Skip to main navigation Skip to main content
  • E-Submission

JKSPE : Journal of the Korean Society for Precision Engineering

OPEN ACCESS
ABOUT
BROWSE ARTICLES
EDITORIAL POLICIES
FOR CONTRIBUTORS
REGULAR

SiC 캐소드 초음파 가공 공정에서의 머신러닝 기반 음향방출 신호 분류

Machine Learning-based Classification of Acoustic Emission Signals in SiC Cathode Ultrasonic Machining Process

Journal of the Korean Society for Precision Engineering 2025;42(6):431-439.
Published online: June 1, 2025

1 한국전자기술연구원 산업데이터융합연구센터

1 Industrial Data Convergence Research Center, Korea Electronics Technology Institute (KETI)

#E-mail: jhchae@keti.re.kr, TEL: + 82-55-716-0372
• Received: February 19, 2025   • Revised: April 14, 2025   • Accepted: April 23, 2025

Copyright © The Korean Society for Precision Engineering

This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

  • 75 Views
  • 4 Download
  • 1 Crossref
prev next

Citations

Citations to this article as recorded by  Crossref logo
  • Flexible Acoustic Emission Sensor Signal Classification Using Convolutional Neural Networks for Pipeline Leak Detection
    Byungjae Park
    Journal of the Korean Society for Precision Engineering.2026; 43(1): 13.     CrossRef

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

Include:

Machine Learning-based Classification of Acoustic Emission Signals in SiC Cathode Ultrasonic Machining Process
J. Korean Soc. Precis. Eng.. 2025;42(6):431-439.   Published online June 1, 2025
Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:
Include:
Machine Learning-based Classification of Acoustic Emission Signals in SiC Cathode Ultrasonic Machining Process
J. Korean Soc. Precis. Eng.. 2025;42(6):431-439.   Published online June 1, 2025
Close

Figure

  • 0
  • 1
  • 2
  • 3
  • 4
  • 5
Machine Learning-based Classification of Acoustic Emission Signals in SiC Cathode Ultrasonic Machining Process
Image Image Image Image Image Image
Fig. 1 Ultrasonic micro hole drilling machine and sensor configuration
Fig. 2 Example of AE signal acquisition triggered by threshold crossing
Fig. 3 State transitions and data collection durations
Fig. 4 Threshold exceedance and data availability across three channels
Fig. 5 Waveform and fft data for different operating states: (a) Machining state, (b) Non-machining state, (c) Idle state, and (d) Power-off state
Fig. 6 Box plots of key parameters from hit data for operating states: (a) CH01, (b) CH02, and (c) CH03
Machine Learning-based Classification of Acoustic Emission Signals in SiC Cathode Ultrasonic Machining Process
Parameter Value Parameter Value
Pre-Amp gain [dB] 40 PDT [µs] 200
Sample rate [kHz] 2,000 HDT [µs] 400
Threshold [dB] 40 HLT [µs] 4,000
Pre-trigger [Samples] 0 MHD [µs] 1000
Count Signal strength [nVs]
Peak amplitude [mV] Signal energy [EU]
Duration [μs] Frequency centroid [kHz]
Signal RMS [mVrms] Peak frequency [kHz]
ASL [mV] Peak magnitude [mV]
Rise-time [μs] 1nVs: 109Vsec
Average frequency [kHz] 1eu: 10-14 V2sec
Classification Description
Machining Slurry Power
Machining state O O O
Non-machining state X O O
Idle state X X O
Power-off state X X X
Variable State CH01 CH02 CH03
Peak amplitude [mV] Machining 99.98 ± 0.37 14.16 ± 2.16 0.61 ± 0.13
Non-machining 1.07 ± 0.22 1.60 ± 0.31 0.60 ± 0.21
Idle 1.38 ± 0.26 1.72 ± 0.45 0.57 ± 0.20
Power-off 0.42 ± 0.05 0.62 ± 0.12 0.59 ± 0.14
Signal RMS [mVrms] Machining 43.38 ± 1.35 4.92 ± 0.44 0.25 ± 0.07
Non-machining 0.64 ± 0.11 1.06 ± 0.22 0.35 ± 0.14
Idle 0.60 ± 0.09 1.13 ± 0.32 0.32 ± 0.14
Power-off 0.22 ± 0.03 0.39 ± 0.09 0.36 ± 0.09
Frequency centroid [kHz] Machining 348.96 ± 3.83 187.24 ± 5.22 152.54 ± 12.02
Non-machining 112.56 ± 8.46 82.74 ± 10.08 152.06 ± 28.71
Idle 120.32 ± 8.13 82.55 ± 15.93 159.15 ± 28.00
Power-off 178.23 ± 9.29 140.04 ± 15.47 151.68 ± 20.93
Signal strength [nVs] Machining 35050.92 ± 1201.98 3934.66 ± 359.05 198.92 ± 62.62
Non-machining 522.70 ± 92.63 922.65 ± 188.21 292.54 ± 134.23
Idle 496.80 ± 71.91 976.32 ± 279.88 270.45 ± 127.45
Power-off 176.84 ± 34.67 341.15 ± 78.35 307.48 ± 83.68
Channel Model Accuracy [%] F1-Score [%]
CH01 XGB 87.63 87.32
KNN 85.87 85.57
LR 85.98 84.91
MLP 85.04 83.86
SVM 87.04 86.20
CH02 XGB 98.35 98.35
KNN 90.11 90.12
LR 92.58 92.48
MLP 97.76 97.77
SVM 94.70 94.67
CH03 XGB 79.15 78.64
KNN 67.02 65.89
LR 73.73 69.62
MLP 71.02 67.57
SVM 78.09 75.46
CH01 M NM I O [%]
M 124 0 60 1 Accuracy
NM 0 438 0 0 90.16
I 34 0 189 0 F1-Score
O 1 0 0 129 90.06
CH02 M NM I O [%]
M 184 0 1 0 Accuracy
NM 0 438 0 0 98.67
I 12 0 211 0 F1-Score
O 0 0 0 130 98.67
CH03 M NM I O [%]
M 81 1 65 38 Accuracy
NM 1 430 7 0 78.59
I 56 5 154 8 F1-score
O 25 1 2 102 78.31
Table 1 AE system parameters
Table 2 Key parameters of hit data
Table 3 Operating states classification for ultrasonic micro hole drilling process
Table 4 Statistical Summary of acoustic emission features across operating states and channels
Table 5 Comparison of machine learning models for hit data
Table 6 Classification Performance Comparison Across Channels in the Additional Experiment