Based on the multivariable technique of angle tolerance
Technical field
The present invention relates to the multivariable technique based on angle tolerance of a kind of method for carrying out noncontact analysis to product, particularly a kind of applicable multicomponent system.
Background technology
The development of metering system is divided into double teacher by process analysis technique: off-line(off-line analysis), at-line(near line analyze), on-line(on-line analysis, on-line analysis), in-line(in-line analysis) and Noninvasive(noncontact analyze), wherein on-line is divided into again intermittent(to be interrupted) and continuous(continuous).The in-line stage refers to measured matter to be taken out and carry out pre-service, noncontact is analyzed and then is more difficult to process measured object and operate when measuring, even usually, the reproducibility analysis environments required by analyzing all may be difficult to meet, and therefore these stages propose very high requirement to analytical approach.Common analysis to environment can reproduction requirements very strict, for photometric analysis, its quantitative basis is langbobier law, requires to stablize through light path, and this is very harsh requirement for noncontact analysis; In recent years, reflection or scatterometry are paid attention in noncontact is analyzed, and compared with transmission measurement, flexible operation degree adds, but environmental impact factor is more, need the sample units such as configuration integrating sphere, add the complexity of instrument; Laser raman is the important means interested to non-contact analysis, but it is quantitatively difficult that the factors such as the variations in refractive index that light source power change, concentration of specimens cause cause directly comparing Raman spectrum band strength, it has been generally acknowledged that adding interior mark is most effective method, but the essence that internal standard method operation and noncontact are analyzed is contradiction.That is, adopt the quantivative approach of band intensity to encounter very large obstacle in noncontact is analyzed, be necessary to propose a kind of new quantivative approach, to adapt to the growth requirement of analytical technology.
Summary of the invention
The technical problem to be solved in the present invention is: provide a kind of adapt to multicomponent system based on angle tolerance multivariable technique, with overcome prior art exist increase instrument complexity, by the larger weak point of such environmental effects.
The ultimate principle that the present invention solves the problems of the technologies described above: each component based on multicomponent system may correspond to as single vector-quantities in space, system is the space that multiple vector is formed.Space also can be described as tested component vector a and not containing the lineoid B of this vector, in system, a tested sample s comprising tested component and other components can be expressed as a vector between tested component vector sum lineoid, as shown in Figure 1.
The plane that a with s opens and B meet at b, as Fig. 2, and the response intensity of the length representative measured object of vector or content; If the content of measured object a is lower in system, the angle theta of tested sample s and a is larger, otherwise less.Therefore, when space can be determined to describe, angle theta can be used as quantitative target.
The angle of vector can use dot product formulae discovery:
,
The matrix description system that current multivariate statistical analysis adopts intensity level to form, when rank of matrix exceedes the space dimensionality of system, system just can be determined to describe.The determination that the matrix that the present invention proposes to adopt angle value to form comes implementation space describes, and when the rank of matrix that angle value is formed exceedes the dimension in system space, the description of this matrix to space is also determined.
Fig. 3 is the spectrum that the intensity level of m wavelength points is formed, and is divided into n interval, if system contains p kind component, to the requirement of m and n is:
,
So, vectorial a and s is split into { a
1..., a
i..., a
nand { s
1..., s
i..., s
n, the angle between each vector:
,
M the intensity level of a and s is converted to n angle value { θ
1..., θ
i..., θ
n.
The technical scheme that the present invention solves the problems of the technologies described above is: a kind of multivariable technique based on angle tolerance, it is characterized in that: the method is qualitative or quantitative apparatus measures tested sample and tested component by material, obtain the multiple spot intensity measurements of tested sample and tested component; And the multiple spot intensity measurements of tested sample and tested component is converted into the angular measure of tested sample and tested component; Choose several modeling samples that quantity is greater than mixed system number of components, in modeling sample tested component in the relative content of whole mixed system be known, the multiple spot intensity measurements of modeling sample and tested component is converted into the angular measure of modeling sample and tested component, then using the tested component concentration in modeling sample as dependent variable, the matrix that the angular measure of modeling sample and tested component is formed, as independent variable, sets up multivariate regression models; The angular measure of tested sample and tested component is substituted into this multivariate regression models, the content of tested component in the whole mixed system of tested sample can be predicted.
Further technical scheme of the present invention is: the method comprises the following steps:
SA1, determine tested sample s and tested component a, and definition cutting interval number n, wherein n > mixed system number of components p;
SA2, by material, the measuring unit of qualitative or quantitative instrument measures tested component a and tested sample s, obtains the multiple spot intensity measurements of tested component a and the multiple spot intensity measurements of tested sample s;
SA3, measurement point is divided into n interval, then the multiple spot intensity measurements of tested component a is split into { a
1..., a
i..., a
n; The multiple spot intensity measurements of tested sample s is split into { s
1..., s
i..., s
n;
SA4, ask for included angle cosine between tested sample s and each tested component a:
;
SA5, ask for the angular measure { θ of tested sample s and each tested component a
1..., θ
i..., θ
n;
B, modeling:
SB1, choose several modeling samples S that quantity is greater than mixed system number of components p, in modeling sample S tested component a in the relative content of whole mixed system be known;
SB2, by material, the measuring unit of qualitative or quantitative instrument measures modeling sample S, obtains the multiple spot intensity measurements of modeling sample S;
SB3, the intensity value of modeling sample S and tested component a is transferred to the angular measure of modeling sample S and tested component a;
SB4, using the content of tested component a in modeling sample S as dependent variable, the matrix formed by the angular measure of modeling sample S and tested component a, as independent variable, adopts multivariate statistical regression method establishment multivariate regression models;
C, prediction:
The angular measure of tested component a and tested sample s is substituted into this multivariate regression models, the content of tested component a in the whole mixed system of tested sample s can be predicted, predict the outcome from the analysis result display unit of the qualitative or quantitative instrument of material and export, operation terminates.
Owing to adopting said structure, the multivariable technique based on angle tolerance of the present invention compared with prior art, has following beneficial effect:
1, the requirement to environment in analysis operation is significantly reduced:
The tolerance conversion method proposed due to the present invention avoids because response intensity change causes the infeasible problem of multivariable analysis, as spectral analysis, as long as the spectral shape obtained is consistent, and due to external factor, the distance of such as measuring distance, the change etc. of the intensity of light source, the Strength Changes caused can not have an impact to multivariate regression models accurately, therefore, present invention significantly reduces the requirement to environment in analysis operation, for completing, contactless on-the site analysis and on-line analysis etc. are significant.
2, can the complexity of lowering apparatus:
Due to the present invention in use, without the need to increasing other operating instrument in addition, also without the need to configuring the sample units such as integrating sphere, thus can the complexity of lowering apparatus.
3, operating process is simple:
Operating process of the present invention is fairly simple, is easy to use when relating to apparatus measures strength fluctuation in chemical analysis, process analysis procedure analysis, instrumental analysis.
Below, in conjunction with the accompanying drawings and embodiments the technical characteristic of the multivariable technique based on angle tolerance of the present invention is further described.
Accompanying drawing explanation
Fig. 1: tested component a in multicomponent system, not containing a lineoid B and comprise tested component a and other components tested sample s between graph of a relation;
Fig. 2: tested component a and graph of a relation between tested sample s and angle theta;
The spectrogram of the intensity level formation of Fig. 3: m wavelength points;
Fig. 4: the FB(flow block) of the multivariable technique steps A based on angle tolerance of the present invention,
Fig. 5: the FB(flow block) of the multivariable technique step B based on angle tolerance of the present invention,
Fig. 6: the Raman spectrogram measured described in embodiment one;
Fig. 7: the model working curve diagram adopting modeling gained of the present invention described in embodiment one,
Horizontal ordinate in Fig. 7 is actual value, and ordinate is predicted value;
Fig. 8: the model working curve diagram directly adopting Raman spectrogram modeling gained described in embodiment one,
Horizontal ordinate in Fig. 8 is actual value, and ordinate is predicted value.
Embodiment
Embodiment one:
Based on a multivariable technique for angle tolerance, the method is qualitative or quantitative apparatus measures tested sample and tested component by material, obtains the multiple spot intensity measurements of tested sample and tested component; And the multiple spot intensity measurements of tested sample and tested component is converted into the angular measure of tested sample and tested component; Choose several modeling samples that quantity is greater than mixed system number of components, in modeling sample tested component in the relative content of whole mixed system be known, the multiple spot intensity measurements of modeling sample and tested component is converted into the angular measure of modeling sample and tested component, then using the tested component concentration in modeling sample as dependent variable, the matrix that the angular measure of modeling sample and tested component is formed, as independent variable, sets up multivariate regression models; The angular measure of tested sample and tested component is substituted into this multivariate regression models, the content of tested component in the whole mixed system of tested sample can be predicted.
The above-mentioned multivariable technique based on angle tolerance is as follows for the concrete steps analyzing ethylene glycol content in ethylene glycol, propylene glycol, alcohol mixture:
The intensity of A, tested sample and tested component is changed to the tolerance of angle, FB(flow block) as shown in Figure 4:
SA1, determine tested sample s and tested component a, and definition cutting interval number n, wherein n > mixed system number of components p:
By ethylene glycol, propylene glycol, ethanol three kinds of components, according to tested component a---ethylene glycol, from 0 to 5.0mL, according to the potpourri of 0.5mL interval preparation cutting interval number n=11 content, with propylene glycol and any proportioning constant volume of ethanol to 10mL, obtains tested sample s;
SA2, by material, the measuring unit of qualitative or quantitative instrument measures tested component a and tested sample s, obtains the multiple spot intensity measurements of tested component a and the multiple spot intensity measurements of tested sample s:
By spectrometer 785nm laser excitation tested sample s, Raman spectrogram is measured by the measuring unit of spectrometer, not fixation measuring distance during mensuration, be as the criterion so that Raman signal can be detected, obtain the multiple spot intensity measurements of tested component ethylene glycol and the multiple spot intensity measurements of tested sample s, the spectrogram of gained as shown in Figure 6;
SA3, measurement point is divided into n interval, then the multiple spot intensity measurements of tested component a is split into { a
1..., a
i..., a
n; The multiple spot intensity measurements of tested sample s is split into { s
1..., s
i..., s
n;
SA3, ask for included angle cosine between tested sample s and each tested component a:
;
SA4, ask for the angular measure { θ of tested sample s and each tested component a
1..., θ
i..., θ
n;
B, modeling, FB(flow block) as shown in Figure 5:
SB1, choose several modeling samples S that quantity is greater than mixed system number of components p, in modeling sample S tested component a in the relative content of whole mixed system be known;
SB2, by material, the measuring unit of qualitative or quantitative instrument measures modeling sample S, obtains the multiple spot intensity measurements of modeling sample S;
SB3, the intensity value of modeling sample S and tested component a is transferred to the angular measure of modeling sample S and tested component a;
SB4, in modeling sample S the content of the serial volume of tested component a ethylene glycol as dependent variable, the matrix formed by the angular measure of modeling sample S and tested component a, as independent variable, adopts the multivariate statistical regression method establishment multivariate regression models of partial least squares regression;
C, prediction:
The angular measure of tested component a and tested sample s is substituted into this multivariate regression models, gained model working curve diagram as shown in Figure 7, the content of tested component a in the whole mixed system of tested sample s can be predicted, predict the outcome from the analysis result display unit of the qualitative or quantitative instrument of material and export, operation terminates.
In order to the multivariable technique based on angle tolerance of the present invention is contrasted with adopting existing Raman spectrogram modeling and analysis methods, the Raman spectrogram of inventor also with reference to the accompanying drawings shown in 6 carries out the modeling of direct Raman spectrogram, the model working curve diagram of gained as shown in Figure 8, as can be seen from this Fig. 8, the modeling of direct Raman spectrogram can reflect the variation tendency of content, but corresponding relation is poor, accurate quantitative analysis can not be realized; And in Fig. 7, adopt angle of the present invention to measure modeling, show good corresponding relation, the requirement of accurate quantitative analysis can have been met.
One as the present embodiment one converts, and the multivariable technique based on angle tolerance of the present invention also can be used for the tested component concentration analyzed in other potpourri.