基于SVM辨识的涡扇发动机全包线稳态控制方法

针对涡扇发动机全飞行包线范围稳态最优控制器的设计问题,首先根据不同飞行条件下发动机各工作状态的稳态“小偏差”线性模型,采用线性二次型调节器(LQR)分别设计得到相应的发动机最优线性控制器参数,然后将所得到的线性控制器用支持向量机方法进行非线性逼近,得到控制器参数的支持向量机辨识模型,以满足发动机全包线、全状态稳态控制的需要.支持向量机模型的输入为飞行高度、马赫数和稳态转速,输出为线性控制器参数.应用实例表明:该方法在全包线范围内对发动机最优稳态控制器的逼近误差均在2%以内,能较好满足控制精度要求.

Identification method based on support vector machines for turbofan engine steady state control in full flight envelop and various working conditions

To accomplish steady state optimal control of turbofan engine in full flight envelop and in different working conditions,nonlinear estimation methods based on linear steady state models of turbofan in all operation points were presented.The principle and algorithm realization of support vector machines for nonlinear controller approximation were improved.Then,the method used the improved support vector machines to establish the identification model of many linear optimization controllers designed for turbofan engine in some representative flight conditions.The controller design results for other operation points in the flight envelope can be conveniently educed by the model.For the identification model,the inputs were the steady rotation speed,the flight altitude and the flight Mach number,and the outputs were the parameters of the synthetic controller.The application examples show that the identification errors of the method in the whole flight envelop and various working conditions can meet the control precision need.