基于数据驱动的声源表面振速稀疏恢复方法

A data⁃driven sparse recovery method for surface velocity of the sound source

准确描述声源结构的表面振速具有重要意义,振速描述的准确性主要依赖于声源表面的测点数,而增加测点数会导致测量成本增加.为了解决这一问题,提出一种基于数据驱动的声源表面振速的稀疏恢复方法.首先,通过等效源法利用数值仿真获得振速的训练样本;然后,通过K-SVD字典训练方法训练出声源表面振速的稀疏基;最后,通过稀疏正则化实现从有限的测量数据中恢复整个声源表面的振速.为了验证方法的有效性,给出了简支板的数值仿真,并在消声室内进行了实验验证.仿真与实验的结果表明,在测量点数较少的情况下,使用数据驱动的声源表面振速稀疏恢复方法相较于常规等效源法的恢复精度更高,且该方法的性能更加稳定,为声源表面振速的测量提供了新的方案.

The accurate description of surface velocity of the sound source is of great significance.The accuracy of the description of velocity mainly depends on the number of sampling points,and the increasing number of sampling points leads to the high cost of measurement.To overcome the aforementioned issue,a data⁃driven sparse recovery method for surface velocity of the sound source is proposed in this study.In the method,the velocity data sample is generated by numerical simulations by taking advantage of equivalent source method.Then the sparse basis of surface velocity of the sound source is constructed based on K⁃SVD dictionary learning method and sparse regularization is applied to realize an accurate reconstruction of the surface velocity with limited number of sampling points.To validate the effectiveness of the proposed method,the simulation of simply supported plate is conducted and the experiment is carried out in the anechoic chamber.The results of the simulation and the experiment indicate that compared with the conventional equivalent source method,the proposed method provides a more accurate reconstruction of the surface velocity.Meanwhile,the performance of the proposed method is more stable,which can provide a new scheme for the measurement of the surface velocity of the sound source.