DETECTION OF DYNAMIC PROPERTIES OF STRUCTURAL SYSTEMS BY THE AUTOREGRESSIVE MOVING AVERAGE METHOD

This paper is concerned with the detection of natural periods and damping factors from uniformlysampled data of the response of a linear structural system subjected to stationary random excitation.The analitical procedure is based on the least-square method (LS) for synthesizing the discrete timeseries. The discrete time series of uniformly spased samples from the output of a linear differentialequation excited by the stationaly white noise is represented as an autoregressive moving averagetime series (AR-MA). And the parameters of the AR-MA model is computed by a tow stageleast-square method (2SLS). In the first LS stage, the autoregressive parameters are estimated fromthe sequence of data by solving the Yule-Walker equation and by using the concept of maximumentoropy proposed by Burg. Numerical examinations show that the computational expense for theestimation of structural properties by the former method is much more than the latter method.This paper is concerned with the detection of natural periods and damping factors from uniformlysampled data of the response of a linear structural system subjected to stationary random excitation.The analitical procedure is based on the least-square method (LS) for synthesizing the discrete timeseries. The discrete time series of uniformly spased samples from the output of a linear differentialequation excited by the stationaly white noise is represented as an autoregressive moving averagetime series (AR-MA). And the parameters of the AR-MA model is computed by a tow stageleast-square method (2SLS). In the first LS stage, the autoregressive parameters are estimated fromthe sequence of data by solving the Yule-Walker equation and by using the concept of maximumentoropy proposed by Burg. Numerical examinations show that the computational expense for theestimation of structural properties by the former method is much more than the latter method