摘要
针对电液伺服系统存在的非线性动态导致的跟踪精度不足与收敛速率受限问题,提出一种融合神经网络与改进预设性能函数的滑模控制策略。通过建立液压系统的非线性状态方程,设计一种改进的有限时间预设性能函数,通过幂函数与指数项的分段耦合,实现跟踪误差的动态分阶段调节,有效改进传统函数收敛速率单一与稳态漂移问题;为克服误差变换过程中潜在的奇异问题,采用泰勒多项式对误差变换函数进行截断近似,确保系统输出的平滑性与稳定性;最后,结合RBF神经网络在线逼近系统非线性,降低模型简化误差,并设计滑模控制律与神经网络自适应律,利用Lyapunov稳定性理论证明闭环系统的有界收敛性。通过对比仿真实验与性能指标计算结果,所提控制器的均值误差比智能PID控制器降低了85.6%~85.7%,比传统预设性能滑模控制器降低了32.2%~81.5%,证明了该控制策略在电液伺服系统中的有效性。
Aiming at the problems of insufficient tracking accuracy and limited convergence rate caused by nonlinear dynamics in electro-hydraulic servo system,a sliding mode control strategy was proposed combining neural network and improving preset performance function.The nonlinear state equation of hydraulic system was established,and an enhanced finite-time preset performance function was designed.By piecewise coupling of power function and exponential term,the dynamic adjustment of tracking error could be realized in stages.This could effectively improve the inherent single convergence speed characteristics and steady-state drift problems of traditional functions.In order to overcome the potential singularity in error conversion,Taylor polynomial was used to truncate and approximate the error conversion function,which could ensure the smoothness and stability of the system output.Finally,a RBF neural network was combined to approximate the system nonlinearity online to reduce the model simplification error.The sliding mode control law and neural network adaptive law were designed,and the bounded convergence of the closed-loop system was proved by Lyapunov stability theory.By comparing the simulation experiments with the performance index calculation results,the mean error of the proposed controller is reduced by 85.6%to 85.7%compared with the intelligent PID controller,and by 32.2%to 81.5%compared with the traditional prescribed performance sliding mode controller,which proves the effectiveness of this control strategy in the electro-hydraulic servo system.
作者
陈胜友
孙春耕
韩世杰
CHEN Shengyou;SUN Chungeng;HAN Shijie(Faculty of Mechanical and Electrical Engineering,Kunming University of Science and Technology,Kunming Yunnan 650500,China)
出处
《机床与液压》
北大核心
2025年第23期125-133,共9页
Machine Tool & Hydraulics
基金
云南省科技厅重大科技专项计划项目(202202AC080008)。
关键词
电液伺服
滑模控制
预设性能
误差转化
神经网络
electro-hydraulic servo
sliding mode control
preset performance
error transformation
neural network
