基于神经网络的电液转台非线性积分滑模控制

针对高精度电液飞行仿真转台具有高度非线性、参数不确定和不确定非线性等特点，提出了一种基于RBF（Radial Basis Function）神经网络的非线性积分滑模鲁棒控制方法.采用自适应RBF神经网络对该系统存在的参数不确定性和不确定非线性进行补偿，从而降低滑模控制器对切换项的增益的需求，进而减小系统抖振幅值.积分滑模面的设计能消除外部干扰对系统带来的稳态误差.根据积分滑模变结构控制器的特点，将控制律分为等效控制律和到达控制律.等效控制律使系统运动于滑模面附近，到达控制律可使处于状态空间内任意初始位置的系统趋近于滑模面，并进一步通过Lyapunov方法证明了系统的渐近稳定性.实验结果表明，所提出的非线性控制器不仅能满足电液转台的高精度跟踪性能的要求，且对参数不确定性和不确定非线性具有一定的鲁棒性.

Integral sliding mode nonlinear controller of electrical-hydraulic flight simulator based on neural network

For the feature that high-accuracy electrical-hydraulic flight simulator (EHFS) is highly nonlinear and contains parametric uncertainties and uncertain nonlinearities, an integral sliding mode nonlinear robust controller based on radial basis function (RBF) neural network was proposed. The adaptive RBF neural network was adopted to eliminate the effect of parametric uncertainties and uncertain nonlinearities. By reducing the gain of switching function in sliding mode controller, chattering phenomenon could be minimized significantly. The steady state error from external disturbances could be eliminated by integral sliding control law, which was divided into an equivalent control law and a hitting control law. Equivalent control law was designed to keep the system sliding along the sliding surface. Hitting control law was applied to drive the representation point of the state space onto the sliding surface. The globally asymptotic stability of developed controller was proven via Lyapunov analysis. Comparative experimental results demonstrate the effectiveness of the proposed algorithm.