In order to overcome the defect of traditional engine condition monitoring only depending on single parameter, this paper establishes the five-level condition monitoring alert system with fuzzy neural network (FNN) which is good at settling uncertain and complicated problems. Firstly, optimal monitoring parameters are selected from three aspects of ferrography, vibration and performance parameters. With sufficient historical data, limit values of parameters and reliable five-level condition monitoring standards are maintained and established by statistical analysis. Then fuzzy membership functions are applied to transform practical data into fuzzy data. Finally the structure of neural network is designed and trained by sample data. The model is tested with original data and proved to be more effective and reliable to engine condition monitoring.