基于优化终端滑模模糊神经网络的板形模式识别

Flatness Pattern Recognition Based on the Optimized Terminal Sliding Mode Fuzzy Neural Network

针对板形模式识别问题,将板形信号离散化、归一化,作为终端滑模模糊神经网络的学习样本,建立识别模型。在模糊神经网络的基础上,利用终端滑模权值调整律代替梯度下降法的权值调整律,提高网络的精度。为了进一步提高识别的精度以及收敛速度,引入布谷鸟算法优化模糊神经网络的模型参数。仿真结果表明,提出的识别模型对训练样本和未训练样本的平均最小方差分别为0.000 5和0.011 0,比模糊神经网络(FNN)和径向基神经网络(RBF)的值都小。对某冷轧厂宽度1 040 mm带材的一组实测板形数据识别结果表明,相比于FNN和RBF网络,CS-TSMFNN的识别效果更好。

Aiming at the problem in flatness pattern recognition,flatness signals were firstly discretized and normalized,which was taken as learning samples for terminal sliding mode fuzzy neural network（ TSMFNN） in the establishment of the recognition model. Based on fuzzy neural network（ FNN）,the weight adjustment law of the terminal sliding mode,instead of the gradient descent,was adopted to improve the accuracy of the network. In order to further improve the accuracy and speed of convergence,cuckoo algorithm was introduced to optimize the parameters of the fuzzy neural network. The simulation results showed that the minimum mean square errors of the trained and untrained samples were 0. 000 5 and 0. 011 0,respectively,smaller than that of FNN and radial basis function（ RBF）. The recognition of a set of measured flatness data for the plate strip with width of 1 040 mm in a cold strip mill factory indicated that CS-TSMFNN recognition result is better than that of FNN and RBF.