基于深度学习的车辆前围声学包隔声性能预测方法  

Prediction Method for Acoustic Package Sound Insulation Performance for Vehicle Front End Based on Deep Learning

作　　者：李振超 乔博磊 刘烁 吴红雷 冯晓明 LI Zhenchao;QIAO Bolei;LIU Shuo;WU Honglei;FENG Xiaoming

机构地区：[1]长城汽车哈弗技术中心,保定071000

出　　处：《上海汽车》2024年第12期6-9,共4页Shanghai Auto

摘　　要：文章提出了一种基于深度学习的车辆前围声学包隔声性能预测方法,利用卷积神经网络(CNN)和循环神经网络(RNN)构建模型,以提高预测精度并缩短设计周期。通过对大量声学包样本数据进行预处理,训练出的混合模型在独立测试集中表现优秀,平均绝对误差小于3%。未来研究将关注数据集的扩展、模型的解释性、集成学习和多任务学习的应用,以及构建实时预测系统以进一步优化声学包设计。这项工作为汽车噪声控制提供了新的思路,有望提升车辆的舒适性和设计效率。A prediction method for the sound insulation performance of vehicle front end acoustic package based on deep learning is proposed,utilizing convolutional neural networks(CNN)and recurrent neural networks(RNN)to construct the model,aiming to improve prediction accuracy and shorten the design cycle.By preprocessing a large amount of acoustic package sample data,the trained hybrid model performs excellently on an independent test set,with an average absolute error of less than 3%.Future research will focus on expanding the dataset,enhancing model interpretabil⁃ity,applying ensemble learning and multitask learning,and building real-time prediction systems to further optimize acoustic packaging design.New perspectives are provided for automotive noise con⁃trol and is expected to enhance vehicle comfort and design efficiency.

关 键 词：深度学习 卷积神经网络 循环神经网络 声学包 隔声性能预测 

分 类 号：U462[机械工程—车辆工程]