Feature Extraction and Classification of Operational Data for Diagnosis of Hot Strip Mill Looper Control

In these days, mechanical systems are becoming more complex and highly automated. So, there exist wide variety of demands for reliable diagnostic technology. A reliable data analysis and quantitative diagnosis method of mechanical system is necessary for the purpose. In this paper a quantitative diagnosis method for looper height control system has been developed based on neural network technologies. The wavelet transformation is used for pre-processing to analyze characteristics of looper height control system. And, self organizing map neural network is used for the purpose of classification based on the pre-processed data. After that, the classified results are used for quantitative diagnosis in hierarchical neural network.