汽车工程 ›› 2021, Vol. 43 ›› Issue (12): 1858-1864.doi: 10.19562/j.chinasae.qcgc.2021.12.016

• • 上一篇    下一篇

电动汽车关门声品质预测模型研究

刘哲,高云凯(),解馥荣   

  1. 同济大学汽车学院,上海  201804
  • 收稿日期:2021-08-02 修回日期:2021-08-29 出版日期:2021-12-25 发布日期:2021-12-24
  • 通讯作者: 高云凯 E-mail:gaoyunkai@tongji.edu.cn
  • 基金资助:
    上海市科委项目(20511104601);国家自然科学基金(51575399);国家重点研发计划(2016YFB0101602)

Study on Predictive Models for Door Slamming Sound Quality of an Electric Vehicle

Zhe Liu,Yunkai Gao(),Furong Xie   

  1. School of Automotive Studies,Tongji University,Shanghai  201804
  • Received:2021-08-02 Revised:2021-08-29 Online:2021-12-25 Published:2021-12-24
  • Contact: Yunkai Gao E-mail:gaoyunkai@tongji.edu.cn

摘要:

鉴于汽车关门声品质直接影响消费者的购买意愿,本文中对某一电动汽车的关门声品质的预测模型进行研究。首先进行了多工况车门关闭试验,采集了驾驶员耳旁的多组噪声样本,接着,提出并测定了6项客观声品质评价指标,同时进行烦躁度的主观评定,分析了6项客观声品质评价指标与主观评定烦躁度的相关性。然后利用遗传-反向传播(GA-BP)神经网络建立了其主观声品质预测模型,并在上述分析的相关性基础上,建立了反映烦躁度与客观声品质指标之间关系的多元线性回归预测模型。最后,利用5个随机的噪声样本对两种预测模型进行对比验证。结果表明, GA-BP神经网络的预测精度优于多元线性回归模型。

关键词: 电动汽车, 关门噪声, 声品质评价, 预测模型

Abstract:

In view of that the sound quality of vehicle door slamming noise directly affects the purchase intention of customer, the predictive models for the door slamming sound quality of an electric vehicle are studied in this paper. Firstly, a multi-condition door slamming test is carried out with several groups of noise samples near driver’s ear collected. Then, six objective sound quality evaluation indicators are proposed and measured, meanwhile the subjective evaluation on the degree of annoyance is conducted, with the correlation between objective sound quality evaluation indicators and subjective annoyance evaluation analyzed. Next, a subjective sound quality predictive model is created by using GA-BP neural network on one hand, while a multiple linear regression predictive model, reflecting the relationship between subjective annoyance and objective sound quality indicators is established based on the correlation analysis mentioned above on the other hand. Finally, by utilizing five random noise samples, a comparative verification on two predictive models is performed. The results show that the prediction accuracy of GA-BP neural network is higher than that of multiple linear regression model.

Key words: electric vehicles, door slamming noise, sound quality evaluation, predictive models