A welding damage identification method of ship composite joint based on Convolutional Neural Network (CNN) is proposed. The welding process is simulated with the Ansys software, 5 simulated operation conditions with different damage positions and damage degrees are considered, and the strain response data of composite joints under different operation conditions are collected. A one-dimensional CNN is constructed, and the data are divided into the training set and test set and are put into the neural network for training and testing, so as to verify the applicability of the method to identify different damage positions and damage degrees during the composite joint welding. The results show that the method is of good testing performance in the structural welding damage testing.