In a previous paper, we applied multivariate Autoregressive Moving-Average (ARMA) models coupled with a basis functions approach to perform the modal identification of time-varying structures. Thanks to the multivariate modelling, it is possible to get modal parameters, including mode shapes, at any time instant. The drawback is that, in order to get the varying modal parameters, an eigenvalue decomposition of a varying companion matrix is required for each time step. To tackle this problem, an alternative parameterization is chosen to switch from the ARMA domain to the modal domain. There are several advantages to use such a parameterization. First, the number of parameters may be reduced in some cases, which is in line with the parsimony principle. Next, because the modal parameters appear explicitly in the modal domain parameterization, there is no need to perform an eigenvalue decomposition to obtain them. The present paper proposes to perform that change in parameterization in the time-varying framework, always using the basis functions approach. The method is applied on an experimental laboratory structure.
展开▼