基于神经网络的缠绕过程张力积分鲁棒控制

纤维缠绕系统是典型的非线性系统，缠绕过程张力控制精度决定了缠绕制品成型品质，然而系统非线性特性、摩擦及外干扰等严重制约了缠绕过程张力控制性能的提升。目前现有方法主要以收/放卷两轴同步控制为基础，通过解耦等复杂操作展开张力控制研究，计算量大且不利于控制算法的应用。为了避免上述问题并准确描述缠绕系统张力产生机理和实际的摩擦特性，建立简化的缠绕系统非线性数学模型。同时以自适应作为神经网络权值训练方法，基于自适应神经网络设计出干扰量的逼近函数，在基于连续积分鲁棒（RISE）控制器设计的控制律中补偿扰动的影响，并基于Lyapunov稳定性理论证明了系统的渐近稳定性。最后，给出所提出控制器与现有方法的对比验证实例，结果表明所提出基于神经网络扰动补偿的积分鲁棒控制显著增强了系统抑制外干扰的能力，提升了张力控制精度。

Neural Network Based RISE Control of Winding Tension

Filament winding system is a typical nonlinear system. Tension control accuracy of the winding process determines the quality of the winding product. However, the nonlinear characteristics, friction and external interference of the system severely restrict the improvement of the tension control performance of the winding process. At present, the existing methods are mainly based on the synchronous control of the rewinding and unwinding axes, and the tension control research is carried out through complex operations such as decoupling. The calculation is large and is not conducive to the application of control algorithms. In order to accurately describe the tension generation mechanism and actual friction characteristics of the winding system, a uncomplicated nonlinear mathematical model of the winding system is established. Meanwhile, taking adaptive method as the neural network weights training procedure and the approximation function of the disturbance based on the adaptive neural network is designed. Therefore, the disturbance can be compensated at the control law of the continuous robust integral of the sign of the error (RISE) controller. Based on the Lyapunov stability theory, the asymptotic stability of the system is proved. Finally, a comparison verification example between the proposed controller and the existing methods is given. Results show that the proposed method of RISE controller based on adaptive neural network disturbance compensation significantly enhances the system's ability to suppress external disturbances, and improves the system control accuracy.