Detection of Blast Furnace Stockline Based on a Spatial–Temporal Characteristic Cooperative Method

Abstract

The accuracy and continuity of stockline detection in blast furnaces (BFs) are useful for improving the burden shape and material charging efficiency. However, conventional mechanical stock rods and radar sensors exhibit problems, such as weak anti-interference ability, large fluctuations in accuracy, poor stability, and discontinuity. Therefore, a spatial–temporal characteristic cooperative method for stockline detection is proposed. The temporal dimensional periodic variation feature and spatial dimensional motion feature of the BF stockline are extracted using a time series joint partitioning method and a piecewise nonlinear polynomial regression model, respectively. Subsequently, by considering the mechanical detection data as the standard, combined with a sliding window operation, a variable structure self-renewing fusion neural network based on a fuzzy clustering with an uncertain membership degree algorithm is developed for the data cooperative fusion of the mechanical stock rod and radar sensor. Real-time continuous stockline information can be accurately obtained. Both the simulation results and industrial validation indicate that the proposed method can provide real-time and effective stockline information and has practical value for industrial production.