本研究提出Wasserstein GANs的改進訓練(WGAN-GP）結合殘差神經網路（Residual network, ResNet），透過基於ResNet而建置的生成模型及鑑別模型進行對抗式訓練，以生成具有相似特徵數據分佈的新樣本，解決數據不足之問題，再使用生成的數據集訓練ResNet-50，完成感應馬達故障診斷模型。首先，由實驗室團隊研製的感應馬達故障模型，量測五種感應馬達運轉狀態(健康、定子匝間短路、轉子斷條、軸承外環損傷及不對心故障)，且於不同負載下之振動加速度訊號。其次，進行原始數據之前置處理案例分析，以決定時頻域轉換所使用的軸向及窗口間距。再者，運用不同時頻轉換及二維時域訊號處理，可觀察到時頻分析於負載不同時具有故障診斷的優勢，若使用基礎CNN模型架構，其分類準確率可達88 %。最後，本研究使用人工智慧之過採樣平衡方法及下採樣平衡方法，分析不同模型之診斷結果，於實驗結果可得知本研究所提出的WGAN-GP結合殘差神經網路優於其它機器學習及深度學習方法，其平均準確率高達96 %，且最高準確率達99 %，已接近實際平衡資料集的準確率，表示生成模型有效地學習到真實樣本分佈，驗證了本研究之可行性。

This thesis proposes WGAN-GP (Improved training of Wasserstein GANs) combined with residual neural network (Residual neural network, ResNet) to conduct adversarial training through the generative and discriminative model built based on ResNet structure. Generate new samples with similar feature data distribution to solve the problem of insufficient data and then use the generated data set to train ResNet-50 to complete the induction motor fault diagnosis model. First, the laboratory team developed by the induction motor fault model measures five induction motor operating states (healthy, stator inter-turn short circuit, broken bar, bearing outer ring damage, and misalignment fault), and the vibration under different working conditions acceleration signals. Second, the raw data preprocessing case study is performed to determine the axis and number of sampling points used for the time-frequency domain transformation. Furthermore, using different time-frequency conversion and 2-dimensional time-domain signal processing, it can be observed that time-frequency analysis has the advantage of fault diagnosis when the load is different, and the classification accuracy of the general CNN model architecture can reach 88 %. Finally, this study uses the artificial intelligence oversampling balance methods and downsampling methods balance to analyze the diagnostic results of different models. The experimental results show that the WGAN-GP combined with the residual neural network proposed in this study is superior to other machine learning methods. And the deep learning method, the average accuracy rate is as high as 96 %, and the highest accuracy rate is 99 %, which is close to the accuracy of the actual balanced data set, indicating that the generative model effectively learns the real sample distribution, which verifies the feasibility of this study.

參考文獻
[1] L. Saidi, J. B. Ali and F. Fnaiech, “Bi-spectrum based-EMD applied to the non-stationary vibration signals for bearing faults diagnosis,” ISA Trans., vol. 53, no. 5, pp. 1650-1660, Sep. 2014.
[2] H. Liu, D. Li, Y. Yuan, S. Zhang, H. Zhao and W. Deng, “Fault diagnosis for a bearing rolling element using improved VMD and HT,” Appl. Sci., vol. 9, no. 7, pp. 1439, Apr. 2019.
[3] M. A. S. K. Khan, T. S. Radwan and M. A. Rahman, “Real-time implementation of wavelet packet transform-based diagnosis and protection of three-phase induction motors,” IEEE Trans. Energy Convers., vol. 22, no. 3, pp. 647-655, Sep. 2007.
[4] S.-Y. Shao, W.-J. Sun, R.-Q. Yan, P. Wang and R. X. Gao, “A deep learning approach for fault diagnosis of induction motors in manufacturing,” Chin. J. Mech. Eng., vol. 30, no. 6, pp. 1347-1356, Oct. 2017.
[5] D. T. Hoang and H. J. Kang, “A motor current signal-based bearing fault diagnosis using deep learning and information fusion,” IEEE Trans. Instrum. Meas., vol. 69, no. 6, pp. 3325-3333, Jun. 2020.
[6] S. Shao, R. Yan, Y. Lu, P. Wang and R. X. Gao, “DCNN-based multi-signal induction motor fault diagnosis,” IEEE Trans. Instrum. Meas., vol. 69, no. 6, pp. 2658-2669, Jun. 2020.
[7] C.-S. Wang, I.-H. Kao and J.-W. Perng, “Fault diagnosis and fault frequency determination of permanent magnet synchronous motor based on deep learning,” Sensors, vol. 21, no. 11, pp. 3608, May 2021.
[8] T. A. Shifat and J.-W. Hur, “ANN assisted multi sensor information fusion for BLDC motor fault diagnosis,” IEEE Access, vol. 9, pp. 9429-9441, 2021.
[9] R. Razavi-Far, M. Farajzadeh-Zanjani and M. Saif, “An integrated class-imbalanced learning scheme for diagnosing bearing defects in induction motors,” IEEE Trans. Ind. Informat., vol. 13, no. 6, pp. 2758-2769, Dec. 2017.
[10] Q. Hang, J. Yang and L. Xing, “Diagnosis of rolling bearing based on classification for high dimensional unbalanced data,” IEEE Access, vol. 7, pp. 79159-79172, 2019.
[11] J. Mathew, C. K. Pang, M. Luo and W. H. Leong, “Classification of imbalanced data by oversampling in kernel space of support vector machines,” IEEE Trans Neural Netw Learn. Syst., vol. 29, no. 9, pp. 4065-4076, Sep. 2018.
[12] H. Yi, Q. Jiang, X. Yan and B. Wang, “Imbalanced classification based on minority clustering smote with wind turbine fault detection application,” IEEE Trans. Ind. Informat., vol. 17, no. 9, pp. 5867-5875, Sep. 2021.
[13] M. Zareapoor, P. Shamsolmoali and J. Yang, “Oversampling adversarial network for class-imbalanced fault diagnosis,” Mech. Syst. Signal Process., vol. 149, Feb. 2021.
[14] P. R. Bal and S. Kumar, “ WR-ELM: Weighted regularization extreme learning machine for imbalance learning in software fault prediction,” IEEE Trans. Rel., vol. 69, no. 4, pp. 1355-1375, Dec. 2020.
[15] Y. Fan, X. Cui, H. Han and H. Lu, “Chiller fault diagnosis with field sensors using the technology of imbalanced data,” Appl. Thermal Eng., vol. 159, Aug. 2019.
[16] Y. Zhuo and Z. Ge, “Gaussian discriminative analysis aided GAN for imbalanced big data augmentation and fault classification,” J. Process. Control, vol. 92, pp. 271-287, Aug. 2020.
[17] J. Zhang, Y. Wang, K. Zhu, Y. Zhang and Y. Li, “Diagnosis of inter-turn short circuit faults in permanent magnet synchronous motors based on few-shot learning under a federated learning framework,” IEEE Trans. Ind. Informat., vol. 17, no. 12, pp. 8495-8504, Dec. 2021.
[18] Y. Feng, J. Chen, T. Zhang, S. He, E. Xu and Z. Zhou, “Semi-supervised meta-learning networks with squeeze-and-excitation attention for few-shot fault diagnosis,” ISA Trans., vol. 120, pp. 383-401, Jan. 2022.
[19] J. Wu, Z. Zhao, C. Sun, R. Yan and X. Chen, “ Few-shot transfer learning for intelligent fault diagnosis of machine,” Measurement, vol. 166, Dec. 2020.
[20] N. Lu, H. Hu, T. Yin, Y. Lei and S. Wang, “ Transfer relation network for fault diagnosis of rotating machinery with small data,” IEEE Trans. Cybern., Jun. 2021.
[21] D. Yang, H. R. Karimi and K. Sun, “ Residual wide-kernel deep convolutional auto-encoder for intelligent rotating machinery fault diagnosis with limited samples,” Neural Netw., vol. 141, pp. 133-144, Sep. 2021.
[22] Y. Azzoug, M. Sahraoui, R. Pusca, T. Ameid, R. Romary and A. J. Marques Cardoso, “High-performance vector control without AC phase current sensors for induction motor drives: Simulation and real-time implementation,” ISA Trans., vol. 109, pp. 295-306, 2021.
[23] S. Yin, J. Xia, Z. Zhao, L. Zhao, W. Liu, L. Diao, et al., “Fast restarting of free-running induction motors under speed-sensorless vector control,” IEEE Trans. Ind. Electron., vol. 67, no. 7, pp. 6124-6134, Jul. 2020.
[24] H. Acikgoz, “Real-time adaptive speed control of vector-controlled induction motor drive based on online-trained type-2 fuzzy neural network controller,” Int. Trans. Electr. Energy Syst., vol. 30, no. 12, pp. e12678, 2020.
[25] A. Choudhary, D. Goyal, S. L. Shimi and A. Akula, “Condition monitoring and fault diagnosis of induction motors: A review,” Arch. Comput. Methods Eng., vol. 26, no. 4, pp. 1221-1238, Sep. 2019.
[26] X. Liang, “Temperature Estimation and Vibration Monitoring for Induction Motors and the Potential Application in Electrical Submersible Motors,” Canadian Journal of Electrical and Computer Engineering., vol. 42, no. 3, pp. 148-162, Jul. 2019.
[27] S. Kumar et al., “A comprehensive review of condition based prognostic maintenance (CBPM) for induction motor,” IEEE Access, vol. 7, pp. 90690-90704, 2019.
[28] N. Yassa and M. Rachek, “Modeling and detecting the stator winding inter turn fault of permanent magnet synchronous motors using stator current signature analysis,” Math. Comput. Simul., vol. 167, pp. 325-339, 2018.
[29] B. A. Vinayak, K. A. Anand and G. Jagadanand, “Wavelet-based real-time stator fault detection of inverter-fed induction motor,” IET Electr. Power Appl., vol. 14, no. 1, pp. 82-90, Jan. 2020.
[30] P. S. Kumar, L. Xie, M. S. M. Halick and V. Vaiyapuri, “Stator end-winding thermal and magnetic sensor arrays for online stator inter-turn fault detection,” IEEE Sens. J., vol. 21, no. 4, pp. 5312-5321, Feb. 2020.
[31] O. Guellout, A. Rezig, S. Touati and A. Djerdir, “Elimination of broken rotor bars false indications in induction machines,” Math. Comput. Simul., vol. 167, pp. 250-266, Jan. 2020.
[32] Y. Chen, T. Zhang, Z. Luo and K. Sun, “A novel rolling bearing fault diagnosis and severity analysis method,” Appl. Sci., vol. 9, 2019.
[33] M. S. Patil, J. Mathew and P. K. RajendraKumar, “Bearing signature analysis as a medium for fault detection: A review,” J. Tribol., vol. 130, no. 1, pp. 1-7, Jan. 2008.
[34] M. Amarnath and I. P. Krishna, “Experimental investigations to assess surface contact fatigue faults in the rolling contact bearings by enhancement of sound and vibration signals,” J. Nondestruct. Eval., vol. 38, no. 1, p. 34, 2019.
[35] N. Bessous, S. Sbaa and A. C. Megherbi, “Mechanical fault detection in rotating electrical machines using MCSA-FFT and MCSA-DWT techniques,” Bull. Polish Acad. Sci. Tech. Sci., vol. 67, no. 3, pp. 571-582, Jun. 2019.
[36] M. Arkan, H. Callis and M. E. Tagluk, “Bearing and misalignment fault detection in induction motors by using the space vector angular fluctuation signal,” Elect. Eng., vol. 87, no. 4, pp. 197-206, Jun. 2005.
[37] 林芝以，「運用電氣訊號於感應馬達狀態監測與故障診斷之研究」，碩士論文，國立臺灣科技大學，2018年。
[38] National Instrument, “NI PXI-8115 Intel Core i5-2510EDual-CorePXI Controller,” June 2013,http://sine.ni.com/nips/cds/view/p/lang/en/nid/204583.
[39] National Instrument, “NI PXIe-4300DataAcquisitionwith Integrated Signal Conditioning for High-Voltage Measurements,” June 2013, http://sine.ni.com/ds/app/doc/p/id/ds-207/lang/zht.
[40] Y. Park, H. Choi, J. Shin, J. Park, S. B. Lee and H. Jo, “Airgap flux based detection and classification of induction motor rotor and load defects during the starting transient,” IEEE Trans. Ind. Electron., vol. 67, no. 12, pp. 10075-10084, Dec. 2020.
[41] J. Tanha, Y. Abdi, N. Samadi, N. Razzaghi and M. Asadpour, “Boosting methods for multi-class imbalanced data classification: An experimental review,” J. Big Data, vol. 7, no. 1, pp. 1-47, Dec. 2020.
[42] V. Sampath, I. Maurtua, J. J. A. Martín and A. Gutierrez, “A survey on generative adversarial networks for imbalance problems in computer vision tasks,” J. Big Data, vol. 8, no. 1, pp. 1-59, Dec. 2021.
[43] I. J. Goodfellow et al., “Generative adversarial nets,” Proc. Int. Conf. Neural Inf. Process. Syst., pp. 2672-2680, 2014.
[44] Y.-J. Cao et al., “Recent advances of generative adversarial networks in computer vision,” IEEE Access, vol. 7, pp. 14985-15006, 2019.
[45] M. Arjovsky, S. Chintala and L. Bottou, “Wasserstein GAN,” arXiv:1701.07875, 2017.
[46] C´edric Villani. Optimal Transport: Old and New. Grundlehren der mathematischen Wissenschaften. Springer, Berlin, 2009.
[47] I. Gulrajani, F. Ahmed, M. Arjovsky, V. Dumoulin and A. C. Courville, “Improved training of wasserstein gans,” in Proc. Adv. Neural Inf. Process. Syst., 2017, pp. 5767-5777.
[48] A. Radford, L. Metz and S. Chintala, ‘‘Unsupervised representation learning with deep convolutional generative adversarial networks,’’ 2015, arXiv:1511.06434.
[49] S. J. Pan and Q. Yang, ‘‘A survey on transfer learning,’’ IEEE Trans. Knowl. Data Eng., vol. 22, no. 10, pp. 1345-1359, Oct. 2009.
[50] Z. Chen, J. Cen and J. Xiong, ‘‘Rolling bearing fault diagnosis using time-frequency analysis and deep transfer convolutional neural network,’’ IEEE Access, vol. 8, pp. 150248-150261, 2020.
[51] G. Vrbancic and V. Podgorelec, ‘‘Transfer learning with adaptive fine-tuning,’’ IEEE Access, vol. 8, pp. 196197-196211, 2020.
[52] K. He, X. Zhang, S. Ren and J. Sun, ‘‘Deep residual learning for image recognition,’’ in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2016, pp. 770-778.
[53] G. Koch, R. Zemel and R. Salakhutdinov, ‘‘Siamese neural networks for one-shot image recognition,’’ Proc. ICML Deep Learn. Workshop, vol. 2, pp. 1-8, 2015.
[54] T. Figarella and M. H. Jansen, ‘‘Brush wear detection by continuous wavelet transform,’’ Mech. Syst. Signal Process., vol. 21, no. 3, pp. 1212-1222, Apr. 2007.
[55] M. -Q. Tran, M. -K. Liu, Q. -V. Tran and T. -K. Nguyen, ‘‘Effective Fault Diagnosis Based on Wavelet and Convolutional Attention Neural Network for Induction Motors,’’ IEEE Trans. Instrum. Meas., vol. 71, pp. 1-13, 2022.
[56] R. G. Stockwell, L. Mansinha and R. P. Lowe, ‘‘Localization of the complex spectrum: the S transform,’’ IEEE Transactions on Signal Processing, vol. 44, no. 4, pp. 998-1001, April 1996.
[57] T. Chakraborty, ‘‘Imbalanced ensemble classifier for learning from imbalanced business school dataset,’’ Int. J. Math. Eng. Manage. Sci., vol. 4, no. 4, pp. 861-869, 2019.
[58] H. Liu, M. Zhou and Q. Liu, ‘‘An embedded feature selection method for imbalanced data classification,’’ IEEE/CAA J. Automatica Sinica, vol. 6, no. 3, pp. 703-715, May 2019.
[59] N. Mamdouh and A. Khattab, ‘‘YOLO-Based Deep Learning Framework for Olive Fruit Fly Detection and Counting,’’ IEEE Access, vol. 9, pp. 84252-84262, 2021.