Research on Fault Diagnosis Model Method of AGC Hydraulic System Based on Transfer Learning

The research focuses on the AGC hydraulic system of a cold rolling mill in a certain steel plant.Abstract-We propose a model that combines transfer learning and CNN for fault diagnosis of AGC hydraulic systems. This model effectively solves the diagnostic difficulties caused by data scarcity. By using simulated fault data as the source domain and real fault data as the target domain, we utilize Hilbert to reduce data differences. Subsequently, CNN learned the fault features in the source domain, and then fine tuned the target domain data to construct a transfer learning fault diagnosis model, which was successfully applied to the fault diagnosis of AGC hydraulic systems.