基于CARS波段筛选的甲醇汽油中甲醇含量中红外光谱检测 下载: 946次
Mid-Infrared Spectroscopy Detection of Methanol Content in Methanol Gasoline Based on CARS Band Screening
华东交通大学机电与车辆工程学院, 江西 南昌 330013
图 & 表
图 1. 甲醇汽油中甲醇含量的中红外检测模型的流程图
Fig. 1. Flow chart of mid-infrared detection model for determining methanol content of methanol gasoline
下载图片 查看原文
图 2. 汽油和无水甲醇样品的原始中红外光谱。(a)汽油中红外光谱;(b)无水甲醇中红外光谱
Fig. 2. Original mid-infrared spectra of gasoline and absolute methanol. (a) Mid-infrared spectrum of gasoline; (b) mid-infrared spectrum of absolute methanol
下载图片 查看原文
图 3. 不同甲醇体积分数的甲醇汽油的原始中红外光谱图
Fig. 3. Original mid-infrared spectra of methanol gasoline with different methanol volume fractions
下载图片 查看原文
图 4. 采用UVE方法筛选得到的波段。(a)在全谱范围内进行波段选择;(b)在全谱范围内挑选的波段点
Fig. 4. Band selected by UVE method. (a) Band selection in full spectrum; (b) band points selected in full spectrum
下载图片 查看原文
图 5. 基于CARS算法的波长变量筛选过程。(a)变量个数与采样次数的关系;(b) RMSECV值与采样次数的关系;(c)回归系数与采样次数的关系
Fig. 5. Wavelength variable screening based on CARS algorithm. (a) Number of variables versus sampling times; (b) RMSECV value versus sampling times; (c) regression coefficient versus sampling times
下载图片 查看原文
图 6. 全谱范围内采用CARS方法挑选的波段点
Fig. 6. Band points selected by CARS method within full spectrum
下载图片 查看原文
图 7. 每个波长点被“入选”的频次
Fig. 7. Frequency of each wavelength being selected
下载图片 查看原文
图 8. RMSECV值随选取变量数的变化的趋势
Fig. 8. RMSECV varying with number of variables
下载图片 查看原文
图 9. CARS-PLS预测甲醇汽油中甲醇含量的散点图。(a)建模集;(b)预测集
Fig. 9. Scatter plots of methanol content in methanol gasoline predicted by CARS-PLS model. (a) Calibration set; (b) prediction set
下载图片 查看原文
表 1建模集和预测集样品中甲醇体积分数分布
Table1. Methanol volume fractions of samples in calibration and prediction sets
Data set | N | Minimum /% | Maximum /% | Mean /% | RSD /% |
---|
Total | 120 | 0.60 | 18.60 | 9.41 | 5.37 | Calibration | 90 | 0.60 | 18.60 | 9.43 | 5.25 | Prediction | 30 | 0.60 | 18.60 | 9.33 | 5.45 |
|
查看原文
表 2甲醇汽油的中红外光谱校正处理PLS建模效果
Table2. Calibration effect of PLS corrected by middle infrared spectra of methanol gasoline
Pretreatmentmethod | PC | Rc | RMSEC | Rp | RMSEP |
---|
Original | 9 | 0.981 | 1.029 | 0.933 | 1.925 | S-G | 13 | 0.997 | 0.375 | 0.944 | 1.772 | MSC | 10 | 0.996 | 0.450 | 0.940 | 1.843 | SNV | 3 | 0.962 | 1.440 | 0.873 | 2.871 |
|
查看原文
表 3不同波长筛选方法结合PLS模型的建模效果对比
Table3. Comparison of calibration effects of different wavelength screening methods combined with PLS method
Method | Number of wavelengths | PC | Rc | RMSEC | Rp | RMSEP |
---|
PLS | 1557 | 13 | 0.997 | 0.375 | 0.944 | 1.772 | UVE-PLS | 135 | 10 | 0.972 | 1.232 | 0.948 | 1.777 | CARS-PLS | 118 | 9 | 0.994 | 0.563 | 0.978 | 1.177 | GA-PLS | 130 | 11 | 0.983 | 0.994 | 0.956 | 1.527 |
|
查看原文
胡军, 刘燕德, 欧阳爱国, 刘洪量. 基于CARS波段筛选的甲醇汽油中甲醇含量中红外光谱检测[J]. 激光与光电子学进展, 2019, 56(23): 233002. Jun Hu, Yande Liu, Aiguo Ouyang, Hongliang Liu. Mid-Infrared Spectroscopy Detection of Methanol Content in Methanol Gasoline Based on CARS Band Screening[J]. Laser & Optoelectronics Progress, 2019, 56(23): 233002.