In order to reduce ship weight, center of gravity, improve speed, ship using aluminum alloy as superstructure material has become the future trend of the industry. However, the problem of stratification often appears in the actual construction, which affects the quality of the ship. Therefore, a damage recognition method for composite joint welding based on convolutional neural network is proposed. Based on the thermal excitation in the welding process, the strain response data of the composite joint under different conditions were collected by considering five simulated conditions with different damage locations and different damage degrees, including undamage, single damage and multiple damage, and one-dimensional convolutional neural network was constructed. The data was divided into training sets and test sets and put into the neural network for training and testing. The results show that the one-dimensional convolutional neural network can identify the damage position and damage degree of the composite joint in the welding process.