Time-Domain Signal Reconstruction of Vehicle Interior Noise Based on Data-Driven Method

Abstract

Background: In the process of high-speed driving, there are many source signals that affect the ear noise of passengers in the car. It is important to obtain the reference signal of Active Noise Control (ANC) of the vehicle at high speed condition. Objective: This paper introduces a method to study the time-domain signal reconstruction of interior noise based on data-driven method. Methods: Based on the noise signal collected in a car, the key point signals affecting the interior noise are determined by the acoustic transfer path analysis method. Considering the time-varying characteristics of the noise signal and the complex nonlinear relationship of interior noise, a noise reconstruction model based on wavelet decomposition Radial Basis Function (RBF) neural network is established. And the BP neural network noise reconstruction model is set up to compare the reconstruction effect. Results: According to the reconstruction comparison, the average absolute error (0.0072) of the proposed algorithm model is smaller than the average absolute error of the noise reconstruction BP network model based on wavelet decomposition (0.0280), and the accuracy is improved by 74.29%. The average absolute error between the reconstructed value of RBF neural network and the real-value is smaller than that of BP neural network, and the error of the proposed model is less than 0.01. The method proposed in this paper can reconstruct the interior noise signal of vehicle accurately and effectively. Conclusion: This paper proposes a reconstruction model of vehicle interior noise signal based on wavelet decomposition RBF neural network algorithm driven by data-driven, and verifies the effectiveness of the algorithm with the real vehicle test data. The reconstruction method of RBF neural network based on data-driven wavelet decomposition provides a certain reference value for ANC to obtain high-precision reference signal.