JP2007187594A - Correction method of calibration curve - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a correction method of a calibration curve which fixes a reference spectrum to predict the optical path length from a sample spectrum and corrects the spectrum, using the predicted optical path length that can neglect the fluctuations due to the individual differences of vials. <P>SOLUTION: On the basis of the sample spectrum measured by the vial or cell known in its optical path length, the ratio of the specific wave number ranges of the reference spectrum, and the measured spectrum is calculated by the method of least squares to determine the measuring optical path length, while the measured spectrum is converted into a spectrum measured by using the same optical path length as the reference spectrum to form the calibration curve. The ratio of the specific wave number range of the spectrum of an unknown sample to that of the reference spectrum is calculated by the method of least squares, to deduce the optical path length; and after the optical path length has been converted into the spectrum of a reference optical path length, the concentration and properties of the sample are deduced by using a preliminarily produced calibration curve. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a calibration curve correction method for a near-infrared spectrometer, a calibration curve is created using a spectrum obtained from a sample collected in a vial, and then the concentration or property value of an unknown sample collected in a vial The present invention relates to a method for correcting the variation in the predicted value using the calibration curve due to the variation in the spectrum due to the individual difference of the vials.

In recent years, studies have been made to introduce or shift manufacturing process control and quality control to near-infrared spectroscopy instead of conventional chemical analysis methods in a wide variety of fields such as chemical, petroleum, biochemical or pharmaceutical, and even food. Yes.
Conventionally, manufacturing process management and quality control have been performed by off-line management by chemical analysis (titration method) in an analysis chamber.

  However, the quantitative analysis by the titration method has a problem that it takes time for pretreatment and requires high accuracy, so that the titrator requires skill and experience, and further uses a reagent for analysis, leading to environmental pollution.

Sample measurement of a near-infrared spectrometer using a disposable vial is simple and inexpensive, and since it is easy to transport a sample, it has the possibility of becoming the mainstream of sample measurement.
FIG. 16 is a diagram showing an example for measuring the spectrum of a sample collected in a vial using a near-infrared spectrometer.

  In the figure, reference numeral 1 denotes a measurement holder, and a vial insertion hole 1a for inserting a vial (herein, a small glass bottle into which a measurement sample is put; hereinafter simply referred to as a vial) is formed near the center of the holder. Then, a vial containing a sample is inserted into the hole to perform measurement. A preheating vial insertion hole 1b for inserting a plurality (six in the figure) of preheating vials is formed around the hole 1a.

  Reference numeral 2 denotes a temperature controller for maintaining the holder 1 at a predetermined temperature. Further, a light transmission hole 1c for transmitting near infrared light (hereinafter simply referred to as light) is formed on the side surface of the holder 1, and light incident from the hole 1c penetrates the central portion of the vial. And reaches the detector 3.

The signal photoelectrically converted by the detector 3 is sent to the signal converter 4 for A / D conversion, and sent to the data processing unit (personal computer) 5 by Ethernet or the like for spectrum display.
FIG. 16 (b) shows the shape of the vial 10 inserted into the vial insertion hole 1a. The outer diameter (D) 8, the inner diameter (d) is about 6, and the height 40 (mm) is made of Pyrex (registered trademark) glass. ) Formed with a polypropylene lid 13.

  Such a vial is for collecting a small amount of the sample 11 and conducting a sample test. It is easy to handle and only needs to be inserted into the vial insertion hole shown in FIG. 16a. It can be measured quickly and accurately.

  By the way, in such a measuring apparatus, a large number of vials are used for measuring a sample. Therefore, there is a problem that the calibration curve output fluctuates due to variations in vials. FIG. 16c is a cross-sectional view for explaining the effective optical path length of the vial. The vial shown above has an outer diameter of 8 mm and an inner diameter of about 6 mm, but the actually transmitted light has a diameter of a predetermined thickness and cannot be said to have an effective optical path length of 6 mm. . However, it is difficult to mechanically measure the effective length of the vial.

  In the present invention, fluctuations in the calibration curve caused by variations in vials are predicted by estimating the optical path length from the sample spectrum by fixing the reference spectrum, and correcting the spectrum using the predicted optical path length, thereby causing fluctuations due to individual differences in vials. An object of the present invention is to provide a calibration curve correction method that can ignore the above.

  For example, the followings are related to the instrumental error correction method using a vial.

JP 2001-41879 A

In order to achieve such a problem, the invention according to claim 1 of the present invention is:
Based on the sample spectrum measured in a vial or cell with a known optical path length,
Determine the optical path length by calculating the ratio of the specific spectrum of the reference spectrum and the measured spectrum using the least square method, convert the measured spectrum to a spectrum measured with the same optical path length as the reference spectrum, create a calibration curve, and Estimate the optical path length by calculating the ratio of the specific spectrum of the sample spectrum to the specific spectrum using the least square method, convert it to the reference optical path length spectrum, and then use the calibration curve created in advance to determine the concentration and properties of the sample. It is characterized by estimating.

In claim 2, in the calibration curve creation method according to claim 1,
The measurement optical path length obtained from the ratio of the reference spectrum and the measured spectrum is estimated, displayed, and output by adopting the reference spectrum.

The present invention has the following effects.
According to invention of Claim 1, 2,
Based on the sample spectrum measured in a vial or cell with a known optical path length,
Determine the optical path length by calculating the ratio of the specific spectrum of the reference spectrum and the measured spectrum using the least square method, convert the measured spectrum to a spectrum measured with the same optical path length as the reference spectrum, create a calibration curve, and Estimate the optical path length by calculating the ratio of the specific spectrum of the sample spectrum to the specific spectrum using the least square method, convert it to the reference optical path length spectrum, and then use the calibration curve created in advance to determine the concentration and properties of the sample. Since the estimation is made, it is possible to correct the variation in the prediction error of the calibration curve due to the individual difference of the vials, and to make a stable prediction with good reproducibility.

FIG. 1 is a flowchart of a calibration curve correction method showing an example of an embodiment of the present invention.
First, in step a, the spectrum of the sample is measured (the measurement result is stored in a memory).
Next, in step b, the sample is compared with the spectrum of a sample placed in a vial having a standard optical path length, and is correlated with the standard spectrum in a specific wave number range.

Next, in step c, the effective optical path length is calculated from the correlation obtained in step 2.
Next, in step d, normality / abnormality is determined by comparing the calculated effective optical path length with a predetermined optical path length determined in advance (if it is outside a certain threshold, it is determined to be abnormal and displayed).
Next, in step e, the optical path length of the spectrum is corrected.
Next, in step f, a calibration curve is calculated.
Next, in step g, the predicted value of the component / property is output using the calibration curve obtained in step f.

  FIG. 2 uses a near-infrared spectrometer as shown in FIG. 16, using three types (2, 5, 10 mm) of cuvettes having different optical path lengths, and a vial having an outer diameter of 8 mm, an inner diameter of about 6 mm, and a length of 40 mm. The absorbance of 100% toluene is measured. In the figure, the horizontal axis represents the wave number, and the vertical axis represents the absorbance. FIG. 2 (b) shows the shape of the cuvette, and the one with the thickness m as the optical path length changed to 2, 5, and 10 mm is used.

In FIG. 2, when attention is paid to the wave number of 5650 cm ⁻¹ , the absorbance of the 2 mm cuvette is about 0.28, and the absorbance of the 5 mm cuvette is about 0.7, which is 2.5 times that of the 2 mm cuvette. . Further, the absorbance of the 10 mm cuvette is about 1.4, which is five times that of the 2 mm cuvette, and it can be seen that there is a linear proportional relationship. And it turns out that the light absorbency of the vial with an internal diameter of about 6 mm used here is about 0.8.
The effective optical path length of the vial can be estimated from the proportional relationship between the absorbance of the cuvette and the absorbance of the vial.

In this way, the spectrum of a sample is measured using a single cuvette or vial with a known optical path length, a calibration curve is created from the absorbance in a specific wavenumber or a specific wavenumber range, and the characteristics of the sample are measured based on this calibration curve. By doing so, accurate analysis results can be obtained.
However, in actual sample measurement, since a calibration curve is created by collecting several tens to a hundred samples, it takes time to put in and take out the sample, and there is a problem that contamination is likely to occur.

  Therefore, it is only necessary to measure the spectrum in a disposable manner using cheap vials, but there are variations in the optical path length of each vial and individual differences even in the same lot, and the same sample is measured in different vials. There was a problem that the prediction results varied.

3 uses the infrared spectroscopic analysis apparatus shown in FIG. 16 and the relationship between the wave number and the absorbance used in FIG. 2, and extracts one vial from each of five types (A to E) of different lots, As shown in FIG. 3 (b), it is a result of performing a process of estimating the optical path length by measuring three times from four directions twice and examining the correlation. It can be seen that there is a correlation between the two estimations and the estimation of the optical path length is reproduced.
It can be seen that there is a correlation with the straight line i shown by the dotted line, although there are variations due to lots and individual differences.

  FIG. 4 shows 80 samples randomly extracted from each of the lots A to E, measured three times, and examined how often the optical path length varies. According to the figure, it can be seen that there is a center at 6.48 mm and there is a variation of ± about 0.2 mm.

  FIG. 5 shows a spectrum change measured by using 7% isopropyl alcohol in toluene and the near-infrared spectroscopic analyzer shown in FIG. 'The part is enlarged. It can be seen that the absorbance of the spectrum changes according to the concentration of isopropyl alcohol (FIG. 5 corresponds to the spectrum measurement in step (a) of the flowchart shown in FIG. 1).

FIG. 6 shows an example in which a calibration curve was created by measuring each vial three times using lot A, where the horizontal axis represents the reference IPA concentration and the vertical axis represents the predicted IPA concentration.
The reference IPA concentration is% by weight and is calculated from the weight ratio of the solvent (toluene) and IPA.
FIG. 7 shows the predicted results of IPA concentration in vials containing other lots using the calibration curve shown in FIG. Compared to FIG. 6, it can be seen that there is a variation in the prediction result due to the individual vial difference.

  FIGS. 8A to 8D are diagrams showing the calibration curve prediction result and the variation between lots when pre-processing is performed and when pre-processing is not performed when creating a calibration curve. The lot A was used as the vial used for preparing the calibration curve.

  FIG. 8A is a diagram showing the degree of variation when the SEP is 0.0929 without preprocessing. The horizontal axis represents the variation of the predicted value using the calibration curve with respect to the laboratory value of isopropyl alcohol concentration (%), and the vertical axis represents the frequency (number of times measured). Here, SEP (Standard Error Prediction) is the standard deviation of the predicted sample concentration and laboratory value.

  According to Fig.8 (a), the dispersion | variation was large and the standard deviation was 0.0783. Here, the standard deviation is a value obtained by taking an average value of predicted values of each sample and taking a standard deviation of data obtained by subtracting the average value from each predicted value.

FIG. 8B is a diagram showing the degree of variation when the second differentiation is performed as the preprocessing and the SEP is 0.0929. The standard deviation in this case was 0.0727.
FIG. 8 (c) shows the degree of variation when the MSP (Multi-Scattering-Correction ... multiple scattering correction ... hereinafter simply referred to as MSC) is applied as pre-processing, and the SEP is 0.1143. FIG. The standard deviation in this case was 0.0317.

  FIG. 8D is a diagram in which second order differentiation and MSC are performed as preprocessing, and the degree of variation when SEP is 0.0602. The standard deviation in this case was 0.0331. From these figures, it can be seen that the variation obtained by performing the second order differentiation and the MSC as the preprocessing is reduced.

  9 and 10 show the effect of the prediction result including the vials of other lots by the second-order differential calibration curve. FIG. 9 shows the state after the second-order differentiation of FIG. 8B, and FIG. The state which performed differentiation and MSC is shown. The horizontal axis represents the reference IPA concentration, and the vertical axis represents the predicted IPA concentration. According to these figures, it can be seen that the MSC (FIG. 10) has a small variation.

  FIG. 11A shows the relationship between the absorbance of the standard spectrum and the absorbance ratio of each sample, and uses a wave number range that varies greatly depending on the concentration of IPA as shown in FIG. 11B. In FIG. 11B, the horizontal axis indicates the wave number, and the vertical axis indicates the absorbance.

FIG. 12A shows the relationship between the absorbance of the standard spectrum and the absorbance ratio of each sample spectrum, as in FIG.
Comparing FIG. 11 and FIG. 12, it can be seen that FIG. 12 has less variation and good correlation.
Linear equation shown in FIG.
y = 0.9902x + 0.0043
r (correlation coefficient) = 0.9994
And the linear equation shown in FIG.
y = 0.9986x + 0.0033
r (correlation coefficient) = 0.9998
Is the correlation between the sample spectrum and the standard spectrum obtained using the least square method, and the value corresponding to a in y = ax + b represents the optical path ratio between the standard spectrum and the sample spectrum.

  As shown in FIG. 12B, a wave number range in which the change due to the concentration of IPA is small is used here. Compared to FIG. 11, it can be seen that the variation is small and the correlation is good (FIG. 12 corresponds to the step of correlating with the standard spectrum measurement in step (b) of the flowchart shown in FIG. 1).

  From the comparison between FIG. 11 and FIG. 12, when the optical path length is predicted using MSC, the stability of the prediction is better when the spectral range that is not affected by the sample concentration is used. In MSC, the average of the sample spectrum at the time of preparing the calibration curve is used as a standard spectrum.

For this reason, the average spectrum changes every time a calibration curve is created, and there is a drawback that the estimated optical path length is not continuous.
Therefore, a modified MSC method is proposed in which the path length correction amount is calculated based on a sample spectrum whose path length is known in advance.

  FIG. 13 shows the correlation between the optical path length estimated on the basis of a sample spectrum containing 3.283% IPA in toluene contained in a vial having a known optical path length of 6.526 mm and the optical path length estimated from the pure toluene spectrum shown in FIG. It is shown.

FIG. 13 shows the correlation between the optical path lengths obtained by the two methods. The horizontal axis represents the predicted optical path length (mm) obtained from the spectrum varying with the IPA concentration, and the vertical axis represents the predicted optical path length (mm) obtained from the toluene spectrum. ).
FIG. 13 shows the correlation between the optical path length obtained from the sample spectrum obtained by changing the IPA concentration obtained using the method of least squares and the optical path length obtained from the toluene spectrum, and r = 0.912 represents the correlation coefficient. Yes.
The result of FIG. 13 shows that the optical path length can be predicted from the sample spectrum without using toluene.

  FIG. 14 shows the result of correcting the spectrum using the optical path length obtained from the sample spectrum and showing the predicted result. Using the spectrum after optical path length correction, a secondary differential calibration curve was created, and other lot vials were prepared. It is the result of having predicted the optical path length correction spectrum including. It can be seen that the SEP (prediction error) is 0.056 and the variation among lots is greatly reduced.

  FIG. 15 is a diagram showing the effect of the modified MSC. That is, the line indicated by (A) in the upper part of FIG. 15A indicates the optical path length predicted from the sample spectrum, and the lines indicated by (B) and (C) in FIG. 15A use the modified MSC. The comparison of the dispersion | variation in the IPA prediction value when not using is shown. The line indicated by (B) is the predicted value of the IPA concentration without optical path length correction, and the line indicated by (C) is the IPA (isopropyl alcohol) prediction that is applied to the calibration curve after correcting the spectrum using the effective optical path length. It shows the variation when the value is calculated. It can be seen that the variation is reduced.

  FIG. 15B is a frequency table (histogram) for comparing the variation of the calibration curve predicted values. When the modified MSC is used, the variation is less than half compared to the case where it is not used (σ = 0.051). → 0.025) (FIG. 15A corresponds to step (g) in the flowchart shown in FIG. 1).

  The above description merely shows a specific preferred embodiment for the purpose of explanation and illustration of the present invention. Therefore, in the present embodiment, the present invention is not limited to the above embodiment, and includes many changes and modifications without departing from the essence thereof.

It is a flowchart of the calibration curve correction method of the present invention. It is a figure which shows the toluene spectrum and optical path length measurement calibration straight line by different optical path lengths. It is a figure which shows the result of having extracted one vial each from the vial from which a lot differs, and measuring the correlation of the estimated optical path length twice by measuring twice each from 4 directions and estimating the optical path length twice. It is a figure which shows the relationship between effective optical path length and dispersion | variation. It is a figure which shows the result of having put 7% isopropyl alcohol in toluene and measuring a spectrum change. It is a figure which shows an example which measured the spectrum in a vial 3 times using lot A, and created the calibration curve. It is a figure which shows the prediction result of the IPA density | concentration in the vial containing the other lot using the calibration curve shown in FIG. 8 is a diagram showing a calibration curve prediction result when pre-processing is performed and when pre-processing is not performed when creating a calibration curve. It is a figure which shows the prediction result of the sample containing the vial of another lot by a secondary differential calibration curve. It is a figure which shows the state which gave the secondary differentiation and MSC. It is a figure which shows the relationship between the light absorbency of a standard spectrum, and the light absorbency ratio of each sample. It is a figure which shows the relationship between the light absorbency of a standard spectrum, and the light absorbency ratio of each sample spectrum. It is a figure which shows the correlation of the optical path length calculated | required by two methods. It is a figure which shows the result of having corrected a spectrum using the optical path length calculated | required from the sample spectrum, and having estimated. It is a figure which shows the comparison result of the dispersion | variation in the measured value when not estimating with the estimated value of the optical path length using a deformation | transformation MSC method, and the correction | amendment using a deformation | transformation MSC method. It is a figure which shows an example for measuring the spectrum of a sample extract | collected to the vial bottle using a near-infrared spectrometer.

Explanation of symbols

1 Measurement Holder 2 Temperature Controller 3 Detector 4 Signal Converter 5 Personal Computer 10 Vial Bottle 11 Sample

Claims (2)

Based on the sample spectrum measured in a vial or cell with a known optical path length,
Determine the optical path length by calculating the ratio of the specific spectrum of the reference spectrum and the measured spectrum using the least square method, convert the measured spectrum to a spectrum measured with the same optical path length as the reference spectrum, create a calibration curve, and Estimate the optical path length by calculating the ratio of the specific spectrum of the sample spectrum to the specific spectrum using the least square method, convert it to the reference optical path length spectrum, and then use the calibration curve created in advance to determine the concentration and properties of the sample. A calibration curve correction method characterized by estimating. 2. The calibration curve correcting method according to claim 1, wherein the reference optical spectrum is adopted, the measurement optical path length obtained from the ratio of the reference spectrum and the measurement spectrum is estimated, displayed, and output.

Images