低频改进的OMP-DAMAS变压器噪声源定位算法

Improved OMP-DAMAS Noise Source Location Algorithm for Transformer Noise Characteristics

摘要:

针对常规波束形成算法对变压器噪声源定位时低频声源定位的失效问题，采用截断奇异值分解的方法对常规OMP-DAMAS波束形成算法进行改进。首先对常规OMP-DAMAS算法在低频声源定位时效果不好的原因进行分析，可知其主要原因为计算因子筛选过程易受到计算因子间相关性的干扰。针对该问题，通过截断奇异值分解的数据拟合方法对声源分布进行求解，计算出声源分布的粗略先验解，进而指导OMP-DAMAS算法的计算因子筛选，解决了 OMP-DAMAS 算法计算因子筛选受相关性干扰的核心问题，实现低频声源精准定位。最后通过对250Hz、500Hz、750Hz的声源下传统波束形成算法、OMP-DAMSA算法以及改进OMP-DAMSA算法的定位效果进行数值仿真。通过对仿真结果的对比分析，可知低频改进OMP-DAMAS算法可以有效提高对低频声源的定位能力，有效减小主瓣宽度、旁瓣幅值，提高定位精度，在检测距离为5米，采用孔径为2米的16阵元十字型阵列的条件下，可以准确识别低频声源位置，定位误差小于0.1米。

Abstract:

To address the failure of conventional beamforming algorithms in localizing low-frequency noise sources of transformers, the conventional OMP-DAMAS beamforming algorithm is improved using the truncated singular value decomposition method. First, the reason for the poor performance of the conventional OMP-DAMAS algorithm in low-frequency sound source localization is analyzed. It is found that the main cause is that the screening process of computational factors is easily disturbed by the correlation among computational factors. To solve this problem, the sound source distribution is solved by a data fitting method based on truncated singular value decomposition, and a rough prior solution of the sound source distribution is obtained, which further guides the screening of computational factors in the OMP-DAMAS algorithm. This resolves the core issue of correlation interference in computational factor screening for the OMP-DAMAS algorithm and achieves accurate localization of low-frequency sound sources. Finally, numerical simulations are conducted to compare the localization performance of the traditional beamforming algorithm, the OMP-DAMAS algorithm, and the improved OMP-DAMAS algorithm at sound source frequencies of 250 Hz, 500 Hz, and 750 Hz. Comparative analysis of the simulation results shows that the improved low-frequency OMP-DAMAS algorithm can effectively enhance the localization capability for low-frequency sound sources, significantly reduce the main lobe width and side lobe amplitude, and improve localization accuracy. Under the conditions of a detection distance of 5 meters and a 16-element cross array with an aperture of 2 meters, it can accurately identify the position of low-frequency sound sources with a localization error of less than 0.1 meter.