• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2018, Vol. 54 ›› Issue (22): 30-37.doi: 10.3901/JME.2018.22.030

• 仪器科学与技术 • 上一篇    下一篇

基于快速在线支持张量机的柴油机智能诊断方法

许小伟1,2, 张楠1, 严运兵1,2, 秦丽1   

  1. 1. 武汉科技大学汽车与交通工程学院 武汉 430081;
    2. 纯电动汽车动力系统设计与测试湖北省重点实验室 襄阳 441053
  • 收稿日期:2018-01-31 修回日期:2018-07-18 出版日期:2018-11-20 发布日期:2018-11-20
  • 作者简介:许小伟,男,1983年出生,博士,副教授,硕士研究生导师。主要研究方向为汽车电子控制与测试技术、机械设备状态监测与故障诊断技术。E-mail:xxw15@163.com;张楠,女,1994年出生,硕士研究生。主要研究方向为机械设备状态监测与故障诊断技术。E-mail:1361747787@qq.com;严运兵,男,1968年出生,博士,教授,博士研究生导师。主要研究方向为新能源汽车及控制方法、汽车系统动力学分析。E-mail:yanyb@126.com;秦丽,女,1981年出生,博士,副教授,硕士研究生导师。主要研究方向为机械设备动力学耦合理论及方法。E-mail:qinli@wust.edu.cn
  • 基金资助:
    国家自然科学基金(51505345,51509194)、电动汽车动力系统设计与测试湖北省重点实验室基金(HBUASEV2015F005)和湖北省教育厅基金(Q20151105)资助项目。

Intelligent Diagnosis Method of Diesel Engine based on Fast Online Support Tensor Machine

XU Xiaowei1,2, ZHANG Nan1, YAN Yunbing1,2, QIN Li1   

  1. 1. School of Automobile and Traffic Engineering, Wuhan University of Science and Technology, Wuhan 430081;
    2. Hubei Key Laboratory of Power System Design and Test for Electrical Vehicle, Xiangyang 441053
  • Received:2018-01-31 Revised:2018-07-18 Online:2018-11-20 Published:2018-11-20

摘要: 柴油机状态监测信号众多且多为非平稳信号,相互干扰较大且具有非线性和复杂耦合的特征,导致基于向量模式的故障诊断方法难以准确诊断其工作状态。提出一种张量模式下的柴油机智能诊断方法。首先,结合线性支持高阶张量机的学习框架和在线随机梯度下降法的思想,设计带核函数的快速在线支持张量机算法。然后,构建“信号类别×曲轴转角×转速”的三阶张量形式的柴油机状态样本,分别以在线支持向量机、线性支持高阶张量机和快速在线支持张量机三种算法,对某柴油机的失火样本进行故障诊断,以“测试精度”“学习时间”“存储空间”作为评价指标对三种算法进行对比分析。分析结果表明,所设计的快速在线支持张量机算法测试精度较高,学习时间显著降低,所需存储空间很小,解决了超大样本、非线性和高维数据的分类问题,满足了柴油机智能故障诊断的工程应用要求。

关键词: 柴油机, 快速在线支持张量机, 张量模式, 智能诊断

Abstract: It is difficult to establish accurate mathematical models based on vector mode to describe the state of diesel engine due to the non-linear and complex coupling of the monitoring signal sources. Therefore, an intelligent diagnosis method of diesel engine under tensor mode is proposed. First, a fast online support tensor algorithm with the kernel function is developed, which combined the linear support high-order tensor learning framework and the method of online random gradient descent. Second, the diesel engine state samples in the form of third order tensor, "signal type×crank angle×rotate speed", are constructed based on the signal acquired from a diesel engine under different working state. By applying the three algorithms of on-line support vector machine, linear support high order tensor, and fast online support tensor, the diesel engine fire failure samples are analysis to predict whether diesel engine is fire or not. Third, the diagnosed results from the three algorithms are evaluated using the three indicators of test precision, learning time, and storage space. It is found that the fast online support tensor algorithm meets the engineering requirements of the intelligent fault diagnosis for the diesel engine, and it has a higher testing precision, lower learning time and small storage space than the others. This method can solve the classification problem, which involves of super large samples with nonlinear and high dimensions characteristics.

Key words: diesel engine, fast online support tensor machine, intelligent diagnosis, tensor mode

中图分类号: