187 / 2021-03-31 20:18:20

Combination ray wave superposition method for near field acoustic holography and neural network construction of its combination coefficient

Wave superposition method,Near-field acoustic holography,Ray wave function,Neural Networks

Near field acoustic holography based on the wave superposition method has been widely used in the sound source identification, location and sound field analysis in recent years because of its advantages in adaptability and numerical calculation. The wave superposition method generally uses the spherical monopole function which is only related to the distance, and its transfer matrix is usually an ill conditioned matrix, which affects the stability of numerical calculation. In this paper, by adding the directional ray wave function to the monopole function, the ill condition of the transfer matrix is effectively improved. However, the combination coefficient of the ray wave function has a certain influence on the accuracy of sound field calculation. If the combination coefficient is too small, it can not effectively improve the ill condition of transfer matrix. If it is too large, the singularity of transfer matrix will increase. Therefore, it is necessary to cyclically select the optimal combination coefficient according to the sound source model, but this process will greatly reduce the computational efficiency of sound field reconstruction. In order to improve the computational efficiency, RBF (radial basis function) neural network algorithm is adopted to construct the mapping of the parameters of sound source model to the combination coefficients of ray wave function. The results of numerical examples show that RBF neural network can quickly calculate the combined ray wave function corresponding to the model. Compared with the traditional monopole wavelet function, the combined ray wave function not only improves the ill conditioned property of the transfer matrix, but also the reconstruction accuracy of the sound field.