In view of the difficulties in fault detection and lack of maintenance experience of unmanned minesweeping vehicles, a new method with fast detection speed and high diagnostic accuracy was proposed. Based on the ELM algorithm, Levy flight strategy and simulated annealing mechanism are introduced to optimize the condor search algorithm, and the improved Condor search algorithm (IBES) is used to optimize the parameters of the limit learning network. A fault diagnosis model for the dynamic system of unmanned minesweeper is established based on the improved Condor search algorithm to optimize the extreme learning machine. The experimental results show that the fault diagnosis accuracy can reach 98.18%, which is significantly higher than the improved model and other methods, and has theoretical and practical significance.