作者投稿
专家审稿
编辑办公
Rapid Identification of Alcohol-Based Gasoline Types Using Modelling of Effective Chemical Information from Near-Infrared Spectroscopy
-
摘要: 甲醇汽油和乙醇汽油是性质特性不同的两种新型清洁能源，准确识别醇基汽油种类是对于鉴定燃油品质、保证车辆安全行驶具有重要意义。近红外光谱（NIR）是识别醇基汽油种类的重要手段，而利用近红外光谱的有效特征波长建模可以摆脱干扰谱段和噪声谱段对模型识别准确度的影响，本研究使用甲醇分子和乙醇分子的近红外光谱特征谱段建立偏最小二乘法判别（partial least squares discriminant analysis, PLS-DA）模型，用于识别甲醇汽油和乙醇汽油。在相同建模条件下，将全波长光谱模型、变量重要性投影（variable importance in projection, VIP）光谱和特征谱段模型的识别成功率进行对比。结果表明，全波长光谱模型和VIP光谱模型对低含量的醇基汽油样本的识别准确率较差，分别为90%和96.7%；而最佳特征谱段模型，即通过甲醇和乙醇分子全部差异光谱信息（4500～5200+5600～7200+7900～8800）cm−1建立的模型，可准确识别体积分数为0.5%～80%的醇基汽油样本，识别成功率为100%。除最佳特征谱段模型外，其余特征谱段PLS-DA模型的识别成功率普遍优于全波长光谱模型，结果证明通过化学结构筛选特征谱段是一种高效的波长选择方法，利用该特征光谱谱段建模能够提高模型的准确度。综上，成功建立了一种基于化学信息筛选近红外光谱特征谱段的高准确率定性识别醇基汽油种类的模型，有望推广应用在其他燃油的种类识别中。Abstract: Methanol gasoline and ethanol gasoline, as two distinct types of clean energy sources, possess unique properties and characteristics. Accurate identification of these alcohol-based gasoline types is crucial for ensuring fuel quality and vehicle safety. Near-Infrared Spectroscopy (NIR) is a pivotal method for identifying alcohol-based gasoline types, and modelling based on NIR's effective characteristic wavelengths can overcome the effects of interfering and noisy spectral segments on model accuracy. This study employed the characteristic spectral bands of methanol and ethanol molecules in NIR spectra to develop a Partial Least Squares Discriminant Analysis (PLS-DA) model for distinguishing between methanol and ethanol gasoline. The success rates of full-wavelength spectral, Variable Importance in Projection (VIP) spectral, and feature spectral band models were compared under identical modelling conditions. Results indicated that the full-wavelength and VIP spectral models showed lower accuracy in identifying low-content alcohol-based gasoline samples, with success rates of 90% and 96.7% respectively. In contrast, the optimal feature spectral band model, constructed using the complete differential spectral information of methanol and ethanol molecules (4500～5200+5600～7200+7900～8800 cm−1), achieved a 100% success rate in identifying alcohol-based gasoline samples with volume fractions of 0.5% to 80%. The study demonstrates that selecting feature spectral bands based on chemical structure is an effective wavelength selection method, enhancing model accuracy significantly. In summary, this research successfully establishes a highly accurate model for the qualitative identification of alcohol-based gasoline types using selected NIR spectral bands based on chemical information, showing potential for application in other fuel type identifications.
-
图 4 甲醇和乙醇的近红外吸收强度差异
Figure 4. Differences in near-infrared absorption intensities of methanol and ethanol
图 5 不同光谱模型的输入变量和识别结果
Figure 5. Input variables and identification results of different spectral models
图 6 不同光谱模型预测集的混淆矩阵
Figure 6. Confusion matrices of prediction sets for different spectral models
图 7 RMSEP和 $ {R}_{P}^{2} $随主因子数变化曲线
Figure 7. Variation curve of RMSEP and $ {\mathit{R}}_{\mathit{P}}^{2} $ with the number of principal components
图 8 最佳特征谱段模型三维评分图
Figure 8. Three-dimensional scoring plots for the optimal characteristic spectral band model
表 1 醇基汽油样本中醇类含量
Table 1. Alcohol content in alcohol-based gasoline samples
编号 体积分数（v/v, %） 编号 体积分数（v/v, %） 编号 体积分数（v/v, %） 编号 体积分数（v/v, %） 1 0 16 7.5 31 15 46* 25 2* 0.5 17 8 32 15.5 47 26 3 1 18* 8.5 33* 16 48 27 4 1.5 19 9 34 16.5 49* 28 5 2 20 9.5 35 17 50 29 6* 2.5 21* 10 36 17.5 51 30 7 3 22 10.5 37* 18 52 35 8 3.5 23* 11 38 18.5 53 40 9 4 24 11.5 39 19 54* 45 10* 4.5 25 12 40 19.5 55 50 11 5 26 12.5 41* 20 56 55 12 5.5 27 13 42 21 57 60 13 6 28* 13.5 43 22 58 70 14* 6.5 29 14 44 23 59* 80 15 7 30 14.5 45 24 60 90 注: 标有*的样本为预测集。 
下载: 导出CSV
表 2 基于化学结构信息筛选的特征谱段方案
Table 2. Characteristic spectral band selection based on chemical structure information
编号 波数选择/cm−1 选择依据 1 4500～5200 O-H弯曲和伸缩振动的组合频 2 5600～6000 C-H键伸缩振动的一阶倍频 3 6000～7200 O-H键伸缩振动的一阶倍频 4 7900～8800 C-H 键对称伸缩的二阶倍频 5 4500～5200
6000～7200O-H键 6 5600～6000
7900～8800C-H键 7 4500～5200
5600～7200
7900～8800全部特征谱段
下载: 导出CSV
表 3 不同光谱预处理的PLS-DA模型效果
Table 3. Performance of PLS-DA models with different spectral preprocessing
预处理方法 RMSEP $ {R}_{P}^{2} $ 识别准确率 未预处理 0.1590 0.8988 87% VN+SG 1stD 0.1003 0.9597 90% SG+SG 1stD 0.1746 0.8778 87% MSC+SG 1stD 0.2054 0.8555 83% SG 1stD 0.1942 0.8645 83%
下载: 导出CSV
-
[1] 许勤华, 张艳伟. 绿色能源的技术突破和未来能源产业前瞻[J]. 人民论坛, 2023(16): 45-49. [2] YESILYURT M K, EROL D, YAMAN H, et al. Effects of using ethyl acetate as a surprising additive in SI engine pertaining to an environmental perspective[J]. International Journal of Environmental Science and Technology, 2022, 19(10): 9427-9456. doi: 10.1007/s13762-021-03706-3 [3] LI S H, WEN Z, HOU J, et al. Effects of Ethanol and Methanol on the Combustion Characteristics of Gasoline with the Revised Variation Disturbance Method[J]. ACS Omega, 2022, 7(21): 17797-17810. doi: 10.1021/acsomega.2c00991 [4] HO C S, PENG J F, YUN U, et al. Impacts of methanol fuel on vehicular emissions: A review[J]. Frontiers of Environmental Science & Engineering, 2022, 16(9): 121. [5] TIAN Z, ZHEN X, WANG Y, et al. Comparative study on combustion and emission characteristics of methanol, ethanol and butanol fuel in TISI engine[J]. Fuel, 2020, 259: 116199. doi: 10.1016/j.fuel.2019.116199 [6] LEE D M, LEE D H, HWANG I H. Gasoline quality assessment using fast gas chromatography and partial least-squares regression for the detection of adulterated gasoline[J]. Energy & Fuels, 2018, 32(10): 10556-10562. [7] AVILA L M, DOS SANTOS A P F, DE MATTOS D I M, et al. Determination of ethanol in gasoline by high-performance liquid chromatography[J]. Fuel, 2018, 212: 236-239. doi: 10.1016/j.fuel.2017.10.039 [8] YANG Y J, WEI Y J, WANG W R, et al. The Effects of Methanol Fraction on the Azeotropic Behaviors of Methanol/Gasoline Mixtures[J]. International Journal of Green Energy, 2015, 12(11): 1076-1085. doi: 10.1080/15435075.2014.890102 [9] WU Y S, LIU Y S, LI X L, et al. Gasoline octane number prediction from near-infrared spectroscopy with an ANN-based model[J]. Fuel, 2022, 318: 123543. doi: 10.1016/j.fuel.2022.123543 [10] CáMARA A B F, DE CARVALHO L S, DE MORAIS C L M, et al. MCR-ALS and PLS coupled to NIR/MIR spectroscopies for quantification and identification of adulterant in biodiesel-diesel blends[J]. Fuel, 2017, 210: 497-506. doi: 10.1016/j.fuel.2017.08.072 [11] LIU Z, LUO N N, SHI J L, et al. Raman spectroscopy for the discrimination and quantification of fuel blends[J]. Journal of Raman Spectroscopy, 2019, 50(7): 1008-1014. doi: 10.1002/jrs.5602 [12] 李轲, 杜彪, 肖哲, 等. 基于近红外光谱技术的油品快检方法研究进展 [J]. 计量科学与技术, 2022, 66(12): 3-10, 26. [13] GORLA G, FERRER A, GIUSSANI B. Process understanding and monitoring: A glimpse into data strategies for miniaturized NIR spectrometers[J]. Analytica Chimica Acta, 2023, 1281: 341902. doi: 10.1016/j.aca.2023.341902 [14] 胡军, 刘燕德, 郝勇, 等. 甲醇汽油、乙醇汽油定性判别及其醇含量测定模型研究 [J]. 光谱学与光谱分析, 2020, 40(5): 1640-1644. [15] 丁怡曼, 薛晓康, 范宾, 等. 基于便携式拉曼光谱的汽油快速识别模型 [J]. 石油炼制与化工, 2021, 52(11): 64-69. [16] JOHNSTONE I M, TITTERINGTON D M. Statistical challenges of high-dimensional data [J]. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 2009, 367(1906): 4237-4253. [17] LI K, ZHANG C, DU B, et al. Selection of the effective characteristic spectra based on the chemical structure and its application in rapid analysis of ethanol content in gasoline[J]. ACS Omega, 2022, 7(23): 20291-20297. doi: 10.1021/acsomega.2c02282 [18] YUN Y H, LI H D, DENG B C, et al. An overview of variable selection methods in multivariate analysis of near-infrared spectra[J]. Trac-Trend Anal Chem, 2019, 113: 102-115. doi: 10.1016/j.trac.2019.01.018 [19] ZOU X B, ZHAO J W, Malcolm J W P, et al. Variables selection methods in near-infrared spectroscopy[J]. Analytica Chimica Acta, 2010, 667(1-2): 14-32. doi: 10.1016/j.aca.2010.03.048 [20] 李轲, 鲁冰, 杜彪, 等. 汽油中乙醇光谱特征谱段的有效选取及应用 [J]. 计量科学与技术, 2022, 66(5): 19-24. [21] 国家市场监督管理总局, 国家标准化管理委员会. 车用甲醇汽油(M85): GB 23799—2021 [S]. 北京: 中国标准出版社, 2021. [22] 国家质量监督检验检疫总局, 中国国家标准化管理委员会. 车用乙醇汽油E85: GB 35793—2018 [S]. 北京: 中国标准出版社, 2018. [23] 李琪, 杜彪, 张正东, 等. 傅里叶变换近红外光谱仪在汽、柴油分析中的应用 [J]. 计量科学与技术, 2022, 66(10): 20-27. [24] GERRETZEN J, SZYMAŃSKA E, BART J, et al. Boosting model performance and interpretation by entangling preprocessing selection and variable selection[J]. Analytica Chimica Acta, 2016, 938: 44-52. doi: 10.1016/j.aca.2016.08.022 [25] 阎续, 沈丽娟, 胥文彦, 等. 拉曼光谱用于CHO细胞培养液多指标快速分析 [J]. 高校化学工程学报, 2019, 33(4): 872-877. [26] 龚瑞昆, 赵学智, 赵福生. 基于EfficientNetV2-HDCA模型水下鱼类图像分类算法研究[J]. 电子测量技术, 2022, 45(22): 128-134. doi: 10.19651/j.cnki.emt.2209829 [27] 刘楠, 刘翠玲, 徐金阳, 等. 基于极限学习机自编码算法的近红外光谱模型传递的研究 [J]. 食品安全质量检测学报, 2023, 14(5): 30-36. [28] LIU H, XU J P, QU L B, et al. Generalized two-dimensional correlation near-infrared spectroscopy and principal component analysis of the structures of methanol and ethanol[J]. Science China Chemistry, 2010, 53(5): 1155-1160. doi: 10.1007/s11426-010-0172-2 [29] ADACHI D, KATSUMOTO Y, SATO H, et al. Near-infrared spectroscopic study of interaction between methyl group and water in water-methanol mixtures[J]. Applied Spectroscopy, 2002, 56(3): 357-361. doi: 10.1366/0003702021954728 [30] DEVOS O, DOWNEY G, DUPONCHEL L. Simultaneous data pre-processing and SVM classification model selection based on a parallel genetic algorithm applied to spectroscopic data of olive oils[J]. Food chemistry, 2014, 148: 124-130. doi: 10.1016/j.foodchem.2013.10.020 [31] LOTTERING R T, GOVENDER M, PEERBHAY K, et al. Comparing partial least squares (PLS) discriminant analysis and sparse PLS discriminant analysis in detecting and mapping Solanum mauritianum in commercial forest plantations using image texture[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2020, 159: 271-280. doi: 10.1016/j.isprsjprs.2019.11.019 -
点击查看大图
图(8) / 表(3)
计量
- 文章访问数: 134
- HTML全文浏览量: 40
- PDF下载量: 35
- 被引次数: 0
