The hydraulic system of unmanned minesweepers is prone to failure which can be difficult to diagnose. An XGBoost fault diagnosis model based on an improved gold rush optimizer (GRO) algorithm is proposed to improve the optimization efficiency of the GRO by means of Piecewise chaotic mapping and the Cauchy variational strategy. A fault simulation model of the hydraulic system of an unmanned minesweeper is established, and fault data for common fault types, including solenoid valve faults, hydraulic cylinder faults, hydraulic pump internal leakage faults and filter blockages are obtained. These data are then used as the input to the diagnostic model to classify and diagnose the unmanned minesweeper hydraulic system faults. Finally, the models before and after modification are compared, and the results show that the modified diagnostic model affords improved accuracy in unmanned minesweeper hydraulic system fault diagnosis.