Multi-phase linear regression: A Novel method for the identification of base-isolated buildings using seismic response data

Base-isolation is an increasingly applied earthquake-resistant design technique in highly seismic areas. The in-situ identification of physical parameters of the isolation layer and monitoring of the isolation performance has become a critical issue. In this paper, a simplified system identification method is proposed for obtaining insight into the structural property of based-isolated buildings. A bilinear hysteresis model is chosen for modeling isolation bearings. Interstory restoring forcedeformation hysteresis loops are first generated from recorded seismic response. A two-step regression analysis based method is proposed to identify the linear/nonlinear physical parameters of the isolation layer. In the first step, hysteresis loops are split into many loading/unloading half-cycles according to zero velocity points. Multiple linear regression is applied to all half-cycles to yield equivalent linear system stiffness and damping. Both parameters clearly vary with half-cycle displacement increment. A threshold is used to separate linear and nonlinear half-cycles. For linear half cycles, linear regression analysis estimates viscous damping coefficient. For nonlinear half cycles, a multi-phase linear regression analysis method is applied to estimate nonlinear model parameters. A proof-of-concept numerical investigation shows the feasibility of the method. The method is effective and robust even of 5% noise level.