A fault diagnosis scheme based on neural network and adaptive observer is proposed to solve the problem of control surface fault of UAV with multi-operation surface layout. Firstly, a 6-DOF dynamic and kinematic nonlinear model of multi-operating surface layout UAV is established. Then, according to the fault model, an adaptive fault diagnosis observer based on neural network is proposed. The desired output is obtained by a nonlinear observer containing unknown inputs. The errors of the actual output and the expected output are taken as the input of the neural network to detect and identify the faults online, and the errors are fed back to the observer, which constitutes the adaptive term of the observer. The simulation results show that this method can estimate the fault type quickly and ensure the accuracy of the fault diagnosis results.