SVR在切削颤振状态趋势预测中的应用  

Application of Support Vector Regression in Predicting Cutting Chatter Trend

作　　者：胡耀斌[1] 邹湘军[2] 张春良[1] 江涌涛[1] 

机构地区：[1]南华大学机械工程学院,衡阳421001 [2]华南农业大学工程学院,广州510642

出　　处：《系统仿真学报》2007年第17期4086-4089,共4页Journal of System Simulation

基　　金：国家自然科学基金资助项目(50375070);湖南省学位办研究生教研基金资助项目(04B21)

摘　　要：针对当前准确地预测机床切削颤振状态趋势存在的困难,提出了基于支持向量机回归算法(SVR)的切削颤振状态趋势的预测方法,给出了运用支持向量机进行回归分析时的参数选择原则。研究了基于小波包分解的切削信号特征提取方法,首先将切削信号进行小波包分解,计算信号分解到各频带区间内的能量并对其进行归一化,并将其作为切削信号的特征向量输入到支持向量机回归分析模型。在CA6140车床上进行了数据采集和仿真,结果表明,通过这种方法得到的信号在各频带区间内的能量变化曲线能准确地反映切削颤振的过渡过程,并且通过SVR对其进行趋势预测也取得了比较满意的结果。Having been encountered the difficulty of properly predicting the cutting chatter trend at present, a novel cutting chatter trend prediction method based on support vector regression algorithm（SVR） was put forward, meanwhile, the principle about how to select the SVR parameters was discussed. Ways about how to extract features from cutting signal process based on wavelet package decomposition were studied. Gabor wavelet basis was used to decompose the cutting signal to 3-order. The energy in each spectrum section was calculated and scaled. The resultswere input to the SVR model. Sampling and simulating on lathe （CA6140）, the results turned out that the energy transition curve in each spectrum segment retrieved from above mentioned method gave a fine reflection of the cutting process transition from normal to chatter. The trend predicted by SVR shows a satisfactory consequence.

关 键 词：切削颤振 小波包分解 SVR 趋势预测 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]