STATISTICAL MODAL IDENTIFICATION TECHNIQUE BASED ON TIME SERIES ANALYSIS.

This report deals with a new statistical identification method which determines the modal parameters of a dynamic structure system. The method is based on an extension of the AR model or the ARMA model for stationary random processes which represent the relation between the input excitations and the response of the system. After reviewing the general characteristics of various identification methods hitherto commonly used, we propose a technique by which a multivariate ARMA model is introduced. By applying this technique to the general multi-degree-of-freedom systems, not only the natural frequency and the damping ratio, but also the natural mode of vibration are skillfully identified.