拉曼光谱结合偏最小二乘法在柴油正丁醇快速定量分析中的应用

正丁醇具有互溶性好、挥发性低、价格低廉以及腐蚀性低等优势,被认为是理想的柴油添加物。柴油中正丁醇的精准定量分析对其品质快速评价与市场监督具有重要科学意义与实用价值。提出了一种基于拉曼(Raman)光谱结合偏最小二乘(PLS)的柴油中正丁醇快速定量分析方法。首先,采集了40个柴油样品的Raman光谱,并考察了不同预处理方法[一阶导数、二阶导数、多元散射校正、标准正态变换、归一化(Normalization)和小波变换]对PLS校正模型预测性能的影响;其次,采用变量重要性投影(VIP)对Normalization方法预处理后的光谱数据进行特征变量提取,并采用五折交叉验证优化VIP的阈值;最后,基于最优的光谱预处理方法、输入变量和模型参数,构建PLS校正模型对柴油中正丁醇含量进行快速定量分析,结果与基于原始光谱(RAW)和Normalization光谱的PLS校正模型的预测性能进行对比。结果表明：Normalization-VIP-PLS校正模型展现出优异的预测能力(R_CV²和RMSECV为0.998 4和0.236 2%:R²     

快速稳健偏最小二乘回归及其在近红外光谱分析中的应用

现代近红外光谱,作为一种间接分析技术,将建立校正模型,实现对未知样本的定量分析。针对近红外光谱分析灵敏度低、抗干扰性差的弱点,构建一种快速稳健的偏最小二乘回归(RRPLSR)算法。它运用峭度法快速识别离群点,排除它们后,再实施偏最小二乘回归,消除复共线性,建立稳健可靠的定量校正模型。将RRPLSR方法实际应用于鱼类物质的近红外光谱数据分析,实现脂肪含量的定量检测,效果良好。与已有的其他方法相比,它能准确识别离群点,所建模型预测性能良好,且计算省时,效率高,适用于快速检测。     

羊肉嫩度傅里叶变换近红外光谱偏最小二乘法定量分析研究

以从内蒙、宁夏、甘肃、新疆4个肉羊产区筛选的有代表性的98份羊肉样品为试材,应用傅里叶变换近红外光谱技术探讨了羊肉嫩度无损检测的方法。以模型决定系数(r2)、校正标准差(RMSECV)和预测标准差(RMSEP)为模型精度评价指标,采用偏最小二乘法(PLS)对近红外光谱信息与样品的质构仪剪切力值进行了拟合,确定了最佳的光谱预处理方法、主成分数和波段范围。结果表明：所选98个羊肉样品的剪切力值分布范围为1.673～6.631 kg,其中75%以上的样品剪切力值在2～5 kg,基本覆盖了我国现有的肉羊嫩度值分布;在11 995～5 446 cm-1和4 601～4 246 cm-1的波段范围内,最佳主成分数为10,光谱经矢量归一法处理后,建立的羊肉嫩度模型精度最高,r2达到86.2%,RMSECV为0.445;用此模型对预测集29个样品进行预测,预测值与实测值的相关系数r达到0.87,预测平均偏差为0.385,RMSEP为0.524。     

基于偏最小二乘增量式神经网络的近红外光谱定量分析模型

提出了一种基于偏最小二乘增量式神经网络的近红外光谱定量分析模型。该模型采用典型三层反向传播神经网络(BPNN),不同波长吸光度和成分浓度是模型的输入和输出。在使用历史样本训练之前先进行偏最小二乘(PLS)回归,所得自变量和因变量的历史负荷矩阵分别用于确定模型输入层和输出层的初始权值,且自变量的主成分个数作为隐层的节点数。当获得新的样本时,对新数据与历史负荷矩阵组合后进行PLS回归,将所得新的负荷矩阵与历史负荷矩阵融合后作为模型输入层和输出层新的初始权值,接着使用新样本对模型进行训练来实现增量式更新。将所提模型与PLS、BPNN、基于PLS的BPNN、递归PLS在天然气燃烧烟气近红外光谱数据上测定后比较。对于烟气中二氧化碳浓度的预测,所提模型的预测均方根误差(RMSEP)分别降低了27.27%,58.12%,19.24%和14.26%;对于烟气中一氧化碳浓度的预测,所提模型的RMSEP分别降低了20.65%,24.69%,18.54%和19.42%;对于烟气中甲烷浓度的预测,此模型的RMSEP分别降低了27.56%,37.76%,8.63%和3.20%。实验结果表明,所提模型不仅通过PLS对BPNN结构和初始权重的优化,使模型具有较强的预测能力,而且能在已建模型信息的基础上,不访问旧数据而用新增样本即可完成自身的增量式更新,从而使模型具有较好的稳健性和泛化性。     

偏最小二乘法在火焰原子吸收分析中的应用

本文用偏最小二乘法校正了火焰原子吸收分析中Ｐｄ２４７．４３ｎｍ对Ｐｂ２４７．６３８ｎｍ的吸收线重叠干扰，对合成样中Ｐｂ和Ｐｄ的测定，结果令人满意，ＰＬＳ法是谱线重叠干扰的有效校正方法。     

偏最小二乘回归在微波效应预测中的应用

 引入偏最小二乘回归(PLSR)原理和方法应用于微波效应实验数据的预测,得到的预测精度与自适应神经模糊推理网络(ANFIS)结果基本一致,平均相对误差小于3%。实例分析了PLSR方法与ANFIS方法对建模数据样本量的需求,在建模样本数较少条件下,PLSR所建模型的预测精度均高于ANFIS模型。因此PLSR方法更适用于微波效应小样本数据的预测,更有利于实际应用。     

偏最小二乘回归在微波效应预测中的应用

引入偏最小二乘回归(PLSR)原理和方法应用于微波效应实验数据的预测,得到的预测精度与自适应神经模糊推理网络(ANFIS)结果基本一致,平均相对误差小于3%。实例分析了PLSR方法与ANFIS方法对建模数据样本量的需求,在建模样本数较少条件下,PLSR所建模型的预测精度均高于ANFIS模型。因此PLSR方法更适用于微波效应小样本数据的预测,更有利于实际应用。     

偏最小二乘法在傅里叶变换红外光谱中的应用及进展

偏最小二乘法(PLS)是一种应用非常广泛的化学计量方法,它综合了多元线性回归法(MLR)和主成分回归法(PCR)的优势,具有预测能力强和模型相对简单等优点。PLS使傅里叶变换红外光谱的应用范围不断扩大,同时算法也得到了改进和完善。文章介绍了偏最小二乘法在傅里叶变换红外光谱中的应用,对改进算法,如移动窗口PLS(MWPLS)、稳健PLS(RPLS)、加权PLS(WPLS)和非线性PLS等进行了介绍。同时,对应用PLS时数据的预处理、变量的选择、噪声的处理和非线性模型的建立进行了综述。     

基于近红外光谱和稀疏偏最小二乘回归的生物质工业分析

林木生物质能源作为一种新型可再生能源,具有非常广阔的发展前景。基于近红外光谱技术,首次引入稀疏偏最小二乘回归建立木屑生物质的工业分析模型,用于生物质燃料特性的快速分析测定。工业分析总共测定了80种木屑的水分、灰分、挥发分和固定碳含量百分比;按照样品种类和产地将其划分为训练集和测试集,利用近红外光谱仪采集光谱数据并进行小波滤波处理;再利用稀疏偏最小二乘回归建立木屑生物质的定量分析模型,并与主成分回归、偏最小二乘回归、最小绝对收敛及变量筛选方法的建模效果进行比较。结果证明,相对于以上三种建模方法,稀疏偏最小二乘回归能够挑选出有重要影响的波长群组,降低非目标波段的噪声干扰,从而增强数学模型的解释能力并提高定量分析的准确度。利用稀疏偏最小二乘回归算法挑选的波长区间基本覆盖了工业分析中水分的吸收峰,而对于灰分、挥发分和固定碳的吸收峰波段尚无准确定位,需要继续探讨。总体而言,稀疏偏最小二乘回归能够减少无关信息的干扰,提高模型定量分析的准确度,增强模型的解释能力,将会在近红外光谱技术应用领域内起到重要作用。     

二维相关近红外光谱检测牛奶中的三聚氰胺

配置合格的纯牛奶样本及含有三聚氰胺质量浓度范围为0.01g/L～3g/L的掺杂牛奶样本各20个,并采集其近红外光谱。以牛奶中掺杂三聚氰胺浓度为外扰,构建二维相关同步谱,研究其相关谱特性。在此基础上,结合偏最小二乘判别分析法(PLS-DA)建立定性模型,可以实现纯牛奶与掺伪牛奶的定性鉴别,正确识别率达100%。同时,将二维相关近红外同步谱矩阵与偏最小二乘法(PLS)结合起来,建立定量分析牛奶中掺杂三聚氰胺的数学模型。对未知样品的预测相关系数R达到0.98,预测均方根误差(RM-SEP)为0.18g/L,说明基于同步相关谱矩阵建立定量分析的数学模型是可行的。该方法无需样品处理,成本低,为快速检测掺伪牛奶提供了一种新的途径。     

基于近红外光谱和偏最小二乘法的慈竹纤维素结晶度预测研究

将近红外光谱技术和化学计量学相结合分析慈竹纤维素结晶度。通过区间偏最小二乘法(iPLS)、联合区间偏最小二乘法(siPLS)和反向区间偏最小二乘法(biPLS)优化建模区域,建立经多元散射校正后光谱的结晶度分析模型,并与全光谱范围350～2 500nm建立的偏最小二乘(PLS)模型进行比较。结果表明,三种改进偏最小二乘法建立的结晶度模型预测效果均优于PLS模型,并且当采用联合区间偏最小二乘法将全光谱进行30个子区间划分,选择三个子区间[8 12 19]组合时,建立的siPLS模型预测效果最好,相关系数(r)达到0.88,预测标准差(RMSEP)为0.0117。因此,采用联合区间偏最小二乘法可以有效选择建模光谱区域,提高模型预测能力,实现慈竹纤维素结晶度的快速预测。     

氯化钠近红外光谱检测技术研究

氯化钠(NaCl)近红外光谱分析在生物医学上有着重要的意义。钠离子(Na+)是人体血液中电解质的主要成分,而电解质有助于维持身体的酸碱平衡。采用近红外光谱技术测量氯化钠浓度,在分析钠离子近红外光谱检测机理的基础上,选定波长建立了NaCl浓度线性回归预测模型,同时为了减小温度对水吸收的扰动,使用选定光谱区建立偏最小二乘(PLS)非线性回归模型。结果表明所建立的非线性校正模型决定系数(R2)=99.82%,交叉验证均方误差(RMSECV)=14.5,剩余预测偏差(RPD)=23.7。完全满足日常生化检测精度要求,该技术可以应用于医院实验室钠离子浓度定量分析。     

Near-infrared spectroscopy-based methods for quantitative determination of active pharmaceutical ingredient in transdermal gel formulations

Abstract

Semi-solid pharmaceutical forms, like topical and transdermal products, are widely used for medical purposes since they provide convenient and pain-free treatment for patients. In this study, two transdermal gel formulations with different active pharmaceutical ingredients were analyzed with near-infrared spectrophotometers with three distinct optical arrangements (dispersive, Fourier transform, and diode-array spectrophotometers) for the measurement of active pharmaceutical ingredient concentration. The reference concentrations were measured with high-performance liquid chromatography methods.

The developing of calibration for reference concentrations based on near-infrared spectra was performed using partial least squares regression. The spectra from near-infrared spectrophotometer with diode-array optical arrangement presented the best root mean square error of cross-validation results with 0.045% and 0.144% for the two active pharmaceutical ingredients. The root mean square error of cross-validation using dispersive spectra were 0.125% and 0.193% and using Fourier transform spectra were 0.130% and 0.344% for the active pharmaceutical ingredients. The results of partial least squares regression were confirmed using simulated spectra based on the measured datasets.

Near-infrared spectroscopy can be used successfully for the quantitative measurement of active pharmaceutical ingredient concentration in transdermal gel formulations. The type of optical arrangements of spectrophotometers can influence the performance of calibrations and have to be optimized case by case.     

Portable milk product quality analyzer based on spectrometry in the near IR range

We have used near IR spectrometry and multiple regression analysis to construct the calibration models for determining the weight fractions of the major components of milk blends, and we have used the latter to study the feasibility of using a portable spectrophotometer (designed at the Institute of Spectroscopy) based on a silicon CCD linear array as a working tool for monitoring the quality of biotechnological products. __________ Translated from Zhurnal Prikladnoi Spektroskopii, Vol. 75, No. 2, pp. 274–279, March–April, 2008.     

Rapid determination of flavonoids in green tea by synchronous fluorescence spectra coupled with chemometrics

Flavonoids are one of the most important antioxidant compounds of tea. It is necessary to determine flavonoid concentration of tea infusions and to understand how the storage condition and tea brewing processing impact on flavonoid content. The aim of this study is to analyze flavonoids’ properties and predict flavonoid concentration using synchronous fluorescence spectra coupled with chemometrics. Different wavelength intervals were measured to develop an optimal model. Successive projection algorithms and interval partial least square regression were applied to extract informative wavelengths. The optimal interval partial least square regression model was obtained with the determination coefficient of calibration of 0.95, determination coefficient of cross-validation of 0.90.     

玉米近红外光谱在仿生模式识别中的特征提取方法研究

近红外光谱特征提取是近红外光谱定性分析的关键步骤,其质量直接影响定性分析结果。采用漫透射方法测量8个玉米品种的近红外光谱,经预处理后,分别采用PCA,ICA,PLS-DA和小波分解四种方法对光谱进行特征提取,并使用仿生模式识别方法建立了8个玉米品种识别模型,最后使用测试集数据进行模型测试。结论如下:使用PLS-DA方法进行特征提取后建立的模型正确识别率优于使用PCA,ICA和小波分解特征提取后建立的模型。     

偏最小二乘法和THz-TDS在正品大黄鉴别中的应用

太赫兹技术的发展近年来受到广泛的关注并被应用于热点。中草药大黄的品质鉴定对于中药制剂的质量控制具有重要的意义。利用大黄的太赫兹时域光谱结合偏最小二乘法（PLS）模型对基于41个正品和非正品大黄的中草药鉴别模型进行了研究。首先采集大黄样品的太赫兹时域光谱（THz-TDS）信号,然后将化学计量学方法用于这些大黄样品太赫兹光谱的信号处理与建模,再建立基于太赫兹光谱的大黄品质鉴定的偏最小二乘模型方法。应用S-G一阶导数、去趋势、标准正态变换、自标度化、均值中心化等方法对原始时域谱预处理再与未经预处理的结果相比,偏最小二乘（PLS）模型的预测正确率从80%明显提高到90%。在模型建立和模型检验中,采用留一法（LOO）选取训练集和检验集样本。利用留一法交叉验证确定了PLS模型的最佳主因子数。结果表明,当采用均值中心化方法时,PLS模型的RMSECV和RMSEP的值均达到了最小,分别为0.076 6和0.169 0。研究结果表明,THz-TDS技术结合化学计量学方法能够快速、准确的对大黄的真伪进行鉴别,直接使用太赫兹时域光谱而不使用计算后的吸收谱有两个优点： （1）在分频测定和光谱信号处理时无需考虑样品的厚度;（2）使光谱信号处理过程得到简化。该技术也可以对其他中草药进行鉴别和质量控制。该法快速、简单、无污染、无需样品预处理,是一种有发展前景的中草药无损检测方法。     

Characterization of Greenhouse Soil Properties Using Mid-infrared Photoacoustic Spectroscopy

ABSTRACT

Evaluation of soil fertility is required for the sustainable production in greenhouses; however, routine agrochemical analytical methods are not practical since most of them are too time consuming and costly. In this study, Fourier transform midinfrared photoacoustic spectroscopy (FTIR-PAS) was applied as an alternative technique in the fast characterization of greenhouse soils. Nine source clays and 235 soil samples were obtained from different greenhouses in China, and their counterpart open-field soils were also sampled for comparison. The FTIR-PAS spectra of soil clays and soils were recorded with the wave-number range of 600–4000 cm^?1, and clear differences were observed among the spectra; soil organic matter (SOM) and soil clay played the main role in the spectral absorption. Partial least square regression (PLSR) models coupled to optimized spectral pretreatment were used to predict greenhouse soil properties, and FTIR-PAS spectra were closely related to SOM (R² = 0.94) and soil total nitrogen (R² = 0.90). For the other soil properties—such as pH, EC, water extractable P, K, Fe, and Cu—the calibration results were excellent or acceptable. These results suggested that FTIR-PAS could be used to determine most of greenhouse soil properties. Since the technique of FTIR-PAS is rapid and inexpensive, and it requires no sample pretreatment and little sample mass, it can be used in fast characterization of greenhouse soil and is a promising method for the management of greenhouse soil.     

Determination of calcium and magnesium in the Solanaceae plant by near infrared spectroscopy combined with interval combination optimization algorithm

Determination of the content of inorganic nutrients was done as part of a project to improve crop yield by use of fertilizers. Near-infrared spectroscopy coupled with interval combination optimization was used to determine the inorganic nutrients of calcium and magnesium in the Solanaceae plant. Interval combination optimization was employed to choose effective intervals with the merit of assigning smaller weights to less significant variables. Compared with full spectrum partial least squares regression, interval combination optimization improved the predictive accuracy and precision for both calcium and magnesium. The result was a high predictive ability for calcium, with a root mean squared error prediction of 0.159%, and for magnesium, with a root mean squared error prediction of 0.064%. In this study, the superiority of interval combination optimization over full spectrum partial least squares was shown in the aspects of effective variables and accurate results. Near-infrared spectroscopy combined with interval combination optimization was efficient for the determination of inorganic nutrients, calcium and magnesium, in the Solanaceae plant.     

应用近红外光谱进行玉米单籽粒品种真实性鉴定的光谱测量方法研究

从玉米籽粒胚的朝向,中长波漫反射分析和短波透射分析,进样方式三个方面进行了研究,通过比较不同的测量方法对品种识别模型性能的影响来选择最佳的玉米单籽粒光谱测量方式,进而鉴定玉米单籽粒品种真实性。原始光谱经过预处理后,使用判别式偏最小二乘法(PLS-DA)对数据进行特征提取和降维,选取前9个因子建立8个玉米品种的仿生模式识别模型。结果表明,玉米籽粒胚正对光源优于胚背对光源,中长波漫反射分析优于短波透射分析,小样品池效果好于小光阑。采用最优测量方法(籽粒胚正对光源,使用小样品池和中长波漫反射分析)获取玉米单籽粒的近红外光谱,建立的品种识别模型对本品种样品的平均正确识别率达到94.6%,对其他品种样品的平均正确拒识率达到96.5%。