【作者】 付庆波；

【导师】 赵纯；

【作者基本信息】 吉林大学 ， 物理电子学， 2012， 博士

【摘要】 利用近红外光谱可以快速地实现目标检测物的种类、组分和含量的无损分析，该技术已广泛应用于石化、农业、制药等各个领域。提取目标检测物的近红外光谱信息需要综合利用光学传感技术、计算机应用技术和一定的化学计量学方法。近年来短波近红外（780-1100nm）检测技术因为明显的成本优势越来越引起人们的重视。该波段光谱的吸收主要源自含氢基团振动的二级和三级倍频，它能够更深地进入样品内部，并产生较少的热效应，而且受水分吸收的干扰较小。该波段的光谱采集仪器技术具有成本低廉、结构紧凑、快速扫描等特点，是一种重要的过程分析技术。然而，在短波近红外仪器设计和应用方面仍然需要改进以适应新的应用需求。首先，现有的多通道短波近红外光谱采集装置的信噪比需要进一步提高，以满足低含量物质的检测。其次，多通道检测电路的设计缺乏灵活性和智能性。设计灵活性更强的检测电路使其不需硬件重新设计就能匹配不同的检测器，无疑能够降低成本。另外，检测电路参数的智能化调节对光谱采集的长期稳定性和一致性也很有帮助。在短波近红外技术的实际应用方面，受限于分子振动的倍频与合频吸收本身存在吸收较弱、吸收峰宽而重叠的特征，苛刻的检测条件，更容易导致短波近红外光谱复杂化，出现严重的重叠、偏移甚至旋转等非线性的变化，极大限制了该技术的应用。这些极端的检测条件包括：目标检测物含量低、存在外部变量的作用、检测物所在基质复杂等。本论文在短波近红外技术的光谱采集硬件技术、多元校正方法和实际应用方面进行了系统性的研究，做出了必要的改进和有益的尝试，以实现短波近红外技术在复杂光谱条件下的准确信息提取，扩展这一低成本、快速分析技术的应用范围。在短波近红外光谱采集装置硬件方面，使用CCD传感器设计实现了固定光路多通道光谱采集系统。光源采用卤素灯并联LED的复合形式，结合在ARM微控制器上实现的自适应光强调节策略，使仪器在整个谱段的信噪比达到了500:1。所设计的传感器驱动和信号采集电路能够针对两种不同型号的CCD器件。该电路不需改变硬件结构，采用软件编程就可以实现两种CCD的驱动，满足了不同场合光谱测量的特殊需要，提高了电路的通用性，降低了成本。针对低含量物质检测的需要，研究了多元校正模型的检测限计算方法和检测限与建模方法的关系。采用短波近红外结合PLS校正检测了医用葡萄糖注射液中葡萄糖的含量，建立了不同浓度范围下的PLS校正模型。研究发现，对含量较低的物质进行检测时，应该选择包含了该浓度的尽可能窄的校正浓度范围来建立PLS多元校正模型。采用傅里叶变换近红外技术实现的对饲料中磺胺抗生素添加剂的检测进一步证明了上述结论。且进一步发现，在建立浓度范围较宽的模型时，可以通过比较检测限来确定最佳的建模样品个数。研究为低浓度物质检测提供了建模方法上的参考。以短波近红外测量水溶液中酒精的含量为例，考察了不同的建模方法在消除外界变量-温度影响的作用。在不同的温度条件下研究对比了直接校正、全局建模、正交信号校正（OSC）和广义最小二乘加权（GLSW）这四种校正方法。结果表明温度对酒精水溶液的短波近红外光谱有强烈的影响，直接校正带来了很大的预测系统偏差，全局建模、OSC滤波和GLSW滤波，三种方法均能在一定程度上消除温度的影响。OSC和GLSW方法都可以作为一种有效的光谱前处理方法使用。相比而言，GLSW方法使用较少的隐含变量，在四个考察温度下均获得了最佳的预测结果，说明其温度校正能力优于OSC方法。采用磁性分子印迹分离结合近红外分析技术研究了制药厂污水中的磺胺抗生素的快速检测方法，探讨了使用这种方法定量检测复杂基质中的含量较低的有机物的可行性。制备了磁性分子印迹材料，且材料对磺胺类药物有很好的选择吸附性。利用这种材料在设计实现的自动制样装置上把制药废水中磺胺类药物残留快速分离、转移到了单一的基质-乙腈中，然后采用短波近红外光谱法和傅里叶变换长波近红外光谱法分别进行了分析。结论证明了这种新型自动检测方法的可行性。长波近红外光谱分析获得了更为满意的结果，说明分子印迹分离方法与近红外检测技术结合，可以弥补近红外技术检测限高、分辨性差的的缺点，为复杂基质中检测物的近红外分析提供了一条新的途径。本论文针低含量物质、外部变量作用和复杂基质下目标物质检测的关键问题，在仪器技术、建模方法方面的研究为复杂条件下应用短波近红外技术提供了参考，拓展了这一低成本快速检测技术的应用范围。更多还原

【Abstract】 Near infrared spectroscopy (NIR) is a fast and non-destructive analytical method forqualitative and quantitative analysis. It has been widely used in petroleum,agriculture and pharmaceutical industry et al. The NIR technology is quitecomplicated, which involves optical sensing technology, computer technology andchemometric methods. Recently the short-wave NIR (SW-NIR,780-1100nm)technology has drawn researchers’ attention, due to the significantly low cost. TheSW-NIR, which originates from the2or3overtone of molecular vibration, cantransfer more deeply into the sample with less heat effect. Moreover, theinterference from intense water absorbance can be diminished. Instrumentationworking on this region is a sort of important process analytical technology, which isof low cost, compact and fast scanning. However, improvement is still need in termsof instrumentation and application technology in order to meet the new demands.Firstly, to determine the low content analyte, the signal-to–noise ratio of SW-NIRinstruments need to be further improved. Secondly, the electrics in the instrumentsneed to be designed more flexible and smarter. A more flexible electronic circuitwhich can drive more than one kind of light sensor should be helpful to reduce thecost. And a smarter instrument which can regulate the parameters automatically isuseful for the long-term stability and consistent of the SW-NIR spectra. Since theSW-NIR spectra are originated from the over tone of molecular vibration, the spectraare quite overlapped and complicated. Thus, the extreme application conditionsoften lead to the shift, rotation of the spectra which lead to more complicatedanalysis of the spectra. The extreme conditions include low content analyte,influence of external variable and complicated analytical matrix et al. Therefore, theapplication field of SW-NIR is limited and need to be expended by simultaneouslyusing other technologies.In this dissertation research was done in terms of the SW-NIR instrumentaltechnology, multivariable calibration methods and new applications. The objective was to use this low-cost technology in complicated conditions and new applicationfields. And this was achieved through development in the instrumentation andcalibration method as well as the applying some new extraction technology. A newmultichannel SW-NIR spectrometer was designed by using the charge-coupled device(CCD) as the detector. A signal-to-noise ratio of500:1in the whole wavelength regionwas achieved by the compound light source design and using the self-adaptive lightintense regulation strategy. The light source was designed using a halogen lampenhanced by two single wavelength light emitting diodes centered at920nm and1020nm respectively. The electronic circuit designed can drive two different CCDs,which need only change the program in the microcontroller. This electronic circuitwas quite flexible and can be used for different demands with a very low cost. Themultivariate detection limit (MDL) was calculated and the relation between MDL andpartial least squares (PLS) calibration was researched to find a calibration strategy fordetermination of low analyte using the SW-NIR technology. This was done by analysisof the glucose injection using the designed SW-NIR spectrometer. PLS models werebuilt in different concentration ranges. It was found that a narrow concentrationrange which included the concentration of target analyte should be selected in orderto obtain a lower MDL. This conclusion was further proved by determining thesulfonamides additives in pig feed using FT-NIR spectroscopy. Furthermore, it wasfound that the optimal calibration sample number can be determined by comparingthe MDL. The influence of external variable can be diminished by appropriatecalibration methods. And this was approved by determining the alcohol content inaqueous solution at different temperature using the SW-NIR spectra. A significantsystematic error was found when using the direct transfer method which indicated atemperature correction was necessary. For this purpose, global modeling, orthogonalsignal correction (OSC) and generalized least squares weighting (GLSW) were utilizedand compared. It was found that all the three methods can reduce the temperatureinfluence. But the OSC and GLSW methods used less calibration samples that theglobal modeling and could be used as spectra pretreatment. The GLSW method provided the best predictions at all the four temperatures requiring less latentvariables. Thus, the GLSW method was superior to the OSC method. The possibilityof using the SW-NIR method to analyze the target analyte in complicated matrix wasapproved by determining sulfonamides in the pharmaceutical wastewater incombination with magnetic molecular imprinted polymers (MMIP). The MMIP wassynthesized and had good selective adsorption ability towards sulfonamides andstructurally related molecular. By using this MMIP with a designed auto-samplingdevice, sulfonamides in complicated matrix, wastewater, were separated andtransferred to a simple matrix-acetonitrile. Then the extractions were analyzed byusing the FT-NIR and SW-NIR spectroscopy. The SW-NIR did not provide goodpredictions due to the weak absorbance. However, good results were obtained byusing the FT-NIR spectroscopy, which approved the possibility of using this newapproach to fast determine the sulfonamides in pharmaceutical wastewater. Theweakness of NIR spectroscopy, such as relatively high detection limit and poorresolution, can be avoided by simultaneously using the MMIP technology. Theconclusions made in the research of the current dissertation are useful and helpfulfor using the SW-NIR technology in complicated conditions. And by using MMIPtechnology, the application field of SW-NIR was further expended.更多还原