本論文主旨為應用渦電流非破壞檢測技術，建立一套辨別多種硬度的金屬扣件檢測分類系統，結合主成份分析對金屬扣件數據集趨勢進行分析，並使用隨機森林演算法進行多硬度的智慧檢測分類。金屬扣件主要應用於機汽車、航太、大小型機械等，而金屬扣件的硬度與機械機構的穩定與安全性有著重要的關聯性，傳統的金屬扣件硬度檢測是採取抽樣破壞檢測方式，但這樣不僅品質不穩定、耗費成本、效率也差。本文以此為發想設計一套系統，可對同鋼材不同硬度的金屬扣件與同鋼材但未經過熱處理的金屬扣件進行全檢分類的系統，使用交流電橋量測檢測探頭的阻抗值，再透過主成份分析計算出最佳頻率點，利用最佳頻率點量測到的阻抗實部值與虛部值作為機器學習的特徵值，並將隨機森林整合到系統中進行線上訓練與實時智慧檢測分類，最後利用同鋼材不同硬度的金屬扣件與相同鋼材但未經過熱處理的金屬扣件進行實測驗證。結果顯示本文所設計的系統能正確地對多種不同硬度的金屬扣件進行分類，且相較於倒傳遞神經網路，隨機森林演算法在分類檢測中，在確保高準確率的狀況下，其模型訓練速度明顯優於倒傳遞神經網路。

The primary purpose of this paper is to apply eddy current non-destructive testing technology to establish a metal fastener detection and classification system that can identify multiple hardness, combine principal component analysis to analyze the trend of metal fastener data sets, and use random forest algorithm for multi-hardness wisdom classification. Metal fasteners are mainly used in locomotives, aerospace, large and small machinery, etc. The hardness of metal fasteners is closely related to the stability and safety of mechanical mechanisms. The traditional metal fastener hardness testing is to take the sample damage detection method, but not only the quality unstable, costly, and inefficient. This thesis is based on designing a system that can perform a full inspection and classification of metal fasteners of the same steel but with different hardness and metal fasteners of the same steel but has not been heat-treated. The impedance value of the probe is measured using an AC bridge circuit. Then calculate the most suitable frequency through principal component analysis, use the real value and imaginary value of the impedance as the feature values for machine learning. Integrate the random forest into the system for online training and real-time wisdom test classification, and finally use metal fasteners with the different hardness of the same steel and metal fasteners with the same steel but without heat treatment for actual measurement verification. The results indicate that the method proposed in this article can correctly classify metal fasteners of different hardness. Compared with the back propagation neural network, the random forest algorithm in classification detection, under the condition of ensuring high accuracy, its model training speed is significantly better than the inverse transfer neural network.

[1]D. Mercier, J. Lesage, X. Decoopman and D.Chicot, “Eddy currents and hardness testing for evaluation of steel decarburizing”, NDT & E International, Vol. 39, pp. 652-660, 2006.
[2]N. Korde and T. Kundu, “Material hardness and ageing measurement using guided ultrasonic waves”, Ultrasonics, Vol. 53, pp. 506-510, 2013.
[3]S. Ding, G. Tian and R. Sutthaweekul, “Non-destructive hardness prediction for 18CrNiMo7-6 steel based on feature selection and fusion of Magnetic Barkhausen Noise”, NDT & E International, Vol. 107, 2019.
[4]W. Yoshimura, R. Tanaka, T. Sasayama and K. Enpuku, “Detection of Slit Defects on Backside of Steel Plate Using Low-Frequency Eddy-Current Testing”, Transactions on Magnetics, vol. 54, no. 11, pp. 1-5, 2018.
[5]M. S. Amiri and M. Kashefi, “Application of eddy current nondestructive method for determination of surface carbon content in carburized steels”, NDT & E International, Vol. 42, pp. 618-621, 2009.
[6]C. C. Tai, J. H. Rose and J. C. Moulder, “Thickness And Conductivity of Metallic Layers From Pulsed Eddy Current Measurements”, Review of Scientific Instruments. vol. 67, no. 11, pp. 3965-3972, 1996.
[7]劉尹雄，「脈衝式渦電流檢測系統之設計與應用」，國立成功大學電機工程學系碩士論文，1999。
[8]C. C. Tai, “Characterization of coatings on magnetic metal using the swept frequency eddy current method,” Review of Scientific Instruments, vol. 71, no. 8 pp. 3161-3167, 2000.
[9]C. C. Tai, H. C. Yang and Y. H. Lin, “Modeling the surface condition of ferromagnetic metal by the swept-frequency eddy current method”, IEEE Transactions on Magnetics, vol. 38, no. 1, pp. 205-210, 2002.
[10]H. C. Yang and C. C. Tai, “Pulsed eddy-current measurement of a conducting coating on a magnetic metal plate,” Measurement Science and Technology, vol. 13, pp. 1259–1265, 2002.
[11]C. C. Tai and H. C. Yang, “The interaction of eddy current with metal hidden flaws for various probes”, Journal of Technology, vol. 17, no. 4, pp. 525-533, 2002.
[12]C. C. Tai and S. F. Wang, “Time-Domain and Frequency-Domain Eddy Current Simulations by the Finite Element Method”, Key Engineering Materials, vol. 270-273, pp. 585-592, 2004.
[13]Y. L. Pan and C. C. Tai, “Thickness and Conductivity Analysis of Molybdenum Thin Film in CIGS Solar Cells Using Resonant Electromagnetic Testing Method”, Magnetics, IEEE Transactions on, vol. 48, pp. 347-350, 2012.
[14]陳林湋，「低頻渦電流金屬厚度檢測系統研製」，國立成功大學電機工程學系碩士論文，2012。
[15]李柏毅，「非接觸式微米級金屬薄膜檢測系統研製及田口法最佳感測線圈設計」，國立成功大學電機工程學系碩士論文，2017。
[16]林茂盛，「架構於FPGA之掃頻式阻抗分析儀暨金屬薄膜厚度估測應用」，國立成功大學電機工程學系碩士論文，2018。
[17]賴行健，「倒傳遞神經網路及主成份分系應用於金屬扣件硬度篩選檢測:系統設計與評估」，國立成功大學電機工程學系碩士論文，2019。
[18]S. Ghanei, M. Kashefi and M. Mazinani, “Eddy current nondestructive evaluation of dual phase steel”, Materials and Design, vol. 50, pp. 491-496, 2013.
[19]T. Liu, W. Wang, W. Qiang and G. Shu, “Mechanical properties and eddy current testing of thermally agedZ3CN20.09M cast duplex stainless steel”, Journal of Nuclear Materials, vol. 501, pp. 1-7, 2018.
[20]K. V. Rajkumar, B. P. C. Rao, B. Sasi, A. Kumar, T. Jayakumar, B. Raj and K. K. Ray, “Characterization of aging behaviour in M250 grade maraging steel using eddy current non-destructive methodology”, Materials Science and Engineering: A, vol. 464, no. 1, pp. 233-240, 2007.
[21]S. H. Khan, A. N. Khan, F. Ali, M. A. Iqbal and H. K. Shukaib, “Study of precipitation behavior at moderate temperatures in 350 maraging steel by eddy current method”, Journal of Alloys and Compounds, vol. 474, no. 1, pp. 254-256, 2009.
[22]M. Zergoug, S. Lebaili, H. Boudjellal and A. Benchaala, “Relation between mechanical microhardness and impedance variations in eddy current testing”, NDT & E International, vol. 37, no. 1, pp. 65-72, 2004.
[23]M. Kashefi, A. Rafsanjani, S. Kahrobaee and M. Alaee, “Magnetic nondestructive technology for detection of tempered martensite embrittlement”, Journal of Magnetism and Magnetic Materials, vol. 324, pp. 4090-4093, 2012.
[24]S. Kahrobaee and M. Kashefi, “Hardness profile plotting using multi-frequency multi-outputelectromagnetic sensor”, NDT&E International, vol. 44, pp. 335-338, 2011.
[25]S. Kahrobaee and E. Z. Karimi, “Characterisation of work-hardening in Hadfield steel using non-destructive eddy current method”, Nondestructive Testing and Evaluation, vol. 34, no. 2, pp. 178-192, 2019.
[26]S. Konoplyuk, T. Abe, T. Uchimoto, T. Takagi and M. Kurosawa, “Characterization of ductile cast iron by eddy current method”, NDT & E International, vol. 38, no. 8, pp. 623-626, 2005.
[27]L. K. Kogan, A. P. Nichipuruk and L. D. Gavrilova, “Effect of the carbon content on the magnetic and electric properties of thermally treated carbon steels and the possibilities of testing the quality of tempering of articles produced from them via the eddy-current method”, Russian Journal of Nondestructive Testing, vol. 42, no. 9, pp. 616-629, 2006.
[28]B. Lei, P. Yi, Y. Li and J. Xiang, “A Temperature Drift Compensation Method for Pulsed Eddy Current Technology”, Sensors(Basel), vol. 18, 2018.
[29]H. Wang and Z. Feng, “Ultrastable and highly sensitive eddy current displacement sensorusing self-temperature compensation”, Sensors and Actuators A: Physical, vol. 203, pp. 362-368, 2013.
[30]H. Wang, B. Ju, W. Li and Z. Feng, “Ultrastable eddy current displacement sensor working in harshtemperature environments with comprehensive self-temperature compensation”, Sensors and Actuators A: Physical, vol. 211, pp. 98-104, 2014.
[31]Q. Li and F. Ding, “Novel displacement eddy current sensor with temperature compensationfor electrohydraulic valves”, Sensors and Actuators A: Physical, vol. 122, pp. 83-87, 2005.
[32]M. P. Arenas, T. J. Rocha, C. S. Angani, A. L. Ribeiro, H. G. Ramos, C. B. Eckstein, J. M. A. Rebello and G. R. Pereira, “Novel austenitic steel ageing classification method using eddy currenttesting and a support vector machine”, Measurement, vol. 127, pp. 98-103, 2018.
[33]K. Proniewska, A. Pregowska and K.P. Malinowski, “Identification of Human Vital Functions Directly Relevant to the Respiratory System Based on the Cardiac and Acoustic Parameters and Random Fores”, IRBM, 2020.
[34]E. Izquierdo-Verdiguier and R. Zurita-Milla, “An evaluation of Guided Regularized Random Forest for classification andregression tasks in remote sensing”, Int J Appl Earth Obs Geoinformation, vol. 88, 2020.
[35]X. Zhou, P. Lu, Z. Zheng, D. Tolliver and A. Keramati, “Accident Prediction Accuracy Assessment for Highway-Rail Grade Crossings Using Random Forest Algorithm Compared with Decision Tree”, Reliability Engineering & System Safety, 2020.
[36]N. Freitas, “Decision Trees”. University of British Columbia, 2013. https://www.youtube.com/watch?v=-dCtJjlEEgM. Accessed at Jun. 20, 2015.
[37]L. Breiman, “Random Forests. Machine Learning”, vol. 45, pp. 5-32, 2001.
[38]A. Statnikov, L. Wang and C. F. Aliferis, “A comprehensive comparison of random forests and support vector machines for microarray-based cancer classification”, BMC Bioinformatics, vol. 9, 2008.
[39]“Discrete Fourier Transform v3.1”, Xilinx Logicore IP, 2011.
[40]G. F. Zhao, J. Ying, L. Wu and Z. H. Feng, “Eddy current displacement sensor with ultrahigh resolution obtained through the noise suppression of excitation voltage”, Sensors and Actuators A: Physical, vol. 299, 2019.
[41]F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Perrot and É. Duchesnay, “Scikit-learn: Machine Learning in Python”. Journal of Machine Learning Research, vol. 12,pp. 2825-2830, 2011.
[42]J. C. W. Chan and D. Paelinckx, “Evaluation of Random Forest and Adaboost tree-based ensemble classification and spectral band selection for ecotope mapping using airborne hyperspectral imagery”, Remote Sensing of Environment, vol. 112, pp. 2999-3011,2008.
[43]T. L. H. Hu, C. Jiang, G. Yang and Z. Zhao, “A comparative study of decision tree, random forest, and convolutional neural network for spread-F identification”, Advances in Space Research, vol. 65, pp. 2052-2061, 2020.
[44]A. Tharwat, “Classification assessment methods”, Applied Computing and Informatics, 2018.
[45]J. Xu, Y. Zhang and D. Miao, “Three-way confusion matrix for classification: A measure driven view”, Information Sciences, vol. 507, pp. 772-794, 2020.