CN104792752A - Method for determining content of pigments in mixed pigment solutions by adopting three-dimensional fluorescence spectroscopy combined with PARAFAC (parallel factor analysis) algorithm - Google Patents

Abstract

The invention discloses a method for determining the content of pigments in mixed pigment solutions by adopting three-dimensional fluorescence spectroscopy combined with the PARAFAC (parallel factor analysis) algorithm. According to the method, firstly, a correction sample set is formed by preparing multiple mixed pigment solutions according to different concentrations, then three-dimensional fluorescence spectra of the correction sample set and the background are measured, measured signals are pre-processed, a three-dimensional data matrix after pre-processing is analyzed with the PARAFAC algorithm, a correction model is established according to an analysis result, and the content of the pigments in unknown samples is predicted with the correction model. The method overcomes the defects of tedious process, time consumption of detection, higher detection cost and the like of a traditional chemical detection method, the concentration of each pigment in the to-be-detected samples can be determined rapidly, the method has the advantages of rapidness, safety, reliability, simple operation, lower cost and the like, technical support is provided for determination of synthetic edible pigments in food, and supervision of the food industry and protection of consumer interests are facilitated.

Description

Three-dimensional fluorescence spectrum measures the method for pigment content in hybrid pigment solution in conjunction with PARAFAC algorithm

Technical field

The invention provides a kind of three-dimensional fluorescence spectrum measures pigment content in hybrid pigment solution method in conjunction with PARAFAC algorithm, belong to food coloring quantitative detection field.

Background technology

Food-safety problem is related to the health and lives of people, is subject to the attention of country always.But the present situation of food security is also pessimistic, and new food-safety problem constantly occurs, wherein the Misuse of the food colour of Prof. Du Yucang occupies sizable ratio.Food colour is the based food adjuvant regulating food color to be fundamental purpose, also known as colorant, is widely used in food industry and medical and Cosmetic Manufacture always.Food colour has synthetic food color and natural colouring matter two kinds, and wherein, synthetic food color is because of lovely luster, strong coloring force, lower-price characteristic and being widely used.But many fabricated food pigments have certain toxicity, as carmine, lure the pigments such as red, amaranth to exceed standard absorptions, have teratogenesis, carcinogenic possibility, strictly must control use kind, scope and quantity.The method of current detection food colour content mainly contains: reversed-phase high-performance liquid chromatography (RP-HPLC) method, chemoluminescence method, RP-HPLC fluorescence detection etc., generally there is complex operation, detect consuming time, the shortcoming such as testing cost is high, result poor reproducibility in said method, therefore, be extremely necessary to find a kind of quick, convenient, stable, reliable food coloring content measurement technology.

Summary of the invention

The object of the present invention is to provide a kind of three-dimensional fluorescence spectrum to measure the method for pigment content in hybrid pigment solution in conjunction with PARAFAC algorithm, according to technical scheme of the present invention, the present invention is realized by following steps:

A. the preparation of calibration samples collection: two kinds of different food colorings are mixed to be incorporated in certain concentration range in several different ratio add another kind of pigment and obtain calibration samples collection as interfering material;

B. the measurement of the three-dimensional fluorescence spectrum of calibration samples collection and background: it fully shaken up after having prepared all biased samples, makes two kinds of pigments mix sufficiently uniformly, measures the three-dimensional fluorescence spectrum figure obtaining above-mentioned calibration samples collection and background;

C. the pre-service of data: utilize wavelet transformation to carry out wavelet de-noising process to spectroscopic data, subsequently by the three-dimensional light spectrum signal of the three-dimensional light spectrum signal background correction of hybrid pigment;

D. the foundation of calibration model: according to the consistent diagnosis of core (The core consistency diagnostic) determine system because of subnumber, parallel factor (PARAFAC) algorithm is utilized to resolve calibration set sample, and set up Quantifying model, realize the Rapid Simultaneous Determination to two kinds of different food coloring content in the mixed solution having an interfering material;

In described steps d, the consistent diagnosis of core (The core consistency diagnostic) is by the similarity degree between the super diagonal matrix T in calculating parallel factor analysis model and cube battle array G in truck3 model, and namely core homogeneity value estimates number of components:

$\mathrm{core}-\mathrm{consistency}=100\×[1-\frac{\underset{d=1}{\overset{F}{\mathrm{\Σ}}}\underset{e=1}{\overset{F}{\mathrm{\Σ}}}\underset{f=1}{\overset{F}{\mathrm{\Σ}}}{(g_{\mathrm{def}}-t_{\mathrm{def}})}^2}{\underset{d=1}{\overset{F}{\mathrm{\Σ}}}\underset{e=1}{\overset{F}{\mathrm{\Σ}}}\underset{f=1}{\overset{F}{\mathrm{\Σ}}}{t}_{\mathrm{def}}^2}]$

In formula, F be model because of subnumber, gdef is the element of cube battle array G, and tdef is the element of cube battle array T.For desirable PARAFAC model, a cube battle array G for super diagonal matrix T and truck3 model is closely similar, and core homogeneity value now will equal 100%.In practical application, as long as core homogeneity value is more than or equal to 60%, then can thinks and meet trilinear model, but when core homogeneity value lower than 60% time, can think depart from three linear.So number of components can be judged according to the situation of change of core homogeneity value.

In described steps d, the concrete iterative process of parallel factor (PARAFAC) algorithm is as follows:

(1) determine that system is because of subnumber N;

(2) initialization matrix A and B, A and B use respectively with front N row singular vector estimate as its initialization;

(3) by X, A and B computational load Matrix C;

$c_{\left(k\right)}^T={(A^{T}A*B^{T}B)}^{-1}\mathrm{diag}\left(A^T{X}_{..K}B\right),(k=\mathrm{1,2},...,K)$

(4) by X, C and B computational load matrix A;

$A=\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{X}_{..k}\mathrm{Bdiag}\left(c_{\left(k\right)}\right)\right){\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\mathrm{diag}\left(c_{\left(k\right)}\right)B^T\mathrm{Bdiag}\left(c_{\left(k\right)}\right)\right)}^{-1}$

(5) by X, A and C computational load matrix B;

$B=\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{X}_{..k}^T\mathrm{Adiag}\left(c_{\left(k\right)}\right)\right){\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\mathrm{diag}\left(c_{\left(k\right)}\right)A^T\mathrm{Adiag}\left(c_{\left(k\right)}\right)\right)}^{-1}$

(6) second step is repeated to the 5th step, until convergence, that is:

$\left|\frac{{\mathrm{SSR}}^{\left(m\right)}-{\mathrm{SSR}}^{(m-1)}}{{\mathrm{SSR}}^{(m-1)}}\right|\≤1\×{10}^{-6}$

Parallel factor algorithm used by the present invention belongs to second order correction algorithm, and second order correction algorithm is according to statistical theory, and carry out studying and finding rule from one group of known sample data, then assimilated equations is predicted unknown data.This kind of methods analyst ability is strong, has so-called " second order advantage ", under unknown disturbances thing exists, can tell the response signal of multiple character similarity analysis thing simultaneously, and directly can carry out quantitative measurement to interested analysis thing component.The present invention compensate for traditional chemical detection method process loaded down with trivial details, detect consuming time, the more high deficiency of testing cost; the Fast Measurement of pigment concentration in testing sample can be realized; there is the advantages such as quick, safe, reliable, simple to operate, cost is lower; there is provided technical guarantee to the mensuration of synthetic food color in food, be conducive to food service industry supervision and Protection of consumer rights and interests.

Accompanying drawing explanation

Below in conjunction with accompanying drawing, the invention will be further described.

Fig. 1, Fig. 2, Fig. 3, Fig. 4 are respectively cochineal solution, lure red solution, be isogonism three-dimensional projection as the left-half of the amaranth solution of interfering material, the three-dimensional fluorescence spectrum figure of three's mixed solution, figure, have half part to be contour map.

Fig. 5 is to determine the result of system component number to the consistent diagnosis of three-dimensional fluorescence spectrum data matrix core of calibration samples collection.

Fig. 6 is hybrid pigment solution three-dimensional fluorescence spectrum by the famille rose wherein obtained after parallel factor arithmetic analysis and the red fluorescence emission spectrogram of temptation.

Fig. 7 is the process flow diagram of the method for the invention.

Embodiment

Below in conjunction with accompanying drawing and the invention will be further described according to the sequence of operation of process flow diagram:

A. the preparation of calibration samples collection and forecast sample collection.With ultrapure water as a setting by carmine, lure that red interfering material amaranth is arbitrarily added within the scope of 0 ~ 1ug/ml according to the concentration ratio mixing shown in table 1, be mixed with 7 experiment samples composition calibration samples collection, 5 experiment sample predicted composition sample sets;

B. the measurement of the three-dimensional fluorescence spectrum of calibration samples collection, forecast sample collection and ultrapure water.After having prepared all biased samples, it is fully shaken up, two kinds of pigments are mixed sufficiently uniformly.Then measuring and obtaining them in excitation wavelength range is 230 ~ 341nm, and be spaced apart 3nm, emission wavelength ranges is 385 ~ 650nm, is spaced apart the three-dimensional fluorescence spectrum figure of 5nm, and incident and exit slit is all set to 5nm, and integral time is set to 0.01s.Fig. 1, Fig. 2, Fig. 3, Fig. 4 are respectively cochineal solution, lure red solution, be isogonism three-dimensional projection as the left-half of the amaranth solution of interfering material, the three-dimensional fluorescence spectrum figure of three's mixed solution, figure, have half part to be contour map.The fluorescence spectral characteristic parameter of these solution is as shown in table 2.With same arranging, three-dimensional fluorescence spectrum measurement is carried out to ultrapure water simultaneously, in triplicate, average;

C. the pre-service of data.In order to avoid temperature, humidity etc. are on the impact of measuring signal to noise ratio (S/N ratio), need to do noise reduction process to spectroscopic data.Wavelet transformation is widely used in noise reduction.Wavelet transformation is utilized to carry out noise reduction process to spectroscopic data.Utilized by spectroscopic data small echo to do three rank to decompose, then three rank decomposed the high-frequency signal that obtains and make zero, finally do inverse wavelet transform just obtain noise reduction after spectroscopic data.Subsequently by correcting the three-dimensional light spectrum signal of the three-dimensional light spectrum signal deduction ultrapure water of sample set and forecast sample collection, reduce the Raman scattering impact of solvent as much as possible;

D. the foundation of calibration model.Three-dimensional fluorescence spectrum through pretreated calibration samples collection is arranged in three-dimensional matrice, and it is determined to the number of components of system with the consistent diagnosis of core (The core consistency diagnostic), the result obtained is as Fig. 5.Round dot in figure represents 0 element, and annulus represents non-zero element, can find out because of subnumber be 3 time, non-zero element (annulus) is close to 1, and 0 element is close to 0, and core homogeneity value is 89%, and at this moment model is trilinear model, and the accuracy of parallel factor algorithm is higher.Thus determine that the number of components in system is 3.After the optimal set mark determined, with parallel factor (PARAFAC) algorithm, the three-dimensional matrice that the three-dimensional fluorescence spectrum by calibration samples collection is arranged in is resolved, obtain score matrix A and loading matrix B, C, by carmine in the hybrid pigment that obtains after parallel factor arithmetic analysis and that temptation is red fluorescence emission spectrum as Fig. 6.Score matrix A can associate with concentration vector y, and set up quantitative calibration models: y=Ab, regression coefficient b obtains by least square method.For unknown sample (being forecast sample collection here), first loading matrix B, C by obtaining before is needed to calculate the score matrix A ' of its correspondence, calculated the predicted value y '=A ' b of pigment content again by regression coefficient b, table 3 illustrates calibration model and concentrates carmine to forecast sample and lure predicting the outcome of red content;

In described step a, calibration samples collection and forecast sample concentrate the concentration proportioning of hybrid pigment solution to carry out according to table 1:

The concentration proportioning of table 1 hybrid pigment solution

In described step b, cochineal solution, lure red solution, as shown in table 2 as the three-dimensional fluorescence spectrum characteristic parameter of the amaranth solution of interfering material, three's mixed solution:

Table 2 three-dimensional fluorescence spectrum characteristic parameter

In described steps d, the consistent diagnosis of core (The core consistency diagnostic) is by the similarity degree between the super diagonal matrix T in calculating parallel factor analysis model and cube battle array G in truck3 model, and namely core homogeneity value estimates number of components:

$\mathrm{core}-\mathrm{consistency}=100\×[1-\frac{\underset{d=1}{\overset{F}{\mathrm{\Σ}}}\underset{e=1}{\overset{F}{\mathrm{\Σ}}}\underset{f=1}{\overset{F}{\mathrm{\Σ}}}{(g_{\mathrm{def}}-t_{\mathrm{def}})}^2}{\underset{d=1}{\overset{F}{\mathrm{\Σ}}}\underset{e=1}{\overset{F}{\mathrm{\Σ}}}\underset{f=1}{\overset{F}{\mathrm{\Σ}}}{t}_{\mathrm{def}}^2}]$

F be model because of subnumber; g _deffor the element of cube battle array G; t _defit is the element of cube battle array T.For desirable PARAFAC model, a cube battle array G for super diagonal matrix T and truck3 model is closely similar, and core homogeneity value now will equal 100%.In general, as long as core homogeneity value is more than or equal to 60%, then can thinks and meet trilinear model, but when core homogeneity value lower than 60% time, can think depart from three linear.So number of components can be judged according to the situation of change of core homogeneity value.

In described steps d, the concrete iterative process of parallel factor (PARAFAC) algorithm is as follows:

(1), determine that system is because of subnumber N.

(2), initialization matrix A and B, in general, A and B uses respectively with front N row singular vector estimate as its initialization, also can use randomization, the first will be adopted herein as initialization.

(3), by X, A and B computational load Matrix C.

$c_{\left(k\right)}^T={(A^{T}A*B^{T}B)}^{-1}\mathrm{diag}\left(A^T{X}_{..K}B\right),(k=\mathrm{1,2},...,K)$

(4), by X, C and B computational load matrix A.

$A=\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{X}_{..k}\mathrm{Bdiag}\left(c_{\left(k\right)}\right)\right){\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\mathrm{diag}\left(c_{\left(k\right)}\right)B^T\mathrm{Bdiag}\left(c_{\left(k\right)}\right)\right)}^{-1}$

(5), by X, A and C computational load matrix B.

$B=\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{X}_{..k}^T\mathrm{Adiag}\left(c_{\left(k\right)}\right)\right){\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\mathrm{diag}\left(c_{\left(k\right)}\right)A^T\mathrm{Adiag}\left(c_{\left(k\right)}\right)\right)}^{-1}$

(6), second step is repeated to the 5th step, until convergence, that is:

$\left|\frac{{\mathrm{SSR}}^{\left(m\right)}-{\mathrm{SSR}}^{(m-1)}}{{\mathrm{SSR}}^{(m-1)}}\right|\≤1\×{10}^{-6}$

In described steps d, calibration model is as shown in table 3 to predicting the outcome of forecast sample collection:

Table 3 predicts the outcome

In sum, the present invention is by three-dimensional fluorescence spectrum and a kind of second order correction algorithm--parallel factor (PARAFAC) algorithm combines, and realizes the Simultaneous Determination of each pigment content in hybrid pigment solution.The method has the advantages such as quick, reliable, simple to operate, cost is lower; compensate for traditional chemical detection method process loaded down with trivial details, detect consuming time, the more high deficiency of testing cost; for specification food service industry and Protection of consumer rights and interests provide technical guarantee, promote the security control of synthetic food color in food.

Claims (3)

1. three-dimensional fluorescence spectrum measures a method for pigment content in hybrid pigment solution in conjunction with PARAFAC algorithm, it is characterized in that: described method is realized by following steps:

A. the preparation of calibration samples collection: two kinds of different food colorings are mixed to be incorporated in certain concentration range in several different ratio add another kind of pigment and obtain calibration samples collection as interfering material;

B. the measurement of the three-dimensional fluorescence spectrum of calibration samples collection and background: it fully shaken up after having prepared all biased samples, makes two kinds of pigments mix sufficiently uniformly, measures the three-dimensional fluorescence spectrum figure obtaining above-mentioned calibration samples collection and background;

C. the pre-service of signal: utilize wavelet transformation to carry out wavelet de-noising process to spectroscopic data, subsequently by the three-dimensional light spectrum signal of the three-dimensional light spectrum signal background correction of hybrid pigment;

D. the foundation of calibration model: according to the consistent diagnosis of core (The core consistency diagnostic) determine system because of subnumber, parallel factor (PARAFAC) algorithm is utilized to resolve calibration set sample, and set up Quantifying model, realize the Rapid Simultaneous Determination to two kinds of different food coloring content in the mixed solution having an interfering material.

2. three-dimensional fluorescence spectrum measures the method for pigment content in hybrid pigment solution in conjunction with PARAFAC algorithm according to claim 1, it is characterized in that: in described steps d, the consistent diagnosis of core (The core consistencydiagnostic) is by the similarity degree between the super diagonal matrix T in calculating parallel factor analysis model and cube battle array G in truck3 model, and namely core homogeneity value estimates number of components:

$\mathrm{core}-\mathrm{consistency}=100\×[1-\frac{\underset{d=1}{\overset{F}{\mathrm{\Σ}}}\underset{e=1}{\overset{F}{\mathrm{\Σ}}}\underset{f=1}{\overset{F}{\mathrm{\Σ}}}{(g_{\mathrm{def}}-t_{\mathrm{def}})}^2}{\underset{d=1}{\overset{F}{\mathrm{\Σ}}}\underset{e=1}{\overset{F}{\mathrm{\Σ}}}\underset{f=1}{\overset{F}{\mathrm{\Σ}}}{t}_{\mathrm{def}}^2}]$

In formula, F be model because of subnumber, gdef is the element of cube battle array G, and tdef is the element of cube battle array T.

3. three-dimensional fluorescence spectrum measures the method for pigment content in hybrid pigment solution in conjunction with PARAFAC algorithm according to claim 1, it is characterized in that: in described steps d, and the concrete iterative process of parallel factor (PARAFAC) algorithm is as follows:

(1) determine that system is because of subnumber N;

(2) initialization matrix A and B, in general, A and B uses respectively with front N row singular vector estimate as its initialization, also can use randomization, the first will be adopted herein as initialization;

(3) by X, A and B computational load Matrix C;

$C_{\left(k\right)}^T={(A^{T}A*B^{T}B)}^{-1}\mathrm{diag}\left(A^T{X}_{..K}B\right),(k=\mathrm{1,2},...,K)$

(4) by X, C and B computational load matrix A;

$A=\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{X}_{..k}\mathrm{Bdiag}\left(c_{\left(k\right)}\right)\right){\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\mathrm{diag}\left(c_{\left(k\right)}\right)B^T\mathrm{Bdiag}\left(c_{\left(k\right)}\right)\right)}^{-1}$

(5) by X, A and C computational load matrix B;

$B=\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{X}_{..k}^T\mathrm{Adiag}\left(c_{\left(k\right)}\right)\right){\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\mathrm{diag}\left(c_{\left(k\right)}\right)A^T\mathrm{Adiag}\left(c_{\left(k\right)}\right)\right)}^{-1}$

(6) second step is repeated to the 5th step, until convergence, that is:

$\left|\frac{{\mathrm{SSR}}^{\left(m\right)}-{\mathrm{SSR}}^{(m-1)}}{{\mathrm{SSR}}^{(m-1)}}\right|\≤1\×{10}^{-6}$