Analysis of high frequency noise sources on air conditioning variable speed compressors Vitor Almeida 1 Tecumseh do Brasil LTDA Rua Cel Augusto de Oliveira Salles, 478, São Carlos Osmar Pinheiro 2 Tecumseh do Brasil LTDA Rua Cel Augusto de Oliveira Salles, 478, São Carlos

ABSTRACT

The improvement on energy consumption requires application of new technologies in existing products. On the past decades, optimization of power consumption on variable speed rotary compressors for air conditioning applications was achieved by replacing induction mo- tors to permanent magnets (PM) motors driven by a frequency inverter. Besides the positive impact of improving the compressor efficiency changing the electric motor type, the acoustic problems on the compressor became more complex. This article makes a revision of the main noise generation mechanism by PM motors of an air conditioning compressor via numerical finite element and experimental analysis aiming a noise reduction on a tonal component around 8 kHz band. Results indicate harmonic components on the forces exciting some specific reso- nance frequencies of the stator. The harmonic components show up in the force signals via pulse width modulation of the electronic controller. The structure-born noise is radiated through the compressor shell. Among the possibilities to mitigate the noise issues, it was chosen to optimize the electronic control, reducing the excitation levels of the stator and the tonal component of 8 kHz band in about 10 dB improving the sound quality of the compressor.

1. INTRODUCTION

Energy saving have always been a concern on refrigeration industry. On the past decades, the inverter technology based on the variable speed compressors shows up as an efficient solution on refrigeration industry. By an electronic inverter this technology converts the household AC power supply current, with fixed frequency of 50Hz or 60Hz, to a DC current and using a pulse width mod- ulation (PWM), then the inverter generates an AC in a specific frequency [1]. On this design, to apply the inverter technology, the usual induction motors of the compressors are replaced to Permanent Magnet Synchronous Motor (PMSM) based on Interior Permanent Magnets (IPM).

On air conditioning devices, the major amount of energy consumption is due to the rotary compressor. The application of variable speed technology makes possible to vary the rotation of the compressor from low rpms until high speeds, such as 7000 rpms. The electronic controller changes the compressor speed according to the thermal loads, prioritizing low rotations, regions in which the

1 vitor.almeida@tecumseh.com

2 osmar.pinehiro@tecumseh.com

worm 2022

compressor is more efficient. On the other hand, in household appliances, the noise levels generated is a key point for the final customer, making the noise characteristics of the product so important as the energy consumption. Thus, the project of a rotary compressor must aim both characteristics: en- ergy consumption and noise levels.

Besides the several advantages of IPM motors, this type of motor has more possible noise sources than other motor types, being a set of excitations straightly attached to the compressor shell and other compressor parts [2] [3].

This paper brings a case study of an inverter rotary compressor with accentuated noise level on 8 kHz band. Starting from the characterization of the issue and contextualization to theoretical background of electromagnetic noise sources, this paper maps the main noise and vibration transmis- sion paths of rotary compressors. Numerical finite element method (FEM) and experimental data provides guidelines to a deep understanding of the noise source around a frequency of 8 kHz, sup- porting several proposals to mitigate noise sources due electromagnetic phenomena.

2. OVERVIEW OF NOISE ISSUE

During the development of a new rotary variable speed compressor, it was noticed a tonal component in the third-octave band of 8 kHz. This unusual behavior hazards the sound quality of the product and is identified as an issue to be mitigated. Figure 1 shows the sound power levels measured according to ISO3744 standard. Although the middle frequency (between 800 Hz to 1600 Hz) has similar levels compared to 8 kHz band, the tonal behavior of the last one is a key point for product quality.

worm 2022

Figure 1: Sound power level of a rotary compressor. On the rotary compressor, the three main natures of noise sources which make the overall compressor noise are the gas pulsation on the suction and discharge pipelines; the mechanical forces and moments which appear due the movement of the parts during the compression cycle; the driven forces of the motor due to the electromagnetic field on the stator. Figure 2 summarize the main noise generation path on the rotary compressor. Understanding the mechanism of noise generation is crucial on troubleshooting of noise, vibration, and harshness (NVH) issues.

‘swt [2B] 108 PPPEPSORPEEPPEPIEPOEP Frequency)

Figure 2: Noise generation path. Measurements of noise, vibration and suction and discharge pressure under different condi- tions point to a problem related to the compressor motor. 3. THEORETICAL BACKGROUND

worm 2022

The noise sources on IPM motors can be divided in three types: (a) Cogging Torque; (b) Rip- ples of mutual and reluctance torque; (c) Fluctuations of radial magnetic attractive forces between rotor and stator.

3.1. Electromagnetic Forces

The electromagnetic forces on an IPM motor can be calculated by the Maxwell stress tensor method [4]. The time and angle dependence of the forces and flux densities (t, α) are suppressed for the sake of notation.

𝑓 𝑟 = 1

2 − 𝐵 𝜃

2 ൯ ,

൫𝐵 𝑟

(1)

2 𝜇 0

𝑓 𝜃 = 𝐵 𝑟 𝐵 𝜃

,

(2)

𝜇 0

where 𝑓 𝑟 and 𝑓 𝜃 are the radial and tangential forces, B r and B θ are the radial and tangential flux den- sities and µ 0 the permeability of free space. The radial forces take more attention on the NVH analysis of electric motors because they are the main cause of the mechanical deformation and vibration of

the stator [5]. The periods of the exciting radial and tangential forces waveforms are proportional to the number of pole pairs [6]. If the motor consists of a 4 magnetic poles design, it means that the 4th temporal order force frequency multiples are expected to contribute on the motor vibrations. The harmonic frequencies of the radial forces are proportional to the rotation of the motor by following:

𝐹 𝑓 = 𝑅𝑃𝑀 2 𝑝𝑝

60 ℎ ,

(3)

where F f is the frequency of the radial force, pp is the number of pole pairs and h is the harmonic order. This equation indicates that the frequency of the excitation forces is linked to the number of pole pairs of the motor.

3.2. Cogging Torque The cogging torque is a parasitic torque which appears due to the attractive forces between the saliencies of the rotor magnetic poles and the stator iron teeth. It is defined as a torque produced by magnetic forces in the circumferential direction between the stator teeth and the rotor magnetic poles. Since the cogging torque is a fluctuating component of the mean output torque, it is an im- portant characteristic on the noise and vibration performance of the motor [7].

3.3. Torque Ripple The IPM motor electromagnetic torque is a sum of the field torque from the permanent mag- nets and reluctance torque generated by the airgap saliences. The ripples of field and reluctance tor- ques are produced by harmonics of the flux linkages related to magnets, inductances and currents, and are defined as fluctuating components of the output torque and governed by several factors such as the shape of the currents with respect to time, variations of inductances and flux linkages with respect to rotor movement. The amplitude of field and reluctance torques depends on the inductance, currents in the coils and number of pole pairs [8]. 4. MOTOR EVALUATION

Tests have been made on the dynamometer with an operational load of 1.9 N m and the rota- tion fixed in 3000 rpm. A PCB impedance head model 288D01 was used to acquire force responses. In addition, a ½” microphone model 4188-A-23 were placed 1 m far from the motor – as illustrated in figure 3. A schematic showing the positions of the force transducer on the center of stator surface is presented on figure 4, named positions #1 to #4. The motor supply cable was used as reference. The data acquisition was made by B&K LAN-XI type 3160-A-042 and post processed on B&K Connect software.

worm 2022

worm 2022

Figure 3: Measurements at dynamometer test bench.

Figure 4: Measurements points on stator surface.

Figures 5 to 8 brings the FFT of the force signals in linear scale for measurement positions from #1 to #4 respectively.

worm 2022

Figure 5: Force levels at position #1.

Figure 6: Force levels at position #2.

Figure 7: Force levels at position #3.

worm 2022

Figure 8: Force levels at position #4. From the figures 5 to 8, it is noticed a first peak in a frequency of 200 Hz which is the funda- mental frequency of the radial magnetic force. Applying equation (3) for a motor with two pair of poles running at 3000 rpm, the fundamental frequency find is verified. Frequencies of 200 Hz, 400 Hz, 600 Hz and 2000 Hz which corresponds to the 1st, 2nd, 3rd and 10th harmonics, respectively. In addition, on the high frequencies - from 6 kHz to 9 kHz - it was noticed an unusual increase of the force levels with some spikes, mainly on positions #2 and #4 and minor levels at position #3. A zoom on figure 8, shows the responses around 7 kHz – figure 9.

ile

Figure 9: Zoom on figure 8. The frequencies pointed on figure 9 are the more relevant frequencies on the high frequency range. A sweep test starting from 900 rpm to 5400 rpm with steps of 70 rpm was made with the objective of evaluate the behavior of the peaks as the operating rotation changes. Figure 10 highlights that the amplitude peaks around 7 kHz exists along the rotations tested and the amplitudes also in- crease as the rotation increases.

worm 2022

Figure 10: Sweep from 900 rpm to 5400 rpm.

Since the elevated noise levels were identified as a vibration of the stator, a numerical modal analysis was carried out on ANSYS ® commercial software to verify the natural frequencies of the component. The simulation focused until the upper limit of 10 kHz band, i.e., 11220 Hz. Figure 11 presents the mode shapes associated to some high frequency modes. The 9 th to 13 th natural frequencies lies in a range from 6 kHz to 9 kHz and are very close with differences about 2% different between experimental and numerical results.

Figure 11: Numerical mode shapes.

From figure 9 and 11, a resonance effect in high frequencies is happening and generating an important tonal component on the compressor noise spectra. The radial magnetic force is the main cause of the stator excitation and must be reduced to improve the noise levels. The coupling of elec- tromagnetic forces and natural frequencies is generating a forced vibration response of the system. Before thinking in a stator redesign, which involves other important performance parameters such as

efficiency, manufacturing process and high investments, the attempt is to use alternative ways of reducing the radial magnetic forces on the stator.

4.1. PWM Frequency PWM is a modulation process or technique used in most communication systems for encoding the amplitude of a signal right into a pulse width or duration of another signal, usually a carrier signal, for transmission. Although PWM is also used in communications, its main purpose is to control the power that is supplied to various types of electrical devices, most especially to inertial loads such as AC/DC motors. The PWM frequency harmonics of the power supply are carried to the current on the motor and, therefore, these spectra appear on the electromagnetic forces. Some results shows that the PWM frequency effect on the acoustic noise levels for the motor could be negligible [9]. However, the acoustic noise can be significant if the introduced PWM harmonics on the current overlap with the stator’s natural frequencies in a resonance effect. On the current controller design, a typically used PWM frequency used is 8 kHz. Also, due to the nature of the PWM switching sidebands around 8 kHz are observed. These sidebands around 8 kHz can be visualized on figure 10 and are relevant to excite more stator modes. Additional tests at dynamometer were done, changing the PWM switching frequency to 4 kHz and 12 kHz – figure 12.

worm 2022

Figure 12: PWM frequency changes. With a switching frequency of 4 kHz, the force signals also show an increase around this mentioned frequency. The same behavior was observed with 12 kHz. The levels around 8 kHz de- creased with other switching frequencies. The sound pressure levels (SPL) measured are in accord- ance with the observation made on the force signals – figure 13. The 4 kHz switching frequency is not feasible since the same band had an elevation of the SPL measurements. Using the PWM of 12 kHz, the reduction observed on the SPL was 16 dB on 8 kHz band and 6 dB on the overall.

worm 2022

Figure 13: Sound pressure levels with different PWM frequencies on dynamometer tests. 5. COMPRESSOR TEST

Since the dynamometer test presented a potential of noise reduction and mitigation of the tonal component on 8 kHz, three compressors were assembled and tested the SWL according to ISO3744 standard at the same test conditions, changing only the switching frequency. All of them showed the same behavior, with a significant decrease on 8 kHz band and an improvement on the overall level. Figure 14 shows the average levels for the compressors tested in both situations. The average reduc- tion achieved on 8 kHz band lies about 10 dB. Moreover, the tonal behavior does not exist anymore.

The narrow band analysis of the high frequencies of figure 14, presented on figure 15, shows relevant reduction on the noise levels at frequencies associated to the stator natural frequencies, dis- cussed on the dyna test and in the numerical analysis.

it THe 66.6 dB(A) 72.8 48(A) 65.3 dB(A)

Figure 14: Average sound power level of compressor samples tested with different PWM switching

frequency.

‘swe (48a 1048 PPLE P LOS IFEIIFPEPSPEEPES reueney)

The tonality metric indicates the tonal feeling of the sound and is calculated as a product of various weighted functions to the tonal component. The case of the compressor is not a single pure tone but a sum of several tones. On the B&K Connect software, the tonality was calculated showing 0.037 tu for the compressor tested with PWM of 8 kHz and 0.005 tu for the samples with PWM of 12 kHz. In addition, the loudness also reduced from 7.8 sones to 6.8 sones.

worm 2022

Figure 15: Narrow band of compressor noise test. 6. CONCLUSIONS

This paper presented a case of study of how electric motors can be a very relevant noise source on compressor or any other devices. Discussions of possible noise sources due PM motors were made with focus on radial magnetic forces. By understanding the excitations on the stator, it was possible to propose an efficient and simple solution to mitigate noise levels on high frequencies, reducing the radiated noise and improving the sound quality of the equipment.

Future works will explore beyond the magnetic forces, focusing on electromagnetic designs that reduces the noise levels while keeping good motor parameters. 7. ACKNOWLEDGEMENTS

We gratefully acknowledge Tecumseh Company for the support on this research 8. REFERENCES

1. John Mix. Inverter technology . ASHRAE Journal, 56(5):34–36, 2014 2. Hong-Seok Ko and Kwang-Joon Kim. Characterization of noise and vibration sources in interior

permanent-magnet brushless DC motors . IEEE Transactions on Magnetics, 40(6):3482–3489, 2004.

3. Jinah Chung, Uiyoon Lee, Jeongbae Lee, Unseop Lee, Eunsil Han, and Jinhwan Yoon. Rotary

compressor noise analysis using mechanisms and electromagnetics coupled approach , 2016. 4. KH Yim, JW Jang, GH Jang, MG Kim, and KN Kim. Forced vibration analysis of an IPM motor

for electrical vehicles due to magnetic force . IEEE transactions on magnetics, 48(11):2981–2984, 2012. 5. HCM Mai, R Bernard, P Bigot, F Dubas, D Chamagne, and C Espanet. Consideration of radial

magnetic forces in brushless DC motors . In 2010 International Conference on Electrical Machines and Systems, pages 1–6. IEEE, 2010 6. Jae-Woo Jung, Do-Jin Kim, Jung-Pyo Hong, Geun-Ho Lee, and Seong, Min Jeon. Experimental

verification and effects of step skewed rotor type IPMSM on vibration and noise . IEEE transac- tions on magnetics, 47(10):3661–3664, 2011. 7. Chang-Sung Jin, Sol Kim, Youn-Hyun Kim, and Ju Lee. Design of IPMSM to reduce cogging

torque considering tapping the rotor and shape of barrier . In INTELEC 2009-31st International Telecommunications Energy Conference, pages 1–4. IEEE, 2009 8. Thushanthan Mohanarajah, Jamal Rizk, Ali Hellany, M Nagrial, and Alexey Klyavlin. Torque

ripple improvement in synchronous reluctance machines . In 2018 2nd International Conference On Electrical Engineering (EECon), pages 44–50. IEEE, 2018 9. Nathan Emery, Yawei Wang, Dhafar Al-Ani, and Berker Bilgin. Acoustic noise and vibration of

an interior permanent magnet traction motor: PWM effect . In 2021 IEEE Transportation Electri- fication Conference & Expo (ITEC), pages 1–6. IEEE, 2021.

worm 2022