ABSTRACT
In this paper, we propose a neural network approach to identify the structural modal shapes of the output-only data. It takes advantage of the sparseness of the response signal in the time-frequency domain. We use the principle of unsupervised learning and takes advantage of sparseness of the response signal in the time-frequency domain, to get the structural modal shapes. A neural network is designed to get the modal shapes. The mixture signals, i.e., the structural response data, first transformed to the time-frequency domain, and then the most sparse point is found. Then we use a neural network to get the cluster center, which is the mode shapes. A numerical example of a simple structure is made to illustrate the modal shape identification ability of the proposed approach.
