摘要
汽车关门时的声音具有复杂的非线性特征，在对汽车关门声的评价上主要还是依靠具有丰富经验的声品质评价工程师反复倾听采集的关门声信号，然后再逐个打分。更重要的是关门声品质的评价和优化需要在生产出样车后才能够进行，灵活性和时效性较差，整个评价和优化过程冗长繁琐。
本文针对现行的关门声品质评价与优化阶段存在的问题，以某汽车公司车门为研究对象，通过建立关门声品质预测模型，结合仿真的方法，在车门投入制造前对车门的关门声品质做出预判。并对车门密封条结构进行优化，达到改善车门关门声品质的目标。将割裂的声品质和结构两个问题直接联系起来，探索了通过修改结构和材料参数构建良好声品质的过程。本文的主要内容有以下几部分：首先，对关门声样本进行采集，使用成对比较法对关门声样本进行主观评价，基于客观心理声学参数，提出用于对关门声品质进行客观量化的多元线性回归数学模型和BP神经网络模型，将主观的“品质”好坏用量化的数值大小来代表。综合对比建立的多元线性回归数学模型和神经网络模型，采用预测精度更高的BP 神经网络模型对后期仿真得到的关门声信号进行声品质预测；
然后，建立车门的有限元和瞬态边界元模型，利用有限元仿真分析车门关闭时各个零部件的加速度特性，以车门部件加速度为边界条件，利用瞬态边界元法仿真得到车门的声压时间曲线。通过进行关门的振动噪声实验，发现采集到的加速度和声压信号与仿真结果吻合程度较高，分别提取仿真和实验关门声信号的响度和尖锐度输入到已建立的BP神经网络模型中，计算得出仿真与实验预测的声品质得分误差小于10%；
最后，采取调整关门时门板加速度的思路对车门密封条的结构进行优化，选取三种不同截面的车门密封条，仿真分析三种密封条对车门加速度的影响和对关门能量的吸收情况，选取对车门关闭时缓冲效果最好的3号密封条进行关门声学仿真，优化后的车门关门声品质预测得分比优化前的关门声品质预测得分高36分，说明双层缓冲结构的密封条能够有效降低关门时的响度和尖锐度，有利于提高车门的关门声品质。
关键词：关门声；非线性声品质；神经网络；相关分析；密封条
II
Abstract
The sound of a car closing has complex non-linear characteristics.The evaluation of door-closing sound quality mainly relies on experienced engineers to listen to the sound signals, then score them one by one. More importantly, the optimization of the door-closing sound quality needs to be carried out after the production of the prototype car. The flexibility of this method is poor, and the whole optimization process is tedious.
To deal with the problems of door-closing sound quality evaluation and optimization, this paper takes a
car company's door as the research object. The prediction model and simulation method are applied to predict the door-closing sound quality before the vehicles are put into production. By optimizing the structure of the door seal strip to improve the sound quality of the door closure. Connecting the sound quality and structure. The process of building sound quality by modifying structure and material parameters is explored. The main contents of this paper are as follows: Firstly, to collect the signal of door closing. Evaluating the sound quality of door closure with paired comparison method. Multiple linear regression and BP neural network model are proposed to predict the door-closing sound quality based on the objective psychoacoustic parameters. The multi-linear regression model and BP neural network prediction model are compared comprehensively. It’s decided to use the neural network model to predict the sound quality of the simulated sound signal.
Secondly, the finite element and transient boundary element models of the door are established. The acceleration characteristics of each component are analyzed by finite element simulation. Taking acceleration of door components as boundary condition, the sound pressure time curve of door is simulated by transient boundary element method. We release the vibration and noise experiment of door closing. The experimental results are in good agreement with the simulation results. The loudness and sharpness of the simulated and experimental sound signals are extracted and input into
the established BP neural network model. The error between simulation and experimental prediction is less than 10%.
Finally, the acceleration of the door panel is adjusted by optimizing the structure of the door seal strip. Three kinds of door sealing strips with different cross sections are selected to analyze the influence of the car door acceleration and energy absorption. Choosing No. 3 seal strip with the best cushioning effect to optimize door-closing sound quality. The predicted score of door-closing sound quality after optimization is
III
36 points higher than before. The results show that the double cushion sealing strip can effectively reduce the loudness and sharpness and improve the door-closing sound quality.
Key Words: door-closing sound; nonlinear sound quality; neural network; correlation analysis; sealing strip
IV
目录
学位论文原创性声明.......................................................................................... I 学位论文版权使用授权书 .................................................................................. I 摘要 ................................................................................................................ I I Abstract ........................................................................................................... III 第1章绪论 (1)
1.1 选题背景及意义 (1)
1.2 国内外汽车声品质研究现状 (3)
1.3 本文研究内容及章节安排 (5)
第2章关门声品质主观评价与预测 (8)
2.1 关门声品质主观评价 (8)
汽车密封条2.2 关门声品质客观参数 (9)
2.3 关门声品质实验设计 (12)
2.3.1 测试系统 (12)
2.3.2 声样本评价 (13)
2.4 关门声品质多元线性回归建模 (14)
2.4.1 相关分析 (15)
2.4.2 多元线性回归 (17)
2.4.3 模型建立 (18)
2.4.4 模型评价与检验 (18)
2.4.5 模型预测结果 (19)
2.5 基于BP神经网络的关门声品质预测 (20)
2.5.1 BP神经网络学习算法 (21)
2.5.2 BP神经网络结构确定 (23)
2.5.3 BP神经网络模型检验 (26)
2.6 本章小结 (28)
第3章关门声品质仿真分析与验证 (29)
3.1 汽车关门碰撞仿真分析 (29)
3.1.1 关门碰撞有限元理论 (30)
3.1.2 关门碰撞仿真模型建立 (32)
3.1.3 关门碰撞求解设置 (35)
3.2 汽车关门边界元声辐射分析 (36)
3.2.1 声学分析流程 (36)
3.2.2 声学边界元仿真理论 (37)
V
3.2.3 车门边界元模型建立 (38)
3.2.4 瞬态边界元求解设置 (39)
3.3 关门仿真结果验证 (40)
3.4 本章小结 (43)
第4章关门声品质优化 (44)
4.1 车门密封条材料特性研究 (44)
4.2 车门密封条结构变形模拟 (45)
4.3 关门声品质优化效果 (48)
4.4 本章小结 (49)
总结与展望 (50)
参考文献 (52)
致谢 (56)
VI