Journal of Mechanical Science and Technology 26 (7) (2012) 2073~2076 www.springerlink.com/content/1738-494x

DOI 10.1007/s12206-012-0517-1

Coherence technique for noise reduction in rotary compressor† H. C. Kim1, M. G. Cho2, J. Kim3, J. H. Park4 and J. Shim1,* 1

School of Mechanical Engineering, Yeungnam University, Kyeongsan, Kyongbuk, 712-749, Korea 2 Department of Media Engineering, Tongmyong University, Busan, 608-711, Korea 3 Center for Bionics, Korea Institute of Science and Technology, 136-791, Korea 4 Division of Nano Convergence Science and Technology, Korea National Nano Fab Center, 305-701, Korea (Manuscript Received February 22, 2012; Revised March 16, 2012; Accepted April 10, 2012) ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Abstract The noise and vibration of a rotary compressor, a type of multi-input, single output system, are generally studied through frequency analysis. Although this method is effective in analyzing frequency components, using this method to identify the specific source of the noise (4 kHz to 6 kHz) is difficult. Hence, noise source should be studied systematically. In this study, a coherence analysis method based on systems analysis is used to identify the compressor noise source. Compressor noise source is identified through the coherence between the vibration signals on the shell of the compressor and the noise signal at one point near the compressor (1 m away from the compressor). A one-third octave band is employed for frequency analysis. The design of experiment is conducted to identify possible noise factors, such as volume, size, and neck area of the resonator in the compressor cylinder. Analysis showed that noise was generated from the cavity of the cylinder and the muffler inside the rotary compressor. A new type of muffler was applied to the rotary compressor to verify this finding. Noise was dramatically reduced. Keywords: Rotary compressor; MIMO system; Coherence analysis; DOE ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

1. Introduction In an attempt to reduce compressor noise, most researchers investigated noise sources, such as the muffler, piping system, rotating parts, lubrication, and the cavity and shell of a compressor. A variety of methods, such as beam forming, nearfield acoustic holography [1], and vibration intensity methods [2], can be used to identify noise source. Although these methods are powerful in searching for the noise source, the theories that they are based on are very difficult to understand, and their application requires several hardware resources. In the current work, coherence analysis [3-5] is carried out to detect the noise source of a rotary compressor. The main noise source (4 kHz to 6 kHz) is used as the target frequency. Coherence between a 91-point vibration signal on the compressor and a 1-point sound signal are analyzed. Finally, a noise test for the tuned compressor is conducted for validation.

2. Coherence analysis Coherence technique is widely used to compare the relation between two signals. In this study, coherence analysis was *

Corresponding author. Tel.: +82 538102465, Fax.: +82 538104627 E-mail address: [email protected] † This paper was presented at the ICMR2011, Busan, Korea, November 2011. Recommended by Guest Editor Dong-Ho Bae © KSME & Springer 2012

used to obtain the relationship between rotary compressor noise and vibration. Coherence analysis is defined in Eq. (1). This method has physical meaning and can properly provide the physical relationship between input signals. 2 γ my ⋅( m −1)! =

Gmy⋅( m −1)! Gmm⋅( m −1)!G yy⋅( m −1)!

,

Gij ⋅ p ! = Gij ⋅( p −1)! − L pj Gip⋅( p −1)! , L pj =

(1) G pj⋅( p −1)! G pp⋅( p −1)!

(2)

where Gmy is the cross spectrum between the input signal (m) and the output signal (y), and Gmm is the auto spectrum between the input signal (m) and the output signal (m). Eq. (3) is also derived from Eqs. (1) and (2). m ⎛ G yy⋅m ! ⎞ 2 ⎟ =1− 1 −γ iy :(i −1)! ⎜ G yy ⎟ ⎝ ⎠ i =1

γ 2y:m ! = 1 − ⎜

∏

(3)

3. Structure of a rotary compressor Fig. 1 shows the structure of a rotary compressor, which is mainly made up of the motor and bearing sub-system, muffler sub-system, rotor sub-system, stator sub-system, accumulator sub-system, and piping sub-system.

2074

H. C. Kim et al. / Journal of Mechanical Science and Technology 26 (7) (2012) 2073~2076

Table 1. Test specifications. Setup

Specification

Frequency range

20 kHz to 12.5 kHz

Number of channel

Vibration for 4 Ch / Noise for 1 Ch

Test condition

Air-conditioner condition

Fig. 3. Schematic of the coherence measurement.

Fig. 4. Picture of vibration sensor. Fig. 1. Structure of a rotary compressor.

Fig. 2. Noise test result of a normal and abnormal compressor. Fig. 5. Coherence measurement point in detail.

4. Problems Fig. 2 shows the noise spectra of two compressors at onethird octave. The noise level of the abnormal compressor is higher than that of the normal compressor in the 4 kHz to 6 kHz range. In the field, noise test was conducted in only one or two directions. This paper focuses on the +X direction of the noise.

shown in Fig. 3. A total of 102 coherence measurements (16 horizontal points x 6 vertical points) were conducted for the normal and abnormal (faulty) compressors. An accelerometer was used for 102 vibration measurements, and a microphone was used for noise signal measurement at 102 points. Figs. 4-5 show the coherence measurement points.

5. Experiment

6. Results and discussion

Noise test was performed in an anechoic room. A microphone was located 1 m away from the compressor, as shown in Fig. 2. B&K 3560C PULSE was used as analyzer. The measurement conditions are summarized in Table 1. The 1channel vibration sensor was used to detect the trigger signal. Other 3-channel vibration sensors (in the same direction as the 3 accelerometers) were used to gather vibration intensity, as

As shown in Fig. 6, the vibration level detected using a onedirectional accelerometer, as commonly done, shows that the vibration source may be located near the lower part of the compressor, which is high for that frequency. Fig. 7 shows the sound level contour for the same frequency of 4 kHz. This graph was obtained by measuring the noise level around the compressor using a microphone to measure

H. C. Kim et al. / Journal of Mechanical Science and Technology 26 (7) (2012) 2073~2076

2075

Fig. 10. Resonator design for the cavity analysis. Fig. 6. Vibration level contour for 4 kHz.

Fig. 7. Sound level contour for 4 kHz.

Fig. 8. Coherence result of vibration and noise signals in the 4 kHz.

Fig. 9. Cavity modes in the cylinder: (a) is first mode; (b) is second mode.

the vibration (Fig. 4). Noise source in the compressor cannot simply be compared in Figs. 6-7. Coherence analysis was done using the same data. The correlation coefficient was obtained for every measurement point in the frequency range of 4 kHz to 6 kHz using 190 vibration data and 1 sound data. Fig. 8 shows the coherence analysis result based on Eqs. (1) - (3). Vibration and noise signals, represented by the red color in Fig. 8, showed high correlation. Based on the coherence analysis, the positions of the noise source were associated with the cavity of the cylinder and the muffler. As such, cavity simulation was firstly carried out first. Fig. 9 shows the first and second modes of the cavity less than the 2 kHz range, which indicate that the cavity source based on the modes were not related to the 4 kHz to 6 kHz range.

Fig. 11. Main effects plot for volume and area of neck.

Fig. 12. Modification of muffler shape.

Therefore, we can conclude that the noise source is related to the function of the muffler. As the pulse-like discharge pressure is generally very high, the noise pulsation should be diminished by a muffler. Fig. 10 shows the resonator used in the muffler, which was designed at the end of the cylinder to reduce noise level in the 4 kHz to 6 kHz range. Several simulations were conducted to tune the frequencies. The neck area and volume of the resonator were changed. Design of experiment (DOE) was conducted to identify source relations among possible noise factors, such as volume, size, and neck area of the resonator in the compressor cylinder. Results of the DOE showed the main effects of neck volume on noise and vibration, as shown in Fig. 11. The optimization of the neck volume is the key to reducing the noise and vibration levels. The DOE objective is used as transmission loss (TL). The DOE found that the volume of the muffler was important in improving the TL of the muffler. Finally, the muffler shape was changed from hole type to dome type. Fig. 12 shows the modified muffler shape. Fig. 13 shows the TL shift graph according to muffler type. The TL value improved at the 4 kHz to 6 kHz frequency range. Overall noise was significantly reduced by 3 dB in the 4 kHz to 6 kHz range after modifying the muffler type from

2076

H. C. Kim et al. / Journal of Mechanical Science and Technology 26 (7) (2012) 2073~2076

References

Fig. 13. TL shift graph according to muffler type.

Fig. 14. Noise test result of modified compressor.

hole type to dome type, as shown in Fig. 14.

7. Conclusion A simple coherence analysis was carried out in this paper to reduce the noise of a rotary compressor. DOE was likewise conducted to identify the effects of resonator parameters on the cavity mode. Overall noise was reduced by about 3 dB(A) when the modified resonator was applied at the end of the cylinder.

Acknowledgment This research was supported by the Ministry of Knowledge Economy, Korea, under the Information Technology Research Center support program supervised by the National IT Industry Promotion Agency (NIPA-2011-(C-1090-1121-0006)). Further support was obtained from the Nano R&D program of the National Research Foundation of Korea. Funding for this research was provided by the Ministry of Education, Science and Technology (2009-0082696).

Nomenclature-----------------------------------------------------------------------γ2

my •( n −1) !

: Multi-coherence function removed of input (n-1)

Gmy⋅( m −1)! : Cross-power spectrum of input m and y, the

effect of input (m-1) is removed

Gmm⋅( m −1)! : Auto-power spectrum of input m, the effect of L pj

input (m-1) is removed : Optimum transfer function of input p and j, the effect of input (p-1) is removed

[1] F. Deblauwe, K. Janssens and M. Robin, Extending the usability of near-field acoustics holography and beam forming using focalization, ICSV14, 2007. [2] N. Kojima, H. Zhou, M. Mikami and T. Hirayu, Visualization of transient vibrational energy flow in a shell structure, ASVA97 (1997) 543-548. [3] S. M. Price and R. J. Bernhard, Virtual coherence: A digital signal processing technique for incoherent source identification, Proceedings of the 4th International Modal Analysis Conference (1986) 1256-1262. [4] M. E. Wang and M. J. Crocker, On the application of coherence techniques for source identification in a multiple noise source environment, journal of the acoustical society of america, 74 (3) (1983) 861-872. [5] J. Y. Chung, Measurement of frequency responses and multiple coherence function of the noise generation system of a diesel engine, JASA, 58 (3) (1975).

HeuiCheol Kim received his B.S. degree in Mechanical Design Engineering from Pusan National University, Korea in 1991. He then received his M.S. from Pusan National University in 1994. Kim is currently taking his doctorate degree at the School of Mechanical Engineering at Yeungnam University in Gyeongsan, Korea. His research interests include signal processing technology. Migyung Cho received her B.S. degree and Ph.D from the Department of Computer Science, Pusan National University, Pusan, Korea in 1994 and 1998, respectively. She has been an associate professor at Tongmyong University in Pusan since September 2002. Her research interests include design and analysis of computer algorithms, pattern recognition, and computer simulation of nano-bio. Jaesool Shim received his B.S. in Mechanical Engineering degree from Pusan National University, Korea in 1995. He then received his M.S. from Pohang University of Science and Technology and Ph.D from WSU in 1997 and 2007, respectively. Dr. Shim is currently a professor at the School of Mechanical Engineering at Yeungnam University in Gyeongsan, Korea. His research interests include lab-on-a chip and bio/nano technology.