A Novel Fault Diagnosis System for Blast Furnace Based on Support Vector Machine Ensemble

Fault diagnosis plays an important role during the process of blast furnace ironmaking for producing safety. In this paper the important parameters of puddling process are selected as judgments criterions of fault diagnosis by analyzing the changes of these parameters. Support vector machine (SVM) is used to establish the fault diagnosis model for its suitable characters for fault classification. But the stability and accuracy of model based on single SVM could not meet the needs of practical ironmaking. Therefore, a SVM ensemble based on bagging is presented to establish a novel fault diagnosis system. The real-time producing data are collected in 5# blast furnace of a steel enterprise for training and testing the fault diagnosis models with single SVM and SVM ensemble. The experiments about the comparison between single SVM and SVM ensemble and about the SVM ensembles with different number of individual SVM are made. The experimental results demonstrate that the performance of novel fault diagnosis system based on SVM ensemble is better than the one based on single SVM, and the best fault diagnosis system that can meet the practical needs of ironmaking is found.