臺灣博碩士論文加值系統

隨著自動化產業的普及使得機械手臂已成為不可或缺之關鍵設備，其中諧波減速機更是維持機械手臂穩定運行之關鍵零件，而即時偵測諧波減速機健康之重要性，對生產效率與產品品質至關重要。本研究使用了一種新的訊號處理方法希爾伯特-高斯轉換法(Hilbert-Gauss Transform, HGT)，該方法有別於傳統的希爾伯特-黃轉換法(Hilbert-Huang Transform, HHT)使用內插法進行訊號拆解之形式。希爾伯特-高斯轉換法是使用高斯平均濾波將原始故障振動訊號進行濾波，使其具有較優異之抗噪能力，且能保有更多原始訊號之特徵。首先，本研究針對一款減速比為100之諧波減速機建立五種故障運轉訊號之資料集，再將原始振動訊號分別利用HHT和HGT轉換為時頻圖，並結合CNN和預訓練網路(Pretrained Deep Neural Networks)來針對諧波減速機之故障進行分類，且在訓練過程中使用五折交叉驗證來確保訓練之公平性。最後，經由實驗結果證明HGT結合AlexNet對諧波減速機之故障進行分類的準確率可達99.42%，且明顯優於HHT模型的97.26%。

The proliferation of the automation industry has established mechanical arms as indispensable key components. Among them, Harmonic Drive play a critical role in maintaining the stable operation of mechanical arms. The real-time detection of the health status of Harmonic Drive holds paramount importance for both production efficiency and product quality. This research introduces a novel signal processing approach known as the Hilbert-Gauss Transform (HGT). This method diverges from the conventional Hilbert-Huang Transform (HHT) by employing Gaussian average filtering for signal decomposition. The Hilbert-Gauss Transform method employs Gaussian averaging to filter the original fault vibration signals, endowing them with superior noise resistance while retaining more intrinsic signal characteristics. Initially, this study constructs a dataset comprising five types of fault operation signals for a Harmonic Drive with a reduction ratio of 100. Subsequently, the raw vibration signals are transformed into time-frequency spectrograms using both HHT and HGT techniques. Convolutional Neural Networks (CNNs) and Pretrained Deep Neural Networks are combined for the purpose of fault classification of Harmonic Drive. Moreover, a five-fold cross-validation is integrated into the training process to ensure fairness and robustness of the model. The experimental results substantiate that the fusion of HGT with AlexNet achieves a fault classification accuracy of 99.42% for Harmonic Drive, a marked improvement over the 97.26% achieved by the HHT model.

誌謝 i
摘要 ii
Abstract iii
目錄 iv
圖目錄 vi
表目錄 viii
第一章 緒論 1
1.1研究背景 1
1.2文獻回顧 2
1.3研究目的與方法 6
1.4章節介紹 7
第二章 訊號分析與深度學習 8
2.1 訊號分析 8
2.1.1快速傅立葉轉換(Fast Fourier Transform, FFT) 8
2.1.2 希爾伯特-黃轉換(Hilbert-Huang Transform, HHT) 9
2.1.3希爾伯特-高斯轉換(Hilbert-Gauss Transform, HGT) 10
2.2 深度學習 11
2.2.1卷積神經網路(Convolution Neural Network, CNN) 12
2.2.2預訓練網路(Pretrained Deep Neural Networks) 12
2.2.3訓練資料集 14
第三章 實驗設備與架構 16
3.1 諧波減速機(Harmonic Drive) 16
3.1.1諧波減速機介紹 16
3.1.2諧波減速機零件架構 16
3.1.3諧波減速機故障模型 17
3.2 ASD-B31-0421-2伺服驅動器 19
3.3 ECM-B3L伺服馬達 20
3.4 Compact DAQ(ADLINK USB-2405) 21
3.5 電腦硬體設備 23
3.6 實驗架構 23
第四章 實驗結果與分析 25
4.1 振動訊號收集 25
4.2 HHT與HGT時頻圖比較 26
4.3 訓練模型混淆矩陣比較 32
4.4 訓練模型PCA和t-SNE比較 36
4.5 資料縮減訓練模型比較 39
第五章 結論與未來展望 41
5.1 結論 41
5.2 未來展望 41
參考文獻 42

[1] F. J. T. E. Ferreira, G. Baoming and A. T. de Almeida, “Reliability and Operation of High-Efficiency Induction Motors”, IEEE Transactions on Industry Applications, Vol. 52, no. 6, pp. 4628-4637, Nov.-Dec. 2016.
[2] Sandaram Buchaiah, Piyush Shakya, “Bearing fault diagnosis and prognosis using data fusion based feature extraction and feature selection”, Measurement, Vol.188, 2022, 110506.
[3] Gangjin Huang, Yuanliang Zhang, Jiayu Ou, “Transfer remaining useful life estimation of bearing using depth-wise separable convolution recurrent network”, Measurement, Vol.176, 2021, 109090.
[4] Guangxing Niu, Xuan Wang, Michael Golda, Stephen Mastro, Bin Zhang, “An optimized adaptive PReLU-DBN for rolling element bearing fault diagnosis”, Neurocomputing, Vol.445, Pages 26-34, 2021.
[5] H. Z. Han, K. G. Yuan, H. Ma, Z. K. Peng, Z. W. Li, S. T. Zhao and B. C. Wen, “Mesh characteristic analysis and dynamic simulation of spur gear pair considering corner contact and tooth broken fault,” Engineering Failure Analysis, Vol. 143, 2023, 106883.
[6] V. Sinitsin, O. Ibryaeva, V. Sakovskaya, V. Eremeeva, “Intelligent bearing fault diagnosis method combining mixed input and hybrid CNN-MLP model”, Mechanical Systems and Signal Processing, Vol.180, 2022, 109454.
[7] Yue-Der Lin, Yong Kok Tan, Baofeng Tian, “A novel approach for decomposition of biomedical signals in different applications based on data-adaptive Gaussian average filtering”, Biomedical Signal Processing and Control, Vol.71, Part A, 2022, 103104.
[8] Lin, Y.-D.; Tan, Y.-K.; Ku, T.; Tian, B. “A Frequency Estimation Scheme Based on Gaussian Average Filtering Decomposition and Hilbert Transform: With Estimation of Respiratory Rate as an Example”, Sensors, 2023, no.8, 3875.
[9] Y. Yoo and S. Jeong, “Vibration analysis process based on spectrogram using gradient class activation map with selection process of CNN model and feature layer,” Displays, Vol. 73, 2022, 102233.
[10] Qian Zhang, Wenhao Ma, Guoli Li, Jinjin Ding, Min Xie, “Fault diagnosis of power grid based on variational mode decomposition and convolutional neural network”, Electric Power Systems Research, Vol.208, 2022, 107871.
[11] Josué Pacheco-Chérrez, Jesús A. Fortoul-Díaz, Froylán Cortés-Santacruz, Luz María Aloso-Valerdi, David I. Ibarra-Zarate, “Bearing fault detection with vibration and acoustic signals: Comparison among different machine leaning classification methods”, Engineering Failure Analysis, Vol.139, 2022, 106515.
[12] 江翠珊，應用小波轉換與卷積類神經網路於軸承故障診斷之研究，國立雲林科技大學工業工程與管理系，碩士論文，2020。
[13] Can Ding, Zhenyi Wang, Qingchang Ding, Zhao Yuan, “Convolutional neural network based on fast Fourier transform and gramian angle field for fault identification of HVDC transmission line”, Sustainable Energy, Grids and Networks, Vol.32, 2022, 100888.
[14] Leh-Sung Law, Jong Hyun Kim, Willey Y.H. Liew, Sun-Kyu Lee, “An approach based on wavelet packet decomposition and Hilbert–Huang transform (WPD–HHT) for spindle bearings condition monitoring”, Mechanical Systems and Signal Processing, Vol.33, Pages 197-211, 2012.
[15] Lili Chen, Chaoyu Wang, Junjiang Chen, Zejun Xiang, Xue Hu, “Voice Disorder Identification by using Hilbert-Huang Transform (HHT) and K Nearest Neighbor (KNN)”, Journal of Voice, Vol.35, Issue 6, Pages 932.e1-932.e11, 2021.
[16] Ziran Guo, Ming Yang, Xu Huang, “Bearing fault diagnosis based on speed signal and CNN model”, Energy Reports, Vol.8, Supplement 13, Pages 904-913, 2022.
[17] Omaima El Alani, Mounir Abraim, Hicham Ghennioui, Abdellatif Ghennioui, Ilyass Ikenbi, Fatima-Ezzahra Dahr, “Short term solar irradiance forecasting using sky images based on a hybrid CNN–MLP model”, Energy Reports, Vol.7, Supplement 5, Pages 888-900, 2021.
[18] Alex Krizhevsky, Ilya Sutskever and Geoffrey E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks”, 2012.
[19] Pretrained Deep Neural Networks - MATLAB & Simulink。檢自:
https://www.mathworks.com/help/deeplearning/ug/pretrained-convolutional-neural-networks.html(Jun.17,2023).
[20] Forrest N. Iandola and Song Han, Matthew W. Moskewicz, Khalid Ashraf, William J. Dally and Kurt Keutzer, “SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size”, 2016.
[21] Xiangyu Zhang, Xinyu Zhou, Mengxiao Lin, Jian Sun, “ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices”, 2017.
[22] Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, Andrew Rabinovich, “Going Deeper with Convolutions”, 2014.
[23] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun, “Deep Residual Learning for Image Recognition”, 2015.
[24] 諧波減速機傳動原理-盟英科技。
檢自:http://www.maindrive.com.tw/proxy/item/506.html(Sep.15, 2022).
[25] Huimin Dong, Bo Dong, Chu Zhang, Delun Wang, “An equivalent mechanism model for kinematic accuracy analysis of harmonic drive”, Mechanism and Machine Theory, Vol.173, 2022, 104825.
[26] T Tjahjowidodo, F Albender, HV Brussel, “Nonlinear Modelling and Identification of Torsional Behaviour in Harmonic Drives[R]”, 2006.
[27] A. Raviola, A. De Martin, R. Guida, G. Jacazio, S. Mauro, and M. Sorli, “Harmonic Drive Gear Failures in Industrial Robots Applications: An Overview”, PHME_CONF, Vol. 6, no. 1, Page.11, Jun. 2021.
[28] 台達標準型交流伺服系統ASDA-B3 系列。檢自: https://resource.iyp.tw/static.iyp.tw/5782/files/f3bf2654-bdc5-48fb-8248-e008d26a1f64.pdf(Sep.25, 2022).
[29] ADLINK USB-2405|USB資料擷取模組|凌華科技ADLINK。檢自: https://www.adlinktech.com/Products/Data_Acquisition/USBDAQ/USB-2405?lang=zh-hant(Nov.4,2022).
[30] Ruihan Wang, Hui Chen, Cong Guan, Wenfeng Gong, Zehui Zhang, “Research on the fault monitoring method of marine diesel engines based on the manifold learning and isolation forest”, Applied Ocean Research, Vol.112, 2021, 102681.
[31] 淺談降維方法中的PCA與t-SNE。檢自: https://medium.com/d-d-mag/%E6%B7%BA%E8%AB%87%E5%85%A9%E7%A8%AE%E9%99%8D%E7%B6%AD%E6%96%B9%E6%B3%95-pca-%E8%88%87-t-sne-d4254916925b(May.1,2023).