Journal of the Korea institute for structural maintenance and inspection (한국구조물진단유지관리공학회 논문집)

Korea Institute for Structural Maintenance Inspection (한국구조물진단유지관리공학회)
권호 표지
DOI QR코드

DOI QR Code

Construction of Faster R-CNN Deep Learning Model for Surface Damage Detection of Blade Systems

블레이드의 표면 결함 검출을 위한 Faster R-CNN 딥러닝 모델 구축

  • 장지원 (인하대학교 사회인프라공학과) ;
  • 안효준 (인하대학교 사회인프라공학과) ;
  • 이종한 (인하대학교 사회인프라공학과) ;
  • 신수봉 (인하대학교 사회인프라공학과)
  • Received : 2019.10.25
  • Accepted : 2019.11.14
  • Published : 2019.12.01
Download PDF
⟨ Previous Next ⟩

Abstract

As computer performance improves, research using deep learning are being actively carried out in various fields. Recently, deep learning technology has been applying to the safety evaluation for structures. In particular, the internal blades of a turbine structure requires experienced experts and considerable time to detect surface damages because of the difficulty of separation of the blades from the structure and the dark environmental condition. This study proposes a Faster R-CNN deep learning model that can detect surface damages on the internal blades, which is one of the primary elements of the turbine structure. The deep learning model was trained using image data with dent and punch damages. The image data was also expanded using image filtering and image data generator techniques. As a result, the deep learning model showed 96.1% accuracy, 95.3% recall, and 96% precision. The value of the recall means that the proposed deep learning model could not detect the blade damages for 4.7%. The performance of the proposed damage detection system can be further improved by collecting and extending damage images in various environments, and finally it can be applicable for turbine engine maintenance.

컴퓨터 성능 향상으로 다양한 분야에서 딥러닝을 활용한 연구가 활발히 진행되고 있으며 최근에는 구조물 안전성 평가 연구에도 그 적용이 이루어지고 있다. 특히 터빈의 내부 블레이드는 분리가 쉽지 않고 어두운 주변 환경으로 인해 블레이드의 표면 결함 검출은 전문 인력의 경험에 의존하고 있으며, 점검시간도 상당히 소요되고 있는 실정이다. 따라서, 본 연구에서는 딥러닝 기술을 적용하여 터빈 구조의 부재 중 하나인 내부 블레이드에 발생하는 결함을 검출할 수 있는 효율적인 방법을 제시하였다. Faster R-CNN 인공신경망 기법을 활용하여 결함의 이미지 데이터를 학습하였고 부족한 이미지는 필터링과 Image Data Generator를 이용하여 데이터를 확장하였다. 그 결과 블레이드의 결함을 학습한 딥러닝 모델은 평균적으로 약 96.1%의 정확도와 재현율은 95.3%, 정밀도는 96%의 성능을 보였다. 재현율을 통해 제시된 딥러닝 모델이 결함을 탐지하지 못하는 경우는 4.7% 로 나타났다. 재현율의 성능은 여러 환경의 많은 결함 이미지 데이터를 수집하고 확장하여 딥러닝 학습에 적용함으로써 더욱 향상되리라 판단된다. 이러한 실제 블레이드의 결함 이미지 데이터 확보와 학습을 통해 향후 터빈엔진 정비에 적용 가능한 결함 검출 시스템으로 발전할 수 있을 것이다.

Keywords

References

  1. Zaccone, G., Karim, R., and Menshawy, A. (2017), Deep Learning with Tensorflow, Acorn, Seoul, 33-35.
  2. Choi, S., and Do, M. (2018), Prediction of Asphalt Pavement Service Life using Deep Learning, Journal of Highway Engineering, KSRE, 20(2), 57-65. https://doi.org/10.7855/IJHE.2018.20.2.057
  3. Jung, H., Nam, W., Kim, G., Kim, D., Kang, I., and Kim, H. (2018), Deep Learning-based Damage Detection Method for Bridge Condition Evaluation, Magazine of the Korea Institute For Structural Maintenance and Inspection, KSMI, 22(3), 16-22.
  4. Lee, G. (2018), Damage Detection for RC Rahmen Bridge Based on Convolutional Neural Network, Inha University, Incheon, 22-25.
  5. Girshick, R., Donahue, J., Darrell, T., and Malik, J. (2016), Region-based Convolutional Networks for Accurate Object Detection and Segmentation, IEEE Transactions on Pattern Analysis and Machine Intelligence, IEEE, 38(1), 142-158. https://doi.org/10.1109/TPAMI.2015.2437384
  6. Lee, B. (2017), R-CNNs Tutorial, (https://blog.lunit.io/2017/06/01/r-cnns-tutorial).
  7. Ren, S., He, K., , Girshick, R., and Sun, J. (2017), Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks, IEEE Transactions on Pattern Analysis and Machine Intelligence, IEEE, 39(6), 1137-1149. https://doi.org/10.1109/TPAMI.2016.2577031
  8. Lee, B. (2016), Gas Turbine for Aircraft, Kyungmoon, Seoul, 147-149.
  9. An, H., Jang, J., Lee, J.H., and Shin, S. (2019), Damage Detection Method of Blade Systems using Deep Learning Technique, Proceedings of the Korea Institute For Structural Maintenance and Inspection, KSMI, 23(1), 30-31.
  10. Alasdair, M. (2007), Introduction to Digital Image Processing with Matlab, Cengage Learning, Seoul, 112-122.
  11. Sin, J., Jang, S., and Ji, I. (2008), Introduction to Digital Image Processing, HANBIT Media, Seoul, 183-187.
  12. Kim, T., (2017), Data Augmentation for Convolutional Neural Network, (https://tykimos.github.io/2017/06/10/CNN_Data_Augmentation).
  13. Nakai, E. (2016), Deep Learning Getting Started with TensorFlow, Jpub, Paju, 65-68.
  14. Cho, D. (2017), Performance Evaluation of Classification Model,(https://bcho.tistory.com/1206).
  15. Cho, J. (2018), Performance Measurement of Multiple Classification Models - Performance Measure (ACU, F1 score), (https://nittaku.tistory.com/295).
  16. An, Y., and Jang, K. (2017), Deep Learning-Based Structural Crack Evaluation Technique Through UAV-Mounted Hybrid Image Scanning, Journal of the Korean association for shell and spatial structures, KASS, 17(4), 20-26

Cited by

  1. 공공건축물 안전성 평가를 위한 지진가속도 계측자료의 유효성 검증 방법에 대한 연구 vol.24, pp.5, 2020, https://doi.org/10.11112/jksmi.2020.24.5.150
  2. Recognition and detection of aero-engine blade damage based on Improved Cascade Mask R-CNN vol.60, pp.17, 2021, https://doi.org/10.1364/ao.423333