基于改进粒子群优化RBF神经网络的轧制力预报  

Rolling Force Prediction Based on Improved Particle Swarm Optimization-RBF Neural Network

作　　者：杨景明[1,2] 闫晓莹[1] 顾佳琪[1] 车海军[1,2] 

机构地区：[1]燕山大学工业计算机控制工程河北省重点实验室,河北秦皇岛066004 [2]国家冷轧板带装备及工艺工程技术研究中心,河北秦皇岛066004

出　　处：《矿冶工程》2014年第6期110-113,118,共5页Mining and Metallurgical Engineering

基　　金：河北省科技支撑计划项目(13211817);河北省高等学校创新团队领军人才培训计划项目(LJRC013)资助

摘　　要：依据RBF神经网络的非线性逼近能力和自学习特性,提出基于RBF神经网络的建模方法。将最近邻聚类用于RBF神经网络隐层中心向量的确定,并采用改进粒子群算法对最近邻聚类的聚类半径进行优化,合理确定了RBF神经网络的隐层结构,提出了一种基于改进粒子群算法的RBF神经网络(IMPSO-RBF)。将该网络应用于轧制力的预报,与基本粒子群算法优化的RBF神经网络比较,仿真结果表明其在预报精度和收敛速度上都有很大提高。In the light of the nonlinear approximation capability and self-learning characteristics of RBF neural network, a modeling method is put forward based on the RBF neural network. Firstly, the center vector of the RBF neural network hidden layer was determined using the nearest neighbor clustering and clustering radius of the nearest neighbor clustering was optimized adopting an improved particle swarm algorithm. Then, a reasonable hidden layer structure of the RBF neural network was determined. Based on the improved particle swarm algorithm, a RBF neural network （IMPSO-RBF） was finally proposed, which was applied to prediction of rolling force. The simulation results show that the forecasting precision and convergence speed of the network are greatly improved compared with the RBF neural network optimized by basic particle swarm algorithm.

关 键 词：RBF神经网络 改进粒子群算法 轧制力预报 

分 类 号：TG339[金属学及工艺—金属压力加工]