Abstract: Subjective evaluation of vehicle’s exhaust noises is executed by paired comparison jury test, and the sound quality prediction model of vehicle exhaust noise is established based on back-propagation (BP) neural network. The acoustic characteristic of muffler is obtained through the computational fluid dynamics simulation, and the frequency spectrum of the exhaust noise of the vehicle with muffler fitted is derived. The objective parameters of the simulated exhaust noise, such as loudness, sharpness, roughness and so on, are calculated based on psychoacoustics theory. Then, the sensory pleasantness is simulated by importing these objective parameters into the sound quality prediction model which is established previously. Thus, the whole numerical simulation method for sound quality evaluation of vehicle’s exhaust noise is fulfilled. The reliability of the simulation method is verified by comparison of the simulation results with the experimental data.

Key words: acoustics, exhaust noise, sound quality, subjective evaluation, neural network, computational fluid dynamics, psychoacoustics

摘要： 采用成对比较法对车辆排气噪声进行主观评价，使用BP神经网络建立车辆排气噪声声音品质预测模型。采用计算流体动力学方法对消声器的声学特性进行仿真计算，得到车辆安装消声器后的排气噪声频谱结构，以心理声学理论为基础计算出该虚拟排气噪声频谱的响度、尖锐度、粗糙度、波动度等客观参数，并将计算结果输入到所建立的声音品质预测模型中得到满意度仿真计算结果。从而形成一套完整的车辆排气噪声声品质仿真计算方法，并将计算结果与实验结果进行对比，证明计算方法的可靠性。

关键词: 声学, 排气噪声, 声音品质, 主观评价, 神经网络, 计算流体动力学, 心理声学