基于支持向量机的钢水LIBS定性分析  

Qualitative Analysis of Molten Steel Based on Support Vector Machine by LIBS

作　　者：杨友盛 张岩 杨友良[2] 马翠红[2] 

机构地区：[1]唐山赛福特智能控制股份有限公司研发中心,河北唐山063000 [2]河北联合大学电气工程学院,河北唐山063000

出　　处：《激光与光电子学进展》2015年第5期209-214,共6页Laser & Optoelectronics Progress

基　　金：国家自然科学基金项目(61271402);国家自然科学基金(61171058);国家科技部科技人员服务企业行动(2009GJA2000)

摘　　要：激光诱导击穿光谱(LIBS)技术具有快速、非接触、无需制样等特点,适合应用于转炉钢水成分的在线分析。由于转炉终点可由Si、Mn含量和温度来判定,因此提出了钢水成分中Si和Mn的LIBS定性分析方法。通过光谱仪采集激光激发的光谱,经过预处理和寻峰等操作后,以原子光谱数据库(NIST)为参考标准,找出Si和Mn对应的特征谱线波长和光谱强度,利用支持向量机(SVM)强大的分类功能和采集到的245组数据中的210组学习得到支持向量分类(SVC)模型,利用SVC模型预测这245组数据,结果证明该模型的准确率为98%以上,将其应用在相同实验条件的情况下,会大大减少LIBS定性分析时间。Laser induced breakdown spectroscopy（LIBS） technology has the characteristics of speediness, noncontact, no need of sample preparation, which is very suitable for the online analysis of the converter steel composition, because the end-point can be determined by Si and Mn contents and temperature. A qualitative analysis of LIBS is proposed for analyzing the Si and Mn composition in molten steel. Laser excitation spectra are collected by spectrometer, and after the operations of pretreatment and peak searching, with the atomic spectra database（NIST） as the reference standard, the corresponding characteristics of spectral line wavelength and spectral intensity of Si and Mn are found out. Based on the powerful classification function of support vector machine（SVM）, 210 sets of 245 sets of data collecting are used to get the support vector classification（SVC） model, which then predicts that245 groups of data. The accuracy of the model is more than 98%, which can identify the corresponding wavelength of Si and Mn very well, and can be used under the condition of the same experimental conditions with significant reduction to the LIBS qualitative analysis time.

关 键 词：光谱学 激光诱导击穿光谱 定性分析 支持向量机 特征谱线 转炉终点 

分 类 号：O433[机械工程—光学工程]