Probabilistic Output-Only Identification of Soil-Structure Systems Using Shear Beam Model

This paper proposes a novel probabilistic framework for output-only identification of a soil-structure system. the system is modeled by a vertical shear beam resting on the soil representative springs. the proposed framework estimates the unknown parameters of the system and the foundation input motion time history simultaneously, using sparsely measured responses of the system. the unknown parameters of the system include stiffness of the sway and rocking springs, shear modulus of the beam, and modal damping ratios of the system. to this end, the extended kalman filter with unknown-input method is employed. at each time step, an a priori estimate of the identification parameters is produced through dynamic equations of the system. next, the system response is predicted using the a priori estimate. finally, using measured responses of the system, the a priori estimates of the identification parameters are updated to produce the a posteriori estimates. the results include the mean of the identification parameters and the foundation input motion along with an estimate of their uncertainties. using the continuous shear beam as the representative of the structure leads to a computationally-feasible framework for structural health monitoring and damage identification of civil structures at regional scale. the proposed approach is verified through a synthetic example in which the simulated absolute acceleration responses of the system to a horizontal seismic excitation are polluted with noise to produce artificial measurements. these measurements are then employed to identify the unknown parameters of the beam and the soil-foundation system as well as the time history of the ground acceleration. the results show that the parameters of the proposed simplified model are highly identifiable. this holds true even in the presence of high measurement noise and for sparsely instrumented structures.