殘差神經網路於旋轉機械故障診斷之應用

Application of Residual Neural Network to the Diagnosis of Rotating Machinery Defects

陳芃睿(Peng-Rwei Chen)

指導教授 ： 范憶華

中原大學/工學院/機械工程研究所/碩士(2020年)

https://doi.org/10.6840/cycu202000518

摘要

旋轉機械是各領域機械設備之關鍵機構，其設備故障會帶來嚴重的後果，由於傳統的故障檢測方式需要大量的專業知識，因此以Resnet卷積神經網路的方式對旋轉機械缺陷進行診斷，而訓練資料的缺乏在訓練神經網路時容易導致不良的訓練結果，因此在本文中將製作五種不同的缺陷組與一組正常組，並依缺陷嚴重程度的不同分為輕度、中度與重度組，擷取振動信號後經由一連串的信號處理繪出其時域圖、頻域圖、能量譜、相位譜與能量譜密度並對缺陷進行診斷。在本實驗中證明神經網路可學習僅兩組不同缺陷程度的特徵（輕度與重度），並成功對未學習的缺陷程度進行診斷（中度），其準確率可達99.25%，成功的對六種不同的類別進行分類。

關鍵字

機器學習 ； 深度學習 ； 卷積神經網路 ； Resnet ； 振動分析

並列摘要

Rotating machinery is a key mechanism of mechanical equipment in various fields. The failure of equipment will bring serious consequences. Because the traditional method of fault detection requires a lot of expertise, the diagnosis of rotating machinery defects is carried out by Resnet convolutional neural network. The lack of training data is likely to cause poor training results when training neural networks. Therefore, in this paper, five different defect groups and a normal group are made, and they are divided into mild, moderate and severe groups according to the severity of the defects. After capturing the vibration signal, through a series of signal processing to draw its time domain diagram, frequency domain diagram, energy spectrum, phase spectrum and energy spectrum density and diagnose defects. In this experiment, it was proved that the neural network can learn only two groups of features with different defect levels (mild and severe), and successfully diagnose the unlearned defect levels (moderate), and the accuracy rate can reach 99.25%. Successfully categorized six different categories.

並列關鍵字

machine learning ； deep learning ； convolutional neural network ； residual network ； vibration analysis

參考文獻

參考文獻

[1] E. Kuriscak, P. Marsalek, J. Stroffek, and P. G. Toth, "Biological context of Hebb learning in artificial neural networks, a review," Neurocomputing, vol. 152, pp. 27-35, 2015/03/25/ 2015, doi: https://doi.org/10.1016/j.neucom.2014.11.022.

[2] R. Seising, "The Emergence of Fuzzy Sets in the Decade of the Perceptron—Lotfi A. Zadeh’s and Frank Rosenblatt’s Research Work on Pattern Classification," Mathematics, vol. 6, p. 110, 06/26 2018, doi: 10.3390/math6070110.

[3] C. Cortes and V. Vapnik, "Support-vector networks," Machine Learning, vol. 20, no. 3, pp. 273-297, 1995/09/01 1995, doi: 10.1007/BF00994018.

[4] Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition," Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, 1998, doi: 10.1109/5.726791.