基于高斯过程的阀控非对称液压缸模型预测控制  

A Model Predictive Control Method for Valve-controlled Asymmetric Hydraulic Cylinders Using Gaussian Process

作　　者：孙浩 陶建峰[1] 余宏淦 刘成良[1] SUN Hao;TAO Jian-feng;YU Hong-gan;LIU Cheng-liang(School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240)

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《液压与气动》2021年第8期26-33,共8页Chinese Hydraulics & Pneumatics

基　　金：国家重点研发计划(2018YFB1306703);国家自然科学基金(52075320)。

摘　　要：针对阀控非对称液压缸系统在采用PID控制器进行位置闭环控制时,存在液压缸活塞位置超调和由伺服阀阀芯切换导致控制效果差的问题,设计了基于高斯过程的非线性模型预测控制器。采用高斯过程回归训练得到阀控非对称缸系统的离散数学模型,通过求解二次规划问题,控制器输出最优序列。仿真结果表明,该机器学习方法有效估计了位置增量中噪声的标准差,预测输出更接近真实值;在多约束条件下能有效提高系统响应时间,保证活塞位置无超调,位置控制精度满足0.1 mm。For Valve-controlled Asymmetric Cylinder Systems(VACS),there are problems like piston position overshoot and poor control effect caused by valve spool switching when PID is used for position closed-loop control.A Gaussian Process Regression-based(GPR)model predictive method is proposed for VACS.The discrete mathematical model of VACS is trained by GPR.After that,an optimal output is obtained by solving the nonlinear quadratic programming problem.The results show that the standard deviation of the noise in the piston position increment can be well estimated,which make it more approximate to the true value.The proposed method can effectively improve the response time under multi-constraint conditions,and ensure that the piston position has no overshoot and the position control precision reaches 0.1 mm.

关 键 词：阀控非对称液压缸 位置控制 模型预测控制 高斯过程回归 

分 类 号：TH137[机械工程—机械制造及自动化] TH132