Abstract: Flight performance parameters can be used for fault prediction and condition monitoring, which is of great importance for the improvement of flight security and reduction of aircraft maintenance costs. Since the airplane is a complicated system, its performance parameter series are always nonlinear. In addition, affected by the operating environment, driving factors and noises generated by sensors, the performance parameters are often mixed with noises, which leads to uncertainty in prediction results. In order to deal with this problem, a new method is proposed to predict flight parameters by using a nonlinear support vector machine. By adding a new restriction, the uncertainty problem is properly solved. This method can not only enhance prediction precision, but also deal with problems involving large amounts of input data by using sequential minimal optimization. The method is evaluated by simulation data and actual flight performance parameters. Test results show that this new model which takes noise into consideration exhibits an improvement in precision as compared with the original model. Thus, this new method provides better precision for flight malfunction prediction, which is of great significance in enhancing flight safety.

Key words: aircraft flight parameter, prediction, support vector machine, uncertainty, exhaust gas temperature, exhaust gas temperature margin