The invention discloses a composite material damage detection method based on wavelet analysis. The method only monitors damage in a damage monitoring zone related to damage in a composite material structure without monitoring damage of the whole detected composite material structure covered with a whole optical fiber grating sensor network through the whole optical fiber grating sensor network sothat the roll poling frequency of an optical fiber grating sensor for damage monitoring is greatly reduced, the quantity of the signals of damage needing to be collected and processed is reduced andthe damage monitoring efficiency is effectively improved. A composite material damage signal is analyzed and pre-processed through a wavelet packet so that a damage signal energy spectrum is obtained.The method extracts feature vectors and is accurate and effective. Through a BP neural network model and training, a network having the function of the composite material damage identification mode is obtained, thereby realizing damage intelligent identification, solving some nonlinear problems, realizing damage identification and positioning and degree determination and being effective and feasible.