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.