KR102178443B1 - Method for quantification analysis of iron oxides on steel surface using glow discharge optical emission spectrometry with multivariate analysis - Google Patents

Abstract

The present invention is a first step of preparing two or more steel samples, and obtaining a learning label by quantitatively analyzing the surface iron oxide of each of the steel samples; A second step of obtaining learning data by measuring the Fe element concentration profile in the depth direction from the surface of the base metal including the surface oxide layer for each of the steel samples; A third step of establishing a prediction model equation representing a correlation between the amount of iron oxide on the surface of the steel sample and the concentration profile of the Fe element using a multivariate analysis based on the learning label and the learning data; And a fourth step of calculating a predicted value of the amount of surface iron oxide by measuring the Fe element concentration profile in the depth direction from the surface of the steel to be measured, and substituting the measured Fe element concentration profile into the prediction model equation. It provides a quantitative analysis method of iron oxide on the steel surface comprising a.

Description

Quantitative analysis method of iron oxide on steel surface using multivariate analysis of glow discharge optical emission spectroscopy {METHOD FOR QUANTIFICATION ANALYSIS OF IRON OXIDES ON STEEL SURFACE USING GLOW DISCHARGE OPTICAL EMISSION SPECTROMETRY WITH MULTIVARIATE ANALYSIS}

The present invention relates to a method for quantitative analysis of iron oxide on a steel surface capable of quickly monitoring the concentration of oxide on the steel surface in an iron making process.

It is very important to quantitatively evaluate the concentration of steel surface oxides in the steel making process. In general, the concentration of steel surface oxide is quantitatively analyzed by selectively extracting the steel surface oxide by electrochemical method and analyzing the extracted components. More specifically, in the conventional selective extraction method, a non-aqueous electrolyte is usually used so that the steel surface is not oxidized in the extraction process. In the electrolyte, metallic iron (Fe), which is a steel medium, is oxidized and ionized in the electrolyte, and Fe- O-based iron oxide is not ionized by the electrolyte and falls out as a lump. At this time, a constant charge can be applied to accelerate the reaction, and the Fe-O-based iron oxide mass that has been separated can be easily separated from the iron ions in the electrolyte through filtration. The separated iron oxide is decomposed through a chemical pretreatment method and then quantitatively analyzed for Fe using ICP-OES, and the result is converted into oxides to quantitatively evaluate the concentration.

However, this conventional method requires extraction and pre-treatment, which takes a lot of time, and cannot be analyzed quickly, and thus, there is a problem in that it is not possible to quickly feedback the concentration result required for the process process. Therefore, there is a need to develop a method that can quickly monitor oxides on the steel surface.

An object of the present invention is to provide a method for rapidly and accurately quantitatively analyzing oxides such as iron oxide on the surface of steel.

The subject of the present invention is not limited to the above description. Those of ordinary skill in the art to which the present invention pertains will not have any difficulty in understanding the additional subject of the present invention from the general details of the present specification.

An aspect of the present invention is a first step of preparing two or more steel samples, and obtaining a learning label by quantitatively analyzing the surface iron oxide of each of the steel samples; A second step of obtaining learning data by measuring the Fe element concentration profile in the depth direction from the surface of the base metal including the surface oxide layer for each of the steel samples; A third step of establishing a prediction model equation representing a correlation between the amount of iron oxide on the surface of the steel sample and the concentration profile of the Fe element using a multivariate analysis based on the learning label and the learning data; And a fourth step of calculating a predicted value of the amount of surface iron oxide by measuring the Fe element concentration profile in the depth direction from the surface of the steel to be measured, and substituting the measured Fe element concentration profile into the prediction model equation. It is a quantitative analysis method of iron oxide on the surface of the steel comprising a.

The predictive model formula may have a MAPE value of 15% or less represented by Equation 2 below.

[Equation 2]

(Here, y _{reference value} is a quantitative analysis value by a conventional extraction method, y _{predicted value} is a predicted value calculated by the above prediction model equation, and n is the number of samples used for analysis.)

Quantitative analysis in the first step may be performed by selectively extracting the steel surface oxide by an electrochemical method and analyzing the extracted components.

The depth-direction Fe element concentration profile in the second step or the fourth step can be analyzed using GD-OES.

The multivariate analysis may be a PLS-R model.

The present invention has an advantageous effect that it is possible to quickly and quickly analyze the amount of iron oxide on the steel surface even during a process process by obtaining a predictive model equation in advance using the quantitative analysis value of the surface oxide of the steel sample and the Fe element concentration profile in the depth direction.

The various and beneficial advantages and effects of the present invention are not limited to the above description, and will be more easily understood in the course of describing specific embodiments of the present invention.

1 is a flow chart of a method for quantitative analysis of iron oxide on a steel surface according to the present invention.
2 is a schematic diagram of an apparatus for electrolyzing a steel sample in order to quantitatively analyze a surface oxide.
3 is a result of analyzing the concentration change of iron (Fe) and oxygen (O) from the surface of the base metal including the oxide layer of the steel sample in the depth direction using GD-OES.
4 is a graph showing reference values and predicted values of steel samples used in an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described. However, embodiments of the present invention may be modified in various other forms, and the scope of the present invention is not limited to the embodiments described below. In addition, embodiments of the present invention are provided in order to more completely describe the present invention to those with average knowledge in the art.

The present invention analyzes the concentration distribution of elements in the depth direction of the oxide layer on the steel surface and the base material using an analysis tool capable of analyzing the composition in the depth direction such as GD-OES (Partial Least Square Regression) This is a method of quantitatively analyzing the amount of iron oxide in the oxide layer on the steel surface using multivariate analysis such as.

In more detail, by calculating the correlation by multivariate analysis of the element distribution profile of the surface including the oxide layer on the steel surface and the result of quantitative analysis by extracting iron oxide on the surface, a prediction model equation is obtained, and the actual process using the prediction model equation It is characterized in that a quantitative result of the amount of surface iron oxide can be obtained even during the process.

1 is a flow chart of a method for quantitative analysis of iron oxide on a steel surface according to an aspect of the present invention. Referring to FIG. 1, the analysis method according to the present invention includes a first step of preparing two or more steel samples and quantitatively analyzing the surface iron oxide of each of the steel samples to obtain a learning label; A second step of obtaining learning data by measuring the Fe element concentration profile in the depth direction from the surface of the base metal including the surface oxide layer for each of the steel samples; A third step of establishing a prediction model equation representing a correlation between the amount of iron oxide on the surface of the steel sample and the concentration profile of the Fe element using a multivariate analysis based on the learning label and the learning data; And a fourth step of calculating a predicted value of the amount of iron oxide on the surface by measuring the Fe element concentration profile in the depth direction from the surface of the steel to be measured, and substituting the measured Fe element concentration profile into the prediction model equation.

Step 1: getting the learning label

First, in order to establish a predictive model equation, two or more steel samples are prepared, and a learning label is obtained by quantitatively analyzing iron oxide on the steel surface through a selective extraction method, which is a conventional quantitative analysis method for surface iron oxide. In more detail, the iron oxide in the oxide is quantitatively analyzed after selectively extracting the steel surface oxide using an electrochemical method.

As a non-limiting embodiment of the quantitative analysis of iron oxide on the surface of steel, FIG. 2 shows an apparatus for electrolyzing a steel sample in order to quantitatively analyze the surface oxide. Referring to Fig. 2, a steel sample, which is an analysis material, is attached to a galvanostat and electrolyzed using a non-aqueous electrolyte. At this time, the steel sample is used as a working electrode, and a platinum counter electrode is placed on the opposite side to apply current and electric charge according to the material. The electrolyte may be applied without limitation as long as it is a non-aqueous electrolyte generally used in the conventional electrolysis method, and as a non-limiting example, a non-aqueous electrolyte prepared by dissolving 10% of acetylacetone and 1% of tetramethylammonium chloride in methanol may be used. .

When the oxide is sufficiently extracted from the steel sample, it is filtered using a membrane filter. Then, the filtrate is decomposed by performing an appropriate chemical pretreatment method according to a known method. As a non-limiting embodiment, the chemical pretreatment may be performed with hydrochloric acid after decomposing the filtrate at about 950°C through alkali melting, but is not limited thereto.

Afterwards, the pretreated solution was analyzed for Fe concentration using analysis equipment such as ICP-OES and AAS, and the Fe concentration was estimated as the concentration of iron oxide (FeO, Fe ₂ O ₃ , Fe ₃ O _4, etc.) Converted to a quantitative value of iron oxide, and set it as a reference value. In addition, for each steel sample, a quantitative value of surface iron oxide is obtained according to the method described above, and the quantitative value of each surface iron oxide is set as a reference value of each steel sample.

Step 2: obtaining training data

For each of the steel samples, element concentration analysis is performed in the depth direction from the surface of the base metal including the surface oxide layer to measure the element concentration profile to obtain learning data. At this time, since GD-OES can be analyzed with a depth direction spatial resolution of several tens of nanometers or less, it is preferable to measure using a GD-OES device, but is not limited thereto. In addition, the element to be measured may not be particularly limited, but it is preferable to accurately analyze the change in the concentration of Fe, which is closely related to the iron oxide, and more preferably, analyze the change in the concentration of Fe and O together. As an example, FIG. 3 shows the results of analyzing the concentration change of Fe and O using GD-OES for a steel sample.

Step 3: Establishing a prediction model formula

The present invention is characterized in that the iron oxide layer present on the surface and the iron base material inside each have different differences in element concentration profiles in the depth direction, and the correlation is identified and analyzed by multivariate analysis using this difference. To this end, a correlation between the quantitative value of surface iron oxide (reference value) and the element concentration profile is derived through multivariate analysis based on the learning label obtained in the first step and the learning data obtained in the second step. Based on this, a prediction model equation is obtained.

For multivariate analysis, various models such as known PLS-R, SVM, and random forest can be used, and can be easily calculated through commercial programs such as “R”, “Matlab”, and “Unscrambler X.” Although not limited thereto, PLS- It is preferable to obtain a predictive model equation using the R model. In the case of multivariate analysis, the input value (X) is the element concentration profile (learning data) of each steel sample obtained in the second step, and the output value (Y) is the first It becomes the quantitative value (reference value) (learning label) of the surface iron oxide of each steel sample obtained in the step, in which the more the number of steel samples as possible to establish the model formula, the more accurate the predicted value can be obtained, and at least 20 is preferable. .

Meanwhile, when using the PLS-R model, it is important to select the number of optimized PCs (principal components). If too many PCs are used, the linearity of the regression line may be good, but it may be a problem of over-fitting. Therefore, the number of PCs is preferably 3 to 5 in consideration of the error range of the predicted value.

As a non-limiting example, the prediction model equation may be expressed as Equation 1 below.

[Equation 1] Y=XB

Here, X is the matrix of the Fe element concentration profile (learning data) in the depth direction measured by GD-OES for each sample, and Y is the matrix of the reference value (learning label) measured by ICP-OES for each sample. B is a matrix that expresses the correlation between X and Y, and training data (GD-OES result) and label value (ICP-OES reference value) are obtained through the PLS-R algorithm. As the algorithm, various machine learning (multivariate analysis) algorithms such as PCR and SVR can be used.

Step 4: Predicting the amount of iron oxide on the surface of the steel being manufactured based on the prediction model equation

Thereafter, the Fe element concentration profile in the depth direction is measured from the surface of the steel being manufactured to be measured using a device such as GD-OES. In addition, by substituting the measured Fe element concentration profile into the prediction model equation, a predicted value of the amount of surface iron oxide may be obtained.

According to the present invention, the amount of oxides on the steel surface can be predicted through a fast analysis method such as GD-OES, and the analysis time, which previously took more than 8 hours, can be reduced to within 10 minutes.

Validation of predictive model equation

The prediction model equation obtained in the third step can be evaluated for its accuracy by a conventionally well-known verification method.

However, as a non-limiting example, in the present invention, the accuracy of the predicted value was verified through the difference between the _predicted value (y _{predicted value} ) calculated by the predictive model equation and the surface iron oxide quantitative value (y _{reference value} ) obtained by electrolytic extraction, which is a conventional method. .

In the present invention, the performance of the predicted value was confirmed by the MAPE (mean absolute percentage error, mean absolute percentage error) expressed by the following equation 2, not the R-square value of the calibration curve, and the smaller the MAPE result, the better the predicted value matches the reference value. It can be determined, and the MAPE value is preferably 15% or less.

[Equation 2]

As a non-limiting embodiment, a steel sample is analyzed by GD-OES to measure the Fe element concentration profile in the depth direction, and then a predicted value is obtained by substituting it into a prediction model equation, and this is used as test data (y _{predicted value} ). Next, a quantitative analysis of iron oxide using ICP-OES, a conventional method, is performed on the same steel sample, and the result is taken as a test label (y _{reference value} ). After performing the same operation on two or more n steel samples, it is possible to evaluate the performance of the predictive model equation by substituting it into Equation 2 to obtain the MAPE.

Hereinafter, the present invention will be described in more detail through examples. However, it should be noted that the following examples are only for exemplifying the present invention and not for limiting the scope of the present invention. This is because the scope of the present invention is determined by matters described in the claims and matters reasonably inferred therefrom.

(Example)

First, 30 steel samples were prepared, and each steel sample was subjected to quantitative analysis of iron oxide on the steel surface by a conventional electrolysis method, and the results (reference values) are shown in Table 1 below. Afterwards, Fe concentration analysis was performed from the surface to the depth direction with GD-OES for each steel sample to obtain the Fe element concentration profile, and a multivariate analysis (PLS-R) based on the quantitative value of iron oxide on the steel surface and the Fe element concentration profile was performed. The prediction model equation was derived. The predicted value was calculated for each steel sample using the derived predictive model equation, and the MAPE value was calculated using this, and it was confirmed that it satisfies 15% or less.

Steel samples Reference value (g/㎡) Predicted value (g/㎡) MAPE Sample 1 0.378 0.333
5.1% Sample 2 0.311 0.296 Sample 3 0.276 0.302 Sample 4 0.288 0.308 Sample 5 0.243 0.301 Sample 6 0.298 0.287 Sample 7 0.262 0.293 Sample 8 0.307 0.310 Sample 9 0.274 0.276 Sample 10 0.279 0.297 Sample 11 0.218 0.245 Sample 12 0.302 0.282 Sample 13 0.334 0.280 Sample 14 0.312 0.305 Sample 15 0.322 0.293 Sample 16 0.113 0.119 Sample 17 0.162 0.124 Sample 18 0.104 0.126 Sample 19 0.136 0.141 Sample 20 0.095 0.123 Sample 21 0.111 0.138 Sample 22 0.110 0.106 Sample 23 0.080 0.100 Sample 24 0.111 0.120 Sample 25 0.142 0.104 Sample 26 0.112 0.103 Sample 27 0.125 0.117 Sample 28 0.101 0.110 Sample 29 0.088 0.101 Sample 30 0.121 0.124

Next, after preparing four other steel samples, this time, Fe concentration analysis was performed from the surface to the depth direction with GD-OES to obtain the Fe element concentration profile, and then substituting it into the prediction model equation to calculate the predicted value. In addition, after performing the quantitative analysis of the surface oxide by the conventional method was compared with the predicted value, the results are shown in Table 2 below.

Steel samples Predicted value (g/m ² ) Electrolytic
Quantitative value of surface iron oxide
(g/m ² ) Difference
(%, Convenience) Sample 31 0.327 0.342 4.6 Sample 32 0.295 0.268 9.1 Sample 33 0.269 0.276 2.6 Sample 34 0.130 0.126 3.2

As can be seen in Table 2, the MAPE of the predicted values by the predictive model is 3.9%, and according to the quantitative analysis method of the present invention, it can be confirmed that the quantitative value of iron oxide on the steel surface can be predicted quickly and quickly with high accuracy.

Although it has been described with reference to the above embodiments, it will be understood that a person skilled in the art can variously modify and change the present invention without departing from the spirit and scope of the present invention described in the following claims. I will be able to

Claims (5)

A first step of preparing two or more steel samples, selectively extracting a steel surface oxide by an electrochemical method, and analyzing the extracted component, to quantitatively analyze the surface iron oxide of each of the steel samples to obtain a learning label;
A second step of obtaining learning data by measuring the Fe element concentration profile in the depth direction from the surface of the base metal including the surface oxide layer for each of the steel samples;
A third step of establishing a prediction model equation representing a correlation between the amount of iron oxide on the surface of the steel sample and the concentration profile of the Fe element using a multivariate analysis based on the learning label and the learning data; And
A fourth step of calculating a predicted value of the amount of surface iron oxide by measuring the Fe element concentration profile in the depth direction from the surface of the steel to be measured, and substituting the measured Fe element concentration profile into the prediction model equation;
Quantitative analysis method of iron oxide on a steel surface comprising a
The method of claim 1,
The predictive model formula is a method for quantitative analysis of iron oxide on the surface of steel, characterized in that the MAPE value represented by the following Equation 2 is 15% or less.
[Equation 2]

(Here, y _{reference value} is a quantitative analysis value by a conventional extraction method, y _{predicted value} is a predicted value calculated by the above prediction model equation, and n is the number of samples used for analysis.)
delete The method of claim 1,
A method for quantitative analysis of iron oxide on a steel surface, characterized in that the depth-direction Fe element concentration profile in the second step or the fourth step is analyzed using GD-OES.
The method of claim 1,
The multivariate analysis is a method for quantitative analysis of iron oxide on a steel surface, characterized in that the PLS-R model.

Images