AN EFFECTIVE METHOD FOR SELECTING PHYSICAL MODES BY VECTOR AUTOREGRESSIVE MODELS

This paper describes an efficient technique for the selection of relevant modes resulting from a vector autoregressive (ARV) model. The technique is based on plotting the deterministic and stochastic contributions of each mode separately. It helps to identify visually how strong a mode is present in the measured data of the system under consideration. The method is simple to program and applicable to any model based on the data-dependent system (DDS) methodology. In order to demonstrate the applicability of the technique, two examples are presented. In the first example, simulated data of a two-degree-of-freedom system are used, while in the second example, real data on a prestressed concrete bridge are considered.