摘要
在利用太赫兹时域光谱(THz-TDS)技术对丁腈橡胶中的添加剂进行定量分析时,将偏最小二乘法(PLS)推广到多维偏最小二乘方法(N-PLS),解决了高维数据的建模问题,并且具有良好的模型预测能力。为验证本算法的有效性,分别采用PLS算法和N-PLS算法对样本进行建模分析,得出N-PLS模型的均方根误差为2.04%,相关系数为0.98。PLS模型的均方根误差为2.71%,相关系数为0.97,结果表明,两种方法都可以对丁腈橡胶中添加剂的含量进行预测,而N-PLS算法具有更好的预测效果。
When using terahertz time-domain spectroscopy（THz-TDS）technology to perform quantitative analysis of nitrile rubber additives,the paper extending the partial least squares（PLS）to multidimensional partial least squares method（N-PLS）,which can solve the problem of high dimensional data modeling effectively and improve the model accuracy.To verify the effectiveness of the proposed algorithm above,PLS method and N-PLS method are used respectively on sample modeling and analysis.Which obtained the root mean square error and the correlation coefficient of N-PLS model is 2.04% and 0.98.While The root mean square error and the correlation coefficient ofPLS model is2.71%and 0.97.The results show that both methods can forecast the content of some components in the mixture,but the prediction effect of N-PLS model is better.
出处
《电子测量技术》
2017年第12期148-151,共4页
Electronic Measurement Technology
基金
桂林电子科技大学研究生教育创新计划(YJCXS201560
2016YJCX27)
广西自动检测技术与仪器重点实验室基金(YQ17102
YQ14113)
广西自然科学基金(2015GXNSFBA139252)项目资助
关键词
太赫兹时域光谱技术
橡胶
定量分析
多维偏最小二乘回归
terahertz time-domain spectroscopy
rubber
quantitative analysis
N-way partial least squares method
