基于智能优化型径向基神经网络的板形模式识别研究

Study on flatness pattern recognition based on intelligent optimal radial basis function neural network

针对传统基于神经网络的板形模式识别方法具有网络精度较低、在线识别速度慢和网络模型建模复杂等技术问题,提出了一种基于智能优化型径向基神经网络的板形模式识别方法.在基于训练数据进行神经网络建模过程中,采用一种改进的粒子群优化控制算法进行网络架构节点数目和网络参数值的离线优化,因而所得方法具有网络结构简单、泛化能力强等优点.仿真实验结果表明,该方法是一种有效板形模式识别方法,有利于提高板形控制精度.

In order to deal with the problem that the usual flatness pattern recognition methods based on neural network have some flaws which restrict their applications,i, e. ,lower precision for the obtained networks,lower velocity for both on-line recognition and complex network modeling, a kind of flatness pattern recognition based on intelligent optimal radial basis function （RBF） neural network was proposed. In the process of modeling the neural network based on some training data, an improved particle swarm optimization algorithm was proposed to optimize both the number of network nodes and the value of network parameters. Therefore, the approach has simpler structure and better generalization than before. The simulation experiment results showed that the approach was effective and could increase the precision of flatness control.