Method for Generating Potential Difference Prediction Model and Apparatus for Generating Potential Difference Prediction Model

Abstract

To provide a method for generating a model for predicting a potential difference which occurs in an alloy material.SOLUTION: The present invention includes: an acquisition step of acquiring an element concentration distribution for at least one type of element and a potential difference distribution for a mother phase in the surface of an alloy material in a two-dimensional coordinate system; an element region specification step of specifying an element region for an element from each element concentration distribution obtained in the acquisition step; an impurity region specification step of specifying an impurity contained in the alloy material from the element region and specifying the impurity region of the impurity; a potential difference region specifying step of specifying a potential difference generation region in which a potential difference is happening in the mother phase from the potential difference distribution obtained in the acquisition step; and a coordinate system conforming step of making the two-dimensional coordinate systems of the element concentration distribution and the potential difference distribution to each other on the basis of the correlation between the position of one of the element region and the impurity region and the position of the potential difference generation region; and a model generation step of generating a prediction model of which objective variable is a potential difference and explanation variable contains each element concentration from the potential difference between the impurity region and the its corresponding potential difference generation region.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method for generating a potential difference prediction model and an apparatus for generating a potential difference prediction model.

Background Art

[0002] In recent years, due to considerations for the global environment and the like, the demand for weight reduction of vehicles such as automobiles has been increasing. To meet such demands, for example, as a material for automotive members including large body panel structures of automobiles, the application of lighter aluminum alloy materials has been increasing instead of steel materials such as steel sheets.

[0003] When an aluminum alloy material is used as a material for automotive members, corrosion resulting from the environment to which the automobile is exposed during driving becomes a problem. It is generally known that the corrosion of an aluminum alloy material is promoted, for example, by the potential difference between a compound (crystal or precipitate) in the alloy material and the matrix phase. This potential difference can be measured, for example, using a Kelvin force probe microscope (KFM) (see Non-Patent Document 1).

[0004] Since this potential difference varies depending on the position of the alloy material because the compound is unevenly present in the alloy material, this potential difference is measured and used for evaluating the corrosion resistance of the alloy material (see, for example, Patent Document 1).

[0005] In addition, efforts have been made to specify the potential difference from the shape of a specific compound species (Non-Patent Document 2). In the method described in this Non-Patent Document 2, a visual correlation is made between a separately measured scanning electron microscope image (SEM image) and the shape of a compound species specified using energy dispersive X-ray analysis (EDX), and the potential difference generated by the compound species is specified. Thus, the potential difference resulting from the position information of the compound species can be predicted.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Non-Patent Document

[0007]

Non-Patent Document 1

Non-Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0008] In the corrosion resistance evaluation method described in the above Patent Document 1, since the measurement of the potential difference is actually required every time an evaluation is performed, the evaluation efficiency is poor. On the other hand, when using the potential difference prediction method described in Non-Patent Document 2, once the prediction model is established, the measurement of the potential difference can be dispensed with in each evaluation. However, in the potential difference prediction method described in Non-Patent Document 2, although the potential difference of a single compound can be predicted, the potential difference of an actual alloy material containing multiple compounds as impurities cannot be quantified.

[0009] The present invention has been made based on the above circumstances, and an object thereof is to provide a method for generating a potential difference prediction model and a potential difference prediction model generation device that can easily quantify the potential difference based on the element concentrations even when multiple compounds are contained as impurities.

Means for Solving the Problems

[0010] A method for generating a potential difference prediction model according to an aspect of the present invention is a method for generating a potential difference prediction model for predicting a potential difference generated in an alloy material. The method includes: an acquisition step of acquiring, on the surface of the alloy material, an element concentration distribution for one or more types of elements and a potential difference distribution for a matrix phase in a two-dimensional coordinate system; an element region specifying step of specifying an element region for a corresponding element from each element concentration distribution obtained in the acquisition step; an impurity region specifying step of specifying an impurity contained in the alloy material from the element region and specifying an impurity region for the impurity; a potential difference region specifying step of specifying a potential difference generation region where a potential difference is generated with respect to the matrix phase from the potential difference distribution obtained in the acquisition step; a coordinate system matching step of matching the two-dimensional coordinate systems of the element concentration distribution and the potential difference distribution based on the correlation between the positions of the element region or the impurity region and the potential difference generation region; and a model generation step of generating a prediction model in which the target variable is a potential difference and the explanatory variables include the respective element concentrations from the impurity region and the potential difference of the corresponding potential difference generation region.

[0011] In this potential difference prediction model generation method, an element region and an impurity region are specified from the element concentration distribution, and further, the two-dimensional coordinate systems of the element region or the impurity region and the potential difference generation region specified from the potential difference distribution are matched, so that the element region, the impurity region, and the potential difference generation region can be associated with each other. In this potential difference prediction model generation method, by associating these three in this way, a prediction model can be generated in which the target variable is a potential difference and the explanatory variables include the respective element concentrations. Therefore, when the prediction model generated by this potential difference prediction model generation method is used, even when a plurality of types of compounds are included as impurities, the potential difference can be easily quantified based on the contained element concentrations.

[0012] In the above-described element region specifying step, a closed region in which the element concentration is within a concentration range of a predetermined lower limit value and an upper limit value is defined as one element region. In the above-described potential difference region specifying step, a closed region in which the potential difference is within a range of a predetermined lower limit value and an upper limit value is defined as one potential difference generation region. By defining a closed region within a range of a predetermined lower limit value and an upper limit value as one element region and one potential difference generation region in this way, irregular data can be excluded, so that the element concentration distribution and the potential difference distribution can be accurately associated with each other.

[0013] The positions of the above-described element region, the above-described impurity region, and the above-described potential difference generation region may be represented by their respective centroid positions. By representing the positions of the above-described element region, the above-described impurity region, and the above-described potential difference generation region by their respective centroid positions in this way, the prediction accuracy of the potential difference by the prediction model can be improved.

[0014] In the above-described impurity region specifying step, dimensionality reduction may be used for the specification of the above-described impurity. By using dimensionality reduction for the specification of the above-described impurity in the above-described impurity region specifying step in this way, the impurity can be accurately specified.

[0015] The above-described alloy material is preferably an aluminum alloy material. The potential difference prediction model generation method can be suitably used for an aluminum alloy material in which corrosion resistance is likely to be a problem.

[0016] A potential difference prediction model generation device according to another aspect of the present invention is a potential difference prediction model generation device that generates a prediction model for the potential difference generated in an alloy material. The device includes: an acquisition unit that acquires, in a two-dimensional coordinate system, an element concentration distribution for one or more elements and a potential difference distribution for a matrix phase on the surface of the alloy material; an element region specifying unit that specifies an element region for a corresponding element from each element concentration distribution obtained in the acquisition step; an impurity region specifying unit that specifies impurities contained in the alloy material from the element region and specifies an impurity region for the impurities; a potential difference region specifying unit that specifies a potential difference generation region where a potential difference occurs with respect to the matrix phase from the potential difference distribution obtained in the acquisition step; a coordinate system matching unit that matches the two-dimensional coordinate systems of the element concentration distribution and the potential difference distribution based on the correlation between the positions of the element region or the impurity region and the potential difference generation region; and a model generation unit that generates a prediction model in which the target variable is the potential difference and the explanatory variables include the respective element concentrations, based on the potential difference of the impurity region and the corresponding potential difference generation region.

[0017] In the potential difference prediction model generation device, an element region and an impurity region are specified from the element concentration distribution, and further, the two-dimensional coordinate systems of the element region or the impurity region and the potential difference generation region specified from the potential difference distribution are matched, so that the element region, the impurity region, and the potential difference generation region can be associated with each other. In the potential difference prediction model generation device, by associating these three in this way, a prediction model can be generated in which the target variable is the potential difference and the explanatory variables include the respective element concentrations. Therefore, when the prediction model generated by the potential difference prediction model generation device is used, even when a plurality of types of compounds are included as impurities, the potential difference can be easily quantified based on the contained element concentrations.

Advantages of the Invention

[0018] As described above, the potential difference prediction model generation method and the potential difference prediction model generation device of the present invention can easily quantify the potential difference based on the contained element concentrations even when a plurality of types of compounds are included as impurities.

Brief Description of the Drawings

[0019]

Figure 1

Figure 2

Figure 3

Figure 4

[0020] Hereinafter, a potential difference prediction model generating method and a potential difference prediction model generating device according to one embodiment of the present invention will be described with reference to the drawings.

[0021] 1 is a method for generating a prediction model of a potential difference occurring in an alloy material. The method for generating a prediction model of a potential difference prediction model includes an acquisition step S1, an element region identification step S2, an impurity region identification step S3, a potential difference region identification step S4, a coordinate system matching step S5, and a model generation step S6.

[0022] The potential difference prediction model generating method can be performed using, for example, a potential difference prediction model generating device that generates a prediction model Y of the potential difference occurring in an alloy material, as shown in Fig. 2. The potential difference prediction model generating device is itself an embodiment of the present invention, and includes an acquisition means 1, an element region identifying means 2, an impurity region identifying means 3, a potential difference region identifying means 4, a coordinate system matching means 5, and a model generating means 6.

[0023] Examples of the above alloy material include carbon steel materials, stainless steel materials, aluminum alloy materials, etc., among which aluminum alloy materials are preferred. The method for generating the potential difference prediction model can be suitably used for aluminum alloy materials with a wide variety of impurity types and where corrosion resistance is likely to be a problem. Hereinafter, the case of an aluminum alloy material will be taken as an example for explanation, but it does not mean that the alloy material of the present invention is limited to aluminum alloy materials.

[0024] <Acquisition Step> In the acquisition step S1, on the surface of the above alloy material, an element concentration distribution X1 for one or more types of elements and a potential difference distribution X2 for the matrix phase are acquired in a two-dimensional coordinate system. The acquisition step S1 can be performed by the acquisition means 1 having the above-described functions.

[0025] Examples of the types of the above elements are elements constituting impurities that can be contained in the above alloy material, and for example, Fe, Mg, Mn, Si, Cu, Cr, Al, Ni, Ti, C, B, N, rare earths, etc. can be mentioned.

[0026] When the types of elements contained in the target alloy material are known, they can be determined based on that known information. Alternatively, after performing elemental analysis on the above alloy material in a wide atomic weight range to identify the existing elements, they can be determined.

[0027] The element concentration distribution X1 is acquired for each element. That is, for example, if there are two types of the above elements, two element concentration distributions X1 will be acquired. The method for acquiring the element concentration distribution X1 is not particularly limited, but for example, it can be obtained by measurement using an electron probe microanalyzer (EPMA). When there are a plurality of the above elements, a plurality of element concentration distributions X1 are obtained corresponding to the number of types of the above elements, and it is assumed that the coordinate systems are the same among these element concentration distributions X1. When performing continuous measurements with the same measuring device, usually the coordinate systems are the same, but if they do not match, the coordinate systems can be made to match via the potential difference distribution X2 by, for example, the same method as the coordinate system matching step S5 described later.

[0028] One potential difference distribution X2 is obtained for the target alloy material. The method for obtaining the potential difference distribution X2 is not particularly limited, and for example, it can be obtained by measurement using a Kelvin force potentiometric microscope (KFM).

[0029] The coordinate systems of the element concentration distribution X1 and the potential difference distribution X2 do not necessarily coincide, but it is preferable that the overlapping area of the analysis regions is large. For this reason, for example, it is advisable to mark the target analysis region. As such a mark, a Vickers indentation can be used.

[0030] The element concentration distribution X1 and the potential difference distribution X2 can be obtained as numerical data, but it is preferable to obtain them as image data. For image data, known processing functions can be used, and in addition, it is easy to perform visualization and visual confirmation.

[0031] <Element region specifying step> In the element region specifying step S2, from each element concentration distribution X1 obtained in the acquisition step S1, the element region D for the corresponding element is specified (see FIG. 3). The element region specifying step S2 can be performed by the element region specifying means 2 having the above-described functions.

[0032] In this element region determination step S2, a closed region where the element concentration is within a concentration range of a predetermined lower limit value and an upper limit value can be defined as one element region D. The composition of impurities that may exist in the alloy material is generally known, and there is an upper limit to the element concentration that constitutes the impurities. Therefore, when an element concentration exceeding this upper limit is measured, it is possible to estimate that the region is not caused by impurities. Thus, it is considered that by setting the upper limit value, it is possible to avoid the inclusion of regions not caused by impurities. Further, when the element concentration is below a certain value, even if there are impurities, they are in trace amounts, and it may be difficult to distinguish them from trace elements such as dissolved Fe that do not affect the potential difference distribution X2. Therefore, by setting the lower limit value, it is possible to effectively extract impurities that may affect the potential difference distribution X2. By defining a closed region within the range of the predetermined lower limit value and upper limit value as one element region D in this way, irregular data can be excluded, so that the element concentration distribution X1 and the potential difference distribution X2 described later can be accurately associated with each other. Note that the lower limit value may be 0 mass%, and the upper limit value may be 100 mass%. That is, the concentration range may be determined substantially by only the lower limit value or only the upper limit value.

[0033] For example, in the case of an aluminum alloy (6061), the lower limit value for Fe can be set to 0.3 mass% to exclude dissolved Fe. In the case of a 7000 series aluminum alloy, the lower limit value for Zn can be set to 10 mass%. In the case of a steel material, the lower limit value for Mn can be set to 1.5 mass%, and the upper limit value for Ti can be set to 50 ppm.

[0034] The position of the element region D is preferably represented by the centroid position. By representing the position of the element region D by the respective centroid positions in this way, the prediction accuracy of the potential difference by the prediction model Y can be improved.

[0035] Also, it is advisable to obtain the size of the elemental region D. Since the potential difference can be affected by the size of the elemental region D, the prediction accuracy of the potential difference by the prediction model Y can be improved by adding the size of the elemental region D to the explanatory variables of the prediction model Y described later. Examples of the above size include the maximum length, minimum length, area, etc. Here, the "maximum length" and "minimum length" refer to the length of the major axis (= maximum length) and the length of the minor axis (= minimum length) in the ellipse with the smallest area that encloses the region.

[0036] <Impurity Region Identification Step> In the impurity region identification step S3, impurities contained in the above alloy material are identified from the elemental region D, and the impurity regions for the above impurities are identified. The impurity region identification step S3 can be performed by the impurity region identification means 3 having the above-described function.

[0037] The above impurities do not exclude single elements but are generally compounds. For example, when the above compound is of the AlFeSi type, the elemental concentration distributions X1 of Fe and Si, which are the elements constituting the above compound (impurity), are obtained in the acquisition step S1, and the elemental regions D are identified for each element in the elemental region identification step S2. Then, for the region where the above compound is of the AlFeSi type, the elemental region D for Fe Fe and the elemental region D for Si Si overlap in position and have a concentration ratio corresponding to the mass ratio of Fe and Si in the AlFeSi system. Conversely, if the positions of the elemental region D for Fe Fe and the elemental region D for Si Si overlap and the concentration ratio is consistent with the mass ratio of Fe and Si in the AlFeSi system, the impurity in the elemental region D can be identified as AlFeSi.

[0038] When the impurities that may be contained in the above alloy material are known, it can be specified by determining from the element region D whether the impurities are actually contained or not. On the other hand, when the generated impurities are complex and unknown, dimensionality reduction means such as principal component analysis (PCA), linear discriminant analysis (LDA), independent component analysis (ICA), non-negative matrix factorization (NMF), and non-negative matrix factorization with soft orthogonal constraint (NMF-so) can be used to specify the above impurities. Among them, it is preferable to use non-negative matrix factorization with soft orthogonal constraint. In this way, in the impurity region specifying step S3, by using dimensionality reduction, particularly non-negative matrix factorization with soft orthogonal constraint, to specify the above impurities, the impurities can be specified accurately. Furthermore, it is preferable to use machine learning to specify the impurities. By using machine learning in this way to specify the impurities, the impurities can be specified efficiently without relying on human judgment.

[0039] On the other hand, in the example where the above compound is of the AlFeSi type, the element region D of Si Si is not necessarily all due to AlFeSi. For example, those due to AlFeMgSi are included. Therefore, for the element region D of Si Si , different labelings are performed for those due to the AlFeSi type and those due to the AlFeMgSi type. For example, "1" is assigned to those due to the AlFeSi type, and "2" is assigned to those due to the AlFeMgSi type. By performing such labeling, for the impurity FeSi, among the element region D of Si Si , those labeled "1" can be specified as the impurity region for the impurity AlFeSi. It should be noted that for the same impurity caused by different elements, it is preferable that the same label is assigned. That is, for example, for those due to the AlFeSi type, if the label "1" is assigned to the element region D of Si Si , it is preferable that the same label "1" is also assigned to the corresponding element region D of Fe Fe .

[0040] The position of the impurity region may be represented by the centroid position. When the position of the element region D is represented by the centroid position, it is represented by the centroid position by following the position of the element region D. On the other hand, for only the region specified as the impurity region, the centroid position may be obtained and assigned.

[0041] Note that it is conceivable that there may be an element region D that does not belong to any of the impurity regions among the element regions D. When such an element region D occurs, for example, an impurity species is added as a new impurity region. Further, for example, when AlFeSi and AlFeMgSi are adjacent to each other, the element regions D do not match among Fe, Si, and Mg. That is, in some cases, Fe and Si are recognized as one element region D including AlFeSi and AlFeMgSi. In such a case, the one element region D is divided into an element region D1 corresponding to Mg and an element region D2 not corresponding to Mg and associated therewith.

[0042] <Potential difference region specifying step> In the potential difference region specifying step S4, a potential difference generation region E where a potential difference is generated with respect to the matrix phase is specified from the potential difference distribution X2 obtained in the acquisition step S1 (see FIG. 3). The potential difference region specifying step S4 can be performed by the potential difference region specifying means 4 having the above-described function. Note that the potential difference region specifying step S4 can be performed independently of the element region specifying step S2 and the impurity region specifying step S3. Therefore, the order of the potential difference region specifying step S4 and the element region specifying step S2 and the impurity region specifying step S3 does not matter.

[0043] The potential difference generation region E is a region where a potential difference is generated with the base material as the reference potential. This potential difference can occur due to various factors, for example, in a region where impurities are present with respect to the base material.

[0044] In the potential difference region specific process S4, a closed region where the potential difference is within a range of a predetermined lower limit value and an upper limit value can be defined as one potential difference generation region E. The composition of impurities that may exist in the alloy material is generally known, and there is an upper limit to the potential difference caused by such impurities. Therefore, when a potential difference exceeding this upper limit is measured, it can be estimated that the region is not caused by impurities. Thus, it is considered that by setting the upper limit value, it is possible to avoid the mixing of regions not caused by impurities. Also, when the absolute value of the potential difference is less than a certain value, the influence of measurement noise and the like cannot be ignored, and it is considered that sufficient accuracy cannot be ensured. Therefore, it is considered that by setting the lower limit value, the prediction accuracy of the potential difference by the prediction model Y can be improved. By setting a closed region within a range of a predetermined lower limit value and an upper limit value as one potential difference generation region E in this way, irregular data can be excluded, so that the element concentration distribution X1 and the potential difference distribution X2 described later can be accurately associated with each other. The lower limit value and the upper limit value are appropriately determined so as to be correlated with the element concentration distribution X1. For example, the lower limit value can be set to 20 mV or more and 50 mV or less. The upper limit value can be appropriately determined according to the type of alloy material.

[0045] The position of the potential difference generation region E is preferably represented by the centroid position. By representing the position of the potential difference generation region E by the respective centroid positions in this way, the prediction accuracy of the potential difference by the prediction model Y can be improved.

[0046] <Coordinate system matching process> In the coordinate system matching process S5, the two-dimensional coordinate systems of the element region D or the impurity region and the potential difference generation region E are matched based on the positional correlation between them. The coordinate system matching process S5 can be performed by the coordinate system matching means 5 having the above-described functions.

[0047] As described above, the coordinate systems of the element concentration distribution X1 and the potential difference distribution X2 do not necessarily match. Therefore, by matching the coordinate systems of both, it becomes possible to associate the potential difference with the element concentration.

[0048] First, focus on a specific element or a specific impurity, and obtain a distribution map representing the element region D of the element or the impurity region of the impurity. For example, when focusing on a specific element, the element concentration distribution X1 after specifying the element region D of the element may be obtained (Figure 3). Note that it is also possible to use a superposition of the element regions D of multiple elements or a superposition of the impurity regions of multiple impurities. Hereinafter, taking the case of using the element concentration distribution X1 after specifying the element region D of a certain element corresponding to Figure 3 as an example, an explanation will be given.

[0049] Next, a comparison is made between the element concentration distribution X1 after specifying the element region D and the potential difference distribution X2 after specifying the potential difference generation region E. Specifically, the characteristic points of each distribution are extracted to define the feature quantity. For example, the element concentration distribution X1 and the potential difference distribution X2 are represented by smoothed images, and points with large luminance changes are extracted as characteristic points, and the luminance changes at these characteristic points are described by the feature quantity of a multi-dimensional (for example, 64-dimensional) vector (A-KAZE algorithm). Note that other feature point extraction methods such as ORB and SIFT may also be adopted.

[0050] At least 3 points, preferably 10 points or more, with a high similarity of the feature quantity are extracted for these characteristic points (for example, corresponding to the arrows shown in Figure 3). In visual terms, these three pairs have a high similarity between the element concentration distribution X1 and the potential difference distribution X2, which means that they are very likely to be information at the same location.

[0051] For the determination of this similarity, a method of calculating the Hamming distance between the feature quantities at each characteristic point and extracting points with a close Hamming distance using the brute-force method can be adopted. Note that the fast approximate nearest neighbor search method (FLANN) may also be adopted.

[0052] Finally, the element concentration distribution X1 and the potential difference distribution X2 are translated, rotated, and scaled so that the coordinates of these highly similar points match, and the two-dimensional coordinate systems of the element concentration distribution X1 and the potential difference distribution X2 are made to coincide. In the example of FIG. 3, for the region surrounded by the dashed line, if the four corners are adjusted to have the same coordinates, the coordinate systems of the two will coincide.

[0053] <Model generation step> In the model generation step S6, a prediction model Y is generated in which the target variable is the potential difference and the explanatory variables include the concentrations of each element, based on the potential difference between the impurity region and the potential difference generation region E corresponding thereto. The model generation step S6 can be performed by the model generation means 6 having the above-described functions.

[0054] As the above-described explanatory variables, in addition to the concentrations of each element, the above-described size may be included. For the selection of necessary explanatory variables, statistical methods can be used. Machine learning may also be used. As the above-described statistical methods, multiple regression analysis, response surface method, logistic regression, random forest, support vector machine (SVM), elastic net, etc. can be used.

[0055] Selected explanatory variable C i For, a prediction formula for calculating the potential difference ΔV, which is the target variable, is derived. As the above prediction formula, for example, a i is a constant, and the following formula 1 can be adopted.

Equation

[0056] In the above formula 1, the constant a iUsing the potential difference between the impurity region and the potential difference generation region E corresponding thereto as reference data, for example, it can be calculated by the least squares method that minimizes the sum of the squares of the differences between the measured value and the predicted value of the target variable ΔV. At this time, for example, the prediction accuracy can be improved by separately calculating those with positive and negative potential differences. For example, since it is known that a positive potential difference is taken with respect to impurities containing Fe, it is advisable to select and process those containing Fe among the types of impurities.

[0057] <Advantages> In the potential difference prediction model generation method and the potential difference prediction model generation device, the element region D and the impurity region are specified from the element concentration distribution X1, and further, the two-dimensional coordinate systems of the element region D or the impurity region and the potential difference generation region E specified from the potential difference distribution X2 are made to coincide, so that the element region D, the impurity region, and the potential difference generation region E can be associated with each other. In the potential difference prediction model generation method and the potential difference prediction model generation device, by associating these three in this way, a prediction model Y can be generated in which the target variable is the potential difference and the explanatory variables include the respective element concentrations. Therefore, when using the prediction model Y generated by the potential difference prediction model generation method and the potential difference prediction model generation device, even when a plurality of types of compounds are included as impurities, the potential difference can be easily quantified based on the contained element concentrations.

[0058] [Other Embodiments] Note that the present invention is not limited to the above embodiments.

[0059] In the above embodiment, in the acquisition step, the case where the element concentration distribution and the potential difference distribution with respect to the matrix phase are acquired by measurement has been described. However, when measurement data already exists, it is not essential to perform the measurement, and it may be acquired by reading out the measured data. It is also possible to acquire one of the element concentration distribution and the potential difference distribution by measurement and the other by reading out the measured data.

[0060] In the above-described embodiment, the case where the coordinate system matching step is performed after the impurity region specifying step has been described. However, when using only the element concentration distribution after specifying the element region in the coordinate system matching step, if it is after the element region specifying step and the potential difference region specifying step, it is also possible to perform the coordinate system matching step before the impurity region specifying step.

Example

[0061] Hereinafter, the present invention will be described in more detail by way of examples, but the present invention is not limited to these examples.

[0062] An aluminum alloy (6016) was prepared as an alloy material. Vickers indentations were made on the surface of the above alloy material as marks for the target analysis region, and six element concentration distributions of Fe, Mg, Mn, Si, Cu, and Cr were obtained by EPMA (JXA8530F PLUS manufactured by JEOL Ltd.), and the potential difference distribution was obtained as image data by KFM (AFM5300 manufactured by Hitachi High-Technologies Corporation), respectively.

[0063] Focusing on the impurities containing Fe for the purpose of taking the correlation with the positive potential difference distribution, using these six element concentration distribution diagrams (image data), the identification of impurities by NMF-so was performed using machine learning according to the procedures of the above-described element region specifying step and impurity region specifying step. As a result, impurities of the AlFeMgSi system, AlFeSi system, AlFeSiMn system, and AlFeSiMnCr system were identified. It was also found that Cu was in a substantially solid solution state and no impurities were formed. In the above element region specifying step, a closed region with an element concentration in the concentration range of 0.3 mass% or more was regarded as one element region so that the dissolved Fe in Fe could be ignored. In addition, data on the centroid and size (maximum length, minimum length, and area) of the element region were obtained.

[0064] Also, using the potential difference distribution diagram (image data), the potential difference generation region was specified according to the procedure of the potential difference region specification step. In specifying the potential difference generation region, a closed region where the potential difference is in the range of 50 mV or more was defined as one potential difference generation region. By setting this lower limit value, it was visually confirmed that there is a good correlation with the distribution of the impurity region.

[0065] Since the elemental concentration distribution of Fe has a good correlation with the positive potential difference distribution, the coordinate system matching step was performed using the elemental concentration distribution of Fe.

[0066] Specifically, it was carried out according to the following procedure. For the elemental concentration distribution diagram and the potential difference distribution diagram of Fe, image smoothing processing was performed 4 times using the Perona and Malk diffusion method, and the same processing was also performed on the images obtained by reducing the size of the original image to 0.5 times and 0.25 times.

[0067] For the elemental concentration distribution diagram and the potential difference distribution diagram after the smoothing process, points with large luminance changes were extracted as feature points. For this extraction, the A-KAZE algorithm was used, and the above 3 images with different scales were input, and the number of vectors was set to 64. By this process, the luminance change at the feature point was described as a feature quantity of a 64-dimensional vector.

[0068] The Hamming distance between the feature quantities at each feature point was calculated, and 10 points with close Hamming distances were extracted using the brute-force method. Using these 10 points, translation, rotation, and scaling of the elemental concentration distribution diagram and the potential difference distribution diagram were performed to match the two-dimensional coordinate systems of both.

[0069] Using the elemental concentration distribution diagram and the potential difference distribution diagram with the two-dimensional coordinate systems matched, a prediction model was generated according to the model generation step.

[0070] As a result of the multiple regression analysis, seven variables were used as explanatory variables, which were the six elements (mass % of Fe, Mg, Mn, Si, Cu, and Cr) for which the elemental concentration distribution was obtained, plus the area of the elemental region of Fe (number of pixels in the image). As the prediction formula, the above formula 1 was used, and the prediction formula of the following formula 2 was derived by the least squares method. In the following formula 2, the element name means the concentration (mass %) of that element, and Size means the area (number of pixels).

Number

[0071] The correlation between the predicted value calculated from the prediction model represented by the above formula 2 and the measured value is shown in FIG. 4. From the results of FIG. 4, it can be seen that even when a plurality of types of compounds are contained as impurities, the potential difference can be accurately predicted based on the contained elemental concentrations.

Industrial Applicability

[0072] The potential difference prediction model generation method and the potential difference prediction model generation device of the present invention can easily quantify the potential difference based on the contained elemental concentrations even when a plurality of types of compounds are contained as impurities.

Explanation of Signs

[0073] 1 Acquisition means 2 Elemental region specifying means 3 Impurity region specifying means 4 Potential difference region specifying means 5 Coordinate system matching means 6 Model generation means D Elemental region E Potential difference generation region X1 Elemental concentration distribution X2 Potential difference distribution Y Prediction model

Claims

1. A method for generating a potential difference prediction model that generates a prediction model for the potential difference generated in an alloy material, comprising: an acquisition step of acquiring, in a two-dimensional coordinate system, an element concentration distribution for one or more types of elements and a potential difference distribution for a matrix phase on the surface of the alloy material; an element region specifying step of specifying an element region for a corresponding element from each element concentration distribution obtained in the acquisition step; an impurity region specifying step of specifying impurities contained in the alloy material from the element region and specifying an impurity region for the impurities; a potential difference region specifying step of specifying a potential difference generation region where a potential difference is generated with respect to the matrix phase from the potential difference distribution obtained in the acquisition step; a coordinate system matching step of matching the two-dimensional coordinate systems of the element concentration distribution and the potential difference distribution based on the correlation between the positions of the element region or the impurity region and the potential difference generation region; a model generation step of generating a prediction model in which the target variable is the potential difference and the explanatory variables include the respective element concentrations from the impurity region and the potential difference in the corresponding potential difference generation region. A method for generating a potential difference prediction model comprising the above steps.

2. In the element region specifying step, a closed region where the element concentration is within a concentration range of a predetermined lower limit value and an upper limit value is defined as one element region, In the potential difference region specifying step, a closed region where the potential difference is within a range of a predetermined lower limit value and an upper limit value is defined as one potential difference generation region. The method for generating a potential difference prediction model according to claim 1.

3. The method for generating a potential difference prediction model according to claim 1 or claim 2, wherein the positions of the element region, the impurity region, and the potential difference generation region are represented by their respective centroid positions.

4. The method for generating a potential difference prediction model according to claim 1, claim 2, or claim 3, wherein dimensionality reduction is used for specifying the impurities in the impurity region specifying step.

5. The method for generating a potential difference prediction model according to any one of claims 1 to 4, wherein the alloy material is an aluminum alloy material.

6. A potential difference prediction model generation device for generating a prediction model for the potential difference generated in an alloy material, comprising: acquisition means for acquiring, in a two-dimensional coordinate system, an element concentration distribution for one or more types of elements and a potential difference distribution for a matrix phase on the surface of the alloy material; element region specifying means for specifying an element region for a corresponding element from each element concentration distribution obtained in the acquisition step; impurity region specifying means for specifying impurities contained in the alloy material from the element region and specifying an impurity region for the impurities; Potential difference region specifying means for specifying a potential difference generation region where a potential difference has occurred with respect to the parent phase from the potential difference distribution obtained in the above acquisition process; Coordinate system matching means for matching the two-dimensional coordinate systems of the element concentration distribution and the potential difference distribution from the correlation between the positions of the element region or the impurity region and the potential difference generation region; Model generation means for generating a prediction model in which the target variable is the potential difference and the explanatory variables include the concentrations of respective elements from the potential difference between the impurity region and the corresponding potential difference generation region; A potential difference prediction model generation device comprising the above.

Images