고중량 멀티콥터 부품 결함 검출을 위한 감독학습 모델 기반의 표준화된 FC 로그 분석 기법 연구
A study of Standardized FC Log Analysis Techniques for Identifying Defects in High-Weight Multi-copter Components Using Supervised Learning Models
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
As the unmanned aerial vehicle industry grows, unexplained multirotor crashes continue to increase, and existing preventive maintenance methods have limitations in managing multirotor safety. Safety must be the top priority in multi-copter operations. To address this, real-time monitoring of the multi-copter's flight status during operation is required, along with anomaly detection and immediate response based on flight log information. However, limitations exist in processing anomaly data for each flight control log, necessitating the development of standardized technology to overcome this challenge. In this paper we propose a standardized process for collecting multi-copter flight control logs in real time, classifying the log information by message sets, and extracting key defect detection indicators contained in each message set. Furthermore, the extracted defect detection indicators were validated using various supervised learning models. In our experimental results, we collected flight logs from a multi-copter equipped with a defective propeller and conducted experiments using three defect detection models. The results show an accuracy rate of 0.99. This is the F1-score for the defect detection rate.
Abstract
As the unmanned aerial vehicle industry grows, unexplained multirotor crashes continue to increase, and existing preventive maintenance methods have limitations in managing multirotor safety. Safety must be the top priority in multi-copter operations. To address this, real-time monitoring of the multi-copter's flight status during operation is required, along with anomaly detection and immediate response based on flight log information. However, limitations exist in processing anomaly data for each flight control log, necessitating the development of standardized technology to overcome this challenge. In this paper we propose a standardized process for collecting multi-copter flight control logs in real time, classifying the log information by message sets, and extracting key defect detection indicators contained in each message set. Furthermore, the extracted defect detection indicators were validated using various supervised learning models. In our experimental results, we collected flight logs from a multi-copter equipped with a defective propeller and conducted experiments using three defect detection models. The results show an accuracy rate of 0.99. This is the F1-score for the defect detection rate.