- 操作摘要
使用Brimrose AOTF-NIR光譜儀收集不同水雜質(zhì)水平的甲醇樣品的光譜數(shù)據(jù)，結(jié)果是極好的，并且證明了Brimrose AOTF-NIR光譜儀可以用于定量測定甲醇中低含量水分雜質(zhì)。利用Brimrose AOFT–NIR光譜儀收集的數(shù)據(jù)，用部分最小二乘回歸法創(chuàng)建一個預(yù)測模型，該模型的SEP值為每1000PEM中18.8PPM，測量值和預(yù)測值之間的相關(guān)系數(shù)為0.9982。這些數(shù)據(jù)表明，Brimrose光譜儀可以用于定量測量甲醇中的水雜質(zhì)，誤差極低。同時也可以測量和建模甲醇的其他特性，但在這篇文章中只討論測量PPM中的水分含量。Brimrose光譜儀的測量速度和穩(wěn)定性使其成為過程控制的理想工具。AOTF技術(shù)允許快速掃描，沒有移動部件，可以實(shí)現(xiàn)實(shí)時在線過程控制?？梢暂敵雒枋鏊蛛s質(zhì)水平實(shí)時變化的趨勢圖，可以預(yù)先設(shè)定高低警告級別和高低報(bào)警級別，這些參數(shù)可以用于過程控制。實(shí)時分析可以比其他方法更快速的出結(jié)果，并可以用于在很短的時間內(nèi)優(yōu)化過程?？傮w而言，本研究結(jié)果表明，利用Brimrose AOTF-NIR光譜儀和校準(zhǔn)模型獲得的光譜數(shù)據(jù)，用來測量甲醇中的低含量水分雜質(zhì)是非?？尚械摹?/p>
- 簡介
聲光可調(diào)濾波器（AOTF）的原理基于光在各向異性介質(zhì)中的聲折射。裝置由粘在雙折射晶體上的壓電導(dǎo)層構(gòu)成。當(dāng)導(dǎo)層被應(yīng)用的射頻（RF）信號激發(fā)時，在晶體內(nèi)產(chǎn)生聲波。傳導(dǎo)中的聲波產(chǎn)生折射率的周期性調(diào)制。這提供了一個移動的相柵，在特定條件下折射入射光束的部分。對于一個固定的聲頻，光頻的一個窄帶滿足相匹配條件，被累加折射。RF頻率改變，光的帶通中心相應(yīng)改變以維持相匹配條件。
光譜的近紅外范圍從800nm到2500 nm延伸。在這個區(qū)域最突出的吸收譜帶歸因于中紅外區(qū)域的基頻振動的泛頻和合頻。是基態(tài)到第二激發(fā)態(tài)或第三激發(fā)態(tài)的能級躍遷。因?yàn)檩^高能級躍遷連續(xù)產(chǎn)生的概率較小，每個泛頻的強(qiáng)度連續(xù)減弱。由于躍遷的第二或第三激發(fā)態(tài)所需的能量近似于第一級躍遷所需能量的二倍或三倍，吸收譜帶產(chǎn)生在基頻波長的一半和三分之一處。觸簡單的泛頻以外，也產(chǎn)生合頻。這些通常包括延伸加上一個或多個振動方式的伸縮。大量不同合頻是可能的，因而近紅外區(qū)域復(fù)雜，有許多譜帶彼此部分疊加。
現(xiàn)在，NIRS被用作定量工具，它依賴化學(xué)計(jì)量學(xué)來發(fā)展校正組成的參照分析和近紅外光譜的分析的關(guān)聯(lián)。近紅外數(shù)據(jù)的數(shù)學(xué)處理包括多元線性回歸法（MLR）、主成分分析法（PCA）、主成分回歸法（PCR）、偏最小二乘法（PLS）和識別分析。所有這些算法可以單獨(dú)或聯(lián)合使用來得到有價值組成的定性描述和定量預(yù)測。
III.?? Methodology
200ml of HPLC grade methanol was measured and placed in a 250ml beaker.?? Approximately 2 ml of the sample was placed in a 5mm pathlength quartz cuvette for the initial measurement.?? 100 scans were collected per reading and averaged into one spectrum.? Wavelength range was from 1550nm to 2050nm with 2nm resolution.? Data were collected in ratio mode to account for any variations in the power source using a Brimrose AOTF-NIR Luminar spectrometer with fibers attached to a cuvette holder.? The cuvette containing the pure methanol was placed in the cuvette holder and spectral data were collected in transmission mode.? After data collection, the sample that was in the cuvette was placed back in the beaker.? Distilled water was added to the beaker to simulate water impurities in a methanol stream.? 50 ppm (approximately 0.01 ml) of water was added to the beaker and the mixture was scanned.? This process was repeated until the total amount of water in the mixture was 1000 ppm.? A total of 21 samples were scanned.? The transmission data were processed into absorbance and the absorbance data were used to create a calibration model.? In a process setting, fewer scans per spectrum would allow a measurement to be taken every one to two seconds.
IV.Results
- Spectra
Figure 2.? ?Transmittance spectra of methanol with water added at 50 ppm increments
Figure 3.? ?Absorbance spectra of methanol with water added at 50 ppm increments.
Figure 4.?? Enhanced absorbance spectra of methanol and water from 1888nm to 1965nm
Figure 5.? Enhanced absorbance spectra of methanol and water from 1900nm to 1950nm.
The absorbance spectra of the methanol with water added clearly shows changes in the water absorbing region from 1900nm to 1950nm.? This is the area where one would expect to see differences in spectral data due to changes in water content.? The PLS 1 regression model will confirm that the Brimrose spectrometer is able to readily quantify water content in methanol using spectral data and a regression model.
- Regressions and Modeling
Figure 6.? PLS 1 regression model for water in methanol
The results for this model are excellent and show good correlation between the amount of water in methanol and the spectral data.? The correlation coefficient is equal to 0.9986, the SEC is equal to 16.1, and the SEP is equal to 18.7.?? These numbers confirm that it is feasible to measure trace amounts of water in methanol.? The results will be even better when a larger sample set of 100 or more data points is used for the calibration model.
Figure 7.? Regression coefficients for PLS 1 model correlating spectral data to water in methanol.
The regression coefficients indicate the wavelength regions from which the model is taking the relevant information.? The model clearly takes most of the information from the 1950nm region, which is the region where water absorbs.? The regression coefficients confirm that the model is taking the information from the wavelength range where spectral changes occur due to changes in water.
- Conclusions and Recommendations
The results of this study show that it is feasible to quantitatively measure trace amounts of water in methanol using spectral data obtained from the Brimrose AOTF-NIR spectrometer and a calibration model.? The results were especially good considering the small size of the sample set used for the calibration model.? Past experience has shown that using a sample set of 100 or more data points will make a model more accurate and robust.? Past experience has also shown that a calibration model generated from laboratory data can easily be transferred to a real-time, on-line setting.? The numbers for the model were excellent and the regression coefficients confirm that the model takes its information from the water absorbing region around 1950nm.? The signal to noise ratio was very high and a calibration model created from a larger sample set should be able to quantitatively analyze trace amounts of water in methanol to about 10ppm or less.? The Brimrose spectrometer is the ideal tool for real-time on-line process control because of its speed and lack of moving parts.? A multiplexed spectrometer can use up to 16 channels to measure and analyze data at 16 different process points.? Calibration models can be used to measure different parameters in different chemicals and it will be possible to take a reading for each channel within one to two minutes.? It is recommended that a purchase order be placed for a Brimrose AOTF-NIR Luminar spectrometer to allow for further calibration and testing.
