In order to solve the problem that the calibration equation established by linear regression is not ideal for the aircraft landing gear load calibration experiment, considering the nonlinear influence of factors such as landing gear compression stroke and strain gauge position on the calibration load in the experiment, the landing gear load calibration model is constructed by using AdaBoost and XGBoost nonlinear regression methods using feature fusion and ensemble learning theory. Firstly, experimental data are obtained through landing gear load calibration experiments, and the input feature matrix is established using principal component analysis method. Then, a landing gear load calibration model is constructed, using the three directional loading loads of the landing gear as label vectors. The training and testing sets are divided according to the random sampling principle, and the calibration model is trained using AdaBoost and XGBoost methods. Finally, the load is fitted and predicted in the testing set, and the model is evaluated using four evaluation indicators: root mean square error, average absolute error, determination coefficient and time. The experimental results show that compared with the widely used least squares method, the calibration model established by XGBoost method can better fit the loading load. XGBoost algorithm is more advantageous in scenarios without considering timeliness. The research results have important value for improving the accuracy of aircraft landing gear load measurement and further research on aircraft Structural health monitoring.