Good vibes for heat pumps: Early Refrigerant Leakage Detection with Vibration Sensors

The rising demand for heat pumps has exposed bottlenecks in production, installation, and maintenance processes. With the growing number of heat pumps in operation, service requests are expected to increase due to wear. To prevent shifting bottlenecks from installation to maintenance, it is crucial to reduce maintenance efforts now. Predictive maintenance is a widely used method to detect abnormal system behavior before inefficiency or damage occurs, but this requires suitable measurement techniques and evaluation methods.

We developed a specialized heat pump test bench to investigate soft faults and identify effective detection methods for the refrigerant cycle, focusing primarily on refrigerant leakage. Our research explores how refrigerant charge levels affect the performance of a receiver-based air-to-water heat pump, aiming to determine the earliest point at which leakage can be detected. Traditional detection methods, which monitor system pressures and temperatures, are accessible but fail to detect early leakage due to phase separation in the receiver. Although recording the receiver level can provide early detection, this method is costly and requires detailed operational knowledge.

To offer a more cost-effective solution, we propose non-invasive vibration sensors. For our proof-of-concept, vibration sensors were installed on the heat pump’s receiver, and data was recorded under controlled conditions. The experiment was conducted at the operating point A10W50, with superheat set to 10 K and the compressor running at 3000 rpm (50 Hz). After charging the system with 3000 g of propane, the refrigerant charge was gradually reduced.

Traditional methods could only detect leakage when the expansion valve's control variable deviated at approximately 1100 g. However, analysis of the spectral power density of recorded vibrations showed that the receiver’s resonance frequencies and amplitudes vary depending on the refrigerant charge. We demonstrated that a deviation from the optimal charge of 2000 g could be detected as early as 1500 g using vibration sensors.

This method provides an earlier, more cost-effective approach to detecting refrigerant leakage, enhancing predictive maintenance strategies and allowing technicians to diagnose potential issues sooner — bringing “good vibes” to heat pump maintenance.