The fault diagnosis of an aero-engine always plagues the industry due to its complex internal structure. In light of this problem， this paper proposes a fault diagnosis method based on the deep self-coding network. First， the monitoring data is preprocessed， and the structure of the network is constructed according to the data's features. Then， the unlabeled data samples train the network for the initial values of its parameters， and the labeled samples work for a slight adjustment. Thus， the aeroengine fault diagnosis model based on deep self-encoding neural network is established. Finally， the proposed model presents its advantages over common fault diagnosis methods， including back propagation neural network and radial basis neural network， in accuracy according to the tested of labeled samples.