In order to improve the accuracy of barrel life prediction, a barrel life prediction algorithm based on extreme gradient boost (XGBoost) model was proposed by taking the wear of barrel inner diameter as the life prediction index. Taking the number of artillery shells as the input and the barrel inner diameter wear as the output, the strong learner was generated by integrating multiple weak learners to fit the residual between the predicted value and the actual value of the previous weak learner, and the risk of model overfitting was reduced by adding a regularization term after the loss function and using pruning technology. The model is verified based on the measured data of a certain type of artillery, and the results show that the model not only effectively solves the mapping relationship between the artillery ejection quantity and the barrel inner diameter wear, but also significantly improves the barrel life prediction accuracy compared with the existing algorithms such as support vector machine, BP neural network and gray model.