摘要
针对阿胶珠炮制对电磁炒药机温度控制精度和响应速度的要求,提出了一种基于干扰观测器的改进粒子群优化(PSO)径向基函数神经网络(RBFNN)PID的控制方法。根据阿胶珠电磁炒药机结构,建立阿胶珠电磁炒药机温度控制系统数学模型,通过对控制系统结构分析,构建RBF神经网络结构。利用RBF神经网络的自学习能力,采用梯度下降法对自身参数进行适当调整,实现PID参数动态调整,使系统惯性和时滞性有效抑制。为降低外部干扰影响,分析并构建干扰观测器模型,使干扰量得到实时观测和有效补偿。为弥补RBF神经网络模型参数精度不足,以系统误差瞬时值为适应度函数,利用改进PSO算法对RBF神经网络模型参数寻优,获取最佳控制性能。仿真结果表明:与传统PID控制方法和RBFNN-PID控制方法相比,所提控制方法使调节时间分别减少35 s和19 s,超调量分别降低19.2%和13.1%;与无干扰观测器相比,所提控制方法对外部干扰抑制能力平均提高50%;所提控制方法满足阿胶珠炮制工艺的要求。
In view of the requirements of temperature control accuracy and response rate of electromagnetic stir-frying machine for producing the Chinese drug beads,an improved particle swarm optimization(PSO)radial basis function neural network(RBFNN)PID control method is proposed based on interference observer.According to the structure of the electromagnetic stir-frying machine of Asini Coii Colla beads,the mathematical model of temperature control system of the special electromagnetic stir-frying machine is established.Analyzing the structure of the control system,the RBF neural network structure is constructed.Adopting the self-learning ability of the RBF neural network,the gradient descent method is chosen to adjust its own parameters adequately to realize the dynamic adjustment of PID parameters,thus the system inertia and time lag are suppressed efficiently.Analyzing and constructing a disturbance observer model,the interference is real-time observed and effectively compensated to reduce the influence of external interference.To obtain the best control performance,the RBF neural network model parameters are optimized by the improved particle swarm optimization algorithm with the system error instantaneous values for the fitness function to make up for the RBF neural network model parameters’accuracy.Simulation results show that the regulating time of this control method is reduced by 35 s and 19 s,and the overshoot is reduced by 19.2%and 13.1%,respectively,compared with the traditional PID control method and RBFNN-PID control method.The external interference suppression ability is increased by 50%on average compared with the case without interference observer.
作者
丁威
杜钦君
赵龙
宋传明
罗永刚
毕胜
王彬
DING Wei;DU Qinjun;ZHAO Long;SONG Chuanming;LUO Yonggang;BI Sheng;WANG Bin(School of Electrical and Electronic Engineering,Shandong University of Technology,Zibo,Shandong 255022,China;Shandong Hongjitang Pharmaceutical Group Co.,Ltd.,Jinan 250100,China)
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2021年第5期133-142,共10页
Journal of Xi'an Jiaotong University
基金
国家重点研发计划资助项目(2018YFC1707104,2018YFC1707103)
山东省重点研发计划资助项目(2019GGX101031)。
关键词
电磁炒药机
温度控制
干扰观测器
改进粒子群算法
径向基函数神经网络
electromagnetic stir-frying machine
temperature control
disturbance observer
improved particle swarm optimization
radial basis function neural network
