A sensing network was constructed by a Fiber Bragg Grating (FBG) sensor, and a mode identification algorithm for the damage of carbon composite materials was researched by combining the wavelet decomposition and reconstruction, frequency spectrum analysis and support vector classifier algorithm. Firstly, according to the impact test on a composite material structure with different damage modes, the relationship between damage modes and signal characteristics was explored. Then, the wavelet decomposition and reconstruction were used to remove the signal baseline interference, and the identification method of damage mode for composite materials was constructed based on extracting signal amplitude frequency characteristic by Fourier transform analysis. Finally, the extracted signal amplitude frequency characteristic was used as an input and the damage mode as an output, the damage modes of composite materials were identified based on support vector classifiers. An experimental area with 200 mm×200 mm was selected in a 500 mm×500 mm×2 mm carbon fiber reinforced plastic plate center. The 30 damage modes are tested, and the result shows that 29 damage modes are identified accurately with identification accuracy of 96.7%. The research provides a reliable method for the identification of damage modes of composite materials.