Abstract: At present, numerical simulation is the main method for predicting residual stress produced by selective laser melting. However, due to different factors such as equipment, environment and powder, it is difficult to establish a practical numerical model. The accuracy of numerical simulation needs to be verified. Based on the strong ability of neural network in predicting multivariable and complex linear information processing, a model for predicting residual stress in selective laser melting (SLM) is established. SLM technology is used to print a considerable number of samples with different process parameters, and ultrasonic wave is used to detect the internal residual stress. Training samples of neural network are used to train the neural network model, and the neural network with predictive function is obtained. The process parameters of the validated samples are input into the neural network, and the predicted residual stress value is calculated, which is in accordance with the actual situation. The detection values were compared. The experimental results show that the deviation between the predicted value and the measured value is small and the accuracy is high, which verifies the effectiveness of the proposed method.The method of predicting residual stress by neural network can quickly determine the residual stress corresponding to laser melting process parameters in different selection areas, avoid setting process parameters with high residual stress, effectively shorten the cycle of preparing high-quality workpiece samples and reduce the cost.