Research on mechanical rotor condition monitoring based on VCNN

• Published in: Energy procedia Vol. 158; pp. 6393 - 6398
• Main Authors: Xiaoxun, Zhu, Jianhong, Zhao, Dongnan, Hou, Zhonghe, Han
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.02.2019
• Subjects: Condition monitoring; Deep learning; Feature extraction; Vbration signal; Vector convolutional Neural network; Vector convolutional Neural network; Condition monitoring; Deep learning; Vbration signal; Feature extraction
• ISSN: 1876-6102; 1876-6102
• DOI: 10.1016/j.egypro.2019.01.208

A vector convolutional neural network based on convolutional neural network was proposed to realize the recognition of one-dimensional vectors. Aiming at the lack of deep learning and subjectivity in rotor condition monitoring feature extraction, a feature extraction method for rotor condition monitoring based on vector convolutional neural network was proposed. The method can directly input the vibration signal into the network, extract features through the convolution layer and the sub-sampling layer, and finally realize the vibration signal identification through the output layer of the network. This method avoids the subjectivity of feature extraction, and the depth of the vibration signal is now learned. The feasibility of the method is verified by experimental research on the method.