Treatment of arbitrarily autocorrelated load functions in the scope of parameter identification

Ambient vibration tests are output-only tests based on the structural response recorded under natural excitation, such as wind or traffic loads. In contrast to forced vibration test, they do not require service interruption and expensive devices to excite the structure. To this aim, we combine the recently proposed H-fractional spectral moments representation of stationary processes with a modification of the Kalman filter to the scope of structural parameter identification. This paper shows that the method is particularly suited to dealing with long-correlated loads, i.e. stochastic processes with inverse power-law correlation, where many existing methods are not applicable or insufficiently accurate.

► Applicable for arbitrarily correlated loads without restriction to the functional form of the PSD.
► It is efficient also in case of long-memory processes.
► The parameters of the model for the load process are known in analytical form.
► It is easily extendible to multivariate loads.
► It is efficiently combined to the EKF used for system identification.