Hybrid Approach Based on Temporal Representation and Classification Techniques Used to Determine Unstable Conditions in a Blast Furnace

摘要

This paper discusses the analysis of differential pressure signals in a blast furnace stack, by a hybrid approach based on temporal representation of process trends and classification techniques. The objective is to determine whether these can be used to predict unstable conditions (slips). First, episode analysis is performed on each individual trend. Next, using the obtained episodes and variable magnitudes, the classification tool is trained to predict and detect the fault in a blast furnace. The proposed approach has been selected in this application, due to the best results obtained using the qualitative representations of process variables instead of only raw data.