基于近紅外光譜的偏最小二乘回歸建模分析

1.8萬(wàn)字
本人今年最新原創(chuàng)的畢業(yè)設(shè)計(jì)，僅在本站獨(dú)家提交，大家放心使用

摘要 近紅外光譜區(qū)是 Herschel 1800 年發(fā)現(xiàn)的，其波長(zhǎng)范圍為 780～2526nm，波數(shù)范圍為 12820～3959 cm-1。近紅外光譜分析技術(shù)屬于弱光譜信號(hào)分析技術(shù)。近紅外光譜(Near Infrared ReflectanceSpectroscopy 簡(jiǎn)稱 NIRS) 主要反映了分子中有機(jī)官能團(tuán)( C-H，N-H，O-H等) 的倍頻與合頻的振動(dòng)吸收，其振動(dòng)吸收強(qiáng)度與官能團(tuán)的含量之間存在著密切的關(guān)系。近紅外光譜綜合運(yùn)用了光譜技術(shù)、計(jì)算機(jī)技術(shù)和化學(xué)計(jì)量學(xué)等多個(gè)學(xué)科的最新研究成果，并且以其獨(dú)特的優(yōu)點(diǎn)在各個(gè)領(lǐng)域得到了日益廣泛的應(yīng)用?，F(xiàn)代近紅外光譜是 90 年代以來(lái)發(fā)展最快、最引人注目的光譜分析技術(shù)，是光譜測(cè)量技術(shù)與化學(xué)計(jì)量學(xué)學(xué)科的有機(jī)結(jié)合。
由于物質(zhì)在近紅外譜區(qū)的倍頻和合頻吸收信號(hào)弱，譜帶重疊，解析復(fù)雜，所以作為化學(xué)計(jì)量學(xué)中最有力工具之一的偏最小二乘法(Partial Least Square ，簡(jiǎn)稱PLS)是研究其最好的方法。偏最小二乘法有較強(qiáng)的提信息的能力，并且能夠有效地降維,并消除自變量間可能存在的復(fù)共線關(guān)系，明顯改善數(shù)據(jù)結(jié)果的可靠性和準(zhǔn)確度。因此，本文主要運(yùn)用了間隔區(qū)間偏最小二乘法（ interval Partial Least Square ，簡(jiǎn)稱iPLS）， 反向區(qū)間偏最小二乘法（Back interval Partial Least Square ，簡(jiǎn)稱BiPLS），移動(dòng)窗口偏最小二乘法(moving window Partial Least square ，簡(jiǎn)稱mwPLS)三種方法，以啤酒近紅外光譜（NIRbeer）為例，檢測(cè)啤酒中的原麥汁濃度，并作定量分析，來(lái)建立其偏最小二乘回歸的數(shù)據(jù)模型。
關(guān)鍵詞： 近紅外光譜 PLS iPLS BiPLS mwPLS 啤酒

Based on partial least-squares regression modeling of near infrared spectral analysis
Abstract Near infrared spectral region is discovered by Herschel in 1800, the wavelength range of 780 ~ 2526nm, wavenumber range of 12820 ~ 3959 cm – 1. Near infrared spectroscopy analysis technique belongs to the weak spectrum signal analysis technology. Near infrared spectroscopy (Near Infrared ReflectanceSpectroscopy referred to as NIRS) which mainly reflects the functional groups in organic molecules (C-H,N-H,O-H) frequency and vibration frequency of the vibration absorption, absorption and there is a close relationship between the strength and the functional group content. Near infrared spectroscopy combined with the latest research achievements of many disciplines spectrum technology, computer technology and chemometrics etc., and with its unique advantages in various fields has been widely applied. Modern near infrared spectroscopy analysis technique is the fastest, most development spectrum attract sb.'s attention since the 90's, is the organic combination of spectroscopy and chemometrics discipline.
The material in the near infrared spectral region frequency doubling and frequency absorption signal is weak, band overlap, complex, so as the partial least squares method is one of the most powerful tools in chemometrics (Partial Least Square, referred to as PLS) is the best studied its. Partial least squares method has strong ability of information, and can effectively reduce dimensionality, and eliminate the multicollinearity of possible variables significantly improved, reliability and accuracy of results. Therefore, this paper mainly uses the interval partial least squares (interval Partial Least Square, referred to as iPLS), backward interval partial least squares (Back interval Partial Least Square, referred to as BiPLS), moving window partial least squares (moving window Partial Least square, referred to as mwPLS) three methods, to make beer near infrared spectroscopy (NIRbeer) as an example, content detection in beer, and quantitative analysis, data model to build partial least squares regression.
Keywords: Near infrared spectroscopy PLS iPLS BiPLS mwPLS Beer

目 錄
第1章 緒論 1
第2章 近紅外光譜技術(shù) 4
2.1 近紅外光譜的發(fā)展歷程 4
2.2 近紅外光譜分析的基本原理 4
2.3 近紅外光譜常規(guī)的分析方法 5
2.4 近紅外光譜的分析 5
2.4.1 近紅外光譜的定性分析 6
2.4.2近紅外光譜的定量分析 6
2.5 近紅外光譜分析儀器 6
2.6 近紅外光譜分析技術(shù)的綜合評(píng)價(jià) 8
2.6.1 近紅外光譜分析技術(shù)的優(yōu)越性 8
2.6.2 近紅外光譜分析技術(shù)的缺陷 9
2.6.3 近紅外光譜分析技術(shù)存在的問題 9
第3章 算法理論基礎(chǔ) 10
3.1 偏最小二乘法（PLS） 10
3.1.1 偏最小二乘法的基本原理 10
3.1.2 模型主因子數(shù)的確定方法 12
3.1.3 偏最小二乘法（PLS）建模的優(yōu)化參數(shù).................................................................12
3.2 間隔區(qū)間偏最小二乘法（iPLS） 14
3.3 移動(dòng)窗口偏最小二乘法（mwPLS） 15
3.4反向區(qū)間偏最小二乘法（BiPLS） 15
3.5 偏最小二乘法（PLS）的特點(diǎn) 16
第4章 實(shí)驗(yàn)過(guò)程及結(jié)果 17
4.1 實(shí)驗(yàn)的運(yùn)行環(huán)境 17
4.2 近紅外光譜的數(shù)據(jù)處理 18
4.2.1 光譜數(shù)據(jù)的預(yù)處理 18
4.2.2 光譜數(shù)據(jù)的分析方法 18
4.3 實(shí)驗(yàn)方法 19
4.4 近紅外光譜的定量分析步驟及模型評(píng)價(jià) 19
4.4.1 近紅外光譜定量分析步驟 19
4.4.2 模型的評(píng)價(jià)參數(shù) 20
4.5 實(shí)驗(yàn)結(jié)果及分析 21
4.5.1 間隔區(qū)間偏最小二乘法（iPLS） 22
4.5.2 反向區(qū)間偏最小二乘法（BiPLS） 27
4.5.3移動(dòng)窗口偏最小二乘法（mwPLS） 30
總 結(jié) 31
致 謝 32
參考文獻(xiàn) 33