Abstract: Nowadays, either wind tunnel tests or CFD (Computational Fluid Dynamics) simulations are time-consuming and inefficient to identify bridge deck flutter derivatives, since a large number of tests or simulations must be performed repeatedly at various reduced wind speeds. In the work reported herein, a method, based on the CFD discrete-time model, is presented, which can effectively extract flutter derivatives of a bridge deck at arbitrary reduced wind speeds among full bandwidth of reduced wind speeds. First, unsteady aerodynamic forces acting on a bridge deck are obtained by using the Arbitrary-Lagrangian-Eulerian (ALE) descriptions in combination with finite volume method and multigrid algorithm. In CFD simulations, forced vertical or torsional displancements in form of exponential pulse time series which defined on interesting range of frequencies are applied to the bridge deck. Then, based on the obtained aerodynamic forces and the displacement inputs, discrete time aerodynamic models can be developed which can represent the unsteady aerodynamic behaviors of the bridge deck. Finally, the aerodynamic models are simulated to obtain unsteady aerodynamic forces to simple-harmonic displacement inputs. With those input and obtained aerodynamic forces, bridge deck flutter derivatives can be identified through a system identification algorithm. The discrete-time aerodynamic models can be developed by the presented method which only employs one run of CFD simulation in heaving or pitching directions respectively, leading to a significant decrease in computing time. Flutter derivatives and flutter onset wind speeds of the Sanchaji Bridge are investigated. Reasonable agreements between results from the present method and those from wind tunnel tests demonstrate the reliability and efficiency of the method.