A lumped parameter model is constructed to accurately predict the dynamic degradation characteristics of the fiber-reinforced composite sheet based on the thermal vibration measurement data. In the modeling process, a multiple-lumped-mass division technique is adopted along the longitudinal and width directions with consideration of thermal degradation on the connective stiffness and damping matrix. By solving the natural frequencies and vibration responses of the structure, a determination method for the key parameters of the model, such as the additional thermal stiffness matrix, the connective stiffness, and the damping coefficients, is proposed. In addition, the constructed model is verified by comparing the theoretical natural frequencies and vibration responses with the experimental results at different degradation time points and temperatures, and the calculation errors are within an acceptable range. Both the theoretical and experimental results reveal a gradual weakened downtrend of natural frequencies and a firstly increasing and then decreasing resonant response phenomenon. Also, some possible reasons for the above complex dynamic degradation phenomenon are provided.