The failure mechanism of gun barrel is a rather complex comprehensive process, and the current theoretical models based on mechanism derivation is less accurate and cannot meet the actual use of combat troops. In order to predict the life of gun barrel more accurately, the life prediction algorithm of gun barrel based on the extreme gradient boosting model is proposed based on the wear of the inner diameter of gun barrel as the life prediction index. The algorithm takes the amount of artillery shell shot as the input, takes the inner diameter wear of the barrel as the output, and continuously integrates multiple weak learners, repeatedly trains to fit the residual difference between the predicted value and the actual value of the previous weak learner, thereby generating a strong learner, and reduces the risk of model overfitting by adding regularization terms after the loss function and using pruning technology. Based on the measured data of a certain type of artillery, the results show that the proposed model not only effectively solves the mapping relationship between the amount of shell shot and the wear of the inner diameter of barrel, but also significantly improves the prediction accuracy of the life of barrel compared with the existing algorithms such as support vector machine, BP neural network and gray model.