基于PSO优化LS-SVM的刀具磨损状态识别  

Tool wear state recognition based on LS-SVM with the PSO algorithm

作　　者：刘成颖[1,2] 吴昊 王立平[1,2] 张智[4] 

机构地区：[1]清华大学机械工程系,北京100084 [2]清华大学精密超精密制造装备及控制北京市重点实验室,北京100084 [3]电子科技大学机械电子工程学院,成都611731 [4]海军航空工程学院飞行器工程系,烟台264001

出　　处：《清华大学学报（自然科学版）》2017年第9期975-979,共5页Journal of Tsinghua University(Science and Technology)

基　　金：国家科技重大专项(2013ZX04001-021)

摘　　要：为监测刀具的磨损状态,该文建立了一个基于声发射的刀具磨损状态监测系统。在刀具磨损状态监测实验中,采集加工过程中的声发射信号,提取方根幅值、绝对值均值、均方根、最大值作为反映刀具磨损的时域特征值。针对人工神经网络容易陷入局部极小值、结构难以确定、学习收敛速度慢等缺点,提出最小二乘支持向量机(least square support vector machine,LS-SVM)的刀具磨损状态识别方法。针对LS-SVM性能依赖于惩罚因子和核参数,利用粒子群优化(particle swarm optimization,PSO)算法对LSSVM参数进行自动寻优,建立PSO优化LS-SVM模型进行刀具磨损状态识别。结果表明:与LS-SVM识别模型相比,优化后的LS-SVM模型具有更高的识别率。A tool wear state monitoring system was developed based on acoustic emissions to monitor the tool wear state. Typical acoustic signals were analyzed to determine the square root amplitude, absolute mean, mean square error and maximum sound level from the time domain to characterize the tool wear. Neural networks can easily fall into a local minimum and have slow learning convergence rates so a tool wear state recognition method was developed based on a least square support vector machine （LS-SVM）. The LS-SVM performance depends on the penalty factor and the kernel parameter, so a particle swarm optimization algorithm was used to automatically optimize the LS-SVM parameters. The optimized LS-SVM model is then shown to be more accurate than the basic LS-SVM model.

关 键 词：刀具状态识别 时域特征值 最小二乘支持向量机(LS-SVM) 粒子群优化(PSO)算法 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]