主办：陕西省汽车工程学会
ISSN 1671-7988  CN 61-1394/TH
创刊：1976年
汽车实用技术

汽车实用技术 ›› 2023, Vol. 48 ›› Issue (20): 39-45.DOI: 10.16638/j.cnki.1671-7988.2023.020.009

• 设计研究 • 上一篇    

基于 GS-SVM 的发动机异响在线检测平台

杨兴国，余 瑶   

  1. 重庆工商职业学院 智能制造与汽车学院
  • 出版日期:2023-10-30 发布日期:2023-10-30
  • 通讯作者: 杨兴国
  • 作者简介:杨兴国（1986－），男，博士研究生，讲师，研究方向为汽车噪声与振动控制，E-mail:yanglixgy@163.com。

The Online Detection Platform of Engine Abnormal Noise Based on GS-SVM

YANG Xingguo, YU Yao   

  1. College of Intelligent Manufacturing and Automobile, Chongqing Technology and Business Institute
  • Online:2023-10-30 Published:2023-10-30
  • Contact: YANG Xingguo
  • Supported by:
    重庆市教育委员会科学技术研究项目（KJQN202004005）；重庆工商职业学院重点科研项目（NDZD2020-02）。

PDF

摘要： 为了解决人工听诊法进行发动机异响识别时产生的劳动强度大、工作效率低与准确率 波动等问题，文章提出一种基于网格搜索-支持向量机（GS-SVM）的发动机异响在线检测技 术。该平台主要包括转速监测、信号采集、信号去噪、特征提取和模式识别等功能，LabVIEW 软件负责发动机转速监测和信号采集，并将信号传输至 MATLAB 接口。在 MATLAB 软件中， 首先利用小波相关系数滤波法去除背景噪声；然后分别利用小波包变换和双谱估计提取信号 特征，经归一化处理的信号特征作为支持向量机进行模式识别的输入向量；接着选择分类器- 支持向量机（C-SVC）和径向基核函数（RBF），并采用改进的网格搜索法优化参数 c 和 g， 建立分类模型；最后利用训练成熟的模型预测发动机异响类型。经过测试，该方法的准确率 在 90%以上，具有一定的工程意义。

关键词: 发动机异响；小波包；双谱估计；GS-SVM；在线检测平台

Abstract: In order to solve the problems of high labor intensity, low work efficiency and fluctuating accuracy rate in engine abnormal noise recognition using manual auscultation, this paper proposes an online detection technology for engine abnormal noise based on grid search- support vector machine (GS-SVM). The platform mainly includes functions such as speed monito- ring, signal acquisition, signal denoising, feature extraction, and pattern recognition, and the LabVIEW software is responsible for monitoring engine speed and collecting signals, and transmitting the signals to the MATLAB interface. Firstly, the wavelet correlation filtering method is used to remove background noise in MATLAB software; Then, wavelet packet transform and bispectral estimation are used to extract signal features, and the normalized signal features are used as input vectors for support vector machine pattern recognition; Next, select the classifier-support vector machine (C-SVC) and radial basis function (RBF), and use an improved grid search method to optimize the parameters c and g to establish a classification model; Finally, use the trained and mature model to predict the type of abnormal noise. After testing, the accuracy of this method is over 90%, which has a certain engineering significance.

Key words: Engine abnormal noise; Wavelet packet; Bispectrum estimation; GS-SVM; Online dete- ction platform