基于L—M网络的涡扇发动机动态过程辨识

Dynamical Process Identification of Turbofan Engine Using L-M Feedforward Neural Network

介绍了多层前馈网络的Levenberg -Marquardt(L -M)算法 ,对涡扇发动机动态过程的数学模型进行了分析和简化 ,运用多层前馈网络L -M算法对涡扇发动机进行加力推力变换时的非线性MIMO动态过程的数学模型进行了辨识 ,辨识所用样本数据和测试数据均为试飞实测数据。辨识结果表明 ,这种方法收敛快精度高 ,非常适合于涡扇发动机非线性动态过程的建模。

In this paper,nulti-layer feedforward network using Levenberg-Marquardt algorithm is enployed in nonlinear dynamical process identification of a turbofan engine.L-M algorithm for neural metwork use is briefly introduced.The nonlinear model of the engine is analyzed and simplified.The data used in the paper are from the flight test.It is shown that L-M algorithm is much more efficient than BP algorithm and is very suitable for dynamical process identification of the engine investigated.