) neural network is proposed to predict the temperature in tuyere raceway of blast furnace. The proposed method integrates the MeanShift-Kmeans algorithm and the immune algorithm to improve its prediction accuracy. The MeanShift-Kmeans is introduced to solve the center vector of radial basis function, in this way, the initial center is determined in accordance with the density distribution of data, reducing the instability of clustering results and further enhancing the prediction accuracy of the temperature in tuyere raceway. We adopt the immune algorithm to optimize the network hyperparameters, effectively avoiding the generation of local optimal solutions. Considering that there is available image data in the tuyere raceway, we establish the soft-sensing model of temperature based on physical data and image data to enhance the prediction accuracy. This paper uses the data of a 2500m
³
blast furnace to conduct experiments, and the results show that the proposed method is effective.