Anomaly detection of ladle driving force based on Temporal Convolutional Network

Abstract

The driving force of ladle is an important index to reflect the normal operation of ladle sliding nozzle. At present, a steel plant uses sliding nozzle to regulate the flow of molten steel, so that the molten steel in the ladle can be safely, stably and reliably injected into the tundish or vibration exciter. In order to diagnose the normal operation of the ladle sliding nozzle, the driving force of the ladle is collected to analyze whether the driving force is abnormal to diagnose the operation state of the ladle sliding nozzle, so as to avoid the occurrence of leakage and other accidents. Considering that the driving force of ladle has strong time-series data characteristics, this paper uses temporal convolution network to extract data features and detect anomalies. The detection effect has been tested in the actual production, with good detection effect.