An improved unscented Kalman filter approach is implemented to estimate induced displacements and changes in structural properties (stiffness, frequencies and damping) during the forced dynamic response of multistory buildings to seismic excitations. The methodology is validated using a fiber-based nonlinear model of a 4-story 4-bays reinforced concrete (RC) frame building subjected to a set of earthquakes causing different levels of inelastic demand on the structure. The variation of the dynamic properties is successfully estimated by iterative updating the filter parameters. The estimated peak values of stiffness and damping reached during the seismic excitation agree with peak inelastic demand values and seem appropriate for detection and damage diagnosis of RC structures.