近年來，無人機已成為各行業中不可或缺的角色，但隨著無人機在實際應用時，其性能和可靠性問題也日益突出，導致無人機無法順利執行任務，甚至造成安全風險。本研究建立階層式即時健康診斷系統針對無人機之健康狀態診斷失效狀態種類和嚴重程度，以無人機結構為基底並定義馬達底座螺絲脫落和螺旋槳破損兩種失效狀態做為主要探討因素。首先蒐集無失效和失效狀態之振動數據，對其進行數據前處理和特徵工程，其中特徵選擇使用SBS從特徵集中挑選出最有用之特徵子集，接著利用最近鄰居法和支援向量機建立模型，並比較分析模型之性能，SVM模型性能相對較好，最終模型由SVM模型成為階層式健康診斷模型，再結合滑動視窗演算法完整建立無人機即時健康診斷系統，並成功完成實際測試。
階層式健康診斷模型進行預測測試資料時，第一層模型與第二層模型之招回率皆表現於93%以上，然而在預測新資料時性能有所下降，可能原因為飛行環境不同所導致，因此採用滑動視窗演算法，並預測同一筆新資料後性能提升，第一層模型無失效招回率提升至87.65%，失效狀態招回率則為76.61%，第二層模型之招回率提升，螺旋槳輕微破損之招回率為84.62%、螺旋槳嚴重破損之招回率為80.43%、馬達底座螺絲脫落一顆之招回率為84.86%和馬達底座螺絲脫落兩顆之招回率為87.56%；最後階層式健康診斷系統應用於實際任務情境時能夠即時診斷無人機自身的健康狀態，不會過於頻繁的發送錯誤診斷，同時又能診斷出無人機目前失效狀態和嚴重程度。

In recent years, drones have become an essential component in various industries. However, as drones face practical challenges, their performance and reliability issues have become increasingly prominent, leading to potential safety risks. The purpose of this study is to develop a hierarchical real-time diagnostic system for drones, capable of diagnosing the types and severity of failure states. This study begins by collecting flight vibration data related to failure states. The data undergoes preprocessing and feature engineering, including feature selection using the wrapper method to select the most useful subset of features from the feature set. Subsequently, a comparative analysis is conducted to assess the performance of each model using the k-Nearest Neighbor(k-NN) and Support Vector Machine(SVM), Ultimately, the SVM model is selected to form the hierarchical health diagnosis model. The sliding window algorithm is employed to create a complete real-time health diagnosis system for drones, which successfully undergoes practical testing.
The hierarchical health diagnosis model achieves recall rates of over 93% for both the first-layer and second-layer models when predicting test data. However, when predicting new data, the performance slightly declines, potentially due to variations in flight environments. To address this problem, the sliding window algorithm is utilized. After predicting the same new data, the recall rate of the first-layer model improves to 87.65% for the healthy state and 76.61% for the failure state. The second-layer model also exhibits improved recall rates, with 84.62% for slight propeller damage, 80.43% for severe propeller damage, 84.86% for one detached motor base screw, and 87.56% for two detached motor base screws. Ultimately, the hierarchical health diagnosis system can be applied to real mission scenarios to promptly diagnose the drone's health status, minimizing excessive false diagnoses while accurately determining the current failure state and its severity.

[1] Dourado, E., Hagemann, R., & Thierer, A. (2015). Operation and Certification of Small Unmanned Aircraft Systems. Public Interest Comment, Mercatus Center at George Mason University, Arlington, VA.
[2] Kandaswamy, G., & Balamuralidhar, P. (2017). Health monitoring and failure detection of electronic and structural components in small unmanned aerial vehicles. International Journal of Mechanical and Mechatronics Engineering, 11(5), 1081-1089.
[3] Misra, P., Kandaswamy, G., Mohapatra, P., Kumar, K., & Balamuralidhar, P. (2018, June). Structural health monitoring of multi-rotor micro aerial vehicles. In Proceedings of the 4th ACM Workshop on Micro Aerial Vehicle Networks, Systems, and Applications (pp. 21-26).
[4] 梁益榕(2022)，「應用支援向量機於無人機即時健康診斷系統」，國立成功大學民航研究所碩士論文。
[5] Lai, W. H., Tsai, S. T., Cheng, D. L., & Liang, Y. R. (2021). Application of wavelet scattering and machine learning on structural health diagnosis for quadcopter. Applied Sciences, 11(21), 10297.
[6] Lu, H., Li, Y., Mu, S., Wang, D., Kim, H., & Serikawa, S. (2017). Motor anomaly detection for unmanned aerial vehicles using reinforcement learning. IEEE internet of things journal, 5(4), 2315-2322.
[7] Cheng, D. L., & Lai, W. H. (2019). Application of self-organizing map on flight data analysis for quadcopter health diagnosis system. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 42, 241-246.
[8] Pourpanah, F., Zhang, B., Ma, R., & Hao, Q. (2018, October). Anomaly detection and condition monitoring of UAV motors and propellers. In 2018 IEEE SENSORS (pp. 1-4). IEEE.
[9] Lei, Y., Yang, B., Jiang, X., Jia, F., Li, N., & Nandi, A. K. (2020). Applications of machine learning to machine fault diagnosis: A review and roadmap. Mechanical Systems and Signal Processing, 138, 106587.
[10] Montalvao, D., Maia, N. M. M., & Ribeiro, A. M. R. (2006). A review of vibration-based structural health monitoring with special emphasis on composite materials. Shock and vibration digest, 38(4), 295-324.
[11] Cooley, J. W., & Tukey, J. W. (1965). An algorithm for the machine calculation of complex Fourier series. Mathematics of computation, 19(90), 297-301.
[12] Radkowski, S., & Szulim, P. (2014, September). Analysis of vibration of rotors in unmanned aircraft. In 2014 19th International Conference on Methods and Models in Automation and Robotics (MMAR) (pp. 748-753). IEEE.
[13] Tahir, M. M., Khan, A. Q., Iqbal, N., Hussain, A., & Badshah, S. (2016). Enhancing fault classification accuracy of ball bearing using central tendency based time domain features. IEEE Access, 5, 72-83.
[14] Kohavi, R., & John, G. H. (1997). Wrappers for feature subset selection. Artificial intelligence, 97(1-2), 273-324.
[15] https://zhuanlan.zhihu.com/p/348698669
[16] Zhang, S., Li, X., Zong, M., Zhu, X., & Cheng, D. (2017). Learning k for kNN classification. ACM Transactions on Intelligent Systems and Technology (TIST), 8(3), 1-19..
[17] Bentley, J. L. (1975). Multidimensional binary search trees used for associative searching. Communications of the ACM, 18(9), 509-517.
[18] Liu, R., Yang, B., Zio, E., & Chen, X. (2018). Artificial intelligence for fault diagnosis of rotating machinery: A review. Mechanical Systems and Signal Processing, 108, 33-47.
[19] https://wisdomml.in/breast-cancer-prediction-using-machine-learning/
[20] Raza, A. (2015). Glucose sensing using photonics waveguide based evanascent Raman spectroscopy. Photonics Research.
[21] https://store.tmotor.com/goods.php?id=727
[22] Chintanadilok, J., Patel, S., Zhuang, Y., & Singh, A. (2022). Mission Planner: An Open-Source Alternative to Commercial Flight Planning Software for Unmanned Aerial Systems: AE576/AE576, 8/2022. EDIS, 2022(4).
[23] Xiong, Z., Cui, Y., Liu, Z., Zhao, Y., Hu, M., & Hu, J. (2020). Evaluating explorative prediction power of machine learning algorithms for materials discovery using k-fold forward cross-validation. Computational Materials Science, 171, 109203.