항공우주시스템공학회지 (Journal of Aerospace System Engineering)

  • 제18권5호
  • /
  • Pages.66-72
  • /
  • 2024
  • /
  • 1976-6300(pISSN)
  • /
  • 2508-7150(eISSN)
항공우주시스템공학회 (The Society for Aerospace System Engineering)
권호 표지
DOI QR코드

DOI QR Code

복합 임베디드 시스템 시계열 데이터를 활용한 딥러닝 이상 탐지 방법 비교 연구

A Comparative Study of Deep Learning-Based Anomaly Detection Methods for Time-Series Data in Complex Embedded Systems

  • 투고 : 2024.07.26
  • 심사 : 2024.09.29
  • 발행 : 2024.10.31
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

비행체 같은 복합 임베디드 시스템은 고장이 발생하면 심각한 위험을 초래할 수 있다. 본 논문에서는 복합 임베디드 시스템에서 출력되는 시계열 데이터 셋과 LSTM, 1차원 CNN과 같은 딥러닝 알고리즘을 활용하여 이상 탐지 모델을 생성하고 추론 결과를 비교했다. 그 결과 1차원 CNN 모델이 좋은 성능을 보였다. 이전 연구(합성곱 신경망을 활용한 항공 시스템의 이상 탐지 모델 연구)에서 생성한 2차원 CNN 모델의 추론 성능을 비교한 결과 정확도와 재현율은 2차원 CNN 모델이 높았지만, 추론 속도는 1차원 CNN 모델이 빨랐다. 실시간 이상 탐지가 필요한 복합 임베디드 시스템의 이상 탐지 모델에는 1차원 CNN 모델이 적합한 것으로 판단된다.

Complex embedded systems such as aircraft can lead to serious hazards when failures occur. This paper presents an anomaly detection model using deep learning techniques such as LSTM and 1D CNN on time-series datasets generated from complex embedded systems and compares inference results. Results showed that the 1D CNN model outperformed the LSTM model. Compared with the inference performance of a two-dimensional CNN model created in a previous study (Anomaly Detections Model of Aviation System by CNN), the two-dimensional CNN model had higher accuracy and recall. However, the 1-dimensional CNN model had faster inference speed. We can conclude that the 1D CNN model is more suitable than the LSTM model for anomaly detection in complex embedded systems that require real-time anomaly detection.

키워드

참고문헌

  1. How Airbus Detects Anomalies in ISS Telemetry Data Using TFX, https://blog.tensorflow.org/2020/04/how-airbus-detects-anomalies-iss-telemetry-data-tfx.html, April 2020.
  2. L. Moddemann, H. Steude, O. Niggemann, P. Grashorn, "Automated Anomaly Detection and Diagnosis of the Environmental Control System of the ISS," the Helmut Schmidt University, dtec.bw - band 1, pp.123-128, 2022.
  3. R. Chalapathy, S. Chawla, "Deep Learning For Anomaly Detection: A Survey," arXiv preprint arXiv:1901.03407, Jan 2019.
  4. J. Seo, J. Park, J. Yoo, H. Park, "Anomaly Detection System in Mechanical Facility Equipment: Using Long Short-Term Memory Variational Autoencoder," J Korean Soc Qual Manag, vol. 49, no. 4, pp. 581-594, Dec 2021.
  5. K. Choi, J. YI, C. Park, S. Yoon, "Deep Learning for Anomaly Detection in Time-Series Data: Review, Analysis, and Guidelines," IEEE Access, vol. 9, pp. 120043-120065, 2021.
  6. Understanding LSTM Networks, https://colah.github.io/posts/2015-08-Understanding-LSTMs, Aug 2015.
  7. J. Rostovski, M. Ahmadilivani, A. Krivosei, A. Kuusik, M. Alam, "Real-Time Anomaly Detection Using 1D-CNN and LSTM," Nordic Conference NCDHWS 2024, pp. 260-278, May 2024.
  8. H. Im, T. Kim, J. Song, B. Kim, "Anomaly Detections Model of Aviation System by CNN," Journal of Aerospace System Engineering, vol. 17, No. 4, pp. 67-74, Aug 2023.
  9. Y. Jung, E. Park, J. Kim, "Detection of Anomalities in Major Components of Unmanned Vehicles Using AI Algorithms Based on Autoencoder," SASE 2023 Fall Conference, pp. 43-45, Oct 2023.
  10. K. Hundman, V. Constantinou, C. Laporte, I. Colwell, T. Soderstrom, "Detecting Spacecraft Anomalies Using LSTMs and Nonparametric Dynamic Thresholding," 24th ACM SIGKDD international conference on knowledge discovery & data mining, pp 387-395, 2018.