Abstract

Algorithms are realized for localizing vibration activity sources in complex mechanoacoustic systems using finite element modeling. For the case of one source, the localization method was used in the least squares form applied to the first eigenvector of the sample correlation matrix. It is shown that frequency summation can significantly increase the localization accuracy in this case. Considerable attention is paid to assessing the number of simultaneously operating vibration sources and the possibility of correctly localizing them. The efficiency of two source localization methods was compared: the maximum likelihood method calculated for a single source and the MUSIC (Multiple Signal Classification) method specially modified for this problem. It is shown that the modified MUSIC method is more efficient in solving the localization problem with two sources compared to the maximum likelihood method calculated for a single source. It is noted that the maximum likelihood method calculated for two or more sources cannot be applied because of the too high computational complexity.

中文翻译：

---

具有多个自由度的机械声系统中定位多个振动源的可能性的研究

摘要

实现了使用有限元建模在复杂机械声系统中定位振动活动源的算法。对于一个源，以最小二乘形式将定位方法应用于样本相关矩阵的第一个特征向量。结果表明，在这种情况下，频率求和可以显着提高定位精度。要特别注意评估同时运行的振动源的数量以及正确定位振动源的可能性。比较了两种信号源定位方法的效率：为单个信号源计算的最大似然方法和为此问题专门修改的MUSIC（多信号分类）方法。结果表明，与为单个信号源计算的最大似然方法相比，改进的MUSIC方法在解决两个信号源的定位问题上效率更高。请注意，由于计算复杂度过高，无法应用针对两个或多个源计算的最大似然法。