Introduction
With the increased use of hydrocarbon refrigerants such as R290 (propane) to respond to the need to reduce global warming emissions, there is a desire to use greater refrigerant charge sizes for application in larger systems. Many of the flammability risk mitigation concepts to enable larger charge quantities are dependent upon leak detection (e.g., Colbourne et al., 2012). For example, closing solenoid shut off valves or to initiate airflow for extract ventilation or dilution are included in safety standards (IEC 60335-2-40: 2022 and draft prEN 378: 2023). Conventional leak detection is by gas concentration sensors, although detection through sensing system parameters such as pressures and temperatures have already been developed for some types of systems. Ultrasonic leak detection has been commercialised for industrial process applications, but the technology is seldom discussed for refrigeration, air conditioning and heat pump (RACHP) equipment.
Gas concentration sensors are used extensively, such as within the process industries and also for larger refrigeration installations. For smaller, cost-sensitive RACHP equipment using flammable refrigerants they may not be practical. Cost for reliable sensors can be prohibitive, they usually require regular re-calibration (at 3 – 24-month intervals) and depending upon the equipment configuration may require more than one sensor to capture all potential leak points. The dependability of gas detection in general is also questionable, for example, according to Royle et al. (2007) and McGillivray and Hare (2008), fixed gas detection systems (for industrial gas applications) detect only 30 – 65% of flammable gas releases. Further, gas sensors are susceptible to poisoning and other contamination that can negatively affect their performance, particularly with the lower cost technologies.
Using system parameters for leak detection avoids several of these pitfalls; thermocouples are low cost and pressure transducers may be used on the refrigeration system anyway, transducer response does not drift appreciably with ageing and neither are vulnerable to contamination. The main drawback is the relatively long response time to a leakage event especially in compressor-on mode; for instance, previous work found a reliable indication of charge deficit can be in the order of minutes rather than seconds (Colbourne et al., 2013, 2016).
Ultrasonic transducers (functioning as receivers/sensors) for leak detection can overcome many of these drawbacks. They are low cost (often less than €1 for transducers that function at frequencies around 40 – 60 kHz). By comparison, infra-red gas sensors are in the order of tens to hundreds of Euros and catalytic sensors from €10 upwards, whilst components of sensor signal processing would be broadly similar, regardless of the sensing technology. Ultrasonic transducers are not susceptible to usual poisoning or contamination. Ageing and thus drifting occurs relatively slowly (over decades) and they are barely susceptible to water and corrosion damage, as with other such solid-state electronic components. A further advantage is that an ultrasonic signal (via a second transducer) can be used to regularly prove the function of the sensor/detection system, which is currently not available (based on consultation with several manufacturers) or impractical with gas sensors.
Ultrasound is classified as sound waves exceeding frequencies usually audible to humans, i.e., above 20 kHz and extending to MHz or GHz. Ultrasound detectors receive airborne soundwaves and transpose them into electrical signals. Due to the noise from a leak travelling at sonic velocity, acknowledgement of a potential leak is effectively instantaneous, as opposed to gas sensors that rely on significantly slower mass transfer processes for the gas to reach the sampling point.
Noise produced by flow exiting an orifice is quantified by the mechanical power that is converted due to the pressure drop and subsequent turbulence. Noise level is a function of the pressure difference, mass flow, inlet and outlet densities, hole size and the sonic velocity of the fluid (Singleton, 1999; Baumann and Singleton, 2008). A greater pressure difference, mass flow, outlet density and sonic velocity produces a higher sound pressure, as does a lower inlet density or smaller hole size (for the same mass flow).
Ultrasonic leak detection is commercialised for industrial process applications (Sizeland, 2014), production testing (Moon et al., 2009) as well as for handheld detectors targeted at the RACHP industry. Elaborate application has been described, for example, Wang et al. (2018) who propose an array of ultrasonic receivers and computations to pinpoint sources of leaks. A similar approach by Eret and Meskell (2012) report on a detection system using audible frequencies (< 20 kHz), again requiring computational treatment of the input noise. In general, off-the-shelf ultrasonic transducers are widely available and their characteristics well described in the literature, providing useful insights. For example, Wolstencroft and Neale (2008) made extensive observations regarding directionality, effect of distance from leaks, orifice shapes and so on. Both air pressure (altitude) and air humidity affect the rate of ultrasound attenuation (Jakevičius and Demčenko, 2008), which could pose implications in certain climates and altitudes. Conversely, a search using IIF-IIR Fridoc1 reveals a vast number of studies relating to gas sensor-based leak detection, yet there are no studies listed concerning acoustic methods for refrigerant leak detection.
The present study involves development and testing an experimental ultrasonic leak detector (ULD). Trials are described using a wall-mounted indoor unit (IDU) of a split air conditioner and for comparative purposes, experiments with gas sensors positioned within the same IDU. Details of design and construction of the experimental ULD are given along with performance objectives. The experimental ULD was fitted within the IDU and tested with various sensor positions and potential leak locations to identify simulated leak release mass flow rates that generate a detection response. Further consideration has been given to possible interference from various environmental noise sources.