JPH05133849A - Time series correcting method of analysis value - Google Patents

Abstract

PURPOSE:To ensure the analytical accuracy of a continuous analysis by correcting the time series drift of the analysis value of a component based on a specific expression with taking note of the fact that the standard values of standard samples interposed in the regular interval are known. CONSTITUTION:A high (low) concentration standard sample is denoted by H(L), and the known concentration thereof is YH0 (YL0). A unit of lots as a subject of the time series correction includes from a pair of analyzing values (YH1, YL1) of the initial standard sample to a pair of analyzing values (YH2, YL2) of a next standard sample, with the analyzing values of unknown samples halfway between the initial and the next samples. The time series correction is repeated for every unit of lots, so that the analyzing values of all the samples are corrected based on the expression. In the expression, (a) and (b) are an inclination and an intercept when the standard value Y and the analyzing value Y' of the standard sample of the initial lot are represented by a linear expression Y=aY'+b; (c) and (d) are an inclination and an intercept when the standard value Y and the analyzing value Y' of the standard sample in the final lot are represented by a linear expression Y=cY'+d; N is the whole number of measuring times of the lot; and (n) is the measuring order in the lot.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a time series correction method for analysis values in component analysis, which improves the accuracy of analysis and the number of continuous analysis by correcting a time series drift occurring in a device for component analysis. The present invention relates to a correction method capable of achieving an improvement in work efficiency by enlarging the.

[0002]

2. Description of the Related Art In the prior art, a calibration curve is created from the relationship between the measurement intensity of a standard sample and a known standard value at the beginning of measurement, and the measurement intensity of an unknown sample is substituted into this calibration curve to obtain an analytical value. It was

[0003]

[Problems to be Solved by the Invention] With the conventional method of obtaining an analysis value from a calibration curve in the initial stage of measurement in component analysis,
Since the time-series drift during the measurement that occurs due to sample clogging of the measurement circuit system, fogging of the optical lens, temperature change of the device, etc. deteriorates the analysis accuracy, the analysis lot should be reduced to reduce the measurement time. It had to be shortened or repeated from the preparation of the calibration curve to the measurement again, which was a serious impediment factor in improving analysis accuracy and work efficiency. An object of the present invention is to solve such a problem, correct the time series drift of the analysis value, and guarantee the analysis accuracy by the continuous analysis.

[0004]

The present invention focuses on the fact that the standard values of standard samples sandwiched at regular intervals are known in the component analysis, and the time series drift of the component analysis values is expressed by the following equation. A method for correcting an analysis value time series, which is characterized in that the correction is performed based on the analysis value to improve the analysis accuracy. Corrected analysis value (Y _n ) = (a × Y ′ + b) + {(c−a) × Y ′ + (d−b)} × n / N (1) where a and b are the beginning of the analysis lot The slope a and the intercept b when the standard value Y and the analysis value Y ′ of the standard sample are represented by the linear expression Y = a × Y ′ + b. c, d: slope c and intercept d when the standard value Y and the analytical value Y ′ of the standard sample at the end of the analysis lot are represented by the linear expression Y = cY ′ + d. N: Total number of measurements in analysis lot n: Measurement order in analysis lot

The present invention will be described below with reference to the drawings.
FIG. 1 is a diagram showing a time series drift and a method of sandwiching a standard sample. The time transition is shown by taking the concentration on the vertical axis and the number of measurement samples on the horizontal axis. The high-concentration standard sample is H, and its known (true) concentration is Y _H0 . The low-concentration standard sample is L, and its known (true) concentration is Y _L0 .
The unit of lot for time series correction is the analytical value of the next standard sample pair (the analytical value of the first standard sample pair (Y _H1 , Y _L1 ) on the left side of the figure with the analytical value of the unknown sample in between). Y _H2 , Y _L2 ). The time series correction is repeated for each analysis lot to correct the analysis values of all the samples.

The cause of the time-series drift is mainly due to the clogging of the sample, fluctuations in the sample amount due to the generation of oxides and salts, fluctuations in the measurement system due to clouding of the optical lens and temperature changes. Occur. The occurrence of this drift has a great influence on the deterioration of the analysis accuracy, which is a serious obstacle particularly to performing unattended analysis continuously for a long time.

In the case of the example of FIG. 1, the standard value of the high-concentration standard sample is Y _{HO when} Al: 100 ppm is 97 ppm (Y _H1 ) in the measurement at the beginning of the analysis lot, and thereafter, the sensitivity decreases as the measurement is repeated. A time-series change occurred, and it was reduced to 90 ppm (Y _H2 ) in the last measurement of the analysis lot. Similarly, the low-concentration standard sample L has a standard value (Y _L0 ) of 50.
ppm is 48 ppm (Y _L1 ) in the first measurement, and is reduced to 40 ppm (Y _L2 ) in the last measurement. Therefore, paying attention to the fact that the standard value of the standard sample is known, quantify the effect of time series baseline drift on the analytical value from the change in the analytical value of the standard sample, and remove the effect of the drift. It was investigated.

FIG. 2 is a diagram showing changes when a standard sample is analyzed in time series. The analytical concentration is plotted on the vertical axis and the known (true) concentration of the standard sample is plotted on the horizontal axis. If the analysis concentration of the standard sample of the analysis lot to be time-series corrected is a high concentration sample,
Originally Y _H0 indicates Y _H1 and changes to Y _H2 at the end of the lot. In the case of a low-concentration sample, Y _L0 originally shows Y _L1 and changes to Y _L2 at the end of the lot. At this time, the first Y _H0 ,
The relational expression of Y _H1 , Y _L0 , and Y _L1 is _calculated from the known (true) concentration Y,
When the analyzed concentration is Y ', it can be expressed as Y = aY' + b (2).

Next, Y _H0 , Y _H2 , Y at the end of the analysis lot
The relational expression of _L0 and Y _L2 is similarly obtained. Y = cY '+ d (3) The relationship when changing from this equation (2) to equation (3) is
Assuming that the slope a changes by δα and the intercept b changes by δβ every time one sample is analyzed, all samples (N
Since the slope has changed to c and the intercept has changed to d at the time of analysis, they are related by the following equation. c = a + N × δα (4) d = b + N × δβ (5) Therefore, the true analysis value (Y _n ) of the unknown sample (nth sample) in the middle of the analysis lot is Y _n = (a + n × δα) × Y ′ + (b + n × δβ) (6) Substituting the equations (3) and (4), Y _n = (a × Y ′ + b) + {(c−a) × Y ′ + (d− b)} × n / N (7) is obtained.

Here, each coefficient of a, b, c and d is calculated by the following equations (8) to (11). _{a = (Y H0 -Y L0)} / (Y H1 -Y L1) (8) b = (Y L0 × Y H1 -Y H0 × Y L1) / (Y H1 -Y L1) (9) c = (Y _{H0 -Y L0) / (Y H2} -Y L2) (10) d = (Y L0 × Y H2 -Y H0 × Y L2) / (Y H2 -Y L2) (11)

In the case of the example of FIG. 2, Y _H0 , Y _H1 , and
The relation of Y _{L1 is} as follows: First, when the coefficients a and b are obtained, a = (Y _H0 −Y _LO ) / (Y _H1 −Y _L1 ) = (100−50) / (97−48) = 1.020408 b = _{(Y LO × Y H1 -Y H0} × Y L1)) / (Y H1 -Y L1) = ( get 50 × 97-100 × 48) / ( 97-48) = 1.020408.

From this, the relationship between the true density Y and the analyzed density is Y = aY '+ b = 1.020408 × Y' + 1.020408 (12)

For example, by substituting Y _H1 = 97 ppm into the equation (12), the true value Y _H0 : 100 ppm can be obtained. Similarly, the relationship of Y _H0 , Y _LO , Y _H2 , and Y _{L2 at} the end of measurement is as follows: c = (Y _H0 −Y _LO ) / (Y _H2 −Y _L2 ) = (100−50) / (90−40) = 1.0000000 d = (Y _LO × Y _H2- Y _H0 × Y _L2 )) / (Y _H2- Y _L2 ) = (50 × 90-100 × 40) / (90-40) = 10.000000 As a result, Y = cY ′ + d = 1.000000 × Y ′ + 10.000000 (13) is obtained. By substituting Y _L2 : 40 ppm into the equation (13), the true value Y _H0 : 50 ppm is obtained.

Next, when 10 unknown samples are measured on the way, the total number of measurements is N = 12, including the number of times of measurement of the standard sample.
And Therefore, the formula for obtaining the _n - _th true value is: Y _n = (a × Y ′ + b) + {(c−a) × Y ′ + (d−b)} × n / N = (1.020408 × Y '+ 1.020408) + {(1.00000 0-1.020408) * Y' + (1.00000-1.020408)} * n / 12 = (1.020408xY '+ 1.020408) + (0 0.020408 × Y ′ + 8.979592) × n / 12 (14) is obtained.

Table 1 shows an example in which a standard value 70 ppm Al solution was repeatedly analyzed 10 times and subjected to time series correction. Items in the table are, from the left, the number of measurements n, standard value Y ₀ = 70 ppm,
An uncorrected analysis value Y ′, a corrected analysis value Y, a difference between the standard value and the uncorrected analysis value: Y ₀ −Y ′, and a difference between the standard value and the corrected analysis value: Y ₀ −Y are shown. In this example, when the analysis value before correction and the analysis value after correction are compared, Y'has a maximum difference of 9 from the standard value.
Although ppm is generated, it is ± of the standard value after correction.
The analysis accuracy was improved because it was within 0.5ppm. Although the present correction method takes Al analysis as an example in the embodiment, the present method can be applied to all other component analyses.

[0016]

[Table 1]

[0017]

EFFECTS OF THE INVENTION With the conventional method of obtaining an analysis value from a calibration curve at the initial stage of measurement in the component analysis, the measurement circuit system may be clogged with a sample, the optical lens may be fogged, or the temperature of the device may change during measurement. In order to reduce the analysis accuracy due to the drift in time series, it was necessary to reduce the analysis lot to shorten the measurement time, and to repeat the procedure from calibration curve creation to measurement again. It is possible to improve the analysis accuracy and work efficiency by correcting the.

[Brief description of drawings]

FIG. 1 is a diagram showing a time series drift and a method of sandwiching a standard sample.

FIG. 2 is a diagram showing changes when a standard sample is analyzed in time series.

Claims (1)

[Claims] 1. In the component analysis, focusing on the fact that the standard values of standard samples sandwiched at regular intervals are known, the time series drift of the component analysis values is corrected based on the following formula to improve the analysis accuracy. A method for correcting an analysis value time series, characterized by: Corrected analysis value (Y _n ) = (a × Y ′ + b) + {(c−a) × Y ′ + (d−b)} × n / N (1) where a and b are the beginning of the analysis lot The slope a and the intercept b when the standard value Y and the analytical value Y ′ of the standard sample are represented by the linear expression Y = a × Y ′ + b. c, d: Slope c when the standard value Y of the standard sample and the analysis value Y ′ at the end of the analysis lot are represented by the linear expression Y = cY ′ + d
And intercept d. N: Total number of measurements in analysis lot n: Measurement order in analysis lot