Conference Agenda

Variable-speed heat pumps have become more popular for residential buildings due to their higher energy efficiency and improved thermal comfort. For these high-tier products, real-time performance monitoring for fault detection and diagnosis (FDD) is necessary. However, the FDD toolkits usually require embedding additional sensors which may significantly increase the product cost. To reduce the sensor cost, the state-of-the-art virtual sensor-based FDD technology provides a possible solution. It only uses low-cost and non-invasive sensors and estimates the properties of a system by combining sensor measurements with the physics and characteristics of the system. This paper aims to extend the application of a previously developed virtual refrigerant charge (VRC) sensor to variable-speed heat pumps. To incorporate the influence of compressor speed on system thermodynamics, a stepwise procedure is proposed to select the best set of features for the extended VRC sensor. In this procedure, domain expertise and statistical tests are used to assess the importance of several physical quantities in the estimation of refrigerant charge. Also, multiple sets of features are compared and analyzed. Finally, the rigorous feature selection process showed that the subcooling temperature, condensing temperature, and refrigerant quality at the evaporator inlet are more important features than others. The performance of the modified VRC sensors is evaluated by experimental data collected from a residential variable-speed heat pump containing several operating conditions and compressor speeds. With the selected best set of features, the extended VRC sensor is capable of identifying the actual charge level with a mean absolute percentage error below 8%.

Suction superheat is an important monitored variable in vapor compression refrigeration cycles. Superheat is the difference between actual temperature and saturation temperature. Positive superheat is used to ensure that no liquid enters the compressor. A common approach of metering refrigerant is the use of a thermostatic expansion valve (TXV), which mechanically assesses the amount of superheat and modulates the metering valve to maintain a superheat setpoint. As electronic expansion valves become more common, alternative methods for measuring superheat are needed. In addition, many fault diagnostic methods use superheat as an input variable. The obvious method of determining superheat – measuring pressure and temperature on the suction line – has the drawbacks that pressure measurements are costly, and they provide a potential leakage point for refrigerant. To address this problem, we have developed a virtual superheat sensor (VSS) method that has two characteristics that are important for practical deployment. The first is that the method uses temperature measurements only. The second is that the method uses measurement locations that are all in the outdoor unit of a split system, meaning that it could be applied without access to the interior of the building, and without knowledge of the indoor coil properties.

This paper is focused on demonstrating the VSS. It reviews the principles of operation of the VSS, then demonstrates its application and performance in an air-conditioner that is operating correctly, and also with faults, including undercharge of refrigerant, non-condensable gas, liquid line restrictions, and reduced airflow rates. The air-conditioner was modeled, and also tested experimentally in a laboratory. The VSS makes use of surface temperature measurements from the hot gas and suction lines, and outdoor coil tube bends of an air-cooled air conditioning system. It involves an iterative calculation of the predicted superheat, and an assumption about compressor efficiency. The paper shows that the VSS is effective in determining suction superheat in real-world deployments, even when the compressor efficiency is unknown a priori.

Soft faults in rooftop units (RTUs) degrade the system performance, impacting equipment, economics, and the environment. Automated Fault Detection and Diagnosis (AFDD) using a data-driven approach showed promising results by fitting machine learning classifiers that can predict typical soft faults using suitable inputs. Among several other issues, a challenge for fault detection and diagnosis protocols is getting a reasonable trade-off between the number of false alarms and missed detections. In addition, the practical deployment of machine-learning fault classifiers will require the most practical set of feature inputs for rooftop units. Finally, practical machine-learning classifiers will need to be able to predict faults from a system different from those with which the classifiers were trained. To obtain a more generalizable fault classifier, this study proposes a machine-learning classifier that was trained by using simulation data for multiple rooftop units, with a limited set of input features, addressing the above practical challenges of application. The proposed classifier was tested using: (i) existing laboratory measurement data, and (ii) field data from a faulty RTU. The results are quite good, compared with other data-driven AFDD results. This indicates that the proposed classifier could be generalizable for diagnosing common soft faults from RTUs equipped with fixed orifice (FXO) metering devices. This study focused on FXO systems because of the availability of training data, but the results suggest that the method could be adapted to the more common TXV-equipped systems if training data is available. The paper shows the diagnostic accuracy in terms of false alarms, missed detections, and misdiagnoses, and describes some of the important methods required to achieve good accuracy with machine-learning based diagnostics, such as rebalancing the training dataset and selecting meaningful features.

Given their performance, heat pumps are a key technology to decarbonize the heating production in buildings. Their share in the residential heating market is increasing at a very fast pace, making preventive maintenance and, thus, automatic fault detection and diagnosis always more necessary, especially to maintain an optimum efficiency and lighten the load for installers-maintainers. Soft and progressive faults such as refrigerant leakages must be detected before they deteriorate the performance of the system or even damage the components. Early detection of leakages and real-time quantification of the refrigerant charge is therefore a major challenge.

As a result of literature review, it has become clear that refrigerant leakages only impact the performance of the system after a significant amount of the original charge has been lost. Besides, the quick evolution of heat pump technologies makes the systems more adaptable to the variations of the operating conditions, therefore tending to hide the impacts of a progressive fault. Several methods for detecting and, more rarely, assessing refrigerant leaks have been developed. However, most focus on cooling or air-conditioning devices, which also often lack up-to-date technologies such as variable-speed compressors and electronic expansion valves. However, some methods do stand out due to their interesting approaches and accurate results.

The study presented here is part of a global work dedicated to the development of a method to determine in real time the refrigerant charge of a residential heat pump. Previous work consisted in identifying two promising methods of charge assessment - Li & Braun’s virtual charge sensor and Haddad’s method relying on a digital twin - and then developing and validating a model of a heat pump filled with R134a for their numerical evaluation. The work presented in this article is dedicated to the extensive experimental evaluation of these two methods, carried out in climatic chambers. The system studied is an air-to-water heat pump filled with R32 and equipped with a variable-speed compressor and an electronic expansion valve. Two sets of tests were carried out in heating mode and for different levels of charge. The first consisted of stationary and imposed operating conditions, whereas the second one included scenario-based Hardware-In-the-Loop (HIL) dynamic conditions. A part of the data collected was used for the validation of the numerical model of the heat pump required for Haddad’s method. The other part was used for the evaluation of both methods. They provide rather accurate results for charge assessment with stationary conditions, but only Haddad’s method was able to predict the refrigerant charge when the system was submitted to dynamic operating conditions. Further experimentation will be carried out to confirm its potential in the quantification of a dynamic leakage.

From the viewpoint of global warming prevention, it is becoming increasingly important to detect and repair leaks of fluorinated refrigerants from refrigeration and air-conditioning equipment as quickly as possible. In order to solve this problem, a technology to detect and identify leaks quickly with fewer man-hours was studied, and a remote detection technology using infrared absorption of gas was focused on.
In this study, the absorption wavelengths of currently used HFC refrigerants were researched, and it was found that R-32, which has a relatively simple molecular structure, is an easy refrigerant to measure for reflective infrared laser absorption spectroscopy measurements. And, considering that R-32 and its mixed refrigerants are used around the world among various fluorocarbon refrigerants, a remote detection device for R-32 gas was developed and successfully detected.

Current commercial refrigeration systems use refrigerants with global warming potential (GWP) values ranging from 1250 to 4000. The emergence of low GWP alternatives (GWP <150) is expected to significantly reduce direct emissions in this sector, playing a crucial role in the ongoing electrification and decarbonization initiatives. However, many of these low GWP alternatives pose a flammability risk, necessitating robust sensing solutions to ensure the reliable and safe operation of the equipment. This paper examines various sensing mechanisms suitable for potential applications in systems that employ flammable refrigerants, specifically those designated as A2L class. It provides a summary of A2L refrigerants and their properties, followed by a comprehensive review of sensor classes, covering their working principles, features, advantages, and limitations. Additionally, the article delves into key performance characteristics such as accuracy, selectivity, sensitivity, dynamic characteristics, and durability, among other properties. The article discusses areas for improvement and suggests corresponding approaches for potential sensors in facilitating the successful adoption of flammable refrigerants. Finally, this paper presents the latest findings from experimental evaluation of 5 different sensing principles in detecting the composition variation as a result of various operational conditions. Reliability and sensitivity of the sensor in responding to shifts in true composition and the resultant LFL value is also discussed.