摘要
准确的路面激励信息对汽车行驶安全性和乘坐舒适性有重要影响。针对目前路面不平度等级识别算法存在复杂和准确率低等问题,提出了一种注意力门控循环单元(BiGRU-Attention)网络和车辆振动响应的路面不平度等级识别算法。首先,通过滤波白噪声法建立随机输入路面模型,1/4车辆振动模型平顺性仿真实验获取车辆振动响应信号,选择滑动窗口为1 s截取样本构造数据集;然后,通过BiGRU网络学习样本信号的深层次信息,Attention机制优化时刻特征对辨识结果的贡献率比重,快速准确地识别出路面不平度等级;最后通过消融实验实现算法的验证。实验结果表明,基于BiGRU-Attention路面不平度等级识别算法的准确率可达96.9%,相比基础模型有1%~3%的提升。该算法能够准确识别路面不平度等级,为车辆动力学控制提供有力的理论依据。
Accurate road surface incentive information has an important impact on the safety and ride comfort of vehicles. Aiming at the problems of complexity and low accuracy in the current road roughness level recognition algorithm, a road roughness level recognition algorithm was proposed based on the BiGRU-Attention network and vehicle vibration responses. Firstly, a random input road model was established by filtering white noise method, and a quarter of the vehicle model ride comfort simulation experiment was used to obtain vehicle vibration response signals to establish a data set. Then, the in-depth information of the sample signals is captured through the BiGRU network and the Attention mechanism optimize the contribution rate of the moment feature to the identification result, so that the road roughness level can be accurately identified. Finally, the verification of the algorithm was achieved through ablation study. The results show that the identification accuracy of the road roughness level recognition algorithm based on BiGRU-Attention can reach 96.9%, which is an improvement of 1% ~3% compared with the basic model. The algorithm can accurately identify the level of road roughness and provide a strong theoretical basis for vehicle dynamics control.
作者
薛俊俊
陈双
Xue Junjun;Chen Shuang(Automobile and Traffic Engineering,Liaoning University of Technology,Jinzhou,Liaoning 121000,China)
出处
《机电工程技术》
2021年第11期66-69,共4页
Mechanical & Electrical Engineering Technology
基金
辽宁省教育厅基础研究项目(编号:JJL202015402)。
关键词
反向分析
路面不平度等级辨识
深度学习
门控循环单元
注意力机制
reverse analysis
road roughness level recognition
deep learning
gated recurrent unit
attention mechanism
