阀控非对称缸系统神经网络多逆模型切换控制研究  

Multi-Inverse Model Switching Control for Valve-Controlled Asymmetric Cylinder Systems Based on Neural Network

作　　者：曾乐 谭建平[1] 许文斌 杨俊[3] ZENG Le;TAN Jianping;XU Wenbin;YANG Jun(State Key Laboratory of High Performance Complex Manufacturing,College of Mechanical and Electrical Engineering,Central South University,Changsha 410083,China;Aviation Machinery Manufacturing Department,Changsha Aeronautical Vocational and Technical College,Changsha 410124,China;College of Engineering and Design,Hunan Normal University,Changsha 410081,China)

机构地区：[1]中南大学机电工程学院高性能复杂制造国家重点实验室,长沙410083 [2]长沙航空职业技术学院航空机械制造系,长沙410124 [3]湖南师范大学工程与设计学院,长沙410081

出　　处：《西安交通大学学报》2019年第12期153-160,共8页Journal of Xi'an Jiaotong University

基　　金：湖南省自然科学（2019JJ70023）;高性能复杂制造国家重点实验室自主探索基金资助项目（ZZYJKT2016-07）

摘　　要：为提高存在负叠合量的阀控非对称缸系统的控制性能,提出基于神经网络的逆系统控制方法,利用神经网络逼近的逆模型与原系统复合,将复杂非线性系统转变为线性系统进行控制,建立了阀控非对称缸系统的数学模型,系统在(x 0,u)的邻域内存在相对阶,证明了系统的可逆性;采用基于遗传算法改进的BP神经网络(GA-BP)求解逆模型,并针对伺服阀存在负叠合量,以及流态存在层流和紊流两种状态的问题,建立系统的多个逆模型集,提高了逆系统的求解精度。利用AMESIM和Simulink联合仿真平台,基于参考速度切换的原则,对系统采用比例-积分-微分(PID)闭环控制器。结果表明:普通PID控制的液压缸伸出运动响应和缩回运动响应不一致,伸出运动存在0.20 mm的稳态误差,误差波动范围为0.06 mm,而缩回运动稳态误差较小,约为0.02 mm,但误差波动较大,约为0.09 mm;多逆系统复合控制的伸出缩回运动响应较一致,伸出和缩回运动均存在0.02 mm的稳态误差,误差波动范围为0.04 mm,验证了多逆模型切换控制方法可以消除阀控非对称缸系统的非对称性,降低波动负载干扰影响,提高系统的响应精度。A control method based on a neural network inverse system is proposed to improve the control performance of valve-controlled asymmetric cylinder systems with negative overlap.An inverse model of neural network approximation is used to transform the complex nonlinear system into a linear system.A mathematical model is established for the valve-controlled asymmetric cylinder system,and the system exists relative orders in the neighborhood of(x 0,u),which proves the reversibility of the system.An improved BP neural network based on a genetic algorithm(GA-BP)is used to solve the inverse model.Aiming at the problems of the negative overlap and the laminar and turbulent flow in the servo valve,several inverse models for the system are established to improve the accuracy of the inverse system.Finally,a simulation platform combined with AMESIM and Simulink is used,and a proportional-integral-derivative(PID)closed-loop controller is adopted for the system based on the principle of reference speed switching.Results show that the extending motion response and retracting motion response of the hydraulic cylinder controlled by ordinary PID are inconsistent.There is a steady error of 0.20 mm in the extending motion,and the error fluctuation range is 0.06 mm.There is only 0.02 mm steady error in the retracting motion,and the range of error fluctuation is 0.09 mm.While the responses of both the extending and retracting motion controlled by the multi-inverse system is comparatively consistent.Both the extending and retracting motion has a steady error of 0.02 mm,and the error fluctuation range is 0.04 mm.These results verify that the multi-inverse model switching control method can effectively eliminate the asymmetry of the valve-controlled asymmetric cylinder system,reduce the impact of fluctuating load disturbance and improve the response accuracy of the system.

关 键 词：阀控非对称缸系统 神经网络逆系统 负叠合 多逆模型切换 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]