学术论文

      一种新型多神经元PID伺服控制器研究

      Research on new type of multi neuron PID servo controller

      摘要:
      针对BP神经网络算法的收敛速度慢,权值畸形导致迭代无法进行,非线性多耦合收敛局部极小点的缺陷,本文提出了一种新型多神经元PID神经网络算法.该算法核心通过简化PID神经元网络控制器,改善其权值初始化,在权值迭代计算中运用符号函数解决上述缺陷.通过对该新型多神经元PID控制算法结果分析,该算法在反传运算过程中大为简化,算法收敛速度快,且网络权值灵活.将该算法与传统神经网络算法分别应用到全液压矫直机伺服控制,通过对比分析验证了该算法的有效性,同时为多神经元PID控制器的工程化应用提供了理论支持和技术借鉴.
      Abstract:
      In the light of the slow convergence speed, which leads to neural network weights iteration to solve the nonlinear coupling,multi weights are converged to the local minimum defects points of defects. A new multi neuron PID neural network algorithm is proposed.The core of this algorithm is to simplify the PID neural network controller and to improve the initialization of the weights.Through the analysis and calculation of the new multi neuron PID control algorithm,it found that the proposed algorithm has many advantages such as simplification,fast convergence speed,flexible network weight.The algorithm with the traditional neural network algorithm is applied to hydraulic straightening machine′s servo control. The comparative analysis is used to verify the effectiveness of the algorithm,and application for multi neuron PID controller provides theoretical support and technical reference.
      作者: 岳光 [1] 潘玉田 [1] 张华君 [2] 杜金祥 [3]
      Author: YUE Guang [1] PAN Yutian [1] ZHANG Huajun [2] DU Jinxiang [3]
      作者单位: 中北大学 机电工程学院,山西 太原,030051 太原科技大学 重型机械教育部工程研究中心,山西 太原,030024 太原工业学院 自动化系,山西 太原,030008
      刊 名: 燕山大学学报 ISTICPKU
      年,卷(期): 2017, 41(3)
      分类号: TP273 TH39
      在线出版日期: 2017年8月2日
      基金项目: 山西省重点研发计划(指南)项目