Visualizing Noise Sources on KTX High-Speed Trains With LabVIEW

Youngkey K.. Kim, SM Instruments Co. Ltd.

"Engineering quieter trains requires an in-depth understanding of the noise sources. The ability to visualize the amplitude of the noise sources with the help of LabVIEW provides the tools necessary to implement noise reduction measures."

- Youngkey K.. Kim, SM Instruments Co. Ltd.

The Challenge:

Developing a system for visualizing noise sources on the KTX-Sancheon high-speed train with the Korea Railroad Research Institute to reduce environmental noise.

The Solution:

Developing a moving source beamforming application with NI LabVIEW software and a 144-channel microphone phased array.


Youngkey K.. Kim - SM Instruments Co. Ltd.
Sunghoon Choi - Korea Railroad Research Institute


Introduced into service in 2010 and built using Korean technology, the KTX-Sancheon is a high-speed train operated by Korail. Operating at speeds up to 300 km/h (186 mph), the train is sensitive to environmental noise, including rolling noise such as propulsion or machinery, mechanical noise from wheel-rail interaction, and aerodynamic noise from the airflow around the train. To reduce overall noise, corrective actions have been taken to identify all prominent noise sources. 



The Korea Railroad Research Institute and SM Instruments Co. Ltd., a NI Gold Alliance Partner specializing in sound and vibration measurements, developed a moving source beamforming application in LabVIEW with a phased microphone array to visualize noise sources on a full-scale train in normal operation. The tests’ main purpose was to compare the noise from the two types of trains: the KTX-1, which is derived from the TGV Réseau that began service in 2004, and the new KTX-Sancheon (KTX-II), which is the first commercial high-speed train developed in Korea.



Beamforming is a method of mapping noise sources using an acoustical array. It discerns the direction from which the sound originates through time delays that occur as the sound passes over an array of microphones. A moving source adds some complexity because as the object moves past the array, such as in a pass-by test, the Doppler effect distorts frequency components. This is critical for conventional real-time beamforming. To compensate for this, we continuously adjusted the time delays in software to follow the moving source. This method automatically removes the Doppler effect. It requires more processing time, but we averaged beam power while it was moving. We used trigger sensors to know the moving source position at every time step. In our software, we assumed constant speed. 


The hardware configuration is almost the same as in standard beamforming. One addition is that the moving source beamformer requires trigger sensors. We used two photo-electric sensors to trigger position and calculate the train’s speed.


For the high-speed train tests, we designed a 144-channel microphone array to increase the resolution of the image. We used NI PXI-4496 dynamic signal acquisition modules to acquire measurements and power the ICP/IEPE microphones with a special type of photo sensor to trigger the position of a train. Earlier tests on the KTX were conducted in 2006, shortly after Korail introduced high-speed rail service, with a 48-channel array successfully capturing noise sources on the KTX at 297 km/h.  



The performance of a microphone array is determined by two parameters: (1) the main lobe width of the beamforming power, which determines the image resolution and (2) the maximum side lobe level, which determines the levels of ghost images. Different patterns of arrays have different performance indexes. When we compared four patterns, our spiral pattern showed very balanced results.


For the 144-channel array, we combined three different patterns to increase the performance. Each type of pattern had the same shape but a different diameter. The smaller was for low maximum-side lobe levels at high frequencies and the larger was for high resolution at low frequencies. To reduce wind-induced noise, we applied windscreens to the microphones.


After acquiring the data and performing postprocessing, we verified the position accuracy by examining the last car image of the train moving to the right. Because we used a photo sensor for the trigger, we were concerned about some position shift. Among the many cars of a train, only the last car has a pantograph (device for collecting electricity from overhead lines) on its roof. The first picture shows the pantograph generating 500 Hz noise at an appropriate position due to vortex shedding. We also verified the wheel position at the bottom picture at 2,000 Hz.


At the higher frequencies, noise sources from the wheels were very clear. This shows that each wheel had a different magnitude. This technique can possibly be applied to monitor the conditions of the wheels in operation for maintenance.


Future plans call for increased train speeds, which could lead to increased noise levels particularly from aerodynamic noise. Engineering quieter trains requires an in-depth understanding of the noise sources. The ability to visualize the amplitude of the noise sources with the help of LabVIEW provides the tools necessary to implement noise reduction measures.


Author Information:

Youngkey K.. Kim
SM Instruments Co. Ltd.
DIREC 302, Taplipdong 697, Yusunggu, Daejeon 305-701
South Korea

Figure 1. Korean High-Speed Trains
Figure 2. Program Flow of the SeeSV230 Sound Camera Based on LabVIEW
Figure 4. Design of the Spiral Microphone Array
Figure 5. Noise Source Verification at Different Frequencies
Figure 6. Noise Source Visualization of Wheels