CN101650306A - Method for quickly testing and analyzing quality of sericite - Google Patents

Abstract

The invention relates to a method for quickly testing and analyzing the quality of sericite, which solves the problems that a microscope viewing method or chemical analysis method used in the prior art for testing the quality of sericite wastes time and labor and has low production efficiency. The method comprises the following steps: acquiring spectral information of the sericite sample to be tested by using a Fourier transform near-infrared spectroanalysis instrument, and analyzing the spectral information by using an optimal calibration model to obtain analysis result; the optimal calibration model is obtained by the following method: establishing a spectral library according to the spectral information of the sample, selecting representative spectral information to establish a calibration sample set, and gradually optimizing the quantitative analysis model of the spectral information of the sample by using a cross validation method according to the calibration sample set until obtaining the optimal calibration model. The method is suitable for quickly testing the purity analysis on the sericite in the green ore exploration and the tailing recycle.

Description

The method of fast detecting analyzing quality of sericite

Technical field

The present invention relates to a kind of method of fast detecting analyzing quality of sericite.

Background technology

China's sericite aboundresources, it is wide to distribute, and provinces and regions such as Shaanxi, Anhui, Fujian, Hubei, Heilungkiang, the Inner Mongol, Shandong all have big sericite mica ore to exist.This resources advantage provides necessary condition for sericite exploitation, and the research of sericite is had great importance.Improve the output and the quality of sericite, reduce production costs, protection mine surrounding environment, improving the mine resources comprehensive utilization ratio is the major subjects of Mine Sustainable Development.Sericite exploitation is mainly reclaimed sericite in the exploitation by sericite mica ore and the mine tailing, the quality of exploitation and recovery sericite directly influences the application and the price of sericite, especially in recent years improving constantly of sericite exploitation level also improves constantly the quality requirements of sericite.Therefore, extremely important to the detection of sericite purity.

Sericite is a kind of natural particulate white mica, belongs to the subspecies of white mica, is the silicate of layer structure, and structure clips the compound silica layer that the layer of aluminum oxygen octahedra constitutes by two-layer silicon-oxy tetrahedron.Sericite chemistry formula is KAl ₂[AlSi ₃O ₁₀] (OH) ₂, chemical constitution standard theory value is SiO ₂45.2%, Al ₂O ₃38.5%, K ₂O 11.8%, H ₂O 4.5%.The normal association of natural sericite ore has other mineral such as quartz, feldspar.Also often contain elements such as a spot of Fe, Mn, Cr, V in the lattice.

Main detection method is to adopt the microscopic examination method at present, adopts chemical analysis to detect wherein SiO simultaneously ₂, AL ₂O ₃And K ₂The content of O, the purity of definite sericite at last.But this method, there are operation steps more loaded down with trivial details and measurement needs a large amount of samples of consumption and reagent, sample can not obtain comprehensively, analyze fast, the experimental data that provides often all lags behind actual production, cause the exploitation sericite can not obtain timely, utilize fully, thereby caused the cost of developing raising, influenced surrounding environment.For this fast detecting of sericite purity to exploitation imperative.

Summary of the invention

The present invention is wasted time and energy and the low problem of production efficiency for solving to exist to detect when available technology adopting microscopic examination method and chemical analysis detect quality of sericite, and a kind of method of fast detecting analyzing quality of sericite is provided, and this method is:

Adopt the Fourier transform near infrared spectrum analyser to gather the spectral information of sericite sample to be detected, adopt best calibration model that the spectral information of described sericite sample to be detected is analyzed, obtain the quality analysis result of sericite to be measured;

Described best calibration model obtains by following method:

Step 1: adopting the Fourier transform near infrared spectrum analyser is 10000cm at wavelength ^-1~ 4000cm ^-1Gather the spectrum of a plurality of sericite samples in the scope, obtain a plurality of spectral informations;

Step 2: a plurality of spectral informations that statistic procedure one is obtained, and set up the sample spectra picture library of sericite;

Step 3: in the sample spectra picture library that step 2 obtains, adopt mathematical method to select representative spectral information, set up the correcting sample collection according to the described representative spectral information that obtains;

Step 4: the correcting sample centralized procurement that step 3 is set up is progressively optimized the Quantitative Analysis Model of sample spectra information with the method for cross validation, finally obtains best calibration model.

Advantage of the present invention: one, method operation steps of the present invention is simple and do not need to consume tested sample when detecting, and can obtain the check and analysis result in 30s.Two, check and analysis process of the present invention can be carried out synchronously with production run, realize monitoring comprehensively, fast product quality in process of production, in time scent a hidden danger, thereby improve the quality of production of sericite, to the application and the sale of actual sericite, improve the mine resources comprehensive utilization ratio has directive significance very much.Three, the present invention's spectral information of adopting the analysis of Fourier transform near infrared spectrum analyser to gather tested sericite is measured, and then acquisition testing result, save time, laborsaving, processing ease, tested sericite there is not destructiveness, testing process safety, environmental protection, to be safe from harm environment around also can not influencing of personnel, and in the production run of sericite, can react the variation of the sericite purity that reclaims in raw ore exploitation and the mine tailing fast, in real time.

Description of drawings

Fig. 1 is in the embodiment one, the spectrogram of the sericite sample to be measured that obtains by the Fourier transform near infrared spectrum analyser.

Embodiment

Embodiment one: the process of the method for the fast detecting analyzing quality of sericite described in the present embodiment is:

Adopt the Fourier transform near infrared spectrum analyser to gather the spectral information of sericite sample to be detected, adopt best calibration model that the spectral information of described sericite sample to be detected is analyzed, obtain the quality analysis result of sericite to be measured;

Described best calibration model obtains by following method:

Step 1: adopting the Fourier transform near infrared spectrum analyser is 10000cm at wavelength ^-1~ 4000cm ^-1Gather the spectrum of a plurality of sericite samples in the scope, obtain a plurality of spectral informations;

Step 2: a plurality of spectral informations that statistic procedure one is obtained, and set up the sample spectra picture library of sericite;

Step 3: in the sample spectra picture library that step 2 obtains, adopt mathematical method to select representative spectral information, set up the correcting sample collection according to the described representative spectral information that obtains;

Step 4: the correcting sample centralized procurement that step 3 is set up is progressively optimized the Quantitative Analysis Model of sample spectra information with the method for cross validation, finally obtains best calibration model.

The evaluation index of the Quantitative Analysis Model in the described step 4 of present embodiment comprises coefficient of determination R ²With correction mean square deviation RMSEC, according to described coefficient of determination R ²With proofread and correct mean square deviation RMSEC and can determine the number of principal components in the Quantitative Analysis Model and preprocess method, the optimal spectrum interval of spectrum.

Described coefficient of determination R ²The expression correcting sample is concentrated the degree of correlation of sample predicted value and theoretical value, and it is by formula:

$R^2=\{1-\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_i-{\hat{y}}_i)}^2}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_i-\stackrel{\‾}{y})}^2}\}\×100\%$

Obtain, n represents the quantity of calibration set sample in the formula; y _iThe actual value of representing i sample; Y represents the mean value of actual value;

Represent the predicted value of i near-infrared model to sample.

According to coefficient of determination R ²Computing formula as can be known,

Be the linear positive relation with y, work as R ²=100% o'clock, it was that it is best to predict the outcome on 1 the straight line that all observation stations all drop on slope, promptly measured sericite best in quality.

Coefficient of determination R ²Near 100%, it is good more to predict the outcome more, and the quality of promptly measured sericite is good more.Proofread and correct mean square deviation RMSEC, be used to weigh the accuracy that the calibration set sample predicts the outcome.It passes through formula:

$\mathrm{RMSEC}=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\hat{y}}_i-y)}^2}{n-1}}$

Obtain.

Correction mean square deviation RMSEC is an important parameter in the cross validation, and it is more little to proofread and correct mean square deviation RMSEC, and described Quantitative Analysis Model predicts the outcome more near theoretical value to the calibration set sample.

Adopting the Fourier transform near infrared spectrum analyser in the present embodiment is 10000cm at wavelength ^-1~ 4000cm ^-1The spectrogram of the sericite sample of gathering in the scope to be measured is referring to Fig. 1, and wherein horizontal ordinate is a wavelength, and ordinate is an absorbance log.

Mathematical method in the described step 3 of present embodiment is selected from a kind of in clustering procedure, major component factorization method or the gridding method.

The method of setting up the sample Quantitative Analysis Model of sericite in the described step 4 of present embodiment is a chemometrics method.

Described chemometrics method adopts offset minimum binary method, principal component analysis (PCA) or the polynary linear Return Law, and it is best wherein to adopt the offset minimum binary method to set up the effect of model.

Before the Quantitative Analysis Model of the sample of optimization sericite, each sample spectra that correcting sample is concentrated is carried out pre-service in the described step 4 of present embodiment.

Described pretreated method is selected from a kind of in straight line deduction, vector normalization, minimax normalization, polynary scatter correction, first order derivative method, second derivative method, linear compensation minusing, first order derivative method and straight line deduction, first order derivative method and vector normalization, first order derivative method and minimax normalization, first order derivative method and polynary scatter correction, first order derivative method and second derivative method or first order derivative method and the combination of linear compensation minusing, and the pretreating effect of wherein said first order derivative method and vector method for normalizing is best.

Optimal spectrum interval in the described Quantitative Analysis Model of present embodiment is 5242cm ^-1-4096cm ^-1Or the optimal spectrum value is 7070cm ^-1

The Fourier transform near infrared spectrum analyser of the step 1 described in the present embodiment can adopt the Fourier transform near infrared spectrum analyser in line style, as configuration optical fiber annex and light source noncontact mode.Also can adopt the Fourier transform near infrared spectrum analyser of offline model, as configuration integrating sphere annex and grating annex.

Embodiment two: present embodiment is a specific embodiment of embodiment one described method;

Best calibration model in the present embodiment adopts following method to obtain:

Step 1: employing resolution is 8cm ^-1, scanning times is 64 times Fourier transform near infrared spectrum analyser, is 10000cm at wavelength ^-1~ 4000cm ^-1Gather the spectrum of a plurality of sericite samples in the scope, obtain a plurality of spectral informations;

Step 2: a plurality of spectral informations that statistic procedure one is obtained, and set up the spectrum picture library of sericite sample;

Step 3: adopt mathematical method to select representative spectral information in the sample spectra picture library that in step 2, obtains, set up the correcting sample collection according to the described representative spectral information that obtains;

Step 4: the correcting sample centralized procurement that step 3 is set up is progressively optimized the Quantitative Analysis Model of sample spectra information with the method for cross validation, and described setting analysis model is the coefficient of determination R with the model evaluation index ²Being 0.86, proofreading and correct mean square deviation RMSEC is that 0.72 parameter value is set up, and finally obtains best calibration model;

Adopt the best calibration model that obtains that the spectral information of sericite sample to be measured is analyzed, obtain the quality results of sericite to be measured;

Adopt the sericite purity of Fourier transform near infrared spectrum analyser prediction and the deviation of the actual sericite purity that obtains by traditional microscopic examination and chemical analysis method, as shown in table 1:

Table 1

Adopt the purity of sericite of Fourier transform near infrared spectrum analyser prediction less as can be seen from Table 1 with the deviation of the purity of the actual sericite that obtains by traditional microscope and chemical analysis method, the result who obtains from two kinds of methods does not have significant difference as can be seen, and explanation can be used the purity of the methods analyst sericite of near infrared spectrum.

Claims (8)

1. the method for fast detecting analyzing quality of sericite is characterized in that the method for fast detecting analyzing quality of sericite is:

Adopt the Fourier transform near infrared spectrum analyser to gather the spectral information of sericite sample to be detected, adopt best calibration model that the spectral information of described sericite sample to be detected is analyzed, obtain the quality analysis result of sericite to be measured;

The process of the best calibration model of described foundation is:

Step 1: adopt the Fourier transform near infrared spectrum analyser in wavelength is 10000cm-1 ~ 4000cm-1 scope, to gather the spectrum of a plurality of sericite samples, obtain a plurality of spectral informations;

Step 2: a plurality of spectral informations that statistic procedure one is obtained, and set up the sample spectra picture library of sericite;

Step 3: in the sample spectra picture library that step 2 obtains, adopt mathematical method to select representative spectral information, set up the correcting sample collection according to the described representative spectral information that obtains;

Step 4: the correcting sample centralized procurement that step 3 is set up is progressively optimized the Quantitative Analysis Model of sample spectra information with the method for cross validation, finally obtains best calibration model.

2. the method for fast detecting analyzing quality of sericite according to claim 1, the evaluation index that it is characterized in that the Quantitative Analysis Model in the step 4 comprises coefficient of determination R2 and proofreaies and correct mean square deviation RMSEC, determines the preprocess method of spectrum range, number of principal components and spectrum in the Quantitative Analysis Model according to coefficient of determination R2 and correction mean square deviation RMSEC;

Described coefficient of determination R2 represents that correcting sample concentrates the degree of correlation of sample predicted value and theoretical value, and it is by formula:

$R^2=\{1-\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_i-\widehat{y_i})}^2}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_i-\stackrel{\‾}{y})}^2}\}\×100\%$

Obtain, n represents the quantity of calibration set sample in the formula; Yi represents the actual value of i sample; Y represents the mean value of actual value;

Represent the predicted value of i near-infrared model to sample;

Computing formula by above-mentioned coefficient of determination R2 gets, when R2 more near 100% the time, the quality of measured sericite is just best more;

Proofread and correct mean square deviation RMSEC, be used to weigh the accuracy that the calibration set sample predicts the outcome, it passes through formula:

$\mathrm{RMSEC}=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(\widehat{y_i}-y)}^2}{n-1}}$

Obtain;

It is more little to proofread and correct mean square deviation RMSEC, and described Quantitative Analysis Model predicts the outcome more near theoretical value to the calibration set sample.

3. the method for fast detecting analyzing quality of sericite according to claim 1 and 2 is characterized in that the mathematical method described in the step 3 is selected from a kind of in clustering procedure, major component factorization method or the gridding method.

4. the method for fast detecting analyzing quality of sericite according to claim 3, the method that it is characterized in that setting up described in the step 4 Quantitative Analysis Model of sericite sample is a chemometrics method.

5. the method for fast detecting analyzing quality of sericite according to claim 4 is characterized in that described chemometrics method comprises offset minimum binary method, principal component analysis (PCA) and the polynary linear Return Law.

6. according to the method for claim 1,2,4 or 5 described fast detecting analyzing quality of sericite, it is characterized in that before the Quantitative Analysis Model of the sample of optimization sericite in the step 4 each sample spectra that correcting sample is concentrated being carried out pre-service.

7. the method for fast detecting analyzing quality of sericite according to claim 6 is characterized in that described preprocess method is selected from a kind of in straight line deduction, vector normalization, minimax normalization, polynary scatter correction, first order derivative, second derivative, linear compensation minusing, first order derivative method and straight line deduction, first order derivative method and vector normalization, first order derivative method and minimax normalization, first order derivative method and polynary scatter correction, first order derivative method and second derivative method or first order derivative method and the combination of linear compensation minusing.

8. the method for fast detecting analyzing quality of sericite according to claim 2 is characterized in that interval 5242cm-1-4096cm-1 of being of optimal spectrum or the optimal spectrum value in the described Quantitative Analysis Model is 7070cm-1.

Images