基于小波包和PSO优化神经网络的刀具状态监测  

Study on monitoring of cutting tool conditions with optimized BP neural networks based on wavelet packets and PSO algorithm

作　　者：唐亮[1] 傅攀[1] 李敏[1] 

机构地区：[1]西南交通大学机械工程学院,四川成都610031

出　　处：《中国测试》2016年第3期94-98,共5页China Measurement & Test

基　　金：中央高校基本科研业务费专项资金(SWJTU12CX039)

摘　　要：为改进BP神经网络进行刀具磨损状态识别时训练时间较长、收敛速度较慢、测试精度不够高、甚至完全不能训练等问题,引入一种全局搜索能力强,收敛速度快的算法——粒子群优化算法(PSO),用其来优化BP神经网络参数,改进网络的训练和识别性能。实验证明:经粒子群算法优化后的BP神经网络较原网络有更快的训练迭代收敛速度和更高的测试准确度,达到优化的目标,对实现数控刀具磨损状态的智能化在线监测具有重要意义。An algorithm-particle swarm optimization（PSO） which has stronger global searching capability and faster convergence speed was proposed. It was used to optimize the parameters and improve the training and recognition performance of BP neural networks. Experiments show that the training,iteration and convergence speeds of BP neural networks optimized by PSO are much shorter and their testing accuracy is much higher compared with original ones. This is significant in realizing intelligent online monitoring of the wear conditions of NC cutting tools.

关 键 词：BP神经网络 磨损状态识别 PSO 优化 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]