A novel identification model of aeroengine based on support vector machines

Considering aeroengine system´s being doughtily nonlinear and time-varying, and there were local minima, overfitting and so on in aeroengine model identification via the traditional neural networks. A novel identification model of aeroengine based on support vector machines was minutely introduced. It was based on modern statistical learning theory, adopted structure risk minimization (SRM) principle, which assured good generalization ability. Via computing a quadratic convex programming problem, it got a global optimum network structure automatically. Aeroengine identification model based on support vector regression machines was set up by using the real flight record data as learning samples. Results show that this method owns the virtue of high precision, good robustness and fault-tolerance. It provides a general new way for aeroengine model identification.