Early detection and classification of internal leakage in boom actuator of mobile hydraulic machines using SVM

Abstract

Mobile hydraulic machines are used in various operations like construction, material handling and mining. High powers to weight ratio and manoeuvrability in rough terrains are striking features that help in edging out their electrical and mechanical counterparts. Internal leakage in hydraulic actuators is a faulty condition frequently observed in hydraulic machines. Internal leakage in the actuator affects the system’s dynamic performance and decreases its energy efficiency. Also, internal leakage is apparent only when the leakage is extreme and the actuator stops responding to command signals. Thus, detecting internal leakage in its early stages is a difficult task. Early detection and corrective action save energy, reduce component degradation, and reduce machine downtime. There are many existing techniques for internal leakage detection of hydraulic actuators, but they are intended for actuators working in a laboratory environment. The main focus of this paper is to present a practical method for early detection of internal leakage fault present in boom actuator of mobile hydraulic machines by analysing the machine work-cycle data with minimum hardware. The method trains and validates a Support Vector Machine (SVM) classifier using pressure and boom angle displacement signals. The time-series signals are processed using ’event-based’ feature extraction method. The binary version of Particle Swarm Optimisation is used for feature selection. The trained classifier can detect and classify internal leakage faults with more than 95% accuracy, which is sufficient for taking appropriate preventive maintenance steps on time.