Fault diagnosis of pressure relief valve based on improved deep Residual Shrinking Network

Pressure relief valves are widely used in hydraulic systems but often fail when operating in harsh environments. Developing effective and robust fault diagnosis models is essential for ensuring the reliable operation of hydraulic systems. In that study, we proposed a novel method for diagnosing clamping faults of pressure relief valve and addressing catastrophic forgetting under deep learning multi-task scenarios. Specifically, we combined the Elastic Weight Consolidation (EWC) algorithm with Deep Residual Shrinkage Network (DRSN) to achieve high diagnostic accuracy. Multiple time-series samples of pressure relief valve were collected and tested under hydraulic media with different pollutant particles. The proposed model achieves an average accuracy of 98.8% and an average loss of 0.095, which proves its effectiveness in the fault diagnosis of pressure relief valve. The main contribution of this study is the effective integration of the EWC algorithm with the DRSN, which provides an effective solution for the fault diagnosis of the pressure relief valve.

Automated summary

Pressure relief valves are commonly used in hydraulic systems but can often fail in harsh environments. This study proposes a novel method that combines the Elastic Weight Consolidation algorithm with the Deep Residual Shrinkage Network for diagnosing clamping faults of pressure relief valves and addressing catastrophic forgetting in deep learning multi-task scenarios. Experimental results show that the proposed method achieves high diagnostic accuracy in the fault diagnosis of pressure relief valves.