阀控液压马达位置伺服系统长短时记忆神经网络预测抗扰反步控制  

Long Short-term Memory Prediction Anti-disturbance Backstepping Control of Valve Controlled Hydraulic Motor Position Servo System

作　　者：柴凌云 栾海英 刘增元 沈洲 任翔 CHAI Lingyun;LUAN Haiying;LIU Zengyuan;SHEN Zhou;REN Xiang(Beijing Research Institute of Automation for Machinery Industry Co.,Ltd.,Beijing 100120)

机构地区：[1]北京机械工业自动化研究所有限公司,北京100120

出　　处：《液压与气动》2024年第8期128-136,共9页Chinese Hydraulics & Pneumatics

摘　　要：针对阀控液压马达位置伺服系统中存在的时滞性与摩擦非线性问题,设计了一种长短时记忆神经网络预测抗扰反步控制器。该控制器通过引入长短时记忆神经网络对当前位置轨迹进行预测,并将预测值反馈给控制器对系统时滞进行直接补偿。对于系统中难以建模的摩擦非线性,将其视为扰动,通过设计扩张状态观测器进行估测,并使用反步法对估测得到的总扰动进行补偿。最后,在Simulink中搭建长短时记忆神经网络预测抗扰反步控制算法进行仿真验证,并与径向基函数滑模控制算法、反步控制算法和自抗扰控制算法进行对比,证明其在对含有时滞及摩擦非线性的阀控液压马达位置伺服系统进行控制时,具有较快的响应速度及较好的跟踪性能。Aiming at the time delay and frictional nonlinearity in the position servo system of valve-controlled hydraulic motor,a long short-term memory predictive anti-disturbance backstepping control is designed.Firstly,long short-term memory is introduced into the control to predict the current position trajectory,and the predicted value is fed back to the controller to compensate the delay directly.The frictional nonlinearity which is difficult to model in the system is regarded as disturbance,which is estimated by designing an extended state observer,and the estimated total disturbance is compensated by backstepping control.Finally,the long short-term memory predictive disturbance rejection and backstepping control is built in Simulink for simulation verification.By comparing with radial basis function sliding mode control algorithm,backstepping control algorithm and auto-disturbance rejection control algorithm,it is proved that the control algorithm proposed in this paper is effective in controlling the position servo system of valve-controlled hydraulic motor with time delay and frictional nonlinearity.It has fast response speed and good tracking performance.

关 键 词：阀控液压马达位置系统 长短时记忆神经网络 反步控制 扩张状态观测器 

分 类 号：TH137[机械工程—机械制造及自动化]