In order to maintain stable operation of a blast furnace, deviations from the steady state should be quickly detected and corrected. Because internal physical states are difficult to measure directly, experienced operators play a crucial role in integrating information from various sources such as sensors and visual observations so as to recognize, predict, and react to probable anomalies. Overlooking undesirable symptoms or delayed actions possibly leads to the excessive loss of heat, which, in the worst case, can cause the abnormal shutdown of operation. One of the events highly associated with the risk of heat loss is gas channeling, which sets off an alarm signaling to the operator to decrease the flow rate of input hot gas. Gas channeling is judged to be imminent by the integration of indices designed so as to detect unusual changes in the variations and patterns of sensory data. The precision of our prediction method is evaluated by using actual data.
展开▼