汽油中乙醇光谱特征谱段的有效选取及应用

近红外光谱快速分析技术是检测汽油中乙醇含量的主要方法之一,光谱谱段的选择是影响快检模型预测准确性的重要因素。本研究建立了一种基于有效特征谱段的近红外光谱快速分析方法,提高了汽油中乙醇含量检测的准确度。通过对比不同浓度乙醇含量的汽油近红外光谱图,确定了汽油中乙醇分子的有效特征谱段是4 524.183～5 044.869 cm^-1和5 985.961～7 108.329 cm^-1。选择最优的近红外光谱预处理方法,分别使用近红外光谱全谱段和有效特征谱段进行建模分析。使用特征谱段建立的数据模型相关参数如下：交叉验证均方根误差(RMSECV)是0.584 9,内部交叉验证相关系数(R_CV²)是0.999 1,预测均方根误差(RMSEP)是0.609 0,预测集外部验证相关系数(R _P²)是0.998 9。相较于全波长建模分析,使用特征谱段建立模型的RMSECV降低了30.27%,RMSEP降低了18.58%。综上,使用特征谱段建立的模型准确度较高,能够满足汽油中乙醇...     

近红外光谱技术检测烟叶中的氨基酸

研究应用傅里叶变换近红外光谱法快速测定烟叶中氨基酸含量的可行性,使用偏最小二乘法(PLS)为建模方法,选择3800~8000 cm-1谱段,采用二阶导数和Norris Derivative滤波法进行光谱预处理,建立了烟叶中氨基酸含量的近红外预测模型。采用留一(leave-one-out)交叉验证法进行建模,并以校正集样品的交叉验证相关系数(R)和均方差(RMSECV)为指标优化光谱预处理方法和模型参数,确定最佳预测模型。将近红外光谱技术与常规标准检测方法相比较,结果表明,近红外光谱技术可以较为准确的测定烟叶中氨基酸的含量。     

利用红外光谱技术快速分析油品中的芳烃含量

本文通过红外光谱技术和化学计量学方法建立了汽油中芳烃含量的快速分析方法。采集20种芳烃-汽油样品的红外光谱图进行平滑预处理,使用预处理后光谱的全谱段建立偏最小二乘法校正模型,此模型的交叉验证均方根误差是0.1271,线性相关系数是0.9995。该校正模型成功地实现了4种汽油样品中芳烃含量的准确预测,其预测均方根误差是0.0705,预测整体平均偏差是-0.0418。实验结果表明,基于红外光谱技术的芳烃含量检测方法是一种简便、快速、可靠的分析手段,可以实现油品中芳烃含量的准确测定。     

双基火药储存过程中安定剂含量的快速检测技术

为解决传统检测方法耗时长、操作复杂等问题,研究了近红外光谱法快速定量检测双芳 3双基火药中安定剂含量的可行性。通过分析安定剂的特征光谱区间,得到合适的建模波段。采用不同的光谱预处理方法和选取最佳主因子数优化模型,使用偏最小二乘法建立安定剂的定量校正模型,对模型进行了外部验证。结果表明:使用1 100~ 1 248 nm、1 323~ 1 515 nm波段,采用标准正态变量变化（SNV）预处理原始光谱,主因子数为7时建立的定量校正模型的预测准确性和稳定性较高。校正模型决定系数(R²_C)以及交互验证的决定系数(R²_CV)分别为0.991和0.987;校正标准偏差（R_MSEC）和交互验证的标准偏差（R_MSECV）分别为0.065和0.077。使用预测集样品对建立的最佳校正模型进行外部验证,安定剂含量预测值与参考值的平均误差为0.044%。该方法可用于双芳-3双基火药中安定剂含量的快速检测。维也里试验证明,近红外光谱法可以用于评估双芳-3双基火药安定性的好坏。     

不同光照下的近红外光谱模型传递研究

针对室外光照对近红外光谱检测带来误差的问题,提出基于模型传递来减少检测误差的方法。以圆黄梨为样品,分析样品在室内、室外阴影下的近红外光谱,建立室内光谱的偏最小二乘(PLS)模型。采用直接校正(direct standardization,DS)算法,减小室内外光谱差距,使得室内PLS模型能预测室外光谱。结果表明:在室内建立的模型能预测经DS算法传递后的室外光谱,预测决定系数(r2p)和预测均方根误差(root mean square error of prediction,RMSEP)分别为0.71和0.374,能有效解决室外光照对光谱检测的影响。     

近红外漫反射光谱结合神经网络检测小麦蛋白质含量

为了实现对小麦蛋白质含量的快速检测,提出了基于近红外光谱结合神经网络的小麦蛋白质检测方法.以160个小麦样品为对象,采集其近红外漫反射光谱,并以国标法分析小麦样品蛋白质含量,作为参考值.样品随机分成预测样品集和定标样品集,其光谱经标准归一化、去趋势等预处理后,采用BP神经网络和偏最小二乘法分别建立蛋白质含量定标模型.BP神经网络模型的预测相关系数和预测均方根误差分别为0.98和0.270 4%.而偏最小二乘法模型的预测相关性系数和预测均方根误差分别为0.98和0.303 8%.结果表明,两种方法建立的模型都具有较好的预测相关性和预测效果,其中BP神经网络模型优于偏最小二乘法模型.用非线性BP神经网络结合相应算法建立模型检测小麦蛋白质含量的定标模型可以提高检测准确性.     

光谱预处理对棉涤混纺面料近红外定量模型的影响

以46个棉涤混纺面料样品为研究对象,采集样品的近红外漫反射光谱,光谱范围为12 000~4 000 cm-1,利用偏最小二乘法建立定量校正模型,并用交叉检验法对模型进行检验,以交叉验证均方差RMSECV和决定系数R2作为判断模型优劣的标准.对利用无光谱预处理、一阶导数法、二阶导数法、多元散射校正和矢量归一化五种不同预处理方法所建的模型进行了比较,发现对光谱进行矢量归一化预处理所建模型最优;此外还分析了建立纺织布料的近红外光谱定量分析模型时主要的误差来源及近红外光谱分析技术用于纺织面料定量分析的可行性.     

单基发射药生产过程中驱水棉水分含量的快速检测方法?

由于驱水棉的水分含量对单基发射药成型工艺有着较大的影响,采用近红外光谱分析技术对驱水棉水分含量进行快速检测。通过对比分析纯水、硝化棉(NC)、乙醇以及驱水棉的光谱图,确定了水分含量检测建模区域为5 015.6~5 224.8 cm~(-1)和6 525.9~7 008.7 cm~(-1)。比较不同光谱预处理方法,发现标准正态变量校正(SNV)、一阶导数和平滑的组合方法对驱水棉的光谱进行预处理效果最好。采用偏最小二乘法对水分含量建立定量校正模型,并对预测集样本进行预测和对模型进行重复性验证。试验结果表明:校正集和交互验证相关系数R2分别为0.995 4和0.994 4,预测均方根误差RMSEP值为0.039 0,对预测集的药料样本预测的平均相对误差为0.997%,模型的重复性良好,检测时间小于20 s,能满足单基发射药连续自动化生产工艺的要求。     

绿茶汤中茶多酚近红外定量分析的光程选择

研究了光程对近红外定量分析绿茶汤中茶多酚的影响.以不同光程(1 mm,2 mm,5 mm)的样品池采集50个绿茶汤样品的近红外透射光谱.采用偏最小二乘法(PLS)建立茶汤中茶多酚的定量分析模型,并验证模型的准确度和精密度.结果表明,选择1 mm光程光谱建立的茶多酚定量分析模型最优,模型的校正集相关系数R2和内部交叉验证均方根RMSECV分别为91.08%和0.009 3%;检验集相关系数R2和预测标准差SEP为91.53%和0.008 5%;实际值与预测值配对t检验值为0.224 9,差异不显著.10次重复测量相对标准偏差RSD为0.008 7%,表明方法检测重复性好.     

近红外光谱法测定茶多酚中总儿茶素含量

以高效液相色谱(HPLC)分析结果为参考值,建立了快速测量茶多酚中总儿茶素含量的近红外光谱定标模型.将48份茶多酚样品组成定标样品集,在1000～2500nm(4000～10000cm-1)的近红外漫反射光谱为定标波长范围内,光谱经一阶导数(Firstderivative)、二阶导数(Secondderivative)、标准归一化(Stan-dardnormalvariate,SNV)和多元散射校正(multiplicativesignalcorrection,MSC)处理后结合偏最小二乘回归(PLS)定标.经内部交叉验证表明,光谱经SNV处理后建模结果最佳.模型的相关系数Corr.Coeff=0.997,校正均方根RMSEC=1.71%.比较了经典最小二乘法(CLS)、偏最小二乘法(PLS)和主成分回归(PCR)等方法建模结果,以偏最小二乘回归建模效果最好.     

近红外光谱测量中的多尺度建模新方法

近红外光谱学是近十年来发展最快、最引人注目的光谱分析技术之一,其高精度测量依赖于化学计量学方法以准确提取光谱信息.针对近红外光谱单尺度传统建模方法中存在的信息易丢失问题,发展了一种多尺度建模新方法.多尺度建模可有效协同利用信号的时/频多尺度特性,并将多尺度特性以加权形式统一映射到多元校正空间,有效避免了信息丢失.该算法成功地应用于面粉中掺杂有毒非法添加剂硼砂含量检测,经验证后模型的预测值与真实值的相关系数和预测均方根误差分别为0.974和0.0019,其预测相对误差为1.9%.研究结果表明,多尺度建模方法完全满足近红外光谱高精度测量的要求.     

Prediction of oil and oleic acid concentrations in individual corn (Zea mays L.) kernels using near-infrared reflectance hyperspectral imaging and multivariate analysis

Due to their heterogeneous structure and variability in form, individual corn (Zea mays L.) kernels present an optical challenge for nondestructive spectroscopic determination of their chemical composition. Increasing demand in agricultural science for knowledge of specific traits in kernels is driving the need to find high-throughput methods of examination. In this study macroscopic near-infrared (NIR) reflectance hyperspectral imaging was used to measure small sets of kernels in the spectroscopic range of 950 nm to 1700 nm. Image analysis and principal component analysis (PCA) were used to determine kernel germ from endosperm regions as well as to define individual kernels as objects out of sets of kernels. Partial least squares (PLS) analysis was used to predict oil or oleic acid concentrations derived from germ or full kernel spectra. The relative precision of the minimum cross-validated root mean square error (RMSECV) and root mean square error of prediction (RMSEP) for oil and oleic acid concentration were compared for two sets of two hundred kernels. An optimal statistical prediction method was determined using a limited set of wavelengths selected by a genetic algorithm. Given these parameters, oil content was predicted with an RMSEP of 0.7% and oleic acid content with an RMSEP of 14% for a given corn kernel.     

Evaluation of visible and near-infrared spectroscopy as a tool for assessing fiber fineness during mechanical preparation of dew-retted flax

Flax fiber must be mechanically prepared to improve fineness and homogeneity of the sliver before chemical processing and wet-spinning. The changes in fiber characteristics are monitored by an airflow method, which is labor intensive and requires 90 minutes to process one sample. This investigation was carried out to develop robust visible and near-infrared calibrations that can be used as a rapid tool for quality assessment of input fibers and changes in fineness at the doubling (blending), first, second, third, and fourth drawing frames, and at the roving stage. The partial least squares (PLS) and principal component regression (PCR) methods were employed to generate models from different segments of the spectra (400-1100, 1100-1700, 1100-2498, 1700-2498, and 400-2498 nm) and a calibration set consisting of 462 samples obtained from the six processing stages. The calibrations were successfully validated with an independent set of 97 samples, and standard errors of prediction of 2.32 and 2.62 dtex were achieved with the best PLS (400-2498 nm) and PCR (1100-2498 nm) models, respectively. An optimized PLS model of the visible-near-infrared (vis-NIR) spectra explained 97% of the variation (R(2) = 0.97) in the sample set with a standard error of calibration (SEC) of 2.45 dtex and a standard error of cross-validation (SECV) of 2.51 dtex R(2) = 0.96). The mean error of the reference airflow method was 1.56 dtex, which is more accurate than the NIR calibration. The improvement in fiber fineness of the validation set obtained from the six production lines was predicted with an error range of -6.47 to +7.19 dtex for input fibers, -1.44 to +5.77 dtex for blended fibers at the doubling, and -4.72 to +3.59 dtex at the drawing frame stages. This level of precision is adequate for wet-spinners to monitor fiber fineness of input fibers and during the preparation of fibers. The advantage of visNIR spectroscopy is the potential capability of the technique to assess fineness and other important quality characteristics of a fiber sample simultaneously in less than 30 minutes; the disadvantages are the expensive instrumentation and the expertise required for operating the instrument compared to the reference method. These factors need to be considered by the industry before installing an off-line NIR system for predicting quality parameters of input materials and changes in fiber characteristics during mechanical processing.     

Rapid quantification of carotenoids and fat in Atlantic Salmon (Salmo salar L.) by Raman spectroscopy and chemometrics

Raman spectroscopy (785 nm excitation) was used to determine the overall carotenoid (astaxanthin and cantaxanthin) and fat content in 49 samples of ground muscle tissue from farmed Atlantic salmon (Salmo salar L.). Chemically determined contents ranged from 1.0 to 6.8 mg/kg carotenoids and 36 to 205 g/kg fat. In addition to the raw Raman spectra, three types of spectral preprocessing were evaluated: the first derivative, subtraction of the fitted fourth-order polynomial (POLY), and the intensity normalized versions of POLY (POLY-SNV). Further, variable selection based on significance testing by use of jack-knifing was performed on each spectral data set. Partial least-squares regression resulted in a root mean square error of prediction of 0.33 mg/kg (R = 0.97) for carotenoids for the variable selected versions of all the preprocessed spectral data sets. The fat content was best estimated by the variable selected POLYSNV, resulting in a root mean square error of prediction of 15.5 g/kg (R = 0.95). Both preprocessing and variable selection improved the regression models significantly. The results demonstrate that Raman spectroscopy is a suitable method for simultaneous, rapid, and nondestructive quantification of both pigments and fat in ground salmon muscle tissue.     

Elastic net grouping variable selection combined with partial least squares regression (EN-PLSR) for the analysis of strongly multi-collinear spectroscopic data

In this paper a novel wavelength region selection algorithm, called elastic net grouping variable selection combined with partial least squares regression (EN-PLSR), is proposed for multi-component spectral data analysis. The EN-PLSR algorithm can automatically select successive strongly correlated prediction variable groups related to the response variable using two steps. First, a portion of the correlated predictors are selected and divided into subgroups by means of the grouping effect of elastic net estimation. Then, a recursive leave-one-group-out strategy is employed to further shrink the variable groups in terms of the root mean square error of cross-validation (RMSECV) criterion. The performance of the algorithm with real near-infrared (NIR) spectroscopic data sets shows that the EN-PLSR algorithm is competitive with full-spectrum PLS and moving window partial least squares (MWPLS) regression methods and it is suitable for use with strongly correlated spectroscopic data.     

Assessment of pareto calibration, stability, and wavelength selection

Recent work has shown that ridge regression (RR) is Pareto to partial least squares (PLS) and principal component regression (PCR) when the variance indicator Euclidian norm of the regression coefficients, //p//, is plotted against the bias indicator root mean square error of calibration (RMSEC). Simplex optimization demonstrates that RR is Pareto for several other spectral data sets when //p// is used with RMSEC and the root mean square error of evaluation (RMSEE) as optimization criteria. From this investigation, it was observed that while RR is Pareto optimal, PLS and PCR harmonious models are near equivalent to harmonious RR models. Additionally, it was found that RR is Pareto robust, i.e., models formed at one temperature were then used to predict samples at another temperature. Wavelength selection is commonly performed to improve analysis results such that bias indicators RMSEC, RMSEE, root mean square error of validation, or root mean square error of cross-validation decrease using a subset of wavelengths. Just as critical to an analysis of selected wavelengths is an assessment of variance. Using wavelengths deemed optimal in a previous study, this paper reports on the variance/bias tradeoff. An approach that forms the Pareto model with a Pareto wavelength subset is suggested.     

On-line monitoring of the buffer capacity of particleboard furnish by near-infrared spectroscopy

This paper reports on a study of on-line monitoring of the buffer capacity of particleboard furnish using near-infrared (NIR) spectroscopy and multivariate analysis models (chemometrics). The buffer capacity of wood furnish is known to affect the quality of polymerization and the curing rates of urea-formaldehyde (UF) resins, which may affect the mechanical properties of manufactured panel. The first phase of the study consisted of building multivariate calibration and validation models from NIR spectroscopy data to predict the buffer capacity of particleboard furnish in a laboratory environment. During this phase, a spectrometer (Ocean Optics USB2000) operating in the 550-1100 nm spectral range was evaluated. The second phase of the study took place at a North American particleboard plant over several weeks. Several multivariate calibration models were constructed and tested on-line during a four-day test period. The on-line root mean square error of prediction (RMSEP) and the coefficient of variation (CV) for buffer capacity predictions ranged from 3.45 to 0.92 and 22.4% to 5.8%, respectively.     

Determination of the crystallinity of cephalexin in pharmaceutical formulations by chemometrical near-infrared spectroscopy

Background: Near-infrared (NIR) spectroscopy has gained wide acceptance in the pharmaceutical industry as a rapid and non destructive method for drug identification and the determination of the drug content of preparations. Aim: The crystallinity of cephalexin (CEX) in microcrystalline cellulose (MCC) was determined using a nondestructive NIR reflectance spectroscopic technique. The molecular interaction of a ground amorphous solid of CEX was investigated by the method. Method: Six kinds of standard material with various degrees of crystallinity were prepared by the physical mixing of crystalline, amorphous CEX, and MCC. X-ray powder diffraction profiles and NIR spectra were recorded for standard samples. A chemometric analysis of the NIR spectral data sets was conducted using principal component regression (PCR). Results: The correlation between the actual crystallinity of CEX and that predicted using the conventional X-ray diffraction method showed a straight line with a slope of 1.000, an intercept of ?2.071 × 10^?5 and a correlation coefficient of determination (R²) of 0.974. The NIR spectrum of amorphous CEX showed significantly different peaks at 1176 and 1206 nm because of the CH₃ group from those of CEX. PCR was performed on various kinds of pretransformed NIR spectral data sets of standard samples of CEX. To minimize the SE of cross-validation (SECV), the spectral data sets were subjected to the leave-one-out method. The second derivative treatment in the range of 1176–1206 nm yielded the lowest SECV values. Based on a two-component model, a plot of the calibration data between the actual crystallinity of CEX and that predicted by the NIR method was obtained. The plot showed a straight line (Y = 0.995X + 0.117 and R² = 0.994; n = 18). The mean bias for the NIR and X-ray powder diffraction methods was calculated to be 1.52% and 2.26%, and mean accuracy was 3.06% and 7.14%, respectively. Conclusion: NIR spectral changes of crystalline CEX during grinding suggested that the intermolecular hydrogen bonds between the amino and carboxyl groups are destroyed and the binding of methyl groups is heightened by the resonance effect of carboxyl groups, and the crystals are transformed into amorphous CEX.     

Determination of humic acids content in composts by means of near- and mid-infrared spectroscopy and partial least squares regression models

Humic acids are part of the stable organic matter fraction in soils and composts. Due to their favorable properties for soils and plants, and their role in carbon sequestration, they are considered a quality criterion of composts. Time-consuming chemical extraction of humic acids and the inherent source of errors require alternative approaches for humic acids quantification. Different measurement techniques in the mid-infrared (MIR: KBr pellet technique) and near-infrared (NIR: fiber probe as well as an integrating sphere with a sample rotator) regions were applied. Partial least squares regression (PLSR) models based on infrared spectra were developed to determine humic acids contents in composts. As the wavenumber regions used (NIR: 6105-5380 cm(-1) and 4360-4220 cm(-1), MIR: 1745-1685 cm(-1) and 1610-1567 cm(-1)) represent different molecular vibrations, the importance of the methylene-group-derived vibrations for the NIR models is discussed. The correlation coefficients obtained for the KBr pellet technique, the NIR fiber probe technique, and the NIR integrating sphere (r = 0.94, 0.93, and 0.94) and the root mean square errors of cross-validation (RMSECV = 2.2% organic dry matter (ODM), 2.5% ODM, and 2.2% ODM) make the models appropriate for application in composting practice.     

Fatty acid profiling of soybean cotyledons by near-infrared spectroscopy

Genetically improved soybean grain often contains altered fatty acid profiles. Such alterations can have deleterious effects on seed germination and seedling development, making it necessary to monitor fatty acid profiles in follow-up physiological studies. The objective of this research was to quantify the five fatty acids in soybean (Glycine max) cotyledons using near-infrared (NIR) spectroscopy. Soybean cotyledon samples were dried, ground, and scanned with visible and NIR radiation from 400 to 2500 nm, and reflectance was recorded. Samples were also analyzed by gas chromatography (GC) for palmitic, stearic, oleic, linoleic, and linolenic acids and total oil; GC data, expressed as actual concentration and proportion of total oil, were regressed against spectral data to develop calibration equations. Equation statistics indicated that four of the five fatty acids could be predicted accurately by NIR spectroscopy; the fifth fatty acid could be determined by subtraction. Principal component analysis revealed that most of the spectral variation in this population was due to chlorophyll absorbance in the visible region. Therefore, the spectra were trimmed to include the NIR region only (1100-2500 nm), and a second set of equations was developed. Equations based exclusively on NIR spectra had equal or greater precision than equations based on visible and NIR spectra. Principal component analysis and partial least squares analysis revealed that even after trimming, at least 90% of the spectral variation was unrelated to fatty acid, though variation from fatty acid was identified in the second and third principal components. This research provides an NIR method for complete fatty acid profiling of soybean cotyledons. Equations were achieved with NIR spectra only, so spectrophotometers that analyze both the visible and NIR regions are not needed for this analysis. In addition, equations were possible with a 250 mg sample, which is one-tenth the normal sample size for this analysis.