DOI QR코드

DOI QR Code

인공 신경망을 이용한 실시간 용접품질 예측에 관한 연구

A Study on the Prediction of Welding Flaw Using Neural Network

  • Cho, Jae Hyung (Department of Industrial Engineering, Dankook University) ;
  • Ko, Sang Hyun (Department of Industrial Engineering, Dankook University)
  • 투고 : 2019.03.25
  • 심사 : 2019.05.20
  • 발행 : 2019.05.28

초록

자동차 분야에서 저항 점용접의 결함 및 품질을 실시간으로 예측할 수 있는 연구는 원가절감과 고품질 생산을 위한 필수 불가결한 연구 분야라 할 수 있다. 용접 품질은 전단강도와 너깃의 크기에 의해서 결정되며 여러 가지 독립변수에 따라 결과가 달라진다. 실시간 예측시스템을 개발하기 위하여 다중 회귀분석을 실시하여 3개의 독립변수로 두 가지 종속변수를 충분한 통계적 결과로 구하였으나 회귀식에 의한 품질 예측은 정확도를 보장할 수 없었다. 본 연구에서는 다층 신경망 회로를 구축하였다. 10가지의 동저항 변수에 의한 신경망은 3개의 은닉층을 구축하여 실행 함수와 가중치 행렬을 구하였다. 그러나 이 경우, 입력 변수가 너무 많아 실시간 제어에 어려움이 있을 수 있으므로 회귀분석에 의한 3개의 독립변수로 신경망을 구축하였다. 그 결과 모든 시험데이터를 불량, 부분 불량, 양품으로 구분하는데 성공하였다. 따라서 다중 회귀분석에 의해서 구한 3개의 독립변수에 의한 실시간 용접 품질 판정 시스템을 완성할 수 있었다.

A study in predicting defects of spot welding in real time in automotive field is essential for cost reduction and high quality production. Welding quality is determined by shear strength and the size of the nugget, and results depend on different independent variables. In order to develop the real-time prediction system, multiple regression analyses were conducted and the two dependent variables were obtained with sufficient statistical results with three independent variables, however, the quality prediction by the regression formula could not ensure accuracy. In this study, a multi-layer neural network circuit was constructed. The neural network by 10 dynamic resistance variables was constructed with three hidden layers to obtain execution functions and weighting matrix. In this case, the neural network was established with three independent variables based on regression analysis, as there could be difficulties in real-time control due to too many input variables. As a result, all test data were divided into poor, partial, and modalities. Therefore, a real-time welding quality determination system by three independent variables obtained by multiple regression analysis was completed.

키워드

DJTJBT_2019_v17n5_217_f0001.png 이미지

Fig. 1. A Method of Construction of Neural Network for Real Time Monitoring of Spot Welding.

DJTJBT_2019_v17n5_217_f0002.png 이미지

Fig. 2. Comparisons of Regression Line and Test Variables.

DJTJBT_2019_v17n5_217_f0003.png 이미지

Fig. 3. 10 Inputs Neural Network Structure

DJTJBT_2019_v17n5_217_f0004.png 이미지

Fig. 4. Learning Performance of 10 input Neural Network

DJTJBT_2019_v17n5_217_f0005.png 이미지

Fig. 5. 3 Inputs Neural Network Structure

DJTJBT_2019_v17n5_217_f0006.png 이미지

Fig. 6. Moving Variables of Network Train.

Table 1. The Result of Multi-Regression Anlalysis for Stress of Spot Welding.

DJTJBT_2019_v17n5_217_t0001.png 이미지

Table 2. The Result of Multi-Regression Anlalysis for Nugget Diameter of Spot Welding.

DJTJBT_2019_v17n5_217_t0002.png 이미지

Table 3. The Comparison Results of Forecasting Welding Qualities.

DJTJBT_2019_v17n5_217_t0003.png 이미지

참고문헌

  1. T. T. Han, K. Y. Lee & J. S. Kim. (2009). Recent Developments and Weldability of Advanced High Strength Steels for Automotive Applications, Journal of KWJS, 27(2), 131-132.
  2. K. W. Kang. (2014). Vibration Fatigue Analysis of Spot Welded Component considering Change of Stiffness due to Fatigue Damage, Convergence Society for SMB. (5)1. 1-8
  3. C. S. Son & Y. W. Park. (2012). Lobe Curve Characteristic Analysis of Resistance Spot Welding for Sheet Combination of 783 MPa Steel Sheet Using Simulation, Journal of KWJS, 30(6). 68-73.
  4. B. N. Cho, H. S. Kim & I. S. Kang. (2015). Development of Estimation Model of Construction Activity Duration Using Neural Network Theory, Journal of the Korea Academia-Industrial, 16(5). 3477-3483. https://doi.org/10.5762/KAIS.2015.16.5.3477
  5. K. K. Seo. (2014). Development of a Sales Prediction Model of Electronic Appliances using Artificial Neural Networks, JJournal of Digital Convergence 12(11), 209-214. https://doi.org/10.14400/JDC.2014.12.11.209
  6. S. K Kang & S. H. Chun, (2017). Human Tracking Technology using Convolutional Neural Network in Visual Surveillance, Journal of Digital Convergence, 15(2), 173-181. https://doi.org/10.14400/JDC.2017.15.2.173
  7. E. M. Yang, H. J. Lee & C. H. Seo. (2017). Comparison of Detection Performance of Intrusion Detection System Using Fuzzy and Artificial Neural Network, Journal of Digital Convergence, 15(6), 391-398. https://doi.org/10.14400/JDC.2017.15.6.391
  8. K. T. Kim, J. Y. Choi. (2018). Facial Local Region Based Deep Convolutioal Neural networks for Atomated Face Recognition, Journal of Korea Convergence Society, 9(4). 47-55. https://doi.org/10.15207/JKCS.2018.9.4.047
  9. T. S. Ki & S. H. Lee. (2017). A Prediction Scheme for Power Apparatus using Artificial Neural Networks, Journal of Korea Convergence Society, 7(6). 201-207
  10. Y. S. Yang, T. T. Nguyen & K. Y. Bae. (2010). Prediction of Heating Line for Plate Forming in Induction Heating Process Using Artificial Neural Network, Journal of the KWJS, 28(4). 1-4.
  11. Y. S. Yang, T. T. Nguyen & J. W. Kim. (2013). An artificial neural network system for predicting the deformation of steel plate in triangle induction heating process, International Journal of Precision Engineering and Manufacturing, 14(4), DOI: 10.1007/s12541-013-0075-1
  12. C. H. Kim, H. Y. Yu & S. H. Hong. (2007). Adaption of Neural Network Algorithm for Pattern Recognition of Weld Flaws, The Journal of the Korea Contents Association, 7(1). 65-72. https://doi.org/10.5392/JKCA.2007.7.1.065
  13. J. H. Cho. (2013). Prediction of Arc Welding Quality through Artificial Neural Network, Journal of KWJS, 31(3), 44-48.
  14. Y. B. Cho, H. S. Chang & H. S. Cho. (1993). Estimation of Nugget Size in Resistance Spot Welding Processes Using Artificial Neural Networks, Journal of Mechanical Science and Technology, 17(2), 393-406.
  15. S. W. Shin, J. H. Lee & S. H. Park. (2018). Strength Estimation Model of Resistance Spot Welding of 1.2 GPa Grade Ultra High Strength TRIP Steel for Car Body Applications, Journal of Welding and Joining, 36(1). 82-89. https://doi.org/10.5781/JWJ.2018.36.1.10