Data-Driven Characterization of Leakage Faults in Hydraulic Cylinders for Sustainable Maintenance Planning

This study presents a cost-effective approach for detecting internal leakage faults in hydraulic cylinders by leveraging features extracted from existing control signals, specifically the PID valve input. The key innovation lies in eliminating the need for additional sensors or hardware modifications, making the method suitable for low-cost real-time implementation. Several low-complexity, time-domain features were identified and extracted from the control signal, which reflect changes in system behavior due to internal leakage. These features are designed for edge computing platforms, enabling practical deployment in industrial environments. The proposed method is particularly applicable to systems with known loads and consistent duty cycles, such as hydraulic presses, where deviations in control signal behavior can reliably indicate leakage. However, limitations arise when applied to systems with stochastic or highly variable loading, such as mobile machinery, where external disturbances can obscure leakage effects. This approach enables early fault detection and condition monitoring in hydraulic systems without increasing system complexity or cost. It provides a foundation for predictive maintenance strategies in both stationary and mobile hydraulic equipment, contributing to improved reliability and reduced downtime.