隨著無人機系統蓬勃發展，無人機的飛行安全是無人機發展的重要議題。本研究以無人機動力機構為基底並設定馬達底座鬆脫失效、螺旋槳破損失效與機臂螺絲鬆脫失效這三種失效狀態做為本研究主要探討與解析的因素，為解決混合失效分類困難的課題。首先收集無失效、馬達底座鬆脫失效、螺旋槳破損失效與機臂螺絲鬆脫失效的飛行數據，並擷取飛控感測器中的振動與電流訊號分別做方均根、標準差和取樣熵的時域特徵擷取，以及擷取音頻感測器中的音頻訊號做小波散射的時頻域特徵擷取。接著對八種無失效與失效狀態做分析與分類，藉由最近鄰居法(k-Nearest Neighbor, kNN)進行失效狀態的分類並建立kNN模型，kNN為一種監督式機器學習演算法，其能分類資料與進行迴歸分析。
本研究提出利用時頻域分析飛行數據並建立無人機健康診斷分類系統，分類多重失效種類於四軸無人機中，其中包含無失效、單一失效狀態、雙重失效與三重失效的分類辨識，在分類無失效與三種失效混合情況下，分類系統準確率達90.73%。其方式改善混合失效分類準確率並保持一定水準的分類效用。

The development of unmanned aircraft systems is booming vigorously. Drone brings convenience and benefit but it also makes risk of flight safety. Therefore, how to control the risk of flight safety is a crucial study about drones. This study dedicates to health diagnosis classification system of structure of quadcopter. Loosening of motor mounts, propeller broken and loosening of arm mounts are the mainly discussed fault conditions used in this study. In the beginning of the research, the data of the undamaged, loosening of motor mount, propeller broken and loosening of arm mount are acquired. Then, the features of vibration signal and pulse width modulation signal are extracted by three methods, root mean square, standard deviation and sample entropy respectively. Moreover, the features of audio signal are extracted by wavelet scattering. Next, kNN (k-Nearest Neighbor) model can be trained by using features which are extracted by the vibration signal and it is a supervised machine learning method. After training by kNN model, kNN model can do fault classification and regression analysis.

The study proposes the method of time and frequency domain analysis on flight data analysis and establishment of quadcopter diagnosis classification system. It can classify multiple failure types in quadcopter, which include no failure, single fault, double faults and triple faults. In the case of classifying triple faults, the classification system has an accuracy of 90.73%. This method improves the accuracy of mixed faults classification and maintains a certain level of classification effectiveness.

參考文獻

[1]Huimin Lu, Yujie Li, Shenglin Mu, Dong Wang, Hyoungseop Kim, Seiichi Serikawa, “Motor Anomaly Detection for Unmanned Aerial Vehicles Using Reinforcement Learning,” IEEE Internet of Things Journal, Vol. 5, No. 4, pp.2315-2322, August 2018.
[2]Joakim Anden and Stephane Mallat, “Deep Scattering Spectrum,” IEEE Transactions on Signal Processing, Vol. 62, Number 16, pp.4114–4128, 2014.
[3]James Max Kanter and Kalyan Veeramachaneni, “Deep Feature Synthesis: Towards Automating Data Science Endeavors,” In Proceedings of the International Conference on Data Science and Advance Analytics. IEEE, 2015.
[4]G. Kandaswamy and P. Balamuralidhar, “Health Monitoring and Failure Detection of Electronic and Structural Components in Small Unmanned Aerial Vehicles,” World Academy of Science, Engineering and Technology International Journal of Mechanical and Mechatronics Engineering, Vol.11, No.5, pp.1080-1089, 2017.
[5]T. Kohonen, “The Self-Organizing Map,” New York, NY, USA: Springer, Proc. IEEE, Vol. 78, No. 9, pp.1464–1480, Sep. 1990.
[6]P. Misra, G. Kandaswamy, P. Mohapatra, Kriti Kumar, P. Balamuralidhar, “Structural Health Monitoring of Multi-Rotor Micro Aerial Vehicles,” Embedded Systems and Robotics, TCS Research and Innovation, TATA Consultancy Services Ltd. Bangalore and Chennai, India, 2018.
[7]Padraig Cunningham and Sarah Jane Delany, “k-Nearest Neighbour Classifiers,” Technical Report UCD-CSI-2007-4 MARCH 27, 2007.
[8]Randall Balestriero and Herve Glotin, “Scattering Decomposition for Massive Signal Classification: From Theory to Fast Algorithm and Implementation with Validation on International Bioacoustics Benchmark,” In: 2015 IEEE International Conference on Data Mining Workshop (ICDMW), pp.753–761. IEEE, New York, 2015.
[9]Richman, J. S. and Moorman, J. R., “Physiological Time-Series Analysis Using Approximate Entropy and Sample Entropy,” American Journal of Physiology-Heart and Circulatory Physiology, Vol.278, pp.H2039-H2049, 2000.
[10]Xu Chen and Peter J. Ramadge, “Music Genre Classification Using Multiscale Scattering Sparse Representations,” presented at the Conference on Information Sciences and Systems (CISS), 2013
[11]Yong Keong Yap, “Structural Health Monitoring for Unmanned Aerial Systems,” Technical Report UCB/EECS-2014-70. Electrical Engineering and Computer Sciences University of California at Berkeley, 2014.
[12]黃嘉銘, “聲音及振動訊號之時頻分析,” 國立成功大學航空太空工程研究所碩士論文, 2007.
[13]鄭德力, “應用自組織映射圖於四軸無人機飛行數據分析之診斷系統,” 國立成功大學航空太空工程研究所碩士論文, 2019.
[14]Ardupilot, http://ardupilot.org/
[15]MATLAB^®, https://www.mathworks.com/
[16]Mission Planner, http://ardupilot.org/planner/