An adaptive substructure testing method with unscented Kalman filter on the basis of traditional substructure pseudo-dynamic testing method is proposed in order to deal with the negative influence from the inaccuracy of the resisting force model of numerical substructure. An online identification technology is adopted to estimate hysteretic model parameters of the testing substructure and modify the model parameters of the numerical substructures in the substructure testing. Fast and accurate hysteretic model identification is the key issue of adaptive substructure testing. In this paper, the parameters of hysteretic model of experimental substructure are taken as part of the state vector of experimental substructure and identified online with unscented Kalman filter. The proposed adaptive substructure testing method is studied through numerical simulation. The results show the new method has good accuracy and high computation efficiency, and test results are significantly improved over conventional substructure method with unchanged mathematical model of numerical substructure.