• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2018, Vol. 54 ›› Issue (8): 177-184.doi: 10.3901/JME.2018.08.177

• 运载工程 • 上一篇    下一篇

基于支持向量机模型的复杂非线性系统试验不确定度评定方法

朱大业1, 丁晓红1, 王神龙1, 王海华2, 余慧杰1   

  1. 1. 上海理工大学机械工程学院 上海 200093;
    2. 延锋安道拓座椅有限公司 上海 201315
  • 收稿日期:2017-06-14 修回日期:2017-10-27 出版日期:2018-04-20 发布日期:2018-04-20
  • 通讯作者: 丁晓红(通信作者),女,1965年出生,博士,教授,博士研究生导师。主要研究方向为结构分析及优化设计方法。E-mail:dingxh@usst.edu.cn
  • 作者简介:朱大业,男,1992年出生。主要研究方向为结构优化设计。E-mail:m18817570627@163.com
  • 基金资助:
    上海市科委科研计划资助项目(15110502300)。

Uncertainty Evaluation Method of Complex Nonlinear System Test Based on Support Vector Machine Model

ZHU Daye1, DING Xiaohong1, WANG Shenlong1, WANG Haihua2, YU Huijie1   

  1. 1. School of Mechanical Engineering, University of Shanghai for Science and Technology, Shanghai 200093;
    2. Yanfeng Adient Seating Co., Ltd., Shanghai 201315
  • Received:2017-06-14 Revised:2017-10-27 Online:2018-04-20 Published:2018-04-20

摘要: 复杂非线性系统试验测试结果的不确定度分析对于保证产品的质量可靠性具有重要意义。以评价车辆座椅头枕对乘员颈部保护效果的鞭打试验为例,提出一种基于支持向量机模型的复杂非线性系统试验不确定度评定方法。研究鞭打试验的主要影响因素概率密度分布,采用拉丁超立方抽样对影响因素进行试验设计;利用试验结果,采用最小二乘支持向量机,建立鞭打试验的数学模型,并与BP神经网络建立的数学模型进行对比,结果显示,最小二乘支持向量机建立的数学模型具有更高的预测精度,满足后续不确定度的评定要求;采用蒙特卡罗法实现鞭打试验测量不确定度的评定,并与国家标准《测量不确定度指南》规定的方法得到的不确定度进行了比较。结果表明,对于复杂非线性系统,应用基于支持向量机模型的蒙特卡罗法得到的不确定度评定结果准确可靠,提出的方法可广泛应用于各种产品的复杂试验结果不确定度分析中。

关键词: 测量不确定度, 复杂非线性系统试验, 蒙特卡罗法, 支持向量机

Abstract: Measurement uncertainty analysis of complex nonlinear system test is of great significance to ensure the quality and reliability of products. As a typical complex nonlinear system test, the whiplash test which evaluates the vehicle seat head restraint on the occupant neck protection effect is considered, and an uncertainty evaluation method based on support vector machine model is suggested. The probability density functions of main influence factors of whiplash test is studied, and design of experiments is implemented by using latin hypercube sampling method. The experiment results are utilized to construct the mathematical model of whiplash test based on the least squares support vector machine. The design result is compared with that of Back Propagation artificial neural networks mathematical model, and is shown that the prediction accuracy resulted from the least squares support vector machine is higher, which meets the follow-up assessment requirement. The evaluation of uncertainty of whiplash test is realized by using the Monte Carlo method. By the comparison of uncertainty evaluation based on the method specified by national standard《Guide to the Expression of Uncertainty in Measurement》, the suggested method is more accurate and reliable for complex nonlinear system test. The suggested method can be widely applied to the uncertainty analysis for various complex tests of products.

Key words: complex nonlinear system test, measurement uncertainty, Monte Carlo method, support vector machine

中图分类号: