【正文】 n these applications, the desired position of the samples is known and the analyst is interested in verifying the samples or identifying when impurities are present. Similarity Match methods are also good for prequalifying materials that will be used for quantitative measurements.Note: If you are interested in paring unknown samples to two or more materials and want to use multiple standards to describe each material, use Discriminant Analysis, Distance Match, or QC Compare search instead of Similarity Match. If you want to use many different materials in the parison but have only one standard to describe each one, use Search Standards.What does Similarity Match do?A Similarity Match method pares the spectral information in the specified region or regions of an unknown sample spectrum with that of a known set of standard spectra to determine how closely the sample matches the standards. The result of this parison is called a match value. The match value represents the unexplained variation in the spectrum of the unknown sample.Note: The Match Type parameter on the Other tab determines the format for reporting the match value.2． 距離匹配（Distance Match）What is Distance Match?The Distance Match classification technique can be used to determine how closely an unknown material matches two or more classes of known materials by calculating a conformity spectrum for each class and measuring its distance from the class average. This technique is typically used to screen ining materials, for example, to determine how closely they match pound a, b, or c or to determine degrees of difference between known and unknown materials.We remend using at least five standards to define each class (two standards are required in order to perform the variance analysis during calibration). Multiple regions of the spectrum may be used for the parison. Why use Distance Match?The Distance Match algorithm works well for differentiating materials that contain different amounts of the same ponents. In this type of analysis, the spectra of the standards in each class are very similar. The main difference may occur only in the intensities of the peaks in a few key regions. For example, a lab might set up a Distance Match method to check the dosage of drugs used in a doubleblinded clinical study to confirm that no mixups occurred in the packaging. In this case, the analyst might set up classes that define different levels of an active ingredient, such as salicylic acid. The method could then be used to test ining drugs to see which category they match and to what degree.You can also use Distance Match to detect slight changes in ining raw materials or production samples by paring them to previous samples of the same type. We refer to this procedure as qualification. It can be an effective technique for statistical quality control.What does Distance Match do?During calibration, the software calculates an average spectrum and a standard deviation spectrum for each class. When you use the method to classify the spectrum of an unknown sample, the software subtracts the average spectrum from the unknown sample spectrum to create a residual spectrum and divides by the corresponding standard deviation spectrum to create a conformity spectrum for each class. (The subtraction step shows the differences between the two spectra。注意，這里的模型修正不是模型維護(hù)。s 由于光譜采集方式不恰當(dāng)、光譜噪音太大和人為誤差造成光譜質(zhì)量下降或引入誤差。s 如果Standards窗口中包含有驗(yàn)證（Validation）樣品，則校正結(jié)果窗口中會(huì)計(jì)算出Performance Index和RMSEP，可以以這兩個(gè)指標(biāo)對模型進(jìn)行評價(jià)。定量分析模型計(jì)算完成后，TQ顯示如下界面，同時(shí)狀態(tài)顯示欄由紅變綠，如果Standards窗口中包含驗(yàn)證（Validation）樣品，則會(huì)計(jì)算出性能指數(shù)（Performance Index）。（2） Report窗口 Report窗口中主要設(shè)置的是報(bào)告中要顯示的內(nèi)容，注意，這里的報(bào)告是指TQ產(chǎn)生的分析報(bào)告，與后面章節(jié)中提到的RESULT生成的在線報(bào)告不同。使用此按鈕，顯示各組分的純組分光譜，其功能與Diagnostics菜單中的Pure Component Spectra項(xiàng)類似。該窗口中同樣包含Suggest按鈕，可以隨時(shí)使用該向?qū)?，也包含How To按鈕以隨時(shí)獲得幫助信息。但TQ中包含了許多這方面專家知識(shí)，為我們選擇有效光譜范圍提供了有用的依據(jù)。導(dǎo)數(shù)處理是凈化譜圖較常用的方法，可根據(jù)需要進(jìn)行一階或二階微分處理。如果某個(gè)樣品有錯(cuò)誤，或者不愿意使用它，還可以將其設(shè)為Ignore（忽略），當(dāng)然也可以從Edit菜單中直接將其刪除（Delete Row）。然后按其推薦的分布進(jìn)行配置標(biāo)準(zhǔn)樣品，并采集其NIR光譜，當(dāng)然實(shí)際配置的樣品中各組分的含量不一定和Suggest Standards向?qū)扑]的含量分布完全一致，此時(shí)只需按實(shí)際含量輸入到對應(yīng)的樣品含量表格中即可。（1）標(biāo)準(zhǔn)樣品的準(zhǔn)備所謂標(biāo)準(zhǔn)樣品（Standards），是指為建立NIR定量分析模型收集的具有足夠代表性、并且已知化學(xué)成分含量的樣品。s 另外，定量分析模型的性能指數(shù)（Performance Index）也與此處設(shè)定的濃度范圍有關(guān)。6. 選擇光程類型（Pathlength Type）在Pathlength窗口中可以對光程類型進(jìn)行設(shè)置。 從TQ Analyst軟件的File菜單中選擇Save Method As167。 在可能遇到的所有待分析樣品中，每一種組分的含量范圍？167。 設(shè)置其它參數(shù)167。 給模型定義一個(gè)名稱167。s 光譜信息查看按鈕：按該按鈕，可以獲得當(dāng)前光譜的所有信息，如：s 下拉光譜名稱按鈕：點(diǎn)擊此按鈕，顯示當(dāng)前窗口中所有光譜名稱，并可進(jìn)行選擇。TQ Analyst可以打開和保存下別所列的各種格式的光譜，其中 .CSV為Excel格式。TQ Analyst窗口介紹工具欄菜單欄性能指數(shù)顯示狀態(tài)指示滾動(dòng)條窗口標(biāo)簽頭活性按鈕參數(shù)1． 菜單欄 TQ Analyst菜單欄的使用方法與其它Windows程序一樣。從參數(shù)的設(shè)置到實(shí)驗(yàn)設(shè)計(jì)的每一個(gè)環(huán)節(jié)，使用TQ的向?qū)Чδ?，可以幫助您一步步地走向成功。TQ Analyst的幫助菜單中還包含下列幫助工具：TQ Analyst online tourTQ Analyst主要功能快速演示；TQ Analyst example methods定量、定性方法實(shí)例演示。第二章 TQ Analyst光譜分析軟件TQ Analyst是一個(gè)通用的光譜分析軟件，它可以為中紅外、近紅外、遠(yuǎn)紅外和拉曼光譜分析的應(yīng)用提供各種定性和定量分析工具。通常而言。NIR光譜定量分析中常用的多元校正方法比較方法優(yōu)點(diǎn)缺點(diǎn)適用對象多元線性回歸（MLR）計(jì)算簡單、物理意義明確、易于理解對參加關(guān)聯(lián)的變量（如波長通道）數(shù)目有限制；會(huì)出現(xiàn)共線性問題波長點(diǎn)較少的簡單線性對象，化學(xué)組成較簡單的樣品體系逐步多元線性回歸（SMLR）與MLR相比，回歸之前可以對自變量進(jìn)行篩選對NIR光譜分析來說，篩選自變量的工作是巨大的；仍無法解決共線性問題簡單線性對象 主成分回歸（PCR）致力于提取數(shù)據(jù)群體中的特征信息，辯識(shí)影響系統(tǒng)的主要因素，對眾多變量作綜合簡化，從而在力保有用信息損失最小的前提下，實(shí)現(xiàn)高維數(shù)據(jù)集合的降維；可解決線性回歸分析中經(jīng)常會(huì)遇到的共線性問題和變量數(shù)限制問題；可對由于光散射和其它組分帶來的干擾做出補(bǔ)償計(jì)算速度比MLR慢；對模型的理解沒有MLR直觀；不能保證參與回歸的主成分一定與樣品性質(zhì)相關(guān)可用于組成較復(fù)雜的樣品體系偏最小二乘回歸（PLS）可以使用全譜或部分譜數(shù)據(jù)；數(shù)據(jù)矩陣分解和回歸交互結(jié)合，得到的特征向量直接與樣品性質(zhì)相關(guān)；模型更為穩(wěn)??；可對由于光散射和其它組分帶來的干擾做出補(bǔ)償；可以適用于復(fù)雜的分析體系模型質(zhì)量容易受到奇異點(diǎn)的影響；模型建立過程較復(fù)雜，較抽象，較難理解可用于組成復(fù)雜的樣品體系目前PLS應(yīng)用范圍最廣定量分析模型建立好后，需要對其性能進(jìn)行評估。實(shí)際樣品分布理想的校正集樣品分布歧異樣品一般而言，對于一個(gè)單組分系統(tǒng)，校正集至少由20個(gè)樣品組成。所謂正面的樣品是指與標(biāo)準(zhǔn)樣品一致或接近的樣品，它用來驗(yàn)證NIR定性分析方法能否對合格的樣品做出正確的分析結(jié)果，以確保NIR定性模型不會(huì)將合格的樣品判為不合格。通常，一個(gè)可以被認(rèn)可的光譜庫需要包含多個(gè)批次的多個(gè)樣品的光譜，每一類至少來源于三個(gè)批次。為了改善光譜特征和補(bǔ)償基線偏移，在NIR光譜學(xué)中，經(jīng)常要用到數(shù)學(xué)方法對光譜進(jìn)行預(yù)處理。漫反射光譜與常規(guī)反射光譜存在根本區(qū)別在于后者的光線不與樣品內(nèi)部發(fā)生作用，故不帶有分析樣品的組成信息。前后各有一個(gè)位置可分別用于放置樣品或參比，其上還設(shè)有可拆卸的光闌（Aperture）。NIR光譜的基本采集方式可以分為透射和反射模式。在高波長（低波數(shù)）段的NIR吸收峰相對較強(qiáng)和尖銳，分辨能力也較好，而在高波數(shù)譜段，吸收峰更低，峰形更寬，如圖2。第一節(jié) NIR光譜區(qū)域在電磁波譜圖上，NIR光譜范圍介于可見光與中紅外光譜之間（7802500nm或128204000 cm1）。而最近幾年，它已經(jīng)逐漸被各行業(yè)用于產(chǎn)品生產(chǎn)過程的每一個(gè)環(huán)節(jié)，對各種形式（固體、液體、浸膏、懸濁液、紙張等）的產(chǎn)品