To monitor the working conditions of UAV's actuator system in time, a novel fault-diagnosis approach was proposed based on Ensemble Empirical Mode Decomposition (EEMD) and the multi-class Support Vector Machine (SVM) with posterior probability. The actuator signals obtained under five typical working conditions were taken as the object of study, i.e., normal condition, loose condition, damaged condition, stuck condition and reverse condition.First, EEMD was made to the collected signals, which were decomposed into a series of Intrinsic Mode Functions (IMFs) with simple components. Then, the energy values of the components of each order were calculated, by which the signal feature vector was obtained. Finally, the multi-class SVM based on posterior probability was established according to the feature information, and thus the type of the aircraft actuator system fault was identified.Simulation results show that the proposed approach can be applied to the fault diagnosis of the actuator system.