Aluminum alloy stiffened panels are the key structural form of aerospace equipment such as airplanes， satellites， space stations， and cargo spacecraft. Monitoring and identifying their damage status is the prerequisite and basis for assessing the health status of spacecraft. Aiming at the dependence of traditional machine learning methods on artificial feature extraction， the method of combining experimental simulation and numerical simulation is used to obtain the damage acoustic emission signal of aluminum alloy stiffened plate and calculate its amplitude-frequency characteristics， establish a damage data set. Based on a deep belief network， the impact damage intelligent recognition model is constructed to extract the damage feature adaptively. Combined with the Softmax classifier， the impact damage recognition research is carried out， and the recognition results are compared with the traditional support vector machine and back propagation （BP） neural network. Experimental results show that： 68 test areas have been identified for multiple impact damage on a 2 200 mm×500 mm×10 mm aluminum alloy stiffened plate， and 15 218 times of impact damage have been accurately identified in 15 300 experiments. The correct rate is 99.47%. The research results provide an effective method for damage monitoring of spacecraft structures.