마할라노비스거리를 이용한 군집기반 열간 조압연설비 상태모니터링과 진단

Abstract

Steel industry is faced with cost reduction pressures due to intense competition in the market. In particular, the steel industry is a representative process industry, and it is essential for cost reduction to produce steel products without unscheduled down. The hot strip roughing mill consists of lots of mechanical and electrical units. In fault diagnosis, various and complex units could be failed with unknown reasons. In this paper, we propose an clustering based fault detection method using mahalabnobis distance to figure out early the units with abnormal status to minimize the system downtime. K-means and PAM (partition around medoids) algorithm with euclidean (ED) and mahalanobis distance (MD) are used to detect the abnormal status. We have proposed a clustering based fault detection algorithm using MD considering the correlation between variables. We evaluate the performance of PAM algorithm using MD through actual field data.