Design and Application of a Blast Furnace Intelligent Diagnosis System Based on Mechanism Models and Knowledge Reasoning

Abstract

In blast furnace operation, abnormal conditions are usually judged from pressure drop, probe movement, gas-flow distribution and thermal indicators, but these variables have different time lags and may be interpreted differently across shifts. To improve operation-level diagnosis, an intelligent diagnosis system for Bensteel blast furnaces was developed by integrating mechanism models with knowledge reasoning. The system forms a closed diagnostic chain covering multi-source data access, data cleaning, time-scale alignment, health evaluation, rule matching, cause tracing and operation recommendation. Process data from the furnace body, furnace top, blast system, raw-material and laboratory systems, hot stoves and historical records are integrated, and mechanism modules are linked through an open inference engine, rule base and process knowledge base. Application on Bensteel No.5 blast furnace showed that the system identified suppressed edge gas flow and a tight pressure-permeability relationship during edge adjustment. After the coke charging matrix was changed from 33321 to 23321, the ore-to-coke ratio (O/C ratio) of Ring 3 decreased from 4.27 to about 2.99, and the permeability index increased from about 0.003 to about 0.004. The pressure-drop model provided a one-hour (1 h)-ahead warning with an online hit rate above 90%.