Fan blades are ones of the core components of the gas path of a high bypass ratio turbofan engine， and fan outlet guide vane detachment is a severe failure mode. This failure could potentially damage the aircraft or other engine components， leading to catastrophic accidents. Therefore， early warning of fan outlet guide vane detachment has become an important task. However， due to the subtle early features of this type of failure， existing methods struggle to effectively warn against it. Therefore， to address this issue， a failure warning method based on deep feature extraction and support vector data description （SVDD） is proposed using monitoring data， aiming to achieve early warning of fan outlet guide vane detachment. First， a modeling method based on engine gas path performance identification is used to establish an observation model of specific engine performance parameters for deep feature extraction. The difference between the real state quantity and the model observation quantity is used as the feature of whether the aero engine has a failure. Second， the SVDD algorithm is used to establish a decision boundary， realizing the automatic division of failure data. The threshold generated by the decision boundary can provide an alarm within a certain time before the failure occurs. Finally， after multiple calculations， the results show that in the interval from the early stage of the failure to the occurrence of the failure， the performance parameters characterizing its health status have a large deviation from the observation quantity， indicating the effectiveness of the selected features. The failure simulation data generated using data augmentation methods are compared and verified with real data. Compared with the actual time of the failure， the warning model realizes the warning of the failure on average 3.14 h in advance.