Few past system identification studies have attempted structural model
s with many degrees-of-freedom (DOFs) owing to the numerical difficult
y with regard to convergence and computational speed. Even fewer studi
es have considered a three-dimensional building with the torsional res
ponse taken into consideration. in this respect, an 'improved condensa
tion' method for the system identification of multistorey three-dimens
ional frame buildings is developed. Specifically, reductions in storey
stiffnesses are determined for the quantification of structural damag
e or deterioration in each storey of a building. Both static and kinem
atic condensations are employed to reduce the DOFs of a (complete) mat
hematical model for the building, resulting in a condensed model of mu
ch fewer DOFs. The modelling error is minimized by introducing a remed
ial model whose parameters are identified by using the extended Kalman
filtering technique. A stiffness correction factor is computed to suc
cessively update the complete model. Finally the method yields integri
ty indices to quantify changes in the respective storey stiffnesses. A
numerical simulation example of an asymmetric three-storey frame buil
ding which has a total of 270 DOFs is presented. Three cases of differ
ent damage status of the building are considered. The proposed method
is implemented to identify the storey damages, under the influence of
input and output noise. The numerical results show that this method ac
curately reports the extent of stiffness reduction at each storey of t
he building and is computationally efficient because of a smaller syst
em used for identification in the remedial model.