Large-dip-angle coal seam overlying strata fracture migration prediction method and system based on machine learning

Abstract

The invention discloses a large dip angle coal seam overlying strata fracture migration prediction method and system based on machine learning, and the method comprises the steps: S1, collecting the geological data of a large dip angle coal seam, carrying out the missing value filling and abnormal value detection of the geological data, and obtaining the preprocessing data; s2, static features are extracted from the preprocessed data, mining dynamic features in the mining process are collected, and fracture migration simulation is conducted on the basis of the static features and the dynamic features; and S3, constructing an overlying strata fracture migration prediction model based on a simulation result, inputting to-be-detected fracture migration data into the overlying strata fracture migration prediction model, outputting a prediction result, and completing large-dip-angle coal seam overlying strata fracture migration prediction based on machine learning.

Description

Technical Field

[0001] The present invention relates to the technical field of overburden fracture and migration prediction, and in particular to a method and system for predicting overburden fracture and migration in steeply inclined coal seams based on machine learning. Background Art

[0002] The overburden fracture and migration of steep-angle coal seams refers to the deformation, fracture and displacement of the rock strata covering the coal seams due to the impact of mining as the working face advances during the mining process of steep-angle coal seams.

[0003] Existing prediction models may perform well for data from specific mining areas or under specific conditions, but their generalization to other mining areas or under different mining conditions needs improvement. While complex models such as deep learning and ensemble learning can improve prediction accuracy, they are computationally expensive. In practical applications, a better balance between model complexity and computational efficiency is needed.

[0004] In actual mining, timely and accurate predictions of overburden fracture and migration are necessary to guide mining. However, current prediction models can experience delays in the entire process from data collection to output. Furthermore, while predictions can provide a useful reference for mining, actual mining requires consideration of more practical factors, such as limitations of mining equipment and operator habits. Therefore, integrating predictions with actual mining practices remains a pressing issue. Summary of the Invention

[0005] In order to solve the above problems, the present invention proposes a method and system for predicting the breakage and migration of overburden in steeply inclined coal seams based on machine learning. The method specifically includes:

[0006] Step S1, collecting geological data of steeply inclined coal seams and performing missing value filling and outlier detection on the geological data to obtain preprocessed data;

[0007] Step S2: extracting static features from the preprocessed data, collecting dynamic features of mining during the mining process, and performing fracture and migration simulation based on the static features and the dynamic features;

[0008] Step S3: construct an overburden fracture and migration prediction model based on the simulation results, input the fracture and migration data to be detected into the overburden fracture and migration prediction model, output the prediction results, and complete the overburden fracture and migration prediction of the steep-angle coal seam based on machine learning.

[0009] Optionally, in step S1, the geological data includes coal seam inclination, coal seam thickness, roof lithology, physical and mechanical parameters of roof rock layers, ground stress data and mining depth.

[0010] Optionally, the missing value detection process includes:

[0011] Setting a data set segmentation length and a value missing rate, segmenting the geological data based on the segmentation length, traversing all segmented geological data, and using the geological data with a missing rate less than a preset value missing rate as sample data to generate a sample data set;

[0012] Building a filling model based on the deep learning model, and training the filling model using the sample dataset, MSE and SIV losses to obtain an optimized filling model;

[0013] The optimized filling model is used to fill in the missing value data in the geological data.

[0014] Optionally, the content of the optimized filling model includes:

[0015]

[0016] Among them, α is the combined weight between the loss function MSE and SIV, is the loss value of MSE, is the loss value of SIV.

[0017] Optionally, in step S2, the static feature extraction method specifically includes:

[0018] Performing dimensionality reduction on the preprocessed data using PCA to obtain dimensionality-reduced data;

[0019] Using the MBK-means clustering algorithm to compress a portion of the dimensionality reduction data to obtain compressed data;

[0020] Use the uncertain sampling method in AL to label the other part of the dimensionality reduction data to obtain labeled data;

[0021] Training a CNN using the compressed data and the labeled data;

[0022] Use the trained CNN model to extract static features.

[0023] The present invention also discloses a machine learning-based prediction system for the breakage and migration of overburden in steeply inclined coal seams, the system comprising:

[0024] A data acquisition module is used to collect geological data of steeply inclined coal seams and perform missing value filling and outlier detection on the geological data to obtain preprocessed data;

[0025] A mining simulation module, configured to extract static features from the preprocessed data, collect dynamic features of mining during the mining process, and perform fracture and migration simulation based on the static and dynamic features;

[0026] The migration prediction module is used to construct a rock breakage and migration prediction model based on the simulation results, input the breakage and migration data to be detected into the rock breakage and migration prediction model, output the prediction results, and complete the overburden breakage and migration prediction of high-angle coal seams based on machine learning.

[0027] Optionally, the missing value detection process includes:

[0028] Setting a data set segmentation length and a value missing rate, segmenting the geological data based on the segmentation length, traversing all segmented geological data, and using the geological data with a missing rate less than a preset value missing rate as sample data to generate a sample data set;

[0029] Building a filling model based on the deep learning model, and training the filling model using the sample dataset, MSE and SIV losses to obtain an optimized filling model;

[0030] The optimized filling model is used to fill in the missing value data in the geological data.

[0031] Optionally, the content of the optimized filling model includes:

[0032]

[0033] Among them, α is the combined weight between the loss function MSE and SIV, is the loss value of MSE, is the loss value of SIV.

[0034] Optionally, the static feature extraction method specifically includes:

[0035] Performing dimensionality reduction on the preprocessed data using PCA to obtain dimensionality-reduced data;

[0036] Using the MBK-means clustering algorithm to compress a portion of the dimensionality reduction data to obtain compressed data;

[0037] Use the uncertain sampling method in AL to label the other part of the dimensionality reduction data to obtain labeled data;

[0038] Training a CNN using the compressed data and the labeled data;

[0039] Use the trained CNN model to extract static features.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] This paper fills missing values through a deep learning model and uses CNTS for outlier detection, which solves the problem of a large number of missing values and outliers in geological data, improves the integrity and accuracy of the data, and provides a high-quality data foundation for subsequent modeling.

[0042] The present invention uses a method that combines PCA dimensionality reduction, MBK-means clustering, AL uncertain sampling and CNN training to extract static features. It can extract features that are of great significance for the prediction of overburden fracture and migration from high-dimensional and complex geological data. At the same time, it combines the dynamic characteristics of the mining process to simulate fracture and migration, realizing the effective fusion of static and dynamic features and solving the problem of inaccurate and incomplete feature extraction.

[0043] The present invention verifies the consistency of the prediction results by combining the physical mechanism of overburden fracture and migration, and generates a dynamic change trend diagram based on the final prediction results, which solves the problem of lack of verification means and low credibility of the prediction results, making the prediction results more scientific and practical. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 This is a flow chart of a method for predicting the breakage and migration of overburden in steeply inclined coal seams based on machine learning according to an embodiment of the present invention;

[0046] Figure 2 Schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] Example 1

[0049] A prediction method for the fracture and migration of overburden in steeply inclined coal seams based on machine learning, such as Figure 1 As shown, the method includes:

[0050] Step S1, collecting geological data of steeply inclined coal seams and performing missing value filling and outlier detection on the geological data to obtain preprocessed data;

[0051] In step S1, the geological data include coal seam inclination, coal seam thickness, roof lithology, roof rock stratum physical and mechanical parameters, ground stress data and mining depth.

[0052] The missing value detection process includes:

[0053] Setting a data set segmentation length and a value missing rate, segmenting the geological data based on the segmentation length, traversing all segmented geological data, and using the geological data with a missing rate less than a preset value missing rate as sample data to generate a sample data set;

[0054] Building a filling model based on the deep learning model, and training the filling model using the sample dataset, MSE and SIV losses to obtain an optimized filling model;

[0055] The optimized filling model is used to fill in the missing value data in the geological data.

[0056] Optionally, the content of the optimized filling model includes:

[0057]

[0058] Among them, α is the combined weight between the loss function MSE and SIV, is the loss value of MSE, is the loss value of SIV.

[0059] Based on this embodiment, the specific calculation methods of the two loss functions are:

[0060]

[0061] Among them, x L For the sample set The data in; remove num in each sample obs *γ values, where num obs *γ represents the sum of the number of filled values and observed values in the sample. And the data to be removed is defined as x drop , the remaining data is x obs It is worth noting that x drop The portion of the filled value is defined as The part belonging to the observation is defined as Get x obs After that, we can input it into the model to get x drop Filling result Same Also need to follow x drop The classification method is divided into and Used for subsequent loss calculations.

[0062] In this embodiment, outlier detection is calculated and implemented using CNTS.

[0063] The goal of reconstruction-based anomaly detection models is to learn an optimization model that can accurately reconstruct normal data samples while failing to effectively reconstruct anomalies (Ruff et al., 2021). Using reconstruction models for anomaly detection is based on the assumption that reconstructed data can distinguish between normal and anomaly data.

[0064]

[0065] The effectiveness of the reconstruction model in detecting anomalies depends on its ability to produce small reconstruction errors for normal values and large reconstruction errors for outliers. This will lead to a relatively low objective function result. In practical modeling, the labels of anomalies are always unknown, and the objective function is as follows:

[0066]

[0067] The reconstructor (R) and detector (D) of CNTS work together to achieve similar goals, ultimately achieving their respective objectives. Specifically, the goal of R is to reconstruct normal samples and distinguish them from abnormal samples, while the goal of D is to detect anomalies. There is a certain degree of overlap between these two goals and they can reinforce each other.

[0068] Therefore, leveraging information between the two networks D and R offers the following advantages. R's reconstruction error can serve as anomaly labels for D during training to guide its training. Furthermore, D can provide information about anomalies to R, reducing the impact of potential outliers on R training. This separation of the anomaly detection task from the data reconstruction task allows CNTS to be modeled as a multi-objective programming problem. The objective function is formulated as:

[0069]

[0070] Among them, D(x i ,θ D ) is the parameter θ D D,R(x i ,θ R ) is the parameter θ R R, and Denote the losses of D and R respectively, is a category label, if D(x i ,θ D ) accounts for all D(x i ,θ D ) first k%, then is equal to 1, otherwise it is equal to 0.

[0071] Step S2: extracting static features from the preprocessed data, collecting dynamic features of mining during the mining process, and performing fracture and migration simulation based on the static features and the dynamic features;

[0072] In step S2, the static feature extraction method specifically includes:

[0073] Performing dimensionality reduction on the preprocessed data using PCA to obtain dimensionality-reduced data;

[0074] Using the MBK-means clustering algorithm to compress a portion of the dimensionality reduction data to obtain compressed data;

[0075] Use the uncertain sampling method in AL to label the other part of the dimensionality reduction data to obtain labeled data;

[0076] Training a CNN using the compressed data and the labeled data;

[0077] Use the trained CNN model to extract static features.

[0078] Specifically: Step 1: Randomly select 2% of the training samples from each class as the initial training set and feed it into C for training. Step 2: Based on the class probabilities provided by the trained CNN, select a certain number of the most informative samples from the remaining unlabeled samples (i.e., the samples with the smallest difference between the best and second-best class probabilities predicted by the CNN) and perform expert annotation. Step 3: Add the difficult-to-classify samples selected in Step 2 to the initial training set as a new training set and feed it back into CNN for training. Step 4: Iterate Step 2 and Step 3 until the number of iterations is reached.

[0079] Principal component analysis (PCA) is used to reduce multidimensional feature data to a lower-dimensional space by extracting key characteristic directions. For example, in overburden fracture and migration prediction, over ten characteristic dimensions may be involved, including coal seam inclination, roof lithology, and mining depth. PCA can reduce these to three or four key characteristic dimensions, which can explain over 80 percent of the variability in the original data.

[0080] In this example, it is important to note that steeply inclined coal seams refer to those with inclinations of 35 to 55 degrees, ranging in thickness from thin (1.3 meters) to extra-thick (8 meters). This angle range significantly impacts mining equipment selection and roadway support planning. Coal seam thickness generally ranges from 0.8 to 3 meters, directly determining the choice of mining techniques and equipment. Roof lithology includes various types, such as sandstone and shale, and its compressive strength and stability are crucial to developing roof management plans. Mining depths range from hundreds of meters to thousands of meters. The greater the depth, the greater the ground pressure, and thus the increased safety risk. During the mining process, dynamic characteristics such as gas emission volume and roof activity intensity are combined with static characteristics to form a time-series feature matrix. For example, in a mining area with a coal seam inclination of 40 degrees, a coal seam thickness of 2.5 meters, a sandstone roof, and a mining depth of 800 meters, the corresponding gas emission volume exhibits a periodic variation pattern, with a peak in gas emission occurring approximately every 100 meters of advance in each working face. Clustering algorithms classify the time series feature matrix, employing density-based clustering methods. Analysis revealed that areas with similar mining depths often exhibit similar gas emission patterns, allowing mining areas to be divided into multiple categories based on mining depth. When data distribution characteristics do not meet pre-defined criteria, regression algorithms are employed to correct them. For example, if gas emission levels in a particular area are abnormally high, multivariate linear regression analysis reveals that the high degree of roof fragmentation in that area leads to an increase in gas channels.

[0081] Static simulation is performed using 3-D finite element method, and the static features are input into AnSYS software for simulation to obtain simulation results;

[0082] The dynamic simulation uses the three-dimensional discrete element method software 3DEC to simulate and obtain the simulation results of dynamic characteristics.

[0083] Step S3: construct an overburden fracture and migration prediction model based on the simulation results, input the fracture and migration data to be detected into the overburden fracture and migration prediction model, output the prediction results, and complete the overburden fracture and migration prediction of the steep-angle coal seam based on machine learning.

[0084] In this embodiment, the prediction model uses a vertical prediction method, and the calculation formula is:

[0085] In(D a )=f(M,V,In(D s )

[0086] Among them, D a is the fracture migration data to be detected, M is the abnormality level, V is the fracture migration cycle speed, D s is the fault distance.

[0087] Real-time monitoring data is acquired to extract key features of overburden fracture and migration. An ensemble approach is used to fuse the random forest and neural network models to calculate the optimal value. Predictions are regenerated based on the optimized value, and the error is determined to be within a preset threshold. If the error exceeds the threshold, the neural network parameters are adjusted and the predictions are recalculated. The predictions are then verified to be consistent with the physical mechanisms of overburden fracture and migration. If the consistency meets the preset criteria, the final prediction is output. Based on the final prediction, a dynamic trend diagram of overburden fracture and migration is generated.

[0088] Example 2

[0089] A machine learning-based prediction system for overburden fracture and migration in steeply inclined coal seams, comprising:

[0090] A data acquisition module is used to collect geological data of steeply inclined coal seams and perform missing value filling and outlier detection on the geological data to obtain preprocessed data;

[0091] Geological data include coal seam images, coal seam dip, coal seam thickness, roof lithology and mining depth.

[0092] The missing value detection process includes:

[0093] Setting a data set segmentation length and a value missing rate, segmenting the geological data based on the segmentation length, traversing all segmented geological data, and using the geological data with a missing rate less than a preset value missing rate as sample data to generate a sample data set;

[0094] Building a filling model based on the deep learning model, and training the filling model using the sample dataset, MSE and SIV losses to obtain an optimized filling model;

[0095] The optimized filling model is used to fill in the missing value data in the geological data.

[0096] Optionally, the content of the optimized filling model includes:

[0097]

[0098] Among them, α is the combined weight between the loss function MSE and SIV, is the loss value of MSE, is the loss value of SIV.

[0099] Based on this embodiment, the specific calculation methods of the two loss functions are:

[0100]

[0101] Among them, x L For the sample set The data in; remove num in each sample obs *γ values, where num obs *γ represents the sum of the number of filled values and observed values in the sample. And the data to be removed is defined as x drop , the remaining data is x obs It is worth noting that x drop The portion of the filled value is defined as The part belonging to the observation is defined as Get x obs After that, we can input it into the model to get x drop Filling result Same Also need to follow x drop The classification method is divided into and Used for subsequent loss calculations.

[0102] In this embodiment, outlier detection is calculated and implemented using CNTS.

[0103] The goal of reconstruction-based anomaly detection models is to learn an optimization model that can accurately reconstruct normal data samples while failing to effectively reconstruct anomalies (Ruff et al., 2021). Using reconstruction models for anomaly detection is based on the assumption that reconstructed data can distinguish between normal and anomaly data.

[0104]

[0105] The effectiveness of the reconstruction model in detecting anomalies depends on its ability to produce small reconstruction errors for normal values and large reconstruction errors for outliers. This will lead to a relatively low objective function result. In practical modeling, the labels of anomalies are always unknown, and the objective function is as follows:

[0106]

[0107] The reconstructor (R) and detector (D) of CNTS work together to achieve similar goals, ultimately achieving their respective objectives. Specifically, the goal of R is to reconstruct normal samples and distinguish them from abnormal samples, while the goal of D is to detect anomalies. There is a certain degree of overlap between these two goals and they can reinforce each other.

[0108] Therefore, leveraging information between the two networks D and R offers the following advantages. R's reconstruction error can serve as anomaly labels for D during training to guide its training. Furthermore, D can provide information about anomalies to R, reducing the impact of potential outliers on R training. This separation of the anomaly detection task from the data reconstruction task allows CNTS to be modeled as a multi-objective programming problem. The objective function is formulated as:

[0109]

[0110] Among them, D(x i ,θ D ) is the parameter θ D D,R(x i ,θ R ) is the parameter θ R R, and Denote the losses of D and R respectively, is a category label, if D(x i ,θ D ) accounts for all D(x i ,θ D ) first k%, then is equal to 1, otherwise it is equal to 0.

[0111] A mining simulation module, configured to extract static features from the preprocessed data, collect dynamic features of mining during the mining process, and perform fracture and migration simulation based on the static and dynamic features;

[0112] The static feature extraction methods specifically include:

[0113] Performing dimensionality reduction on the preprocessed data using PCA to obtain dimensionality-reduced data;

[0114] Using the MBK-means clustering algorithm to compress a portion of the dimensionality reduction data to obtain compressed data;

[0115] Use the uncertain sampling method in AL to label the other part of the dimensionality reduction data to obtain labeled data;

[0116] Training a CNN using the compressed data and the labeled data;

[0117] Use the trained CNN model to extract static features.

[0118] Specifically: Step 1: Randomly select 2% of the training samples from each class as the initial training set and feed it into C for training. Step 2: Based on the class probabilities provided by the trained CNN, select a certain number of the most informative samples from the remaining unlabeled samples (i.e., the samples with the smallest difference between the best and second-best class probabilities predicted by the CNN) and perform expert annotation. Step 3: Add the difficult-to-classify samples selected in Step 2 to the initial training set as a new training set and feed it back into CNN for training. Step 4: Iterate Step 2 and Step 3 until the number of iterations is reached.

[0119] Principal component analysis (PCA) is used to reduce multidimensional feature data to a lower-dimensional space by extracting key characteristic directions. For example, in overburden fracture and migration prediction, over ten characteristic dimensions may be involved, including coal seam inclination, roof lithology, and mining depth. PCA can reduce these to three or four key characteristic dimensions, which can explain over 80 percent of the variability in the original data.

[0120] In this example, it is important to note that steeply inclined coal seams refer to those with inclinations of 35 to 55 degrees, ranging in thickness from thin (1.3 meters) to extra-thick (8 meters). This angle range significantly impacts mining equipment selection and roadway support planning. Coal seam thickness generally ranges from 0.8 to 3 meters, directly determining the choice of mining techniques and equipment. Roof lithology includes various types, such as sandstone and shale, and its compressive strength and stability are crucial to developing roof management plans. Mining depths range from hundreds of meters to thousands of meters. The greater the depth, the greater the ground pressure, and thus the increased safety risk. During the mining process, dynamic characteristics such as gas emission volume and roof activity intensity are combined with static characteristics to form a time-series feature matrix. For example, in a mining area with a coal seam inclination of 40 degrees, a coal seam thickness of 2.5 meters, a sandstone roof, and a mining depth of 800 meters, the corresponding gas emission volume exhibits a periodic variation pattern, with a peak in gas emission occurring approximately every 100 meters of advance in each working face. Clustering algorithms classify the time series feature matrix, employing density-based clustering methods. Analysis revealed that areas with similar mining depths often exhibit similar gas emission patterns, allowing mining areas to be divided into multiple categories based on mining depth. When data distribution characteristics do not meet pre-defined criteria, regression algorithms are employed to correct them. For example, if gas emission levels in a particular area are abnormally high, multivariate linear regression analysis reveals that the high degree of roof fragmentation in that area leads to an increase in gas channels.

[0121] The migration prediction module is used to construct a rock breakage and migration prediction model based on the simulation results, input the breakage and migration data to be detected into the rock breakage and migration prediction model, output the prediction results, and complete the overburden breakage and migration prediction of high-angle coal seams based on machine learning.

[0122] Real-time monitoring data is acquired to extract key features of overburden fracture and migration. An ensemble approach is used to fuse the random forest and neural network models to calculate the optimal value. Predictions are regenerated based on the optimized value, and the error is determined to be within a preset threshold. If the error exceeds the threshold, the neural network parameters are adjusted and the predictions are recalculated. The predictions are then verified to be consistent with the physical mechanisms of overburden fracture and migration. If the consistency meets the preset criteria, the final prediction is output. Based on the final prediction, a dynamic trend diagram of overburden fracture and migration is generated.

[0123] Example 3

[0124] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the program, the method for predicting the fracture and migration of overburden in high-angle coal seams based on machine learning as described in any of the above embodiments is implemented.

[0125] Figure 2 10 is a schematic diagram showing a more specific hardware structure of an electronic device provided in this embodiment. The device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are communicatively connected to each other within the device via the bus 1050.

[0126] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0127] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage devices, dynamic storage devices, etc. The memory 1020 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0128] The input / output interface 1030 is used to connect input / output modules to implement information input and output. The input / output modules can be configured as components within the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. Input devices may include a keyboard, mouse, touch screen, microphone, various sensors, etc., and output devices may include a display, speaker, vibrator, indicator light, etc.

[0129] The communication interface 1040 is used to connect to a communication module (not shown) to enable communication between the device and other devices. The communication module can communicate via a wired method (such as USB (Universal Serial Bus), network cable, etc.) or a wireless method (such as mobile network, WIFI (Wireless Fidelity), Bluetooth, etc.).

[0130] The bus 1050 comprises a path for transmitting information between the various components of the device (eg, the processor 1010 , the memory 1020 , the input / output interface 1030 , and the communication interface 1040 ).

[0131] It should be noted that although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040, and the bus 1050, in a specific implementation, the device may also include other components necessary for normal operation. In addition, it will be understood by those skilled in the art that the above device may only include the components necessary to implement the embodiments of this specification, and does not necessarily include all the components shown in the figure.

[0132] The system of the above embodiment is used to implement the corresponding machine learning-based high-angle coal seam overburden fracture and migration prediction method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0133] Example 4

[0134] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present disclosure also provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute the high-angle coal seam overburden fracture and migration prediction method based on machine learning as described in any of the above embodiments.

[0135] The computer-readable media of this embodiment include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0136] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the machine learning-based method for predicting the fracture and migration of overburden in high-angle coal seams as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0137] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present disclosure (including the claims) is limited to these examples. Within the scope of the present disclosure, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present disclosure as described above, which are not provided in detail for the sake of simplicity.

[0138] In addition, to simplify the description and discussion, and so as not to obscure the embodiments of the present disclosure, known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided figures. In addition, devices may be shown in the form of block diagrams to avoid obscuring the embodiments of the present disclosure, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the present disclosure are to be implemented (i.e., these details should be fully within the purview of those skilled in the art). Where specific details (e.g., circuits) are set forth to describe exemplary embodiments of the present disclosure, it will be apparent to those skilled in the art that the embodiments of the present disclosure may be implemented without these specific details or with variations in these specific details. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0139] Although the present disclosure has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0140] Therefore, the units of each example described in the embodiments of this application can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0141] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A method for predicting the breakage and migration of overburden in steeply inclined coal seams based on machine learning, characterized in that: The method comprises: Step S1, collecting geological data of steeply inclined coal seams and performing missing value filling and outlier detection on the geological data to obtain preprocessed data; Step S2: extracting static features from the preprocessed data, collecting dynamic features of mining during the mining process, and performing fracture and migration simulation based on the static features and the dynamic features; Step S3: construct an overburden fracture and migration prediction model based on the simulation results, input the fracture and migration data to be detected into the overburden fracture and migration prediction model, output the prediction results, and complete the overburden fracture and migration prediction of the steep-angle coal seam based on machine learning.

2. The method for predicting the breakage and migration of overburden in steeply inclined coal seams based on machine learning according to claim 1 is characterized in that: In step S1, the geological data includes coal seam inclination, coal seam thickness, roof lithology, physical and mechanical parameters of roof rock layer, ground stress data and mining depth.

3. The method for predicting the breakage and migration of overburden in steeply inclined coal seams based on machine learning according to claim 2 is characterized in that: The missing value detection process includes: Setting a data set segmentation length and a value missing rate, segmenting the geological data based on the segmentation length, traversing all segmented geological data, and using the geological data with a missing rate less than a preset value missing rate as sample data to generate a sample data set; Building a filling model based on the deep learning model, and training the filling model using the sample dataset, MSE and SIV losses to obtain an optimized filling model; The optimized filling model is used to fill in the missing value data in the geological data.

4. The method for predicting the breakage and migration of overburden in steeply inclined coal seams based on machine learning according to claim 3 is characterized in that: The content of the optimized filling model includes: Among them, α is the combined weight between the loss function MSE and SIV, is the loss value of MSE, is the loss value of SIV.

5. The method for predicting the breakage and migration of overburden in steeply inclined coal seams based on machine learning according to claim 1 is characterized in that: In step S2, the static feature extraction method specifically includes: Performing dimensionality reduction on the preprocessed data using PCA to obtain dimensionality-reduced data; Using the MBK-means clustering algorithm to compress a portion of the dimensionality reduction data to obtain compressed data; Use the uncertain sampling method in AL to label the other part of the dimensionality reduction data to obtain labeled data; Training a CNN using the compressed data and the labeled data; Use the trained CNN model to extract static features.

6. A machine learning-based prediction system for the breakage and migration of overburden in steeply inclined coal seams, the system being used to implement the prediction method according to any one of claims 1 to 5, characterized in that the system include: A data acquisition module is used to collect geological data of steeply inclined coal seams and perform missing value filling and outlier detection on the geological data to obtain preprocessed data; A mining simulation module, configured to extract static features from the preprocessed data, collect dynamic features of mining during the mining process, and perform fracture and migration simulation based on the static and dynamic features; The migration prediction module is used to construct a rock breakage and migration prediction model based on the simulation results, input the breakage and migration data to be detected into the rock breakage and migration prediction model, output the prediction results, and complete the overburden breakage and migration prediction of high-angle coal seams based on machine learning.

7. The machine learning-based prediction system for overburden fracture and migration in steeply inclined coal seams according to claim 6 is characterized in that: The missing value detection process includes: Setting a data set segmentation length and a value missing rate, segmenting the geological data based on the segmentation length, traversing all segmented geological data, and using the geological data with a missing rate less than a preset value missing rate as sample data to generate a sample data set; Building a filling model based on the deep learning model, and training the filling model using the sample dataset, MSE and SIV losses to obtain an optimized filling model; The optimized filling model is used to fill in the missing value data in the geological data.

8. The machine learning-based prediction system for overburden fracture and migration in steeply inclined coal seams according to claim 7 is characterized in that: The content of the optimized filling model includes: Among them, α is the combined weight between the loss function MSE and SIV, is the loss value of MSE, is the loss value of SIV.

9. The machine learning-based prediction system for overburden fracture and migration in steeply inclined coal seams according to claim 6, characterized in that: The static feature extraction methods specifically include: Performing dimensionality reduction on the preprocessed data using PCA to obtain dimensionality-reduced data; Using the MBK-means clustering algorithm to compress a portion of the dimensionality reduction data to obtain compressed data; Use the uncertain sampling method in AL to label the other part of the dimensionality reduction data to obtain labeled data; Training a CNN using the compressed data and the labeled data; Use the trained CNN model to extract static features.

Images