中国激光, 2011, 38 (s1): s115006, 网络出版: 2011-12-16  

区间极限学习机在气体FTIR光谱浓度反演中的应用研究

Research on Concentration Retrieval with Gas FTIR Spectra by Interval Extreme Learning Machine Method
作者单位
1 中北大学光电信息与仪器工程技术研究中心, 山西 太原 030051
2 中北大学仪器科学与动态测试教育部重点实验室, 山西 太原 030051
3 中北大学电子测试技术国家重点实验室, 山西 太原 030051
摘要
为准确反演气体浓度,节约建模时间,提出了基于区间极限学习机(ELM)定量分析模型的傅里叶变换红外(FTIR)光谱分析技术。该方法基于区间划分思想,将整个光谱范围划分为若干个子区间,利用ELM分别建立各个子区间的定量分析模型,并根据各个子区间模型的决定系数大小评价其泛化性能,进而筛选出最具代表性的子区间组合。基于上述方法,对NO与NO2气体的红外光谱进行波长筛选,并利用筛选后的特征波长点光谱建立定量分析模型。实验结果表明,NO气体测试集的决定系数R2为0.9999,NO2气体测试集的决定系数R2为0.9997。与区间偏最小二乘法相比,利用区间ELM方法建模速度更快,模型泛化性能更优。
Abstract
To retrieve the gas concentration accurately and rapidly, a new quantitative analysis technique based on Fourier transform infrared spectroscopy of interval extreme learning machine (ELM) model is proposed. Based on the idea of interval division, this approach firstly divides the whole spectrum into several subintervals, secondly establishes quantitative analysis model corresponding to each subinterval with ELM method, and finally selects the best subinterval combinations according to the determination coefficient of each model. Based on the above approach, wavelengths are selected in the spectrum of NO and NO2, and then establishs the quantitative analysis model using the selected spectrum combinations, respectively. The experimental results showed that, the testing set determination coefficient of NO and NO2 are 0.9999 and 0.9997, respectively. The outcome indicates that, compared with Interval partial least squares method, the proposed Interval ELM method can establish quantitative analysis model more rapidly and accurately.

陈媛媛, 张记龙, 王志斌, 赵冬娥, 陈友华. 区间极限学习机在气体FTIR光谱浓度反演中的应用研究[J]. 中国激光, 2011, 38(s1): s115006. Chen Yuanyuan, Zhang Jilong, Wang Zhibin, Zhao Donge, Chen Youhua. Research on Concentration Retrieval with Gas FTIR Spectra by Interval Extreme Learning Machine Method[J]. Chinese Journal of Lasers, 2011, 38(s1): s115006.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!