CN101368905A - Infrared spectrum non-linear modeling quantitative anslysis method - Google Patents

Abstract

The invention discloses an infrared spectrum non-linear modeling quantitative analysis method based on the molecular vibration theory; the method makes use of Matlab software programming and Statistica software data processing to establish a non-linear model and obtain the final quantitative result through the model. Through the establishment of the non-linear model, the invention realizes quantitative analysis and reduces measurement error to improve the measurement accuracy, precision and feasibility.

Description

Infrared spectrum non-linear modeling quantitative anslysis method

Technical field

The present invention relates to quantitative analysis method, be specifically related to infrared spectrum non-linear modeling quantitative anslysis method.

Background technology

Infra-red sepectrometry is generally used for qualitative analysis.Infrared spectrum comes from the vibration of molecule, and molecular vibration causes the variation of dipole moment, and dipole moment changes more greatly, and ratio of infrared absorption intensity is big more.Be attended by bigger dipole moment during the polar group vibration and change, produce stronger infrared absorption.The position difference of different group Absorption Spectroscopies has obvious characteristics.As 1540～2500cm ^-1Frequency zones is generally the vibration frequency district of unsaturated group.Typical carbon-carbon double bond C=C absorption peak is at 1600～1700cm ^-1With 3000～3100cm ^-1Near, typical hydroxyl-OH absorption peak is at 1000～1200cm ^-1With 3000～3700cm ^-1Near, and the hydroxyl absorption band in the carboxyl is positioned at 2500～3330cm ^-1Near.Typical carboxylic acid carbonyl C=O absorption peak is at 1760cm ^-1And 3570cm ^-1Near, and for aromatic carboxylic acid, because the carbonyl and the kinds of aromatic ring conjugation of carboxylic acid, make the position of carbonyl absorption peak at 1660～1690cm ^-1Near.The absorption position of typical case's ester class is at 1000～1300cm ^-1With 1700～1800cm ^-1Near.Phenyl ring is at 1450～1650cm ^-1There is absorption in the zone, and the most outstanding bands of a spectrum of monokaryon aromatic hydrocarbons are positioned at 675～900cm ^-1, or the like.

Infrared spectrum energy provides the abundant information of relevant organic compound molecule structure, becomes the dactylogram of organic compound; The sample that is suitable for is in extensive range, can analyzing organic substance, also can analyze inorganics; What restriction is the sample form do not had, and gas, liquid, suspending liquid, solid sample can both carry out Infrared spectroscopy; Amount of samples is few, can be as small as μ g magnitude; Analysis speed is fast; Do not destroy sample; The charge is small for conventional infrared spectrum instrument, and operating cost is low, is convenient to promote.

Traditional quantitative analysis method is the linear analysis method, be that the typical curve model is a linear equation, be commonly referred to calibration curve method, just choose the specific absorption cutting edge of a knife or a sword, preparation standard items concentration gradient, measure the absorbance that each concentration is located at specific wavelength (wave number), make linear standard curve (working curve), so-called typical curve is actually the linear equation of setting up according to the relation of concentration and absorbance.Being independent variable with measured object concentration normally, is dependent variable with the absorbance, obtains dependent linear equation by methods such as linear regressions.

Because infrared spectrum belongs to the molecular vibration spectrum, molecular vibration crest place is often because due to the molecular resonance, the strong more then peak value that resonates is big more.When forming resonance, the variation tendency that specific wavelength (wave number) is located obviously shows as non-linear, if it is bigger to set up the obvious error of typical curve model with traditional linear equation.In the last few years, IR spectrum quantitative analysis particularly Non-Destructive Testing receive publicity gradually.

Summary of the invention

The objective of the invention is to: a kind of infrared spectrum non-linear modeling quantitative anslysis method is provided, reduces measuring error, improve accuracy of measurement by this analytical approach.

Technical solution of the present invention is: this analytical approach is utilized the Matlab software programming according to the molecular vibration principle, and the Statistica software data is handled, and sets up nonlinear model, obtains final quantitative result by model.

The detailed process of analytical approach of the present invention is as follows:

(1) the Infrared spectrum scanning parameter is set, also promptly set the infrared spectrum instrument parameter, what conventional quantivative approach derived is the collection of illustrative plates of pdf form, contain the information such as peak intensity at Partial Feature peak, different with conventional method is, the needed output data of Nonlinear Modeling of the present invention is a text formatting, comprises the data of all test points, but not the pdf form;

(2) tested standard items concentration gradient is accurately prepared in standard items scanning, respectively each concentration sample is carried out infrared scan, obtain the absorption value in whole wavelength (wave number) scope, each sample scans more than 8 times at least, averages, and concentration gradient increases density as far as possible;

(3) characteristic peak is selected, and different infrared spectrometer scan operation differences according to all types of instrument requirements, is that target is implemented sample scanning with the Non-Destructive Testing; According to the structure of measured object, and there is environment in measured object, chooses the pairing characteristic peak of the group that can represent the measured object feature structure, is independent variable with concentration, and absorbance is a dependent variable, draws scatter diagram, to watch figure trend;

(4) set up nonlinear model, according to the feature and the Changing Pattern of scatter diagram, tentatively selected nonlinear model function is formed, and utilizes the matlab software programming, obtains primary mold; Utilize the statistica software data to handle, determine each parameter value, and the degree of fitting of model;

(5) modelling verification, at first compare by true value, or record the tested components contents of sample by national standard method or official method or the higher methods such as method of accuracy, with this numerical value as with reference to (also can be described as relative true value), calculate characteristic absorption peak place absorption of sample by nonlinear model and be worth pairing sample size, with the measured value comparison of preceding method institute, try to achieve relative error or relative deviation or standard relative deviation or average recovery.

Analytical approach of the present invention realizes quantitative test by setting up nonlinear model, has reduced measuring error, has improved measurement accuracy, accuracy, feasibility.

Description of drawings

Fig. 1 is that the infared spectrum of moisture and anhydrous Aspirin compares.

Fig. 2 is the Aspirin infared spectrum of different moisture content.

Fig. 3 different moisture content Aspirin is at 2950cm ^-1Place's absorbance.

Fig. 4 is the non-linear quantitative model matched curve of Aspirin sheet water cut.

Fig. 5 is that the different content standard glucose is at 2900cm ^-1The absorbance at place.

Fig. 6 is the non-linear quantitative model matched curve of glucose.

Embodiment

Below in conjunction with specific embodiment, the present invention is described in further detail.Should be understood that these embodiment just in order to demonstrate the invention, but not limit the scope of the invention by any way.

Example 1: infrared spectrum Nonlinear Modeling method is measured aspirin medicine water cut

1. sample and reagent: aspirin (Aspirin) enteric coatel tablets (commercially available), KBr analyzes pure;

2. instrument: Nicolet 5700 Fourier transformation infrared spectrometers (Thermo ELECTRONCORPORATION), YP-2 type sheeter (Shanghai lofty mountains scientific instrument company limited), HW-10 infrared drying oven (Ying Yu of Gongyi City gives magnificent instrument plant), the quartzy automatic dual pure water distiller (Danyang, Jintan City door quartz glass factory) of 1810-B type, KQ-250B type ultrasonic cleaner (Kunshan City, Jiangsu Province Dianshan Lake town), AY120 electronic balance (SHIMADZUCORPORATION JAPAN).

3. the parameter of infrared spectrometer: scanning times 32 times; Resolution 4cm ^-1Data break 1.926cm ^-1Light source IR; Scanning optical spectrum scope 4000～400cm ^-1Detecting device DTGSKBr;

4. aspirin (Aspirin) sheet determination of moisture method: Infrared spectroscopy KBr pressed disc method, quantitative result is done reference with pharmacopeia regulation GB oven drying method;

5. the infared spectrum characteristic peak is selected:

The Aspirin sheet belongs to salicylic acid (chemistry 2 hydroxybenzoic acid by name and 2 hydroxybenzoic acid sodium) class medicine, its chemical name is 2-(acetoxyl group) benzoic acid, this is the analgesic analgesics anti-inflammatory of using always, this type of preparation variety is more, volume of production and marketing is bigger, and Aspirin is an acetylsalicylic acid esterification thing, contains characteristic groups such as phenyl ring, carboxyl, ester bond in the molecular structure, corresponding with it, its infared spectrum 1450～1650cm ^-1The strong absorption peak in zone, the peak type is sharp-pointed and absorption peak is more, illustrates wherein to contain phenyl ring; 700～750cm ^-1Near absorption peak shows the adjacent replacement of phenyl ring; 1750cm ^-1Near absorption peak shows that ester bond exists; 1690cm ^-1Near absorption peak shows that containing the carboxyl carbonyl exists;

In the Aspirin sheet, except effective constituents A spirin and auxiliary material, also contain target measured object water, because the existence of moisture, make infrared spectrogram 2500～3700cm ^-1The zone is especially at 3500cm ^-1Near, anhydrous and moisture Aspirin infared spectrum peak height and peak type have notable difference, and go out peak intensity along with the variation of moisture presents regular the variation, see Fig. 1,2.

By Fig. 1,2 as seen, the change of moisture causes the baseline wander of each collection of illustrative plates, and visualize is the change that peak intensity, i.e. the change of absorbance, the peak type of the measured object Aspirin infared spectrum of different moisture content and go out the peak position basically identical; 400～1800cm ^-1The zone go out the peak intensive and collection of illustrative plates moves and peak intensity change all very little, and at 2800～3500cm ^-1The zone has significant change, and this is because H ₂Form due to the intermolecular ydrogen bonding between O molecule and OH; Therefore, select 2950cm ^-1The Aspirin collection of illustrative plates of different moisture content and the relation of absorbance are investigated in the zone, thereby obtain the method for water and basis weight;

6. quantitative test modeling:

Preparation different moisture content test sample, sheet weight and moisture Aspirin sample according to KBr are different with the ratio of KBr, in sample moisture content massfraction 0～20% scope, Infrared spectrum scanning result shows that the spectrogram of different sample moisture contents has very big similarity; With the sample moisture content is horizontal ordinate, 2950cm ^-1Place's absorbance is the ordinate mapping, along with change of moisture content, and 2950cm ^-1Near peak intensity has than obvious variation thereupon, sees Fig. 3; The spectroscopic data of measured object is scanning mean value 8 times among the figure;

From Fig. 3 as seen, along with the variation of Aspirin sample moisture content, 2950cm ^-1The absorbance significant change at place, present the peculiar vibration wave feature of molecular vibration, be characterized in that amplitude decays by index law, vibration frequency increases gradually, its variation tendency obviously shows as non-linear, if it is bigger to set up the obvious error of typical curve model with traditional linear equation;

According to dynamics and mathematical modeling theory, adopt the nonlinear fitting method, at first choose fitting function, by MATLAB programming and STATISTICA data processing,, obtain nonlinear equation through repeatedly optimizing:

$y={0.7460e}^{{-0.3311x}^{0.95}}\sin \left({0.970}^{0.8035x}(1.270x+5.842)\right)-0.03553r=0.9759---\left(1\right)$

The degree of fitting of equation is 97.59%, sees Fig. 4, formula (1);

7. modelling verification:

Recording Aspirin sheet water cut by pharmacopeia regulation national standard method oven drying method is 2.341%, is reference point with this value, calculates the nonlinear model relative error, and in traditional linear model relatively; With 2950cm in the sample ^-1The absorbance 0.2393 of place's characteristic absorption peak is respectively in substitution nonlinear model and the traditional linear model equations, linear equation water cut-49.98% as a result, this numerical value is less than zero, obviously distortion; Nonlinear equation Aspirin sheet water cut as a result is 2.494%, then the relative error of linear model result and GB oven drying method result of calculation is-2231%, nonlinear model result 6.54%, the error that this shows linear model is bigger, and the accuracy of nonlinear model is more much bigger than linear model.

Example 2: sugared content in the infrared spectroscopic determination tomato

1. sample and reagent: tomato, commercially available; Glucose is analyzed pure; KBr analyzes pure;

2. instrument: Nicolet 5700 Fourier transformation infrared spectrometers (Thermo ELECTRONCORPORATION), YP-2 type sheeter (Shanghai lofty mountains scientific instrument company limited), HW-10 infrared drying oven (Ying Yu of Gongyi City gives magnificent instrument plant), the quartzy automatic dual pure water distiller (Danyang, Jintan City door quartz glass factory) of 1810-B type, KQ-250B type ultrasonic cleaner (Kunshan City, Jiangsu Province Dianshan Lake town), AY120 electronic balance (SHIMADZUCORPORATION JAPAN);

3. the parameter of infrared spectrometer: scanning times 32; Resolution 2; Data point is 0.964cm at interval ^-1Light source IR; Scanning optical spectrum scope 4000～400cm ^-1Detecting device DTGS KBr; 4. sugared content assaying method: Infrared spectroscopy KBr pressed disc method, quantitative result with traditional Fehling method measurement result as reference;

5. the infared spectrum characteristic peak is selected: except glucose, also contain components such as starch, sucrose in the tomato, its molecular structure is comparatively approaching, and each component infared spectrum goes out the peak feature and sees Table 1:

Table 1 infared spectrum goes out the peak feature

By table 1 as seen, measured object has and goes out the peak position jointly, is respectively: 3400cm ^-1, 2900cm ^-1And 1000cm ^-1～1200cm ^-1According to document announcement, 2900cm ^-1Neighbouring is that (absorption peak CHO), aldehyde radical are the characteristic group of glucose to aldehyde radical; 3400cm ^-1And 1000cm ^-1～1200cm ^-1Near for hydroxyl (OH) go out the peak position, hydroxyl is comparatively common functional group, and hydroxyl goes out the position difference at peak in the different material; Therefore, select 2900cm ^-1Near aldehyde radical absorption peak is used for quantitative test for the tomato characteristic peak;

6. quantitative test modeling:

Sheet weight and standard glucose sample according to KBr are different with the ratio of KBr, is 1/100～1/320 scope by standard glucose with the KBr ratio, preparation series standard glucuronide fraction, Infrared spectrum scanning result shows, the spectrogram of different sample sizes has very big similarity, along with the reduction of standard glucose concentration, 2900cm ^-1Near peak intensity weakens gradually, has than obvious variation, and the absorbance at its corresponding wave number place also reduces thereupon, and spectrum moves down, and sees Fig. 5, and the spectroscopic data of measured object is scanning mean value 8 times among the figure;

From Fig. 5 as seen, along with the variation of sample size, 2900cm ^-1The absorbance significant change at place presents the peculiar vibration wave feature of molecular vibration, and its variation tendency obviously shows as non-linear, if it is bigger to set up the obvious error of typical curve model with traditional linear equation;

According to dynamics and mathematical modeling theory, adopt the nonlinear fitting method, at first choose fitting function, by MATLAB programming and STATISTICA data processing,, obtain nonlinear equation through repeatedly optimizing:

y＝11299.7+3771.87x-10123x ²+2094.4x ³-13066cosx+31.5767cosx ²-88.364log-31.31logx ² r＝0.9383 (2)

The degree of fitting of equation is 93.83%, sees Fig. 6, formula (2); If with traditional linear model, the equation model degree is 83.15%, nonlinear model obviously is better than linear model;

7. modelling verification:

Recording tomato sugar content 61.18% (butt) by traditional Fehling method, is reference point with this value, calculates the relative error of nonlinear model and linear model respectively; With sample at 2900cm ^-1In near two kinds of model equations of absorbance mean value difference substitution of characteristic absorption peak, get linear equation tomato sugar content 119.69% (butt) as a result; Nonlinear equation is tomato sugar content 59.76% (butt) as a result; Then the relative error of linear model result and Fehling method result of calculation is 95.63%; Nonlinear model result-2.29% this shows that the error of linear model is too big, and the accuracy of nonlinear model is more much bigger than linear model.

Claims (1)

1. infrared spectrum non-linear modeling quantitative anslysis method, it is characterized in that: this analytical approach is utilized the Matlab software programming according to the molecular vibration principle, and the Statistica software data is handled, and sets up nonlinear model, obtains final quantitative result by model; Detailed process is as follows:

(1) the Infrared spectrum scanning parameter is set, also promptly set the infrared spectrum instrument parameter, what conventional quantivative approach derived is the collection of illustrative plates of pdf form, contain the information such as peak intensity at Partial Feature peak, the needed output data of Nonlinear Modeling of the present invention is a text formatting, the data that comprise all test points, but not pdf form;

(2) tested standard items concentration gradient is accurately prepared in standard items scanning, respectively each concentration sample is carried out infrared scan, obtains the absorption value in whole wavelength coverages, and each sample scans more than 8 times at least, averages, and concentration gradient increases density as far as possible;

(3) characteristic peak is selected, and different infrared spectrometer scan operation differences according to all types of instrument requirements, is that target is implemented sample scanning with the Non-Destructive Testing; According to the structure of measured object, and there is environment in measured object, chooses the pairing characteristic peak of the group that can represent the measured object feature structure, is independent variable with concentration, and absorbance is a dependent variable, draws scatter diagram, to watch figure trend;

(4) set up nonlinear model, according to the feature and the Changing Pattern of scatter diagram, tentatively selected nonlinear model function is formed, and utilizes the matlab software programming to obtain primary mold, handle definite each parameter value by the statistica software data, and the degree of fitting of model;

(5) modelling verification, at first compare by true value, or record the tested components contents of sample by national standard method or official method or the higher method of accuracy, with this numerical value as reference, calculate characteristic absorption cutting edge of a knife or a sword place absorption of sample by nonlinear model and be worth pairing sample size, with the measured value comparison of preceding method institute, try to achieve relative error or relative deviation or standard relative deviation or average recovery.

Images