Application of PCA based process monitoring method to ironmaking process

It is quite challenging to monitor an ironmaking process because of its special characteristics such as frequent fluctuations and lack of direct measurements. To tackle these issues, a two-stage PCA based monitoring method was proposed in our previous work. However, only one type of operating anomaly was considered and the historical data of one accident was utilized. To further evaluate the performance of the two-stage PCA based method, four different anomaly types and 25 corresponding historical datasets collected from three real blast furnaces are tested in this paper. The results demonstrate good potential of our proposed method for anomaly detection in ironmaking process.