Online Current Signature Analysis Approach for Cavitation Assessment in Hydraulic Pumps

Non-intrusive load monitoring provides an efficient method for assessing the operational state of various equipment, including pumps, by identifying abnormal behaviour such as cavitation. However, Traditional electromechanical monitoring techniques often rely on fixed operating speeds or require expensive sensors, such as accelerometers, limiting their applicability in industrial environments where machines operate under constant and variable speed conditions. This study proposes a novel non-intrusive condition monitoring methodology capable of detecting cavitation under stable and fuctuating speed conditions. The approach is based on real-time electrical current signature feature extraction combined with a machine learning framework to assess the presence and severity of cavitation in mechanical components. The methodology has been validated through a practical case study involving heating, ventilation, and air conditioning (HVAC) water pumps. The results demonstrate the effectiveness of this approach in enhancing predictive maintenance strategies by enabling early cavitation detection, thereby reducing both energy inefficiencies and potential mechanical damage.