This study presents an unscented Kalman filter (UKF) approach for identification of linear or nonlinear flutter derivatives (FDs) of bridge decks from free vibration or buffeting response time history. The nonlinear FDs, which are dependent on torsional vibration amplitude, are represented in polynomial functions of torsional displacement. The augmented state variables of the two degrees of freedom (2DOF) bridge deck system, which include bridge deck motions and unknown FDs, are estimated simultaneously with the UKF approach based on response measurement data. Firstly, the steady-state vortex-induced vibration and flutter of a streamlined bridge deck section are used to illustrate the performance of UKF approach in extracting nonlinear FDs. The equivalent linear FDs are also identified from the same response data which reveals the deficiency of the linear model. Secondly, the stochastic buffeting responses of the bridge deck contributed from two modal responses with linear FDs are generated, and the performance of UKF approach with unknown excitations is examined. It is pointed out that the buffeting response must be separated into two modal response components, such that the unknown buffeting force excitations can be modeled as white noise processes, and the UKF approach offers satisfactory identification of FDs.

这项研究提出了一种无味卡尔曼滤波器（UKF）方法，用于从自由振动或抖振响应时间历史中识别桥面板的线性或非线性颤动导数（FDs）。非线性FD取决于扭转振动幅度，以扭转位移的多项式函数表示。两种自由度（2DOF）桥面系统的增强状态变量（包括桥面运动和未知FD）是根据响应测量数据与UKF方法同时估算的。首先，利用流线型桥面板截面的稳态涡激振动和颤振来说明UKF方法在提取非线性FD方面的性能。还可以从相同的响应数据中识别出等效的线性FD，这揭示了线性模型的不足。其次，生成了由两个具有线性FD的模态响应贡献的桥面板的随机抖振响应，并研究了具有未知激励的UKF方法的性能。需要指出的是，抖振响应必须分为两个模态响应分量，以便可以将未知的抖振力激励建模为白噪声过程，并且UKF方法可以提供令人满意的FD识别。