Civil engineer usual appear time-varying characteristic, it is very important to identify instantaneous frequency for structural health monitoring and safety evaluation. A method about instantaneous frequency identification of a structure based on variational mode decomposition (VMD) and generalized morse wavelets (GMW) is proposed. Firstly, fast iterative shrinkage-thresholding algorithm is used to reduce noise, which utilize the sparse feature of signal by transforming to wavelet region. And then, an identification theory of time-varying parameter based on VMD and GMW is presented. A numerical stimulation is used to validate the effectiveness of the method. Finally, an experiment model of moving vehicle across a beam is built. The modal test is proceed and the finite element model is calibrated. The measured stain data is used to estimate the position of vehicle on beam at any time, and the accelerate response is used to identify the instantaneous frequency by proposed method. The results show that the proposed method has good validation and accuracy to identify the instantaneous frequencies.