用于近场声全息正则化的共轭梯度法

摘要

Conjugate gradient algorithm is proposed to stabilize the reconstruction process in the distributed source boundary point method (DSBPM) based nearfield acoustic holography (NAH) and to control the influence of measurement errors on the reconstructed results. The optimal iteration number is determined by placing an auxiliary surface between the hologram surface and the source surface and minimizing the relative error between the calculated pressure on the auxiliary surface corresponding to each iterative number and the measured pressure. Compared with the Landweber iterative regularization method,the proposed method can obtain the similar reconstruction precision, however, the iteration time is decreased greatly. Two experiments are investigated to demonstrate the validity and superiority of the method to decrease the reconstruction error.%提出采用共轭梯度正则化方法稳定基于分布源边界点法的近场声全息重建过程,控制测量误差对重建结果的影响.共轭梯度法中的最佳迭代次数通过在全息面与源面之间布置一个小型辅助面,并通过最小化测量值与重建值之间的相对误差来选取.与Landweber迭代正则化方法相比,该方法同样具有较高的重建精度,但迭代时间却大大减少.对实际声源的实验研究验证了采用共轭梯度正则化方法控制近场声全息重建误差影响的有效性及其优越性.