A new multi-modal decomposition algorithm based on frequency spectrum clustering

With frequency spectrum clustering, a new multi-modal decomposition algorithm is put forward in this paper. The spectrum values of frequency response function are looked upon as samples waited for clustering, and the modals are looked upon as classes in clustering. At first using the clustering updated algorithm, a piece of every frequency value is clustered to different modal classes. That is, each modal class obtains itself response frequency values by clustering. Then, the parameters of modals are calculated with the single modal fitting algorithm. The multi-modal decomposition algorithm with frequency spectrum clustering can obtain modal parameters quickly and accurately. And it is proved that the algorithm is a local convergence algorithm at least. There are amount emulation experiments to test the algorithm performance within different noise strength and different influence of wrong modal numbers. The emulation experiments indicate that the performance of the algorithm in this paper is better than the Levy algorithm under different noise strength and different influence of wrong modal numbers.