Z-Freq Hybrid: Signal Analysis Based Car Air Conditioning Compressor Monitoring Technique Using Accelerometer and Piezo-Film Sensor

The efficiency and performance of car air conditioning systems rely heavily on adequately functioning the compressor and its associated components. There are many ways to detect an unfunctioning compressor and one of them is through vibration. Disfunctioning compressors could lead to discomfort, fatigue, stress and fogging windows, especially in long-distance driving. Thus, researchers focus on detecting the nonfunctioning car air conditioning compressor early. This paper introduced a new enhanced statistical method, namely Z-freq Hybrid. Z-freq Hybrid was based on a Z-notch frequency domain filter with a combination input of two different types of sensors introduced to detect functioning compressors. Data were recorded at various compressor speeds using an accelerometer and piezo film sensor with Signal Express 2015 software. The acceleration (m/s2) and voltage (mV) data were analyzed to find the combination degree of scattering data in a Z-freq Hybrid chart. The analyzed data show that the Z-freq Hybrid coefficient increases as the compressor’s speed increases. Then, the value dropped significantly when the compressor was dysfunctional. In conclusion, a Z-freq Hybrid can be employed to detect abnormalities and irregular vibration patterns, which may indicate the impending failure of a compressor.


Introduction
The automotive industry has seen significant changes in recent decades as modern technologies have become essential to vehicle systems.An example of a crucial system is the automobile air conditioning (AC) system, which significantly ensures maximum comfort inside the cabin [1].The central component of this system is the compressor, which plays a critical role in facilitating the refrigerant's circulation and the cooling process's efficiency.Nevertheless, the vehicle's second largest energy consumption device is the air conditioning system, which is up to 58% of energy [2].An inefficient compressor would lead to a tremendous amount of energy consumption [3].
In the past, compressor performance was evaluated with simplistic methodologies that mainly depended on human observations and simple instrumentation.The analysis is based on the air humidity, temperature, cleanliness, gas pressure, and flow meter [4].Although these procedures have previously been effective, they often prove inadequate in the present-day setting when accuracy, efficiency, and proactive diagnostic capabilities are crucial [5].Over time, the deficiencies inherent in these conventional methodologies, such as their absence of instantaneous monitoring, lack of data-driven fault detection, and their incapacity to identify subtle anomalies, become conspicuously apparent [6].The current research trend is integrating artificial intelligence and big data in improving various early detections of any machine failure and large scale HVAC system [7,8].
In recent years, there has been a change in study focus toward using technology to detect and diagnose anomalies, especially in industrial HVAC Systems.Research was done on the chiller, Rooftop Cooling Unit (RTU) on various faults such as flow blockage, poor air quality, water leakage, vibration, fan imbalance, refrigerant undercharge and overcharge [3,5,8].A. Hamad (2019), research on the domestic compressor performance based on varying refrigerant and speeds [10].Yet research still needs to be improved on faulty automobile air conditioning compressors.Typically, mean value, standard deviation, variance, skewness kurtosis, and mean root square are used to analyze vibration [11].The effectiveness of vibration signal analysis can be enhanced by combining data from two distinct sensor types [12].Currently, there is no optimal vibration analysis based on the frequency domain's voltage and acceleration kurtosis combination.This was acknowledged by researcher Ngatiman (2018), who proposed the addition of sensors such as accelerometer sensors and microphones to continue the investigation of Z-freq [13].This sensor was added by Abdullah (2010), who combined the vibrations resulting from the displacement and vibrations from the vehicle suspension system named I-kaz Hybrid [14].The advent of modern sensor technologies, along with the use of data analytics and pattern recognition methods, has facilitated the development of intricate diagnostic approaches.Vibration analysis has gained considerable attention as a prominent field within detection methods.Researchers claim that examining aberrant vibration patterns might provide useful information regarding the operational condition of the compressor since such patterns often accompany mechanical malfunctions or irregularities.
The use of sophisticated sensors, such as accelerometers and piezo-film sensors, has been especially remarkable within this field.Feng Lin (2023) used piezo film and mmWave sensing to recover highquality speech from outside soundproof zones.The researcher achieved over 98% accuracy for digit recognition [15].Meanwhile, Jeong (2019) used piezo film to predict the bolt fastening state [16].When combined with appropriate analysis tools, these sensors can provide comprehensive insights into the vibration patterns shown by compressors across different operating circumstances.This was proved by Mostafavi (2019) in his research on the detection of terminal oscillation patterns in ultrasonic metal welding [17].M.I.Ramli (2020) has also utilized Piezo film to determine the Young Modulus and Poisson Ratio of copper and stainless steel [18].Based on intricate algorithms and statistical techniques, signal analysis enhances the accuracy of various diagnostic procedures.The Z-freq Hybrid approach was established in the context of scholarly investigation and inquiry.The proposed methodology, which integrates a Z-notch frequency domain filter with sensor inputs, has the potential to provide a novel and enhanced strategy for assessing the health of compressors with increased accuracy [19].The objective of this study is to thoroughly investigate this approach, evaluate its suitability, and ascertain its efficacy in proactively detecting compressor problems, paving the way for a new era in automotive air conditioning system diagnostics.

Z-Freq Hybrid Statistical Method
The statistical parameter was used to derive the statistical features from the observed signal.The statistical parameter used for signal classification in research is an indicator of the air conditioning compressor's ability to function.The compilation of random data was first converted to the frequency domain using Fast Fourier Transform (FFT).Then, the kurtosis value was calculated for each data set because of its sensitivity to significant amplitude events.Gaussian distributions have a kurtosis value close to 3.0 [20].Outliers are found when the kurtosis value is more than expected from a Gaussian distribution.The 4 th order statistical moment of kurtosis is susceptible to the spikiness of the signal data as shown in equation (1), where  is number of data,  is the standard deviation,   is the frequency value at instantaneous point dan  ̅ is mean frequency [21].
The data set was further analyzed using a statistical method, namely Z-freq Hybrid.Z-freq Hybrid is an advancement from Z-freq, which was introduced by N.A. Ngatiman et.al in year 2021 [22].Z-freq Hybrid is a new statistical signal analysis from two different types of sensors based on frequency domain.The sensor used is an accelerometer, whose output is acceleration (mm/s2), and a piezo film sensor, which gives a voltage (V) output.Z-freq Hybrid,  ℎ  coefficient value can be calculated using equation (2), where  is the number of data,   and   are kurtosis and standard deviation for acceleration value.In contrast,   and   are kurtosis and standard deviation for voltage value, respectively.Lastly, the calculated kurtosis, standard deviation and Z-freq Hybrid value were presented in a 2D graphical scattered plot. (2)

Experimental Setup
The experiment was set up using a compact car 1.5cc with air conditioning compressor model number XI447260-9990.The compressor was first charged with 320g R134a and 80 ml of compressor oil.
Figure 1 shows that the compact car is being charged with R134a Refrigerant, measured using a 9055 Programmable Refrigerant Meter.Acceleration and voltage signals were measured by attaching the accelerometer and piezo film to the compressor, as shown in figure 2 1.

Result And Discussion
From equation (1) and equation ( 2),  and  ℎ  value have been calculated and represented in the scattered graph in figure 3 for A/C compressor ON and figure 4 for A/C compressor OFF. Figure 3 shows a green scattered graph voltage (V) versus acceleration (m/s 2 ) at varying engine revolution speeds for 320g R134a with the compressor on.The scattered graph ascends as the data points increase in value as the RPM increases [12,21,22].The data points are also tightly clustered as they are very close to each other.Meanwhile, figure 4 shows a yellow scattered graph voltage (V) versus acceleration (m/s 2 ) at various engine revolution speeds for 320g R134a with the compressor off.The data points are also strongly stable as they are close together and remain consistent across the x-axis from an idle position up to 1500 rpm.Then the data points trend rises until 3000 rpm. Figure 5 shows a linear line graph of the Z-freq Hybrid coefficient versus compressor speed.For figure 5(a), the graph trend for A/C On and A/C Off is linearly increasing from idle to 1500 rpm.Then the 320g A/C Off suddenly increases dramatically until 3000 rpm.This is because the engine's vibration was unstable as and transferred to the compressor that runs freely without engaging the magnetic clutch.
Compared to the graph in figure 5(b), a similar Z-Freq Hybrid trend happened from the idle position to 2000 rpm, and then the 340g A/C Off suddenly crossed and increased sharply until 3000 rpm.From figures 5(a) and 5(b), one could conclude that compressor dysfunction could be determined by examining the Z-Freq Hybrid value from idle position up to 1500 rpm.This is generally good for an engine as it does not necessarily ram up more than  For all refrigerant amounts, there is a notable increase in Z-freq Hybrid coefficient with an increase in compressor speed [13].The 360g refrigerant exhibits the highest coefficient of Z-freq Hybrid across the board, particularly noticeable at higher compressor speeds.

Conclusion
In conclusion, we captured data from an accelerometer and a piezo film sensor operating at different compressor speeds.Z-freq Hybrid Chart analysis of acceleration (m/s 2 ) and voltage (mV) data yielded the combination scattering degree.According to the analysis results, the Z-freq Hybrid coefficient rises as compressor speed increases.When the compressor stopped working, the value fell drastically.Finally, the Z-freq Hybrid may be used to monitor compressors for signs of failure, such as unusual vibration patterns.
(a).The data was collected using National Instrument (NI) Signal Express 2015 software through the sound and vibration module NI-9234.The complete schematic diagram of the experiment is shown in figure 2(b).The compressor was at six different speeds (rpm): idle, 1000, 1500, 2000, 2500 and 3000.The detailed parameters are presented in table Figure 1: Refrigerant Charge Process

Figure 3 .Figure 4 .
Comparison of the Z-freq Hybrid for varying rpm with A/C ON. 7th International Conference on Noise, Vibration and Comfort (NVC 2023) Comparison of the Z-freq Hybrid for varying rpm with A/C OFF.

Figure 5 .
Comparison of the Z-freq Hybrid for Different Refrigerant Amounts

Figure 6 .
Figure 6.Comparison of the Z-freq Hybrid for Different Refrigerant Amounts at Compressor ON Figure 6 compares Z-freq Hybrid for different refrigerant amounts (320g, 340g and 360g) during A/C On.For all refrigerant amounts, there is a notable increase in Z-freq Hybrid coefficient with an increase in compressor speed[13].The 360g refrigerant exhibits the highest coefficient of Z-freq Hybrid across the board, particularly noticeable at higher compressor speeds.

Table 1 .
The set of tests at 80ml Refrigerant Oil