CN104713855A - Method of utilizing laser probe to detect iron ore pH value - Google Patents

Abstract

The invention discloses a method of utilizing a laser probe to detect an iron ore pH value. The method comprises the following steps: utilizing the laser probe to acquire spectral information of a target element in a characteristic spectrum section, using the spectral information as an independent variable, using the content of the target element compound as a dependent variable, utilizing a principal component analysis method to extract principal components, removing minor principal components, weakening the matrix effect in LIBS analysis and the influence due to random error, and utilizing a partial least squares regression to establish a steady quantitative analysis model, so as to realize detection of the iron ore pH value. The method disclosed by the invention can reduce the introducing probability of irrelevant information, effectively lowers spectroscopic data handling capacity, and improves the detection efficiency and the detection precision.

Description

A kind of laser probe detects the method for monocalcium acid basicity

Technical field

The invention belongs to atomic emission detection technical field, be specifically related to a kind of method detecting monocalcium acid basicity based on laser probe, be mainly used in the quick detection of monocalcium acid basicity.

Background technology

Iron ore is the important sources that metallic iron is produced, and the control of inspection to iron ore smelting process and product quality of components of iron ore and characteristic parameter is all significant.Wherein, monocalcium acid basicity is the important parameter of iron ore smelting furnace material proportioning material, which determines the type and proportioning of adding flux.Monocalcium acid basicity CaO and content of MgO sum and Al ₂o ₃and SiO ₂the ratio of content sum represents.Conventional detection technique has ICP-AES, atomic absorption spectrography (AAS) and titrimetry etc., but these method sample preparations are complicated, and the cycle is long, can not meet the demand that production scene is detected real-time.Laser probe technology, i.e. Laser-induced Breakdown Spectroscopy (Laser Induced BreakdownSpectroscopy, being called for short LIBS) technology is a kind of Laser-Induced Plasma Spectral Analysis Technique, this technology sample preparation is simple, detection speed is fast, is applicable to field, real-time online and remote detection.But the single argument scaling method that conventional laser probe uses is when the detection for monocalcium acid basicity owing to affecting by factors such as spectra1 interfer-, self absorption effect and matrix effects, and analysis precision is not high, is difficult to reach examination criteria or application demand.Therefore, a kind of method of an urgent demand improves the analysis precision of LIBS technology to monocalcium acid basicity.

Summary of the invention

The invention provides a kind of method that laser probe detects monocalcium acid basicity, object is quick, the accurate analysis that realize monocalcium acid basicity.

For achieving the above object, a kind of laser probe provided by the invention detects the method for monocalcium acid basicity, the method utilizes laser probe to gather object element characteristic spectrum place section spectral information, and using this spectral information as independent variable, dependent variable is made with object element compounds content, principal component analysis (PCA) is utilized to carry out Principle component extraction, remove secondary major component, slacken the impact that in LIBS analysis, matrix effect and stochastic error produce, utilize partial least-squares regression method to set up sane Quantitative Analysis Model, realize the quick detection of monocalcium acid basicity.

The single argument calibrating method used for existing laser probe technology is difficult to the high precision test realizing monocalcium acid basicity, and PLSR model easily introduces the spectral information even problem of noise outside detection elements characteristic spectral line when selecting more major component Modling model, PLSR method and LIBS combine with technique detect monocalcium acid basicity by the present invention, and choose reasonable is carried out to input spectrum, the quick high accuracy realizing monocalcium acid basicity detects.Specifically, the present invention has following technical characterstic:

(1) technical characterstic that the present invention is the most outstanding is combined with laser probe technology by PLSR, choose object element characteristic spectrum place section spectral information and carry out Principle component extraction as independent variable and object element compounds content as dependent variable, remove secondary major component, slacken the impact that in LIBS analysis, matrix effect and stochastic error produce, set up sane Quantitative Analysis Model, realize the high precision test of monocalcium acid basicity;

(2) the present invention utilizes the feature of identity element spectral profile region Relatively centralized, select the one section of spectral information (about 10 ~ 20nm) containing many analytical element characteristic spectral lines as independent variable, probability is introduced to reduce irrelevant information, and effectively reduce spectroscopic data treatment capacity, improve detection efficiency and accuracy of detection;

(3) PLSR that the present invention is used is a kind of multiple linear regression method, effectively can overcome the Matrix Match problem of the harshness required by unit linear regression method such as basic scaling method and internal standard method that laser probe technology generally adopts, have better adaptability;

(4) the present invention is the improvement of laser probe quantitative analysis method, except detecting monocalcium acid basicity and being suitable for, also has good Detection results to the detection of iron ore taste (total iron content) and chemical composition thereof.

Accompanying drawing explanation

Fig. 1 is the FB(flow block) of the monocalcium acid basicity detection method that the present invention is based on laser probe and partial least-squares regression method.

Fig. 2 (a), (b), (c) and (d) are that example of the present invention is to CaO, MgO, Al in iron ore respectively ₂o ₃and SiO ₂standard value and predicted value graph of a relation.

Embodiment

Be described further below in conjunction with accompanying drawing 1 pair of the specific embodiment of the present invention.It should be noted that at this, the explanation for these embodiments understands the present invention for helping, but does not form limitation of the invention.In addition, if below in described each embodiment of the present invention involved technical characteristic do not form conflict each other and just can mutually combine.

As shown in Figure 1, the laser probe that the present invention proposes detects the method for monocalcium acid basicity, and implementation step is as follows:

1st step, using iron series ore sample known for chemical composition as calibration sample, detects them by laser probe technology after compressing tablet, respectively one section, record object element spectral line place spectrum.

In this example, the 1st step can adopt following concrete steps to be achieved:

1.1st step, gets n chemical composition is known and constituent content has certain gradient iron ore standard model does calibration sample, and be generally and prevent Expired Drugs from occurring, n gets more than 20;

1.2nd step, by bottom and the surrounding of boric acid inlaying process parcel iron ore, and is pressed into surfacing, the uniform cylindrical sheet of thickness with 25MPa pressure by each sample sheeter;

1.3rd step, utilizes laser probe technology to detect calibration sample, gathers or intercept one section of spectrum (about 10 ~ 20nm) of at least 3 characteristic spectral lines containing Ca, Mg, Al and Si to each calibration sample respectively;

2nd step, make independent variable with each calibration sample object element spectroscopic data, corresponding element compounds content makes dependent variable, utilizes partial least-squares regression method to set up CaO, MgO, Al ₂o ₃and SiO ₂quantitative Analysis Model, and utilize cross-validation method to carry out School Affairs optimization to the predictive ability of model.

In this example, the 2nd step can adopt following concrete steps to be achieved:

2.1st step, by i-th calibration sample object element spectrum region spectrum segment data _xiobtain do standardization i=1,2 ..., n, and s _xbe respectively x _imean value and standard deviation, then with x ' _ias spectrum matrix X _nmthe i-th row, n is standard model number, and m is the object element spectrum segment number of data points chosen;

2.2nd step, if object element oxide content is y in calibration sample i _i, to y _iobtain do standardization wherein and s _ybe respectively y _imean value and standard deviation, with y ' _ifor element set becomes row vector Y _n;

2.3rd step, with X _nmfor independent variable matrix, Y _nfor dependent variable matrix, to X _nmand Y _ncarry out principal component analysis (PCA), namely convert former variable, new variables is the linear combination of former variable, and these new variables are called major component.Matrix X _nmrow participate in Y _nthe calculating of the major component factor, matrix Y _nrow also participate in X _nmthe calculating of the major component factor, to X _nmand Y _nextraction major component t and u, t and u should carry X as much as possible _nmand Y _nin variation information, and covariance is between the two maximum, obtains:

X _nm＝TP′+E

Y _n＝Uq′+f

Wherein, P and q is X respectively _nmand Y _nthe load matrix of major component t and u, P ', q ' represent the transposed matrix of load matrix P and q respectively, T and U is the corresponding score matrix of major component t and u, E and f utilizes partial least square method matching X _nmand Y _nthe residual matrix produced.

2.4th step, the model of fit of partial least-squares regression method selects a front k major component to be similar to, and the determination of k value utilizes cross-validation method to verify its validity, makes Y _nand predicted value cross validation root-mean-square error (Root-mean-square error ofcross-validation, RMSECV) reach minimum value or increase number of principal components again and do not have clear improvement, at this moment intercept X _nmfront k major component t ₁, t ₂, t _k, set up regression model:

${\hat{Y}}_n=X_{\mathrm{nm}}{B}_m$

Wherein, for Y _nestimated value, B _m=W (P ' W) ^-1the regression coefficient column vector that q ' obtains for multilinear fitting, for characterizing X _nmwith between internal relation, W is offset minimum binary weight matrix;

3rd step, carries out laser probe detection to oxide t in testing sample, according to the 1st step and the same mode of the 2nd step, intercepts k the number of principal components identical with the 2.4th step, and they are substituted into above-mentioned regression model, obtain the predicted value of unknown oxide t content by right inverse standardized transformation, the content of oxide t

4th step, utilizes above-mentioned model to predict oxide CaO, MgO, Al in unknown sample respectively ₂o ₃and SiO ₂content with the potential of hydrogen of measurable unknown ironstone sample

$\hat{S}=\frac{{\hat{y}}_{\mathrm{CaO}}+{\hat{y}}_{\mathrm{MgO}}}{{\hat{y}}_{{\mathrm{Al}}_2{O}_3}+{\hat{y}}_{{\mathrm{SiO}}_2}}$

Embodiment:

The experiment of this example is carried out under air ambient, and experimental provision used is common laser probe device.Adopt Q-switch Nd:YAG pulsed laser (wavelength 532nm, repetition frequency 3Hz, pulse width 8ns), laser energy 30mJ/ pulse.Laser focuses on sample surfaces through catoptron and plano-convex lens (focal length 15cm), and ablation sample produces plasma.For avoiding producing dell and affect spectral characteristic in the ablation of sample same position, sample being placed on three-D displacement platform and making " bow " zigzag motion.For preventing from puncturing air and obtaining best spectral signal-noise ratio, focus is placed in about 4mm below sample surfaces.Plasma resonance light is collected by light collector and is coupled in optical fiber, carries out light splitting, and carry out the opto-electronic conversion of signal by the ICCD be arranged on spectrometer by Optical Fiber Transmission to spectrometer (AndorTechnology, Shamrock 500i).ICCD gate adopts accumulation pattern, and in order to obtain higher spectral intensity and signal-to-background ratio, collection time delay and gate-width are all set to 5 μ s.Computing machine is connected with ICCD, obtains and analyze spectroscopic data.

This example uses 60 kinds of known ironstone samples of composition, and composition is as shown in table 1, and the 50 kinds of samples being wherein numbered 1 ~ 50 set up PLSR model as calibration sample, and all the other are numbered 10 kinds of 51 ~ 60 as test sample.Get each ironstone sample 1 gram, boric acid powder 9 grams, by bottom and the surrounding of boric acid inlaying process parcel iron ore, and with 25MPa pressure, each sample sheeter is pressed into surfacing, the uniform cylindrical sheet of thickness.

Respectively to above-mentioned 60 kinds of sample collection Si, Al, Ca and Mg elemental characteristic spectral line region spectrum, Si chooses 278.7 ~ 291.5nm and (comprises Si I 288.16nm, SiII 290.43nm, the characteristic spectral line of the Si such as Si II 290.56nm), Al and Ca chooses 386 ~ 404nm and (comprises Al II 390.07nm, Al I 394.4nm, the characteristic spectral line of the Al such as Al I 396.15nm, and Ca II 393.37nm, Ca II 396.85nm, the characteristic spectral line of the Ca such as Ca I 397.37nm), Mg chooses 508 ~ 523nm and (comprises Mg I 516.73nm, Mg I 517.26nm, the characteristic spectral line of the Mg such as Mg I 518.36nm).Walk according to embodiment 1-2, utilize 50 kinds of calibration samples to set up oxide CaO, MgO, Al respectively ₂o ₃and SiO ₂pLSR Quantitative Analysis Model.

Table 1. iron ore sample oxidation thing is containing scale

In this example, the PLSR process of spectroscopic data all uses MATLAB ^tMprogram realizes.As shown in Fig. 2 (a)-(d), " △ " represents calibration sample, the robustness linear fit coefficients R of its Modling model ²represent with cross validation root-mean-square error (Root-mean-square error of cross-validation, RMSECV), oxide CaO, MgO, Al in this example ₂o ₃and SiO ₂do not have significant change when RMSECV reaches minimum value or increases number of principal components again when the number of principal components k chosen is followed successively by 11,9,14 and 12, RMSECV is respectively 0.0769,0.0397,0.2797 and 0.5515wt.%, R ²be respectively 0.9967,0.9958,0.9996 and 0.9991, the visible this method that uses can set up the good Quantitative Analysis Model of robustness to 4 kinds of oxides above-mentioned in iron ore.

Under identical experiment condition, gather one section of spectrum that Si, Al, Ca are consistent with standard sample collection region with Mg elemental characteristic spectral line.According to the method for 1-2 step in embodiment, the spectrum of 10 test samples is normalized and principal component analysis (PCA) as independent variable, then according to mode described in embodiment the 3rd step, intercept and calibration sample sets up PLSR Quantitative Analysis Model time identical major component, substitute into PLSR model, obtain CaO, MgO, Al in test sample oxide ₂o ₃and SiO ₂content with as shown in Fig. 2 (a)-(d), " ■ " represents test sample, as seen to oxide CaO, MgO, Al in test sample ₂o ₃and SiO ₂content prediction average relative error (Averagerelative error ofprediction, AREP) is respectively 6.44,5.03,6.63 and 3.52%.

Finally, according to embodiment the 4th step, utilize with the potential of hydrogen of iron ore test sample can be obtained

$\hat{S}=\frac{{\hat{y}}_{\mathrm{CaO}}+{\hat{y}}_{\mathrm{MgO}}}{{\hat{y}}_{{\mathrm{Al}}_2{O}_3}+{\hat{y}}_{{\mathrm{SiO}}_2}}$

By test sample standard composition y each in table 1 _caO, y _mgO, with calculate its potential of hydrogen S:

$S=\frac{y_{\mathrm{CaO}}+y_{\mathrm{MgO}}}{y_{{\mathrm{Al}}_2{O}_3}+y_{{\mathrm{SiO}}_2}}$

and S value is as shown in table 2, utilize the inventive method to the predicted value of ironstone sample potential of hydrogen as seen with standard value S energy quite well, average relative error ARE is 3.65%.

Table 2. tests sample potential of hydrogen standard value S and predicted value (wt.%)

The above is a kind of preferred embodiment of the present invention, and for showing and describing ultimate principle of the present invention, principal character and advantage of the present invention, the present invention is not limited to the content disclosed in this embodiment and accompanying drawing.Without departing from the spirit and scope of the present invention, the present invention also has various changes and modifications, and these changes and improvements all fall in the claimed scope of the invention.Application claims protection domain is defined by appending claims and equivalent thereof.

Claims (4)

1. the method for a laser probe detection monocalcium acid basicity, the method utilizes laser probe to gather object element characteristic spectrum place section spectral information, and using this spectral information as independent variable, dependent variable is made with object element compounds content, principal component analysis (PCA) is utilized to carry out Principle component extraction, remove secondary major component, slacken the impact that in LIBS analysis, matrix effect and stochastic error produce, and utilize partial least-squares regression method to set up sane Quantitative Analysis Model, realize the quick detection of monocalcium acid basicity.

2. laser probe according to claim 1 detects the method for monocalcium acid basicity, and it is characterized in that, the method specifically comprises the steps:

1st step, using iron series ore sample known for chemical composition as calibration sample, detects them by laser probe technology after compressing tablet, records the spectrum segment of calibration sample object element spectrum region respectively;

2nd step, make independent variable with the data of each described spectrum segment, corresponding element compounds content makes dependent variable, utilizes partial least-squares regression method to set up CaO, MgO, Al ₂o ₃and SiO ₂quantitative Analysis Model, and utilize cross-validation method, choose major component and School Affairs optimization is carried out to the predictive ability of Quantitative Analysis Model, obtain regression model;

3rd step, carries out laser probe detection to oxide t in testing sample, according to the 1st step and the same mode of the 2nd step, intercepts identical number of principal components, and they are substituted into described regression model, obtain the predicted value of unknown oxide t content by to predicted value inverse standardized transformation, namely obtains the content of oxide t

4th step, utilizes described regression model to predict oxide CaO, MgO, Al in unknown sample respectively ₂o ₃and SiO ₂content with predict the potential of hydrogen of unknown ironstone sample

$\hat{S}=\frac{{\hat{y}}_{\mathrm{CaO}}+{\hat{y}}_{\mathrm{MgO}}}{{\hat{y}}_{{\mathrm{Al}}_2{O}_3}+{\hat{y}}_{{\mathrm{SiO}}_2}}.$

3. laser probe according to claim 2 detects the method for monocalcium acid basicity, and it is characterized in that, described 2nd step specifically comprises the steps:

2.1st step, by i-th calibration sample object element spectrum region spectrum segment data x _iobtain do standardization i=1,2 ..., n, and s _xbe respectively x _imean value and standard deviation, then with x ' _ias spectrum matrix X _nmthe i-th row, n is standard model number, and m is the object element spectrum segment number of data points chosen;

2.2nd step, if object element oxide content is y in calibration sample i _i, to y _iobtain do standardization wherein and s _ybe respectively y _imean value and standard deviation, with y ' _ifor element set becomes row vector Y _n;

2.3rd step, with X _nmfor independent variable matrix, Y _nfor dependent variable matrix, to X _nmand Y _ncarry out principal component analysis (PCA), namely convert former variable, new variables is the linear combination of former variable, and these new variables are called major component; Matrix X _nmrow participate in Y _nthe calculating of the major component factor, matrix Y _nrow also participate in X _nmthe calculating of the major component factor, to X _nmand Y _nextraction major component t and u, t and u should carry X as much as possible _nmand Y _nin variation information, and covariance is between the two maximum, obtains:

X _nm＝TP′+E

Y _n＝Uq′+f

Wherein, P and q is X respectively _nmand Y _nthe load matrix of major component t and u, P ', q ' represent the transposed matrix of load matrix P and q respectively, T and U is the corresponding score matrix of major component t and u, E and f utilizes partial least square method matching X _nmand Y _nthe residual matrix produced;

2.4th step, the model of fit of partial least-squares regression method selects a front k major component to be similar to, and the determination of k value utilizes cross-validation method to verify its validity, makes Y _nand predicted value cross validation root-mean-square error reach minimum value or increase number of principal components again and do not have clear improvement, now intercept X _nmfront k major component t ₁, t ₂..., t _k, set up regression model:

${\hat{Y}}_n=X_{\mathrm{nm}}{B}_m$

Wherein, for Y _nestimated value, B _m=W (P ' W) ^-1the regression coefficient column vector that q ' obtains for multilinear fitting, for characterizing X _nmwith between internal relation, W is offset minimum binary weight matrix.

4. laser probe according to claim 2 detects the method for monocalcium acid basicity, and it is characterized in that, the 1st step specifically comprises the steps:

1.1st step, gets n chemical composition is known and constituent content has certain gradient iron ore standard model does calibration sample;

1.2nd step, by bottom and the surrounding of boric acid inlaying process parcel iron ore, and is pressed into surfacing, the uniform cylindrical sheet of thickness by each calibration sample;

1.3rd step, utilizes laser probe technology to detect calibration sample, gathers or intercept one section of spectrum of at least 3 characteristic spectral lines containing Ca, Mg, Al and Si to each calibration sample respectively.