【摘要】 针对传统风洞试验、数值模拟等方法计算噪声值费时长、资源消耗大等问题,提出一种基于机器学习的气动噪声预测方法。以后视镜特征参数为数据集输入,对不同特征参数下的后视镜模型进行瞬态流场与声场联合仿真,将计算得到的总声压级值作为数据集输出,分别用不同数量的样本数据训练支持向量回归机,通过建立的预测模型对同一测试集进行预测得到总声压级预测值。结果表明,基于支持向量回归机的预测方法能得到与计算值误差较小的预测结果,在较少样本数据支撑下也具有较高的预测精度,可用于汽车后视镜气动噪声的预测。更多还原

【Abstract】 In view of the problem that the aerodynamic noise calculation in traditional wind tunnel tests and numerical simulation methods was time-consuming and resource intensive, an aerodynamic noise prediction method based on machine learning was proposed. The features of rearview mirror parameters were chosen as input dataset. The total sound pressure level was calculated by the co-simulation of transient flow field and the sound fields of the rearview mirror model with different characteristic parameters were the output dataset. The support vector regression machine was trained with different amounts of sample data and the prediction models established predict the total sound pressure level by using the same test dataset. The results show that the proposed method based on the support vector regression machine can obtain the prediction results with small errors compared with the calculation results, can achieve high prediction accuracy with few training data, and thus can be used for the prediction of aerodynamic noise generated by the automobile rearview mirror.更多还原