Analysis of Split-System Air Conditioner Faults through Electrical Measurement Data

Abstract

This work presents an electrical measurement dataset from a split-system air conditioner in normal operating conditions and with specific faults, such as incrustation in the condenser and evaporator air inlet with different levels of blocking, which often occurs in this type of equipment. We also added compressor capacitor degradation, which is a very common fault in this type of equipment, although it is scarcely addressed in research. The data were obtained through a non-invasive current sensor and a grain-oriented voltage sensor containing the values of the current and voltage of equipment that was installed in the field and tested at different levels for these fault conditions. This work not only explains how the entire data collection process was carried out but also presents two examples of fast Fourier transform (FFT) applications for the detection and diagnosis of faults through the electrical measurements analyzed in our studies, which had good effectiveness.

Dataset: The dataset has been submitted as a Supplementary File.

Dataset License: CC-BY-NC

1. Summary

The fault detection and diagnosis (FDD) in heating, ventilation, and air conditioning (HVAC) equipment is a subject studied in several research works for large and medium-sized applications, such as industrial and commercial buildings. We can cite research [1] in which the faults of a chiller were analyzed using temperature and pressure measurements, the research in [2] where the AHU faults were analyzed using temperature measurements, and the research in [3] where the faults of a multi-chiller plant with air-handling units (AHUs) and a cooling tower were analyzed through various types of measurements, mainly temperature, as examples.

The measurements performed in this equipment generally occur through invasive thermodynamic or manometric means, most of the time utilizing a large number of sensors for data acquisition, which increases the price of the system [4,5]. However, the application of these systems is impractical for small-sized equipment from a financial point of view, such as the split-system air conditioner, which is widely used in residential and small office applications [6]. In addition to the cost issue, another disadvantage to be considered in the application of these FDD systems is that invasive methods cause a need to shut down the equipment [7] for the insertion, adjustment, and calibration of sensors.

Because of the difficulty in carrying out laboratory or field studies due to resource and time constraints, some researchers have opted for the acquisition of data through simulation software for equipment operation. This was carried out in [2,8], in which the data were generated using the building energy simulator EnergyPlus, and in [9], where a simulation algorithm was used for this purpose. Despite the importance of this type of study, it is well known that this form of data acquisition will hardly correctly represent the operation of real equipment faithfully [10].

There are datasets made available by the authors of research where data on HVAC equipment faults were found, such as in [11] for different types of large-sized equipment with various datasets containing simulated measurements of temperature, pressure, air and water flows, among others, generated by the EnergyPlus and HVACSIM+ software; in [10], the data were on industrial AHU installations, with a large volume of data being obtained from temperature, air-flow, and power sensors; in the laboratory research in [12], the data were on heat pumps and air conditioners with various types of temperature, pressure, airflow, and refrigeration capacity measurements; in [13], the data were on field-installed rooftop units (RTUs) with temperature, humidity, pressure, and power data.

The possibility of FDD in a split-system air conditioner through non-invasive short-duration electrical measurements is verified in our studies, and it was not found in other publications. The current and voltage data in fault conditions were also not found in other datasets for this type of equipment. Another novelty in this dataset is the data from an electrical fault, the degradation of the compressor capacitor. This is a fault that is frequently found in the corrective maintenance of this type of air conditioner, which causes the loss of equipment refrigeration capacity and has not yet been studied in other works.

The other faults in this dataset were collected in various situations of incrustation in an air conditioner’s air inlet. In the case of the evaporator unit, these faults block the entry of air to be conditioned by the equipment, which may result in a drop in its energy performance. However, in the case of the condenser unit, these faults increase the value of the electrical current, causing an increase in the electricity consumption of the equipment [14], in addition to increasing its operating noise.

The data were collected for these fault conditions and for the faultless operation of a split-system air conditioner with 5272 Watts (18,000 Btu) that was field-installed for refrigeration at the office of the Research Group on Instrumentation and Control in Energy and Environment Study (GPICEEMA) at the Federal University of Paraiba (UFPB) in Joao Pessoa, Brazil. Data acquisition was repeated three consecutive times to then be validated and made available for the dataset. The files are separated by each operational condition analyzed and made available in the Supplementary Materials.

We chose to apply the dataset in a method used to analyze electrical signals, the fast Fourier transform (FFT), to better demonstrate to readers one of its possible uses.

This article provides all the necessary information about the use of this dataset, the procedures executed for data acquisition and processing, the demonstrations of the proposed method, and the results obtained.

Data Value

• This dataset contains current and voltage data from a split-system air conditioner from a real field installation.
• The data present frequent faults at different levels of occurrence of a split-system air conditioner in real operation, such as incrustation in the condenser’s air inlet, incrustation in the evaporator’s air filter, and electrical degradation of the compressor’s capacitor.
• The dataset can be used for the purpose of improving and developing electrical FDD techniques or studies of energy efficiency by the academic community and refrigeration professionals.

2. Data Description

The dataset has two types of information for each operating condition of the analyzed equipment: electrical current, in Amperes (A), and voltage, in Volts (V). The operating conditions include the faultless condition, single faults, and double faults, which are composed of two single faults in combination. The files are in txt format so that they can be utilized in some types of programming languages or be converted into the csv format, which would enable them to be used by software such as Matlab, Python, and R, among others. Each file has a size of 32 MB.

Table 1. Header and initial data of the CI60 fault dataset file.

Dataset_CI60
Current	Voltage
2.142	−20.050
2.226	−23.435
2.306	−28.089
2.384	−31.520
2.461	−35.915
2.539	−39.840

shows the dataset header, where the first column of the table (Current) contains the data of the electric current acquired in a temporal series for an interval of 60 s, with a sampling rate of 30,000 Hz/s, adding up to 1,800,000 samples. The voltage data are laid out in a similar manner, although in the second column (Voltage) of the table.

The files are separated by each operating condition using the nomenclatures presented in

Table 2. File naming for each analyzed operation of the air conditioner.

File	Operation Condition	%	Implementation Method
FL	faultless	0	not necessary
CI30	condenser incrustation	30	partial blocking of the condenser’s air inlet
CI60	condenser incrustation	60
FI50	air filter incrustation	50	partial blocking of the air filter’s inlet
CD30	compressor capacitor degradation	30	replacement of the compressor’s capacitor
CD30CI30	compressor capacitor degradation and condenser incrustation	30 30	replacement of the compressor’s capacitor and partial blocking of the condenser’s air inlet
CD30CI60	compressor capacitor degradation and condenser incrustation	30 60
CD30FI50	compressor capacitor degradation and air filter incrustation	30 50	replacement of the compressor’s capacitor and partial blocking of the air filter’s inlet
CI60FI50	condenser incrustation and air filter incrustation	60 50	partial blocking of the condenser’s air inlet and air filter’s inlet

. The File column contains the nomenclature of the dataset file for each specific operating condition, which is composed of the characters of the abbreviated name and characteristics of this condition. The Operation Condition column describes the equipment's operational condition, which may be faultless or involve condenser incrustation, air filter incrustation, compressor capacitor degradation, or a combination of two of these faults. The % column has the blocking percentage in the unit’s air inlet for the incrustation fault case, while for the capacitor degradation fault, it has the percentage loss of electrical capacitance of the component. The Implementation Method explains how faults were implemented.

For incrustation faults in the condenser unit, two different percentages of air inlet obstruction were tested, 30% and 60%, while the evaporator’s air filter was obstructed by 50% of its area. The definition of analyzing the different levels of incrustation in the occurrence of faults is based on the principle that the accumulation of dirt in these components can occur in different ways depending on external weather conditions and the lack of cleaning activity of this equipment during preventive maintenance. To generate these conditions, a cardboard plate was used to block the proportional area of the air inlet for the fault to be analyzed (Figure 1).

For the compressor capacitor degradation fault, the loss of component capacitance was tested by replacing the original 35 µF capacitor with a 25 µF one, thus generating approximately 30% lower capacitance (Figure 2). Despite being an electrical component, a fault in the capacitor can affect the performance of the air conditioner, such as a decrease in compressor rotation and, consequently, a change in the flow of refrigerant gas through the system, which can thus affect its cooling capacity.

These faults were selected based on a literature review of articles that researched FDD in HVAC equipment and in consultation with air conditioner installation and maintenance technicians, who corroborated the theoretical information on the topic.

3. Methods

This section describes the equipment utilized for the acquisition and treatment of data, in addition to demonstrating the use of the dataset in these studies.

3.1. Equipment

The equipment used for the study is a conventional split system from the Agratto brand, with 5272 Watts (18,000 Btu), 220 V single-phase voltage, and 7.5 A current according to the technical specifications in the device plate (Figure 3). The air conditioner has a compressor capacitor of 35 µF.

The evaporator and condenser units of the installed air conditioner are shown in Figure 4a,b.

The air conditioner was installed in the professors’ room of the GPICEEMA at UFPB, which measures 28.83 m² (Figure 5).

A circuit was assembled in a circuit breaker box to monitor the performance of the equipment during the data collection in order to better organize the test bench (Figure 6).

Sockets were installed in this box to input the 220 V main voltage and output the same voltage value, and they were protected by a 16 A circuit breaker (recommended by the manufacturer). A digital meter (model PZEM-022) was used together with a non-invasive current sensor (PZCT-02) to monitor the current and voltage of the air conditioner operation during the analysis. The equipment used for the signals’ acquisition was a voltage sensor, with a grain-oriented transformer with an input of 220 V and output of 1.5 V, and a non-invasive current sensor (SCT-013) for an input of 20 A and an output of 1 V (Figure 7). Both of them were connected to the power input signal of the air conditioner and were connected to a signal acquisition device (DAQ NI USB-6215 from National Instruments).

The DAQ was connected to a Dell Latitude E7240 notebook with an Intel Core I7-4600U processor and 16 GB of RAM (Figure 8) with the LabVIEW 2020 20.0f1 (64 bits) and Matlab R2022a software installed.

A simplified block diagram of the applied method is illustrated in Figure 9, which shows the split-system air conditioner powered by the input single-phase 220 V alternating voltage socket, which has phase, neutral, and ground. A non-invasive current sensor connected to the phase wire and a voltage sensor connected to the neutral and ground wires on the input of the equipment were used to generate the electrical current and voltage signals of the air conditioner. Both of these were connected to the DAQ, which was connected to the notebook with the LabVIEW and Matlab software installed.

3.2. Data Acquisition

The algorithm for data acquisition was graphically developed through the LabVIEW software, and data collection was carried out with an acquisition time of 60 s with a sampling rate of 30,000 Hz. Figure 10 presents the model developed in the block diagram, and Figure 11 presents the graphic mode of the program.

Initially, the air conditioner was powered up in Cool mode (refrigeration) and adjusted with an operating temperature of 20 °C, maximum-velocity ventilation, and the output air fins of the evaporator unit in a 45° opening position. A waiting period of 30 min was observed to ensure that the equipment could adequately cool the environment where it was installed. This time period was set as the standard based on previous operational tests.

After preparing the entire environment where the equipment was installed, including setting up the test bench with instruments for signal monitoring and data acquisition, the current and voltage signals during normal operation of the equipment (faultless) were collected by the DAQ. Subsequently, data were collected for each fault condition described in Table 2. This sequence of data collection was repeated three consecutive times in order to check that the data obtained had been validated so that they could be made available in the dataset. In addition, the data were validated by processing the electrical signals using the FFT, which was congruent with the equipment’s fundamental frequency of 60 Hz (Figure 12).

3.3. Dataset Applications

The main application for this dataset is for the FDD of the air conditioner. Therefore, two examples of techniques applied using the acquired electrical current are presented.

An algorithm in Matlab was used to process the data obtained for signal analysis, and it employed the fft function for the application of the FFT. The procedure began with processing the acquired data from the electrical current signal of the system in the time domain during faultless and faulty operation. A Hamming window was performed by employing the hamming function before the FFT’s application. FDD was carried out graphically in the end, as represented in the flowchart in Figure 13.

Firstly, the dataset files in txt format were loaded into the software by the algorithm to plot the signals graphically in the time domain. We decided to use samples within a 30-s time interval of acquisition to optimize data processing, as there were no significant changes throughout the signal. This resulted in a total of 900,000 samples. Figure 14 illustrates a segment of the acquired electrical current signals of single faults from the equipment, displaying their respective amplitudes and waveforms in the time domain.

The Hamming window was initially applied to select only relevant information from the signal before the acquired signal was transformed into the frequency domain, thereby improving the quality of results and optimizing computational resources. Subsequently, processing was performed to apply the FFT technique for single-fault analysis in the frequency domain (Figure 15).

Figure 15 also displays a projection of a green line delineating the region encompassing the frequency range obtained during normal equipment operation, which serves as a reference for FDD. It is observed that the tested faults can be detected through comparison with the faultless operation of the equipment (FL) left of the green line, as they occupy distinct frequency ranges in the graph area. Regarding diagnosis, it is possible to clearly identify the type of fault based on the occupied frequency range having distinct peak points. Even for the same type of fault with different incrustation levels, it is possible to diagnose clearly, as shown in faults of the condenser units CI60 and CI30.

Figure 16 illustrates a segment of the acquired electrical current signals of double faults in the equipment, displaying their respective amplitudes and waveforms in the time domain.

Figure 17 depicts the FFT graph with data processing for double faults. This graph also displays a projection of a green line delineating the faultless region. Both the combination of condenser unit faults, CD30CI60 and CD30CI30, and the combination of the condenser and evaporator unit faults, CI60FI50 and CD30FI50, could be clearly diagnosed.

4. User Notes

The dataset contains electrical current data in the first column of the txt file and electrical voltage data in the second column. The data were collected continuously over a 60-s period and are arranged sequentially in the columns of the spreadsheet. Depending on the purpose of the study, they can be used individually or together.

To demonstrate the FDD technique applied in this article, only current data were required. However, voltage data can also be used if the purpose of the study is to analyze the power of the equipment, which requires these two variables to be combined, such as in energy efficiency studies for buildings or even in FDD techniques that can use voltage information for specific faults to be analyzed.

The fact that a grain-oriented voltage sensor was used for data acquisition resulted in a more accurate measurement of the system.

For the specific application demonstrated in this work, data collection could have been carried out at a lower sampling rate. However, since the dataset is used in other research to develop new electrical measurement techniques, we have chosen to make it available at a sampling rate of 30,000 Hz/s for researchers interested in these studies. Depending on the type of fault analyzed and the method applied, some specific faults can only be identified in high-frequency ranges. We also have a dataset with the same faults at a rate of 5000 Hz/s, which can be made available upon request to the authors of this article.

The files have been made available in txt format so that the data can be easily imported in different programming languages and software depending on the purpose of the research.

5. Conclusions

This work provides a dataset of electrical measurements obtained from fault conditions and the faultless operation of a field-installed split-system air conditioner in order for it to be used in the improvement of existing techniques and algorithms, as well as the development of new techniques for the detection and diagnosis of faults in small-sized equipment, since they are less studied than medium and large HVAC systems.

These dataset files provide electric measurements of current and voltage for frequent fault conditions that occur in this type of equipment for single and simultaneous faults, facilitating other researchers in spending less time and fewer resources for the generation of data in their investigations on the subject.

This dataset was of great importance for the application of an electrical signal processing method for FDD, the FFT. This shows that electrical signal processing techniques can also be used to detect and diagnose faults in a small-sized air conditioner. Also, it may be a better alternative to thermodynamic or manometric techniques, which are widely used by researchers but require more resources. As presented in this work, electrical measurements can be carried out using non-invasive short-duration methods, thus reducing the time and operating costs required for an FDD system.