摘要
为了提高伺服系统瞬态性能,满足工业生产线的高精度加工要求,提出模糊自适应深度强化学习方法,用于永磁同步电机伺服系统的控制性能优化。根据伺服系统瞬态运行的响应快速性和稳定性要求,通过构造奖励函数与Actor-Critic网络,在伺服控制器中引入瞬态响应优化环节,面向实时运算需求设计学习率自适应模糊控制器,构建模糊自适应瞬态性能优化方法;建立伺服系统Simulink仿真模型,根据实际伺服系统拟合仿真模型,通过瞬态性能仿真获得优化参数,并将优化参数应用于西门子840D数控系统。以仿真运算及系统实验对该优化方法进行验证,结果表明,所提出方法使伺服系统调节时间缩短10%以上,显著提高了系统的瞬态响应速度,且未引入明显超调,验证了方法的可行性。提出的方法具有较强的通用性,为伺服系统智能控制优化提出了新的途径。
To improve the transient performance of servo systems and satisfy the high-precision processing requirements of industrial production lines,a fuzzy adaptive deep reinforcement learning algorithm is proposed to optimize the control performance of permanent magnet synchronous motor servo system.According to the rapidity and stability requirements of transient response,the reward function and Actor-Critic networks are constructed,the transient optimization link for servo controller is introduced,and the learning rate fuzzy controller for real-time computing requirements is designed.A servo system Simulink simulation model is established and fitted in with the actual servo system,the optimized parameters are obtained via transient performance optimization simulation and then applied to a Siemens 840D computerized numerical control system.The verification of simulation calculation by system experiment shows that the proposed method of transient performance optimization shortens the servo system adjustment time by more than 10%,significantly heightens the system transient response rate without introducing an obvious overshoot.This method has strong versatility to provide a new way for the intelligent control optimization of the servo system.
作者
魏晓晗
张庆
蒋婷婷
梁霖
WEI Xiaohan;ZHANG Qing;JIANG Tingting;LIANG Lin(School of Mechanical Engineering, Xi’an Jiaotong University, Xi’an 710049, China;Key Laboratory of Education Ministry for Modern Design and Rotor-Bearing System, Xi’an Jiaotong University, Xi’an 710049, China)
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2021年第8期68-77,共10页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(51675405)。
关键词
伺服系统
控制优化
瞬态性能
深度强化学习
模糊控制
servo system
control optimization
transient performance
deep reinforcement learning
fuzzy control
