A two stage approach is proposed for non-parametric identification of strong nonlinear structures with limited measurements in this paper.In the first stage,an equivalent linear system for a nonlinear system is proposed and identified using extended Kalman filter(EKF)with limited output measurements.Then,the locations of nonlinearity are identified by comparing the differences between the identified parameters of the equivalent linear system and the linear system.In the second stage,power series polynomials are utilized to approximate the nonlinear restoring forces in the identified locations of nonlinearity.The parameters of these power series polynomials together with the linear parameters of the rest linear parts of the nonlinear system are identified by unscented Kalman filter(UKF)with the incomplete measurements.