负载口独立阀控缸库普曼模型预测控制方法  

Model predictive control method for independent metering valvecontrolled cylinders by using koopman operators

作　　者：刘恒 陶建峰 孙炜 孙浩 刘成良[1] LIU Heng;TAO Jianfeng;SUN Wei;SUN Hao;LIU Chengliang(State Key Laboratory of Mechanical System and Vibration,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学机械系统与振动国家重点实验室,上海200240

出　　处：《中南大学学报(自然科学版)》2025年第3期911-922,共12页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(52075320)。

摘　　要：负载口独立控制阀控液压缸系统(IMVCS)控制自由度多,系统能效提升空间大,在工程机械等领域有很好的应用前景,然而,控制自由度增加以及阀口节流方案的固有非线性使得实现此类系统的高能效、高精度控制面临挑战。本文提出了一种基于深度神经网络库普曼(Koopman)算子的液压系统模型预测控制方法。首先,通过数据训练得到被控对象的高精度线性预测模型,并将预测模型用于IMVCS的模型预测控制;其次,在控制器的损失函数引入能耗项,分别控制执行器两侧腔室的流量和压力来减少能耗;最后,使用NSGA-II算法来对控制器参数进行调优,实现高能效、高精度控制。研究结果表明:该方法能够保证控制精度,提高节能效率;相较于传统的PID控制,所提出的控制策略降低了至少29%的能量消耗,并且轨迹跟踪误差控制在0.7mm以内。The independent metering valve control system(IMVCS)for hydraulic cylinders offers high control degrees of freedom and significant potential for improving system efficiency,presenting excellent application prospects in fields such as construction machinery.However,the increase of control degrees of freedom and the inherent nonlinearity of the valve metering equations pose challenges for achieving high-efficiency and highprecision control in such systems.A model predictive control(MPC)for hydraulic systems based on deep neural network Koopman operators was proposed.Firstly,through training,a high-precision linear predictive model of the controlled object was obtained and applied to the model predictive control of IMVCS.Secondly,an energy consumption term was introduced into the cost function of the controller.The flow and pressure of both chambers of the actuator were controlled to reduce energy consumption.Finally,the NSGA-II algorithm was used to optimize the controller parameters to achieve high-efficiency and high-precision control.The results show that the proposed method ensures control accuracy and achieves energy efficiency optimization.Compared with conventional PID control,the implemented strategy reduces energy consumption by at least 29% and maintains trajectory tracking errors within 0.7 mm.

关 键 词：负载口独立控制阀控液压缸系统(IMVCS) 模型预测控制(MPC) Koopman算子 深度神经网络(DNN) 

分 类 号：TH137.52[机械工程—机械制造及自动化]