A Data-Driven Approach for Identifying Air Volume Variations in the Blast Furnace

Air volume regulation serves as a fundamental lower-zone control strategy in the blast furnace smelting process, playing a critical role in stabilizing furnace operation and managing the optimal thermal state. In most blast furnace operations, air volume control is predominantly reliant on manual expertise due to the absence of a robust, logically structured regulatory framework and a comprehensive control system. From the perspective of intelligent regulation, it is essential to first accurately determine the air volume status and excavate the interrelationships among regulatory parameters, thereby providing a foundation for feedback and iterative updates to the knowledge system. Consequently, acquiring precise information regarding whether the air volume has been augmented or reduced constitutes a critical prerequisite for achieving intelligent control of the blast furnace. Based on the characteristics of air volume data, this paper proposes an optimized algorithm for identifying air volume augmentation and reduction states. The method employs first-order lag filtering to process raw air volume data, utilizes the average difference of multiple elements as the characteristic parameter, and determines the threshold of this parameter through empirical methods. This data processing algorithm is established as the optimal approach for recognizing air volume adjustment states. The proposed algorithm was implemented on a 1750 m³ blast furnace, using 201,600 sets of air volume data for training and 892,800 sets for testing. Offline validation of the algorithm on the test set achieved a correct recognition rate exceeding 90 pct. Online verification analysis demonstrated that the method performs with high efficiency and accuracy in real-time air volume status identification.