Machine learning-based X-ray diffraction mineral composition analysis system and method using domain knowledge

Abstract

The invention discloses an X-ray diffraction composition analysis system and method based on machine learning by utilizing domain knowledge. The machine learning-based X-ray diffraction composition analysis method using domain knowledge comprises: a step of receiving an input of X-ray diffraction data of a mineral; a step in which the input X-ray diffraction data is normalized and pre-processed for each data sample; a step of generating a clustering model by using the preprocessed X-ray diffraction data; generating a mineral composition analysis model by using the preprocessed X-ray diffraction data; receiving the input of new X-ray diffraction data, performing preprocessing, and classifying the preprocessed new X-ray diffraction data into conventional composition data or abnormal composition data by using the clustering model; and a step of estimating the mineral composition of new X-ray diffraction data classified as conventional composition data by analysis using the mineral composition analysis model.

Description

TECHNICAL FIELD

[0001] The present invention relates to a machine learning-based X-ray diffraction mineral composition analysis system and method using domain knowledge. BACKGROUND

[0002] In view of global warming and energy security issues, the development of gas hydrates, which are widely distributed around the world and contain 3 trillion cubic meters, is increasingly important. In particular, the net import dependency of energy based on 2010 in Korea is 88.1%, and most of the energy is imported from abroad. Therefore, Korea also attaches great importance to the surrounding gas hydrates. The main component of gas hydrates is methane, which emits less carbon dioxide than traditional resources such as coal and oil, and has environmental characteristics. However, since gas hydrates have different characteristics from traditional resources, in order to achieve commercial production, research on the geological information of the reservoir layer, flow characteristics, etc. is required.

[0003] Mineral composition data provides information on the formation environment and origin of sediments. In addition, the composition analysis of constituent minerals is essential for the characterization of oil resource reservoirs. Such mineral composition data can be obtained through X-ray diffraction (XRD) experimental analysis. The X-ray diffraction experiment is an experiment in which X-rays are irradiated to a sample powdered to less than 10 μm, and the intensity of the diffracted X-rays is measured. The X-ray diffraction intensity obtained through the experiment is analyzed by an expert using analysis software. Individual minerals have characteristic intensities depending on the incident angle of X-rays, but if composed of a plurality of minerals such as sediments, the pattern is complex, and the degree of dependence on experts in interpretation is high. In addition, when a large number of samples are analyzed, the time-consuming repetitive work is a great limitation. SUMMARY

[0004] Problems to be Solved by the Invention The present invention aims to provide a machine learning-based X-ray diffraction mineral composition analysis system and method using domain knowledge, which classifies X-ray diffraction data of sediments into regular composition data or abnormal composition data using a clustering model, estimates the composition value using a mineral composition analysis model for regular composition data, and analyzes it by an expert for abnormal composition data.

[0005] Means for Solving the Problems According to an aspect of the present invention, a machine learning-based X-ray diffraction mineral composition analysis method using domain knowledge performed by an X-ray diffraction mineral composition analysis system is disclosed.

[0006] According to one embodiment of the present application, a machine learning-based X-ray diffraction mineral composition analysis method using domain knowledge includes: a step of receiving input of X-ray diffraction data of a mineral; a step of preprocessing the inputted X-ray diffraction data by normalizing the data samples; a step of generating a clustering model using the preprocessed X-ray diffraction data; a step of generating a mineral composition analysis model using the preprocessed X-ray diffraction data; a step of classifying the preprocessed new X-ray diffraction data into regular composition data or abnormal composition data using the clustering model after receiving input of new X-ray diffraction data and preprocessing; and a step of estimating the mineral composition of the new X-ray diffraction data classified as the regular composition data by analyzing using the mineral composition analysis model.

[0007] In the preprocessing step, the intensity values included in the X-ray diffraction data are preprocessed by data samples using min-max scaling.

[0008] The step of generating a clustering model includes: a step of separating the preprocessed X-ray diffraction data into training and validation data and test data; a step of training a clustering model using the training and validation data; a step of selecting an abnormal composition standard for clustering model performance evaluation using the training and validation data; a step of classifying test data into regular composition data and abnormal composition data according to the selected abnormal composition standard using the trained clustering model; a step of generating a confusion matrix using the classification result; a step of selecting precision in a confusion matrix performance evaluation indicator in order to determine the optimal number of clusters; and a step of determining the number of clusters with the highest precision of the generated confusion matrix as the optimal number of clusters, thereby generating an optimized clustering model.

[0009] The step of generating a mineral composition analysis model includes: a step of separating the preprocessed X-ray diffraction data into training data, validation data, and test data; a step of performing deep learning using the training data and the validation data; and a step of optimizing Hyper parameters of a deep learning model generated by the deep learning performance using the validation data and the test data, thereby generating an optimized mineral composition analysis model.

[0010] The new X-ray diffraction data classified as the abnormal composition data is analyzed by an expert and the mineral composition is estimated.

[0011] The X-ray diffraction data includes intensity values and mineral composition values.

[0012] According to another aspect of the present application, there is disclosed a machine learning based X-ray diffraction mineral composition analysis system using domain knowledge.

[0013] According to an embodiment of the present application, a machine learning based X-ray diffraction mineral composition analysis system using domain knowledge includes a memory for storing instructions, and a processor for executing the instructions, the instructions performing an X-ray diffraction composition analysis method including a step of receiving input of X-ray diffraction data of a mineral, a step of preprocessing the inputted X-ray diffraction data by normalizing the data by data sample, a step of generating a clustering model using the preprocessed X-ray diffraction data, a step of generating a mineral composition analysis model using the preprocessed X-ray diffraction data, a step of classifying preprocessed new X-ray diffraction data as normal composition data or abnormal composition data using the clustering model after receiving input of new X-ray diffraction data and preprocessing, and a step of estimating mineral composition of the new X-ray diffraction data classified as the normal composition data by performing analysis using the mineral composition analysis model.

[0014] Effects of the Invention According to the machine learning based X-ray diffraction mineral composition analysis system and method using domain knowledge according to an embodiment of the present application, X-ray diffraction data of a sediment can be classified as normal composition data or abnormal composition data using a clustering model, for the normal composition data, a composition value is estimated using a mineral composition analysis model, and for the abnormal composition data, analysis is performed by an expert. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is a flowchart schematically illustrating a machine learning based X-ray diffraction mineral composition analysis method using domain knowledge performed by an X-ray diffraction mineral composition analysis system according to an embodiment of the present application.

[0016] Figure 2 is a flowchart schematically illustrating detailed steps of step S300 in Figure 1

[0017] Figure 3 is a flowchart schematically illustrating detailed steps of step S400 in Figure 1

[0018] Figures 4 to 11 is a diagram for explaining a machine learning based X-ray diffraction mineral composition analysis method using domain knowledge according to an embodiment of the present application in Figure 1

[0019] Figure 12 ​​​This is a schematic diagram illustrating the configuration of a machine learning-based X-ray diffraction mineral composition analysis system utilizing domain knowledge according to an embodiment of the present invention. Detailed Implementation

[0020] The singular expressions used in this specification include the plural expressions unless the context clearly indicates otherwise. Terms such as “constituting” or “comprising” should not be construed as necessarily including all the various constituent elements or steps described in the specification; some of these constituent elements or steps may be excluded, or additional constituent elements or steps may be included. Furthermore, terms such as “...part” or “module” used in the specification refer to a unit that performs at least one function or operation, which may be implemented in hardware or software, or a combination of hardware and software.

[0021] Various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Figure 1 This is a flowchart schematically illustrating a machine learning-based X-ray diffraction mineral composition analysis method utilizing domain knowledge, performed by an X-ray diffraction mineral composition analysis system according to an embodiment of the present invention. Figure 2 It is a general overview Figure 1 A detailed flowchart of step S300 in the process. Figure 3 It is a general overview Figure 1 A detailed flowchart of step S400 in the process. Figures 4 to 11 It is used for explanation Figure 1 The figure shows a machine learning-based X-ray diffraction mineral composition analysis method utilizing domain knowledge according to an embodiment of the present invention. The following will use... Figure 1 Centered on, with reference Figures 2 to 11 This invention describes a machine learning-based X-ray diffraction mineral composition analysis method that utilizes domain knowledge, according to embodiments of the present invention.

[0023] In step S100, the X-ray diffraction mineral composition analysis system receives input X-ray diffraction data of the mineral.

[0024] Figure 4 The locations of eight boreholes from the UBGH-1 and UBGH-2 projects conducted in the Ulleung Basin of the East Sea are shown. In this embodiment of the invention, X-ray diffraction data from 488 data samples obtained through the large-scale drilling exploration projects UBGH-1 and UBGH-2 conducted in the Ulleung Basin of the East Sea in 2007 and 2010 were used. These data were provided by the Gas Hydrates Project. The data consists of sediment data obtained from a total of eight boreholes, the locations of which are shown below. Figure 4 As shown in the table below, the number of data points for each borehole is as follows.

[0025] [Table 1]

[0026] Of the total 488 data samples, 90%, i.e., 439, were used for model training, and the remaining 10%, i.e., 49, were used for testing.

[0027] The X-ray diffraction data is composed of 3,100 intensity values (input layer) at 0.02° intervals in the range of 3.01° to 64.99° incident angle and 12 mineral composition values (output layer) analyzed from the intensity values. Of the 12 minerals, quartz, albite, opal-A, calcite, muscovite, dolomite, chlorite, kaolinite, illite, pyrite, NaCl, and K-feldspar.

[0028] As an example, the intensity values of the 1st data sample with respect to the incident angle are shown in FIG. 1, and the mineral composition values are shown in the following table. Figure 5

[0029] [Table 2]

[0030] In step S200, the X-ray diffraction mineral composition analysis system pre-processes the input X-ray diffraction data by normalizing the data samples.

[0031] For example, the intensity values of the input X-ray diffraction data can be pre-processed using the following minimum-maximum scaling mathematical formula.

[0032] [Mathematical Formula 1]

[0033] where X ms denotes the intensity value normalized by the minimum-maximum scaling, X max,j and X min,j denote the maximum and minimum values of the intensity in the jth data sample, respectively.

[0034] That is, the X-ray diffraction composition analysis system according to the embodiment of the present application pre-processes the input X-ray diffraction data by data sample. This is because the size and accurate position of the absolute peak are not important for interpretation, and the mineralogical characteristics of the data and experimental errors, etc. cause differences to occur, and the proportion and trend of the intensity values with respect to the incident angle in a single data sample are more important. Therefore, in the present application, the X-ray diffraction data is normalized to between 0 and 1 to make the scale uniform.​

[0035] Figure 6 The X-ray diffraction data pattern according to the pre-processing method is shown. In general, it can be confirmed that the original data pattern of (a) is completely damaged to the pattern of (b) when using the factor-wise pre-processing in machine learning, and when applying the min-max scaler by data sample, it can be seen that it maintains the pattern of the original data as shown in (c) and corrects the scale to 0~1. Therefore, for X-ray diffraction data, it is considered appropriate to apply the min-max scaling by data sample.

[0036] In step S300, the X-ray diffraction mineral composition analysis system generates a clustering model using the pre-processed X-ray diffraction data.

[0037] Hereinafter, a detailed description will be made with reference to Figure 2 A detailed step of step S300 will be described.

[0038] In step S310, the X-ray diffraction mineral composition analysis system separates the pre-processed X-ray diffraction data into training and validation data and test data.

[0039] In step S320, the X-ray diffraction mineral composition analysis system trains a clustering model using the training and validation data.

[0040] Figure 7 The process of one of the clustering algorithms, i.e., a k-means clustering algorithm, is shown. Referring to Figure 7 , the k-means clustering algorithm is a method of deciding the number of clusters k in advance, assigning data to each cluster based on the initial center point, and repeatedly updating the average value of the cluster until the center no longer changes, thereby classifying data closest to the center as one cluster. Since the result value varies depending on the number of clusters, it is important to select the best number of clusters that meet the purpose.

[0041] In step S330, the X-ray diffraction mineral composition analysis system selects an abnormal composition standard for performance evaluation of the clustering model using the training and validation data.

[0042] For example, Figure 8 A boxplot for 12 kinds of mineral composition values and a composition value range of the normal composition data grasped using the training and validation data are shown. When it does not belong to the range, it is classified as abnormal composition data.

[0043] In step S340, the X-ray diffraction mineral composition analysis system classifies the test data into normal composition data and abnormal composition data using the trained clustering model.

[0044] In step S350, the X-ray diffraction mineral composition analysis system generates a confusion matrix using the classification result.

[0045] In step S360, the X-ray diffraction mineral composition analysis system selects precision in the confusion matrix performance evaluation index in order to determine the optimal cluster number.

[0046] As an index for quantitatively analyzing the cluster result, recall and precision of the confusion matrix are used, which can be expressed by the following mathematical formula.

[0047] [mathematical formula 2]

[0048] That is, the recall is the proportion of the number of data that is accurately predicted as the regular composition data among the actual regular composition data, and the precision is the proportion of the number of data that is actually the regular composition data among the data predicted as the regular composition data.

[0049] When considering from the aspect of the object of the present application that it is important to grasp whether there is abnormal composition data in the regular composition cluster, it is reasonable to determine the cluster number based on the precision that represents the proportion of the actual regular composition data in the classified data.

[0050] In step S370, the X-ray diffraction mineral composition analysis system determines the cluster number with the highest precision of the generated confusion matrix as the optimal cluster number, thereby generating an optimized cluster model.

[0051] Figure 9 A confusion matrix with respect to 49 Yushan Basin test data is shown. Referring to Figure 9 , the recall is 100% when the cluster number is 3, and 92.7% when the cluster number is 5, and the precision is 83.7% when the cluster number is 3, and 100% when the cluster number is 5. When considering the recall, the optimal cluster number should be 3, but since the index suitable for the present application is the precision, the optimal cluster number can be 5.

[0052] In step S400, the X-ray diffraction mineral composition analysis system generates a mineral composition analysis model using the preprocessed X-ray diffraction data.

[0053] Hereinafter, a detailed step of step S400 will be described with reference to Figure 3 A detailed step of step S400 will be described.

[0054] In step S410, the X-ray diffraction mineral composition analysis system separates the preprocessed X-ray diffraction data into training data, validation data, and test data.

[0055] In step S420, the X-ray diffraction mineral composition analysis system performs deep learning using the training data and the validation data.

[0056] The following table is a table showing the average and deviation of the coefficient of determination and the mean absolute error according to the pre-processing method using one of the deep learning algorithms, CNN (convolutional neural network). Feature_CNN and Sample_CNN are cases in which factor normalization and data sample normalization are performed, respectively, and the data sample normalization shows a coefficient of determination of 0.757 compared to the factor normalization, thus confirming that the data sample normalization is reasonable.

[0057] [Table 3]

[0058] In step S430, the X-ray diffraction mineral composition analysis system optimizes the hyper parameters of the deep learning model generated by performing deep learning using the validation and test data, thereby generating a mineral composition analysis model.

[0059] Referring again to Figure 1 Steps S500 to S700 will be described.

[0060] In step S500, after the X-ray diffraction mineral composition analysis system receives input of new X-ray diffraction data and pre-processes the same, the pre-processed new X-ray diffraction data is classified as regular composition data or abnormal composition data using the generated clustering model.

[0061] In step S600, the X-ray diffraction mineral composition analysis system estimates the mineral composition of the new X-ray diffraction data classified as regular composition data using the generated mineral composition analysis model.

[0062] In step S700, the new X-ray diffraction data classified as abnormal mineral composition data is analyzed by an expert and the mineral composition is estimated.

[0063] Figure 10 The locations of the UBGH-1, UBGH-2 projects and the Korea Big Earth drilling are shown, Figure 11 The confusion matrix results regarding the Korea Big Earth additional validation data are shown. The validation data were obtained from the Korea Big Earth drilling site as shown in FIG. 1. Figure 10The 54 data samples of the Korean ground obtained from the periphery of the Yeongil basin in the north were shown to have the applicability of the clustering model extended to additional data of the adjacent areas. As shown in Figure 11 As shown in the same as the previous results, the precision of 100% was shown when the number of clusters was 5 compared to when it was 3, and thus it can be confirmed that there is no abnormal composition data in the conventional composition clusters.

[0064] Figure 12 FIG. 1 is a diagram schematically showing a configuration of a machine learning-based X-ray diffraction mineral composition analysis system using domain knowledge according to an embodiment of the present application.

[0065] Referring to Figure 12 The machine learning-based X-ray diffraction mineral composition analysis system using domain knowledge according to an embodiment of the present application includes a processor (10), a memory (20), a communication part (30), and an interface part (40).

[0066] The processor (10) can be a CPU or a semiconductor element that executes processing instructions stored in the memory (20).

[0067] The memory (20) can include various types of volatile or non-volatile storage media. For example, the memory (20) can include a ROM, a RAM, and the like.

[0068] For example, the memory (20) can store instructions for executing the machine learning-based X-ray diffraction mineral composition analysis method using domain knowledge according to an embodiment of the present application.

[0069] The communication part (30) is a device that transmits and receives data with other devices through a communication network.

[0070] The interface part (40) can include a network interface for connecting a network and a user interface.

[0071] On the other hand, the configuration elements of the aforementioned embodiments can be easily grasped from the perspective of the flow. That is, each configuration element can be grasped as each flow. In addition, the flow of the aforementioned embodiments can be easily grasped from the perspective of the configuration elements of the device.

[0072] Further, the aforementioned technical contents can be implemented in the form of program instructions capable of being executed through various computer means, and recorded in a computer-readable medium. The computer-readable medium can include program instructions, data files, data structures, etc., alone or in combination. The program instructions recorded in the medium can be designed and configured specifically for the embodiments, or can be known and used by those skilled in the computer software field. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specially configured to store and execute program instructions, such as ROMs, RAMs, flash memories, etc. Examples of the program instructions include machine language codes generated by a compiler; and high-level language codes executable by a computer using an interpreter, etc. The hardware device can be configured to operate as one or more software modules in order to perform the operations of the embodiments, and vice versa.

[0073] The above-described embodiments of the present application are disclosed for exemplary purposes, and various modifications, alterations, and additions can be made by those skilled in the art in light of the idea and scope of the present application, and such modifications, alterations and additions should be interpreted as falling within the scope of the following claims.

Claims

1. A machine learning-based X-ray diffraction mineral composition analysis method utilizing domain knowledge, executed by an X-ray diffraction mineral composition analysis system, characterized in that... The X-ray diffraction mineral composition analysis method based on machine learning that utilizes domain knowledge includes: The steps for receiving X-ray diffraction data of minerals as input; The step of preprocessing the input X-ray diffraction data by normalizing it according to the data sample; The step of generating a clustering model using the preprocessed X-ray diffraction data; The step of generating a mineral composition analysis model using the preprocessed X-ray diffraction data; The steps include receiving new X-ray diffraction data input, preprocessing it, and then using the clustering model to classify the preprocessed new X-ray diffraction data into normal composition data or abnormal composition data; and... The step of estimating the mineral composition of new X-ray diffraction data classified as conventional composition data by using the mineral composition analysis model.

2. The X-ray diffraction mineral composition analysis method based on machine learning utilizing domain knowledge according to claim 1, characterized in that, In the preprocessing step, the intensity values ​​contained in the X-ray diffraction data are preprocessed according to the data samples using min-max scaling.

3. The X-ray diffraction mineral composition analysis method based on machine learning utilizing domain knowledge according to claim 1, characterized in that, The steps for generating the clustering model include: The step of separating the preprocessed X-ray diffraction data into training and validation data and test data; The steps for training a clustering model using the training and validation data; The step of selecting outlier composition criteria for evaluating the performance of the clustering model using the training and validation data; The steps of classifying test data into normal composition data and abnormal composition data according to the selected abnormal composition criteria using the trained clustering model; The step of generating a confusion matrix using the classification results; To determine the optimal number of clusters, the step of selecting precision as a performance evaluation metric for the confusion matrix is ​​as follows; and The step of determining the optimal number of clusters by selecting the number of clusters that yields the highest precision in the generated confusion matrix, thereby generating an optimized clustering model.

4. The X-ray diffraction mineral composition analysis method based on machine learning utilizing domain knowledge according to claim 1, characterized in that, The steps for generating the mineral composition analysis model include: The step of separating the preprocessed X-ray diffraction data into training data, validation data, and test data; The steps of performing deep-learning using the training data and the validation data; and The steps involve using the validation data and the test data to optimize the hyperparameters of the deep learning model generated through the deep learning process, thereby generating an optimized mineral composition analysis model.

5. The X-ray diffraction mineral composition analysis method based on machine learning utilizing domain knowledge according to claim 1, characterized in that, The new X-ray diffraction data classified as anomalous composition data were analyzed by experts to estimate the mineral composition.

6. The X-ray diffraction mineral composition analysis method based on machine learning utilizing domain knowledge according to claim 1, characterized in that, The X-ray diffraction data includes intensity values ​​and mineral composition values.

7. A machine learning-based X-ray diffraction mineral composition analysis system utilizing domain knowledge, characterized in that, The X-ray diffraction mineral composition analysis system based on machine learning that utilizes domain knowledge includes: Memory for storing instructions; and The processor used to execute the instructions. The instruction executes a machine learning-based X-ray diffraction composition analysis method that utilizes domain knowledge. This machine learning-based X-ray diffraction composition analysis method includes: The steps for receiving X-ray diffraction data of minerals as input; The step of preprocessing the input X-ray diffraction data by normalizing it according to the data sample; The step of generating a clustering model using the preprocessed X-ray diffraction data; The step of generating a mineral composition analysis model using the preprocessed X-ray diffraction data; The steps include receiving new X-ray diffraction data input, preprocessing it, and then using the clustering model to classify the preprocessed new X-ray diffraction data into normal composition data or abnormal composition data; and... The step of estimating the mineral composition of new X-ray diffraction data classified as conventional composition data by using the mineral composition analysis model.

Images