Abstract
The Independent Metering Valve-Controlled (IMVC) hydraulic cylinder system utilizes a twin spool structure to provide independent control of the load. This system overcomes the limitations of the coupling mechanical structure in traditional valve-controlled cylinder system with a single spool, while providing superior accuracy, flexibility, and energy efficiency. However, fault information representation is similar, and fault component identification is difficult for IMVC hydraulic cylinder system. This paper proposes a fault diagnosis method for IMVC hydraulic cylinder system which employs a deep neural network model utilizing 1DLCNN-ResNet to identify specific fault components via multi-sensor information fusion. The model captures global information using 1DLCNN and gains deeper insight with ResNet, enabling accurate diagnosis of detailed fault problems in specific components, such as pilot valves, main valves, displacement sensors, and hydraulic cylinder. A combination of simulation and experimentation was employed to discuss 16 detailed fault problems under different feature spaces. The results show that, detailed fault problems in specific component can be accurately diagnosed, particularly in the 9-dimensional feature space. The overall diagnostic accuracy of the system can reach 96.71%, and the detailed fault of one component can be effectively diagnosed which leads to a diagnostic coverage of 99%.
