计算机应用 ›› 2017, Vol. 37 ›› Issue (3): 786-790.DOI: 10.11772/j.issn.1001-9081.2017.03.786

• 人工智能 • 上一篇    下一篇

基于无监督学习卷积神经网络的振动信号模态参数识别

方宁, 周宇, 叶庆卫, 李玉刚   

  1. 宁波大学 信息科学与工程学院, 浙江 宁波 315211
  • 收稿日期:2016-08-19 修回日期:2016-10-29 出版日期:2017-03-10 发布日期:2017-03-22
  • 通讯作者: 周宇
  • 作者简介:方宁(1992-),女,湖北黄冈人,硕士研究生,主要研究方向:振动信号处理、深度学习;周宇(1960-),男,山东威海人,教授,硕士,主要研究方向:振动信号处理、网络多媒体通信、信息安全;叶庆卫(1970-),男,浙江衢州人,副教授,博士,主要研究方向:振动信号处理、智能优化;李玉刚(1991-),男,安徽合肥人,硕士研究生,主要研究方向:振动信号处理。
  • 基金资助:
    国家自然科学基金资助项目(61071198);浙江省自然科学基金资助项目(LY13F010015);浙江省科技创新团队资助项目(2013TD21);宁波大学科研基金(学科项目)资助项目(xkx11417)。

Modal parameter identification of vibration signal based on unsupervised learning convolutional neural network

FANG Ning, ZHOU Yu, YE Qingwei, LI Yugang   

  1. College of Information Science and Engineering, Ningbo University, Ningbo Zhejiang 315211, China
  • Received:2016-08-19 Revised:2016-10-29 Online:2017-03-10 Published:2017-03-22
  • Supported by:
    This work is partially supported by the National Natural Sciences Foundation of China (61071198), the Natural Science Foundation of Zhejiang Province (LY13F010015 ), the Science and Technology Innovation Team of Zhejiang Province (2013TD21), the Scientific Research Foundation of Ningbo University (xkx11417).

摘要: 针对现有的时域模态参数识别方法大多存在难定阶和抗噪性差的问题,提出一种无监督学习的卷积神经网络(CNN)的振动信号模态识别方法。该算法在卷积神经网络的基础上进行改进。首先,将应用于二维图像处理的卷积神经网络改成处理一维信号的卷积神经网络,其中输入层改成待提取模态参数的振动信号集合,中间层改成若干一维卷积层、抽样层,输出层得到的为信号对应的N阶模态参数集合;然后,在误差评估中,对网络计算结果(N阶模态参数集)进行振动信号重构;最后,将重构信号和输入信号之间差的平方和作为网络学习误差,使得网络变成无监督学习网络,避免模态参数提取算法的定阶难题。实验结果表明,当所构建的卷积神经网络应用于模态参数提取时,与随机子空间识别(SSI)算法及其局部线性嵌入(LLE)算法对比,在噪声干扰下,构建的卷积神经网络识别精度要高于SSI算法与LLE算法,具有抗噪声强、避免了定阶难题的优点。

关键词: 卷积神经网络, 模态参数, 无监督学习, 学习误差, 随机子空间识别, 局部线性嵌入

Abstract: Aiming at the problem that most of the existing time-domain modal parameter identification methods are difficult to set order and resist noise poorly, an unsupervised learning Convolution Neural Network (CNN) method for vibration signal modal identification was proposed. The proposed algorithm was improved on the basis of CNN. Firstly, the CNN applied to two-dimensional image processing was changed into the CNN to deal with one-dimensional signal. The input layer was changed into the vibration signal set of modal parameters to be extracted, and the intermediate layer was changed into several one-dimensional convolution layers, sampled layers, and output layer was the set of N-order modal parameters corresponding to the signal. Then, in the error evaluation, the network calculation result (N-order modal parameter set) was reconstructed by the vibration signals. Finally, the squared sum of the difference between the reconstructed signal and the input signal was taken as the network learning error, which makes the network become an unsupervised learning network, and avoids the ordering problem of modal parameter extraction algorithm. The experimental results show that when the constructed CNN is applied to modal parameter extraction, compared with the Stochastic Subspace Identification (SSI) algorithm and its Local Linear Embedding (LLE) algorithm, the convolutional neural network identification accuracy is higher than that of the SSI algorithm and the LLE algorithm under noise interference. It has strong noise resistance and avoids the ordering problem.

Key words: Convolutional Neural Network (CNN), modal parameter, unsupervised learning, learning error, Stochastic Subspace Identification (SSI), Locally Linear Embedding (LLE)

中图分类号: