簡易檢索 / 詳目顯示

回結果列表
研究生： 卡爾思
Cristales Cardona, Carlos René
論文名稱： 應用音頻信號結合人工神經網絡於無人機結構即時故障檢測系統
Application of Audio Signal and Artificial Neural Networks for Real-Time Structural Fault Detection System in UAV
指導教授： 賴維祥
Lai, Wei-Hsiang
學位類別： 碩士
Master
系所名稱： 工學院 - 航空太空工程學系
Department of Aeronautics & Astronautics
論文出版年： 2023
畢業學年度： 111
語文別： 英文
論文頁數： 69
中文關鍵詞： 無人機（UAV） 、故障檢測 、音頻信號 、梅爾頻率倒譜係數（MFCCs） 、人工神經網絡（ANN）
外文關鍵詞： Unmanned Aerial Vehicle (UAV), fault detection, audio signals, Mel Frequency Cepstral Coefficients (MFCCs), Artificial Neural Network (ANN)
相關次數： 點閱：87  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 無人機（UAV）融入人類環境中帶來了新的挑戰，例如UAV故障可能對附近的人造成潛在風險。故障的檢測可以提供給UAV使用者提前警示，減少事故風險。為了降低空中載具故障的風險，通常會使用容錯控制系統。這些系統在故障發生時可以通知使用者或將載具的性能降至最低。
    本論文提出了一種基於音頻信號處理和人工神經網絡（ANN）的UAV即時故障檢測系統。該系統旨在飛行過程中檢測和定位UAV的結構故障，以提高安全性。該系統通過記錄UAV飛行過程中馬達的音頻信號，提取梅爾頻率倒譜係數（MFCCs）作為音頻信號的特徵，並使用ANN將信號分類為代表正常或異常的UAV。ANN是通過使用至少有一個螺旋槳受損的飛行的音頻信號紀錄以及UAV處於正常狀態的飛行的紀錄進行訓練的。
    本論文的主要貢獻是開發和測試一個能夠即時運作的故障檢測系統，該系統使用機載電腦和Wi-Fi連接到地面站。該系統能夠在飛行過程中檢測故障，並正確識別UAV中的損壞螺旋槳。結果顯示，該系統有潛力提高UAV操作的安全性和可靠性，平均實時個別故障定位的準確率達到85%，對於診斷UAV的整體健康狀態的準確率達到97%，特別是在視覺檢查或其他故障檢測方法可能具有挑戰性的情況下。

    The integration of unmanned aerial vehicles (UAVs) into human environments introduces new challenges, such as the potential risks posed by UAV malfunctions, which could endanger nearby individuals. Detection of faults may provide early warning to UAV users and reduce accident risk. To mitigate the risk of faults in aerial vehicles, fault-tolerant control systems are frequently employed. These systems can either notify the user of the fault or reduce the performance of the vehicle to a minimum in the event of a fault occurrence.
    This thesis presents a real-time fault detection system for UAVs using audio signal processing and an artificial neural network (ANN). The system is designed to detect and localize a structural fault in UAVs during flight to improve safety. The system works by recording the audio signals of the UAV's motors during flight, extracting Mel frequency cepstral coefficients (MFCCs) as the characteristic features of the audio signals, and using an ANN to classify the signals as representing a healthy or unhealthy UAV. The ANN is trained on audio signal recordings from flights in which at least one propeller was damaged, as well as recordings from flights in which the UAV was in a normal state.
    The main contribution of this thesis is the development and testing of a fault detection system that can operate in real-time, using an on-board computer and a Wi-Fi connection to a ground station. The system is capable of detecting faults, correctly identifying a damaged propeller in the UAV during flight. The results demonstrate the potential of the system to enhance the safety and reliability of UAV operations with an average accuracy of 85% for real-time individual fault localization and 97% for diagnosing the overall health status of the UAV, particularly in situations where visual inspection or other methods of fault detection may be challenging.

    中文摘要 I Abstract II Acknowledgement III List of Tables VI List of Figures VII List of Abbreviations X List of Symbols XI Chapter 1 Introduction 1 1.1 Background 1 1.2 Research Motivation and Objective 2 1.3 Literature Review 2 1.4 Thesis Overview 6 Chapter 2 Basic Theory 8 2.1 Feature Extraction 8 2.2 Mel Frequency Cepstral Coefficients Extraction 8 2.2.1 Pre-emphasis 8 2.2.2 Frame Blocking 9 2.2.3 Windowing 10 2.2.4 Discrete Fourier Transform 10 2.2.5 Triangular Bandpass Filters 10 2.2.6 Discrete Cosine Transform (DCT) 12 2.3 Artificial Neural Networks 12 2.3.1 Feed-Forward Neural Networks 13 2.3.2 Stochastic Gradient Descent (SDG) 14 2.3.3 The Activation Function 16 2.3.4 Loss Function 18 2.4 Cross-Entropy 19 Chapter 3 Methodology 20 3.1 Experimental Setup 20 3.1.1 Quadrotor UAV 21 3.1.2 Flight Controller 21 3.1.3 GPS Module 22 3.1.4 UAV Remote Control 23 3.1.5 Power Configuration 24 3.2 Embedded System 26 3.2.1 Raspberry Pi 26 3.2.2 Respeaker 4-Mic Array 27 3.2.3 Raspberry Pi + Microphone Array Module 28 3.3 Definition of Faults 28 3.4 Flight Audio Recording 30 Chapter 4 Fault Classification Model 33 4.1 Audio Dataset 34 4.2 MFCCs Feature Extraction 35 4.2.1 Feature Scaling 36 4.3 ANN Classification Model 39 4.3.1 ANN Architecture 39 4.3.2 ANN Training Parameters 40 4.4 ANN Classification Model Results 41 4.4.1 Number of MFCCs 42 4.4.2 Audio Recording Length for Diagnosis 43 Chapter 5 Experimental Results and Discussion 47 5.1 Real-Time Classification 47 5.2 Real-Time Flight Experiments 49 5.3 Real-Time Fault Detection Results 55 5.3.1 Real-Time Fault Detection 55 5.3.2 Real-Time Fault Localization 59 Chapter 6 Conclusions and Future Work 64 6.1 Conclusions 64 6.2 Future Work 65 References 67

    [1] Tvaryanas, A. P., Thompson, W. T., & Constable, S. H. (2006). Human factors in remotely piloted aircraft operations: HFACS analysis of 221 mishaps over 10 years. Aviation, space, and environmental medicine, 77(7), pp.724-732.
    [2] Fourlas, G. K., & Karras, G. C. (2021). A survey on fault diagnosis and fault-tolerant control methods for unmanned aerial vehicles. Machines, 9(9), pp.197.
    [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] Altinors, A., Yol, F., & Yaman, O. (2021). A sound based method for fault detection with statistical feature extraction in UAV motors. Applied Acoustics, 183, 108325.
    [5] Zhang, J., Jiang, Q., & Chang, F. (2018, August). Fault diagnosis method based on MFCC fusion and SVM. In 2018 IEEE International Conference on Information and Automation (ICIA) (pp. 1617-1622). IEEE.
    [6] Bondyra, A., Kołodziejczak, M., Kulikowski, R., & Giernacki, W. (2022). An Acoustic Fault Detection and Isolation System for Multirotor UAV. Energies, 15(11), 3955.
    [7] 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。
    [8] Rao, K.S. and Manjunath, K.E. (2017). Speech Recognition Using Articulatory and Excitation Source Features, Springer Briefs in Speech Technology, DOI 10.1007/978-3-319-49220-9
    [9] Picone, J. W. (1993). Signal modeling techniques in speech recognition. Proceedings of the IEEE, 81(9), 1215-1247.
    [10] Deller, Jr, J. R. (1993). Discrete-time processing of speech signals. In Discrete-time processing of speech signals (pp. 908-908).
    [11] Benesty, J., Sondhi, M. M., & Huang, Y. (2008). Introduction to speech processing. Springer Handbook of Speech Processing, 1-4.
    [12] Jang, A. E. (2023). Mel Frequency Cepstral Coefficients (MFCCs). In Audio Signal Processing and Coding (2nd ed.).
    [13] Haykin, S. (1999). “Neural Networks, A Comprehensive foundation”, 2* edition. Prentice Hall.
    [14] Sazli, M. H. (2006). A brief review of feed-forward neural networks. Communications Faculty of Sciences University of Ankara Series A2-A3 Physical Sciences and Engineering, 50(01).
    [15] LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.
    [16] Holbrook, R. (2023). Stochastic Gradient Descent [Course module]. In Intro to Deep Learning. Kaggle.
    [17] Holbrook, R. (2023).Deep Neural Networks [Course module]. In Intro to Deep Learning. Kaggle.
    [18] Holbrook, R. (2023) Stochastic Gradient Descent [Course module]. In Intro to Deep Learning. Kaggle.
    [19] Holbrook, R. (2023) Binary Classification [Course module]. In Intro to Deep Learning. Kaggle.
    [20] Librosa. (2023). Documentation - Librosa 1.0.1 documentation. Retrieved from https://librosa.org/doc/latest/index.html (Accesed May 4, 2023).
    [21] McFee, B., Raffel, C., Liang, D., Ellis, D. P., McVicar, M., Battenberg, E., & Nieto, O. (2015, July). librosa: Audio and music signal analysis in python. In Proceedings of the 14th python in science conference (Vol. 8, pp. 18-25).

    下載圖示 校內：2025-07-15公開
    校外：2025-07-15公開
    QR CODE