Coal seam three-dimensional stress prediction and visualization method, system, equipment and medium

Abstract

The invention relates to the technical field of coal seam stress prediction, in particular to a coal seam three-dimensional stress prediction and visualization method, system and device and a medium, and the method comprises the steps: setting an artificial seismic source and a detection point, exciting a seismic source signal, and collecting a seismic wave response signal and a stress value; carrying out wave field forward modeling simulation through a time fractional order viscoelastic wave equation to obtain a forward simulation wave field, and carrying out full waveform inversion to output three-dimensional wave velocity data; performing reverse time migration imaging on the seismic wave response signal to obtain a three-dimensional coal seam structure mask pattern; performing slicing processing on the three-dimensional wave velocity data and the three-dimensional coal seam structure mask pattern along the same spatial dimension, then combining the three-dimensional wave velocity data and the three-dimensional coal seam structure mask pattern into an input tensor, inputting the input tensor into a convolutional neural network, and outputting a two-dimensional stress prediction slice; and splicing the two-dimensional stress prediction slices to construct a three-dimensional stress body, generating three-dimensional stress point cloud data, and further generating a three-dimensional stress grid model and a structure-stress integrated visual three-dimensional cloud picture. According to the invention, high-precision three-dimensional stress prediction and real-time early warning of the coal seam can be realized.

Description

Technical Field

[0001] The present application relates to the technical field of coal seam stress prediction, and in particular to a method, system, equipment and medium for three-dimensional stress prediction and visualization of coal seams. Background Art

[0002] As coal mining continues to increase in depth and intensity, the geological structure within mines is becoming increasingly complex, with abnormal structures such as faults, fracture zones, and weak interlayers becoming commonplace. These geological anomalies can easily trigger dynamic hazards such as rock bursts, coal and gas outbursts, and roof collapse, seriously threatening miners' lives and the stability of coal mine production. Therefore, developing high-precision, intelligent coal seam geological detection technology to achieve real-time identification and early warning of disaster sources has become a core requirement for ensuring coal mine safety.

[0003] Existing technologies primarily use seismic wave reflection / transmission methods for geological imaging, inferring the regional stress field based on data from anchor stress monitoring points, and employing spatial interpolation methods such as Kriging to predict coal seam stress distribution. For example, seismic wave inversion based on the elastic wave equation is used to obtain wave velocity data, and stress prediction maps are generated by manually calibrating the linear or power function relationship between wave velocity and stress. Furthermore, some existing technologies also use techniques such as reverse time migration imaging to identify geological boundaries and assist in the construction of static risk zone display systems to guide mining operations.

[0004] However, the existing prediction technology for coal seam stress distribution has significant shortcomings: first, the existing wave velocity and stress mapping method relies on a simple empirical model, which cannot fully capture the nonlinear coupling characteristics of the coal rock medium, resulting in low stress prediction accuracy; second, the existing seismic wave simulation method is based on the assumption of unattenuated elasticity, ignoring the viscoelastic dissipation and dispersion effects of coal rock, resulting in high-frequency attenuation and phase delay not being modeled, and severe wave velocity inversion distortion; third, the stress prediction process does not integrate geological boundary information, resulting in the "cross-fault misinsertion" phenomenon, which destroys the continuity of the stress field in complex tectonic areas and affects detection reliability. Summary of the Invention

[0005] In order to address the technical problems that the existing coal seam stress distribution prediction technology relies on a simple empirical model of wave velocity-stress, ignores the viscoelastic dissipation effect of coal rock, and does not integrate geological boundary information, resulting in serious stress prediction distortion and cross-fault misinterpretation, the present application provides a coal seam three-dimensional stress prediction and visualization method, system, equipment and medium. Through viscoelastic wave field forward modeling, deep learning mapping with structural mask constraints and a three-dimensional visualization full-process closed loop, high-precision three-dimensional stress prediction and real-time early warning are achieved, solving the problem of detection distortion.

[0006] In a first aspect, the present application provides a method for three-dimensional stress prediction and visualization of coal seams, comprising the following steps: S1. Set up an artificial seismic source and multiple detection points in the coal seam to be detected, stimulate the source signal at the artificial seismic source, and simultaneously collect seismic wave response signals and stress values ​​at each detection point; S2. Based on the source signal, the wave field is forward modeled using the time fractional viscoelastic wave equation to obtain the forward simulated wave field. , in which a short-memory fast algorithm is used to solve the fractional derivative terms of the time fractional-order viscoelastic wave equation; Perform full waveform inversion based on forward simulation wave field and seismic wave response signal to output three-dimensional wave velocity data of the area to be detected ; in, is the spatial position vector; is the time variable; S3. Perform reverse time migration imaging on the seismic wave response signal based on the 3D wave velocity data to obtain a 3D coal seam structure mask map, including: Time-reverse the seismic wave response signals of each detection point to generate virtual source signals; Using the virtual source signal as the excitation source, the wavefield is back-propagated based on the three-dimensional wave velocity data to generate the back-propagation wavefield; Perform time-space cross-correlation calculation on the forward simulation wave field and the reverse propagation wave field to obtain the three-dimensional imaging value; Perform gradient enhancement and threshold segmentation on the three-dimensional imaging values ​​to generate a three-dimensional coal seam structure mask map that identifies the geological boundary; S4. Slice the 3D velocity data and the 3D coal seam structure mask image along the same spatial dimension and combine them into an input tensor. This is then fed into a pre-trained convolutional neural network to output a 2D stress prediction slice. It should be noted that the "prediction" here refers to the inference of the stress spatial distribution in the unmeasured area based on wave velocity data and structural masks through convolutional neural networks (CNN). It is a data reconstruction in the spatial dimension, not a prediction of the future state in the temporal dimension. S5. Construct a 3D stress volume by stitching all 2D stress prediction slices together, and generate 3D stress point cloud data based on the 3D stress volume. The 3D stress point cloud data is converted into a 3D stress mesh model through a 3D reconstruction algorithm; Use mechanical analysis methods to delineate dangerous areas in a three-dimensional stress grid model; The three-dimensional stress grid model and the three-dimensional coal seam structure mask map are combined and then rendered with stress value-color mapping to generate a structure-stress integrated visual three-dimensional cloud map, and the dangerous areas are marked in the visual three-dimensional cloud map.

[0007] It should be further explained that, in step S1, a channel wave geophone and a stress monitoring anchor are configured at each detection point, the channel wave geophone is used to collect seismic wave response signals, and the stress monitoring anchor is used to collect stress values.

[0008] It should be further explained that in step S2, the time fractional-order viscoelastic wave equation is:

[0009] in, is the coal seam medium density; is the displacement vector field; is the displacement vector field in The direction component, ; are the stress tensor components, where represents the normal direction of the stress action plane, represents the direction of stress action, ; for The spatial coordinate component of the direction; is the source term, representing The body force in the direction satisfies:

[0010] is the coordinate of the artificial earthquake source; is the spatial Dirac function; is the time function of the source signal; is the viscoelastic constant, calibrated by laboratory dynamic mechanical testing; is the Kronick function, hour =1, otherwise =0; is the fractional order, ∈[0.3,0.8]; is the displacement divergence; is the strain tensor, ; The formula for solving the fractional derivative term using the short memory fast algorithm is:

[0011] in, is the current time step; The step length is 1 time step; L is the finite memory window length; is the fractional order The weight coefficient of the decision; The wave field forward simulation uses the finite difference method to discretize the wave equation and calculate the forward simulation wave field in the full range of the spatial domain. .

[0012] It should be further explained that step S3 includes: S301. Based on seismic wave response signal Generate a virtual source signal, the expression is:

[0013] in, For the Detection points at time seismic wave response signal; is the total recording time of the seismic wave response signal; For the Detection points at time The virtual source model of the monitoring point; Will As a virtual source applied to the position of the corresponding detection point, the wave field is reversely propagated using the same time fractional viscoelastic wave equation based on the three-dimensional wave velocity data to obtain the reverse wave field ; S302. Calculating 3D Imaging Values : ; S303. After performing gradient enhancement and high-pass filtering on the three-dimensional imaging value, set the threshold , by threshold Generate a structural mask map, the expression is: .

[0014] It should be further explained that, in step S3, the geological boundaries include geological faults and interlayer boundaries.

[0015] It should be further explained that step S4 includes: S401. 3D wave velocity data With 3D coal seam structure mask map Slice along the Y axis at a fixed step size to obtain a wave velocity data slice sequence and coal seam structure mask slice sequence ; in, is the total number of slices, Indicates the slices; S402. Combine the velocity data slice and the corresponding coal seam structure mask slice into an input tensor: ; S403. Input the input tensor into the pre-trained convolutional neural network and output a 2D stress prediction slice. .

[0016] It should be further explained that the convolutional neural network is a U-Net encoding-decoding structure, including an encoder and a decoder, with jump connections between the encoder and the decoder, where: The encoder consists of multiple sequentially connected encoding blocks. Each encoding block contains two 3×3 convolutional layers. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function. At the end of each encoding block, a 2×2 maximum pooling layer with a stride of 2 is connected to perform spatial downsampling. The decoder consists of multiple sequentially connected decoding blocks. Each decoding block contains a 2×2 transposed convolutional layer with a stride of 2 for spatial upsampling. The transposed convolutional layer is followed by a feature concatenation layer, which concatenates the feature maps output by the transposed convolutional layer with the encoder feature maps of the corresponding level in the channel dimension. After concatenation, two 3×3 convolutional layers are connected. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function. The skip connection connects the encoder feature map output by each encoding block in the encoder to the feature concatenation layer of the corresponding level in the decoder.

[0017] It should be further explained that the structural boundary attention mechanism is used in the skip connection process. After obtaining the encoder feature map, the mask feature map after downsampling the structural mask map is first fused with the encoder feature map to obtain a fused feature map, and then the fused feature map is input into the feature splicing layer. The specific fusion method of the structural boundary attention mechanism is: For each encoder depth level , perform the calculation:

[0018] Indicates the Feature maps output by the layer encoder; Structure mask image The mask feature map after downsampling; is a trainable guidance factor, with a value range of [0.5, 2.0]; Represents element-wise multiplication operation; Represents the fused feature map.

[0019] It should be further explained that the training samples used by the convolutional neural network come from the three-dimensional wave velocity data obtained by inverting the seismic wave response signal through the time fractional viscoelastic wave equation and the three-dimensional coal seam structure mask image extracted based on the reverse time migration imaging technology; The convolutional neural network uses measured stress interpolation maps as labels for semi-supervised training and is jointly trained on multiple geological samples. The measured stress interpolation map is a coarse-resolution stress distribution map generated by interpolation based on the measured stress values ​​at each detection point. The loss function of the convolutional neural network is:

[0020] in, For The measured stress interpolation diagram composed of the stress values ​​of all detection points with equal height on the Y axis; 、 、 is the weight coefficient,

[0021] MAE is the mean absolute error; SSIM is the structural similarity index; Edgeloss is the boundary alignment loss, which is the difference between the predicted stress gradient and the measured stress gradient at the geological boundary identified by the 3D coal seam structure mask.

[0022] It should be further explained that, in step S5, the step of generating three-dimensional stress point cloud data includes: S501. Slice all two-dimensional predictions Splice back to the three-dimensional coordinate axis in the original spatial order to form a complete three-dimensional stress body: ; S502. Perform voxel sampling on the 3D stress volume, extract voxel coordinates and stress values, and generate 3D stress point cloud data:

[0023] in, The first The spatial coordinates of the sampling points; The first The predicted stress value of each sampling point; Displays the lower threshold value for stress.

[0024] It should be further explained that, in step S5, the mechanical analysis method includes the finite element method (FEM), the discrete element method (DEM), and the boundary element method (BEM).

[0025] It should be further explained that in step S5, during the stress value-color mapping rendering process, the corresponding relationship between stress value and color is: .

[0026] In a second aspect, the present application provides a three-dimensional coal seam stress prediction and visualization system for implementing the above-mentioned three-dimensional coal seam stress prediction and visualization method, comprising: The data acquisition module is used to set an artificial seismic source and multiple detection points in the area to be detected in the coal seam, stimulate the source signal at the artificial seismic source, and synchronously collect the seismic wave response signal and stress value of each detection point; The wavefield forward modeling module is used to perform wavefield forward modeling based on the source signal using the time fractional viscoelastic wave equation to obtain the forward simulated wavefield; Full waveform inversion module, used to perform full waveform inversion based on the forward simulation wave field and seismic wave response signal, and output three-dimensional wave velocity data of the area to be detected; A reverse time migration imaging module is used to perform reverse time migration imaging on the seismic wave response signal based on the three-dimensional wave velocity data to obtain a three-dimensional coal seam structure mask map; The 2D stress prediction module is used to combine the 3D wave velocity data and the 3D coal seam structure mask image into an input tensor after slicing along the same spatial dimension. The input is fed into a pre-trained convolutional neural network and the 2D stress prediction slice is output. 3D visualization module for: All 2D stress prediction slices are spliced ​​together to construct a 3D stress body, and 3D stress point cloud data is generated based on the 3D stress body; The 3D stress point cloud data is converted into a 3D stress mesh model through a 3D reconstruction algorithm; Use mechanical analysis methods to delineate dangerous areas in a three-dimensional stress grid model; The three-dimensional stress grid model and the three-dimensional coal seam structure mask map are combined and then rendered with stress value-color mapping to generate a structure-stress integrated visual three-dimensional cloud map, and the dangerous areas are marked in the visual three-dimensional cloud map.

[0027] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to implement the steps of the above-mentioned coal seam three-dimensional stress prediction and visualization method when executing the computer program.

[0028] In a fourth aspect, the present application provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned coal seam three-dimensional stress prediction and visualization method.

[0029] It can be seen from the above technical solutions that this application has the following advantages: 1. This application sets up an artificial seismic source and multiple detection points in the area to be detected in the coal seam, synchronously collects seismic wave response signals and stress values, and performs wave field forward simulation based on the time fractional viscoelastic wave equation. It uses a short-memory fast algorithm to solve the fractional-order derivative terms, thereby solving the wave velocity distortion problem caused by the existing technology ignoring the viscoelastic characteristics of coal rock, achieving more accurate three-dimensional wave velocity data inversion, and effectively simulating the energy dissipation and phase delay of seismic waves in coal seams.

[0030] 2. This application performs reverse time migration imaging based on three-dimensional wave velocity data, including time reversal to generate virtual source signals, wave field backpropagation and cross-correlation calculation, and generates a three-dimensional coal seam structure mask map through gradient enhancement and threshold segmentation. It solves the problem of cross-fault misinterpretation caused by the lack of structural information in existing stress prediction technology, realizes the precise identification of geological faults, and ensures the spatial continuity of stress field modeling.

[0031] 3. This application solves the low accuracy problem of the existing simple velocity-stress mapping model by combining three-dimensional velocity data and three-dimensional coal seam structure mask slices into input tensors, inputs them into a convolutional neural network, and supervises training with stress values. It realizes end-to-end nonlinear stress prediction and improves the accuracy of two-dimensional stress slice output.

[0032] 4. This application constructs a three-dimensional stress body by splicing all two-dimensional stress prediction slices, generates three-dimensional stress point cloud data, converts the three-dimensional stress point cloud data into a three-dimensional stress grid model, and then combines it with a three-dimensional coal seam structure mask map for stress value-color mapping rendering to generate a structure-stress integrated visual three-dimensional cloud map, and marks dangerous areas in the visual three-dimensional cloud map. This solves the problems of poor interactivity and update delays in existing visualization methods, realizes real-time structure-stress integrated three-dimensional cloud map display, and can support second-level early warning in high-risk areas of coal mines.

[0033] 5. This application solves the problems of error accumulation and poor real-time performance caused by step-by-step independent processing in existing technologies through integrated processing of the entire process from viscoelastic wave field forward modeling, full waveform inversion, reverse time migration imaging to deep learning stress prediction, and realizes a high-precision and high-efficiency closed loop for disaster source detection and early warning. It can dynamically update three-dimensional stress cloud maps on coal mining faces, thereby improving disaster response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0035] Figure 1 This is a flow chart of a method for three-dimensional stress prediction and visualization of coal seams in one embodiment of the present application.

[0036] Figure 2 It is a schematic block diagram of a coal seam three-dimensional stress prediction and visualization system in one embodiment of the present application.

[0037] Figure 3 It is a schematic diagram of the hardware structure of an electronic device in one embodiment of the present application. DETAILED DESCRIPTION

[0038] In order to make the application objectives, features, and advantages of this application more obvious and easy to understand, the technical solutions protected by this application will be clearly and completely described below using specific embodiments and drawings. Obviously, the embodiments described below are only part of the embodiments of this application, not all of them. Based on the embodiments in this patent, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this patent.

[0039] The following describes in detail the coal seam three-dimensional stress prediction and visualization method involved in this application. Specific details, such as specific system structures and techniques, are provided for illustrative purposes, not for limitation, to facilitate a thorough understanding of the embodiments of this application. However, those skilled in the art will appreciate that this application may also be implemented in other embodiments without these specific details.

[0040] In the coal seam three-dimensional stress prediction and visualization method involved in this application, the term "comprising" is used to indicate the presence of the described features, entities, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, entities, steps, operations, elements, components, and / or their collections. The terms "including," "comprising," "having," and their variations all mean "including but not limited to," unless otherwise specifically emphasized.

[0041] To facilitate the clear description of the technical solutions of this application, the words "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or order of execution, and the words "first" and "second" do not necessarily mean different.

[0042] The phrases "one embodiment" or "some embodiments" described in this application mean that the particular features, structures, or characteristics described in the embodiment are included in one or more embodiments of the application. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in other embodiments," etc. that appear in different places in this application do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized.

[0043] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application.

[0044] The coal seam three-dimensional stress prediction and visualization method provided in the embodiment of the present application is executed by a computer device, and accordingly, the coal seam three-dimensional stress prediction and visualization system runs in the computer device.

[0045] Figure 1 This is a flow chart of a method for three-dimensional stress prediction and visualization of coal seams according to an embodiment of the present application. Figure 1 The execution subject may be a coal seam three-dimensional stress prediction and visualization system. According to different requirements, the order of the steps in the flow chart may be changed, and some steps may be omitted.

[0046] like Figure 1 As shown, the coal seam three-dimensional stress prediction and visualization method includes: Step S1: an artificial seismic source and multiple detection points are set in the area to be detected in the coal seam, a source signal is excited at the artificial seismic source, and seismic wave response signals and stress values ​​of each detection point are synchronously collected.

[0047] By setting up artificial seismic sources and detection points, stimulating source signals, and synchronously collecting seismic wave response signals and stress values ​​at each detection point, the synchronous acquisition of original seismic wave and stress data is achieved, which can ensure the time alignment and spatial coverage of the data, provide basic input for subsequent wave field forward simulation and full waveform inversion, and avoid error accumulation caused by asynchronous data acquisition. In some specific embodiments, a channel wave geophone and a stress monitoring anchor are configured at each detection point, the channel wave geophone is used to collect seismic wave response signals, and the stress monitoring anchor is used to collect stress values.

[0048] By configuring slot wave detectors at each detection point to collect seismic wave response signals and using stress monitoring anchors to collect stress values, the synchronous and high-precision collection of seismic wave signals and stress values ​​is achieved. This ensures data consistency and reliability, avoids data deviations caused by equipment mismatch, and provides an accurate input basis for subsequent wave field forward simulation and full waveform inversion.

[0049] Step S2: Based on the source signal, the wave field is forward modeled using the time fractional viscoelastic wave equation to obtain the forward simulated wave field. , in which a short-memory fast algorithm is used to solve the fractional derivative terms of the time fractional-order viscoelastic wave equation; Perform full waveform inversion based on forward simulation wave field and seismic wave response signal to output three-dimensional wave velocity data of the area to be detected ; in, is the spatial position vector; is the time variable.

[0050] By using the time fractional-order viscoelastic wave equation to perform wave field forward simulation, and combining the forward simulated wave field with the seismic wave response signal to perform full waveform inversion to output three-dimensional velocity data, high-precision inversion of the coal seam velocity characteristics is achieved, which can accurately capture the viscoelastic behavior of the medium and provide a reliable three-dimensional velocity data foundation for reverse time migration imaging.

[0051] In some specific embodiments, the time fractional viscoelastic wave equation is:

[0052] in, is the coal seam medium density; is the displacement vector field; is the displacement vector field in The direction component, ; are the stress tensor components, where represents the normal direction of the stress action plane, represents the direction of stress action, ; for The spatial coordinate component of the direction; is the source term, representing The body force in the direction satisfies:

[0053] is the coordinate of the artificial earthquake source; is the spatial Dirac function; is the time function of the source signal; is the viscoelastic constant, calibrated by laboratory dynamic mechanical testing; is the Kronick function, hour =1, otherwise =0; is the fractional order, ∈[0.3,0.8]; is the displacement divergence; is the strain tensor, ; The formula for solving the fractional derivative term using the short memory fast algorithm is:

[0054] in, is the current time step; The step length is 1 time step; L is the finite memory window length; is the fractional order The weight coefficient of the decision; The wave field forward simulation uses the finite difference method to discretize the wave equation and calculate the forward simulation wave field in the full range of the spatial domain. .

[0055] By adopting the time fractional-order viscoelastic wave equation and using a short-memory fast algorithm to solve the fractional-order derivative terms (the specific formula is the approximate summation weight coefficient and the finite memory window length), combined with the finite difference method to discretize the wave equation for wave field forward simulation, accurate simulation of the viscoelastic characteristics of coal seam media is achieved, and the forward simulation wave field in the full range of the spatial domain can be efficiently calculated, thereby improving the accuracy of the three-dimensional wave velocity data of the full waveform inversion.

[0056] Step S3, performing reverse time migration imaging on the seismic wave response signal based on the three-dimensional wave velocity data to obtain a three-dimensional coal seam structure mask map, including: Time-reverse the seismic wave response signals of each detection point to generate virtual source signals; Using the virtual source signal as the excitation source, the wavefield is back-propagated based on the three-dimensional wave velocity data to generate the back-propagation wavefield; Perform time-space cross-correlation calculation on the forward simulation wave field and the reverse propagation wave field to obtain the three-dimensional imaging value; The three-dimensional imaging values ​​are subjected to gradient enhancement and threshold segmentation to generate a three-dimensional coal seam structure mask map that identifies the geological boundary.

[0057] By performing reverse time migration imaging on the seismic wave response signal based on three-dimensional wave velocity data, the clear extraction of geological boundaries is achieved, which can enhance the boundary recognition ability of the coal seam structure mask map and provide structural constraint information for the convolutional neural network input. In some specific embodiments, step S3 includes: S301. Based on seismic wave response signal Generate a virtual source signal, the expression is:

[0058] in, For the Detection points at time seismic wave response signal; is the total recording time of the seismic wave response signal; For the Detection points at time The virtual source model of the monitoring point; Will As a virtual source applied to the position of the corresponding detection point, the wave field is reversely propagated using the same time fractional viscoelastic wave equation based on the three-dimensional wave velocity data to obtain the reverse wave field ; S302. Calculating 3D Imaging Values : ; S303. After performing gradient enhancement and high-pass filtering on the three-dimensional imaging value, set the threshold , by threshold Generate a structural mask map, the expression is: .

[0059] By generating a virtual source signal (expressed as a time reversal function) based on the seismic wave response signal, using the virtual source signal as the excitation source to perform wave field backpropagation based on three-dimensional wave velocity data to generate a reverse wave field, calculating the time-space domain cross-correlation integral of the forward simulation wave field and the reverse propagation wave field to obtain a three-dimensional imaging value, and performing gradient enhancement and high-pass filtering on the three-dimensional imaging value and setting a threshold to generate a structural mask map, a clear identification of the geological boundary of the coal seam is achieved, which can enhance the boundary recognition ability of reverse time migration imaging and reduce noise interference. In some embodiments, the geological boundaries include geological faults and interlayer boundaries.

[0060] By defining geological boundaries as including geological faults and interlayer boundaries, targeted processing of structural mask maps in reverse time migration imaging is achieved, which can focus on the detection of key geological structures and improve the practicality of coal seam structural mask maps and the accuracy of geological interpretation.

[0061] In step S4, the three-dimensional wave velocity data and the three-dimensional coal seam structure mask image are sliced ​​along the same spatial dimension and then combined into an input tensor, which is input into a pre-trained convolutional neural network to output a two-dimensional stress prediction slice.

[0062] By using convolutional neural networks to output two-dimensional stress prediction slices, effective fusion of wave velocity and structural characteristics and deep learning modeling are achieved, which can improve the local accuracy and generalization of stress prediction and provide high-quality slice data for the construction of three-dimensional stress bodies. In some specific embodiments, step S4 includes: S401. 3D wave velocity data With 3D coal seam structure mask map Slice along the Y axis at a fixed step size to obtain a wave velocity data slice sequence and coal seam structure mask slice sequence ; in, is the total number of slices, Indicates the slices; S402. Combine the velocity data slice and the corresponding coal seam structure mask slice into an input tensor: ; S403. Input the input tensor into the pre-trained convolutional neural network and output a 2D stress prediction slice. .

[0063] By slicing the three-dimensional velocity data and the three-dimensional coal seam structure mask map along the Y-axis at a fixed step size, a velocity data slice sequence and a coal seam structure mask map slice sequence are obtained, and the velocity data slices and the corresponding coal seam structure mask map slices are combined as input tensors to input the convolutional neural network to output two-dimensional stress prediction slices. This achieves efficient data preprocessing and feature fusion, optimizes the input structure of the convolutional neural network, and improves the local accuracy and computational efficiency of stress prediction.

[0064] In some specific embodiments, the convolutional neural network is a U-Net encoding-decoding structure, including an encoder and a decoder, and a jump connection is used between the encoder and the decoder, wherein: The encoder consists of multiple sequentially connected encoding blocks. Each encoding block contains two 3×3 convolutional layers. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function. At the end of each encoding block, a 2×2 maximum pooling layer with a stride of 2 is connected to perform spatial downsampling. The decoder consists of multiple sequentially connected decoding blocks. Each decoding block contains a 2×2 transposed convolutional layer with a stride of 2 for spatial upsampling. The transposed convolutional layer is followed by a feature concatenation layer, which concatenates the feature maps output by the transposed convolutional layer with the encoder feature maps of the corresponding level in the channel dimension. After concatenation, two 3×3 convolutional layers are connected. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function. The skip connection connects the encoder feature map output by each encoding block in the encoder to the feature concatenation layer of the corresponding level in the decoder.

[0065] By adopting the U-Net encoding-decoding structure, multi-scale extraction and fusion of feature maps are achieved, which can enhance the convolutional neural network's ability to model the relationship between coal seam structure and stress and improve the accuracy of two-dimensional stress prediction slices.

[0066] In some specific embodiments, a structural boundary attention mechanism is used in the skip connection process. After obtaining the encoder feature map, the mask feature map after downsampling the structural mask map is first fused with the encoder feature map to obtain a fused feature map, and then the fused feature map is input into the feature splicing layer. The specific fusion method of the structural boundary attention mechanism is: For each encoder depth level , perform the calculation:

[0067] Indicates the Feature maps output by the layer encoder; Structure mask image The mask feature map after downsampling; is a trainable guidance factor, with a value range of [0.5, 2.0]; Represents element-wise multiplication operation; Represents the fused feature map.

[0068] By adopting the structural boundary attention mechanism in the jump connection process, the dynamic weighted fusion of the encoder feature map and the structural mask information is realized, which can enhance the feature expression of the geological boundary area and improve the stress prediction accuracy of the convolutional neural network at the boundary.

[0069] In some specific embodiments, the training samples used by the convolutional neural network are derived from three-dimensional wave velocity data obtained by inverting seismic wave response signals using a time fractional viscoelastic wave equation, and three-dimensional coal seam structure mask images extracted based on reverse time migration imaging technology; The convolutional neural network uses measured stress interpolation maps as labels for semi-supervised training and is jointly trained on multiple geological samples. The measured stress interpolation map is a coarse-resolution stress distribution map generated by interpolation based on the measured stress values ​​at each detection point. The loss function of the convolutional neural network is:

[0070] in, For The measured stress interpolation diagram composed of the stress values ​​of all detection points with equal height on the Y axis; 、 、 is the weight coefficient,

[0071] MAE is the mean absolute error; SSIM is the structural similarity index; Edgeloss is the boundary alignment loss, which is the difference between the predicted stress gradient and the measured stress gradient at the geological boundary identified by the 3D coal seam structure mask.

[0072] By setting the loss function, multi-objective optimization of convolutional neural network training is achieved, which can balance the overall stress prediction accuracy, structural similarity and geological boundary alignment, and improve the robustness and generalization ability of two-dimensional stress prediction slices. Step S5: All two-dimensional stress prediction slices are spliced ​​together to construct a three-dimensional stress body, and three-dimensional stress point cloud data is generated based on the three-dimensional stress body; The 3D stress point cloud data is converted into a 3D stress mesh model through a 3D reconstruction algorithm; Use mechanical analysis methods to delineate dangerous areas in a three-dimensional stress grid model; The three-dimensional stress grid model and the three-dimensional coal seam structure mask map are combined and then rendered with stress value-color mapping to generate a structure-stress integrated visual three-dimensional cloud map, and the dangerous areas are marked in the visual three-dimensional cloud map.

[0073] By splicing two-dimensional stress prediction slices to construct a three-dimensional stress body, three-dimensional stress point cloud data is generated. After being converted into a three-dimensional stress grid model and the dangerous area is delineated, stress value-color mapping rendering is performed in combination with the three-dimensional coal seam structure mask map. A visual three-dimensional cloud map of structure-stress integration is further generated, and the dangerous area is marked, realizing a three-dimensional intuitive display of stress distribution, which can support the visual analysis of the coupling relationship between geological structure and stress, and facilitate coal mine safety assessment and decision-making.

[0074] In some specific embodiments, the step of generating three-dimensional stress point cloud data includes: S501. Slice all two-dimensional predictions Splice back to the three-dimensional coordinate axis in the original spatial order to form a complete three-dimensional stress body: ; S502. Perform voxel sampling on the 3D stress volume, extract voxel coordinates and stress values, and generate 3D stress point cloud data:

[0075] in, The first The spatial coordinates of the sampling points; The first The predicted stress value of each sampling point; Displays the lower threshold value for stress.

[0076] In some specific embodiments, the mechanical analysis method includes finite element method (FEM), discrete element method (DEM), and boundary element method (BEM).

[0077] In some specific embodiments, the steps of performing the finite element method include: Model discretization: Divide the three-dimensional stress body mesh model into tetrahedral or hexahedral unit meshes; Constitutive model loading: assigning an elastic-plastic constitutive relationship (such as the Mohr-Coulomb criterion or the Drucker-Prager model) to the coal rock mass; Boundary condition setting: applying displacement constraints (such as bottom fixation and lateral pressure) according to mining conditions; Stress solver: solves the equilibrium equations and outputs the stress tensor of each unit grid; Danger judgment: If the stress long span burst of the unit grid meets the uniaxial compressive strength calibrated for the laboratory, the unit grid will be marked as a dangerous area.

[0078] In some specific embodiments, the steps of performing the discrete element method include: Particle model construction: The coal rock mass is discretized into a collection of rigid particles, and stress is transmitted between particles through contact force chains; Joint network import: embed fault / crack geometry parameters based on 3D stress body mesh model; Dynamic analysis: simulate particle motion under mining disturbances and calculate contact force distribution; Danger judgment: The average contact force in a local area is calculated. If it exceeds the preset threshold, the area is determined to be an instability risk zone.

[0079] In some specific embodiments, the steps of performing the boundary element method include: Boundary discretization: only the surface of the three-dimensional stress body mesh model is divided into elements; Integral equation solution: Calculate boundary stresses via Green’s function Fault activation analysis: Calculate the stress intensity factor K on the fault plane. If K>K1 (preset fracture toughness), mark it as a potential water / gas outburst hazard area.

[0080] In some specific embodiments, during the stress value-color mapping rendering process, the corresponding relationship between stress value and color is: .

[0081] By setting the correspondence between stress values ​​and colors, the graded color expression of the visual three-dimensional cloud map is achieved, which can intuitively distinguish low, medium and high stress areas, facilitating the rapid identification of potentially dangerous geological structures.

[0082] In a specific embodiment, the method for predicting and visualizing three-dimensional stress in coal seams includes the following steps: Step S1, set up an artificial seismic source and multiple detection points in the area to be detected in the coal seam, excite the source signal at the artificial seismic source, configure a slot wave detector and a stress monitoring anchor at each detection point, use the slot wave detector to synchronously collect the seismic wave response signal of each detection point, and use the stress monitoring anchor to synchronously collect the stress value of each detection point.

[0083] Step S2: Based on the source signal, the wave field is forward modeled using the time fractional viscoelastic wave equation to obtain the forward simulated wave field. , in which a short-memory fast algorithm is used to solve the fractional derivative terms of the time fractional-order viscoelastic wave equation; Perform full waveform inversion based on forward simulation wave field and seismic wave response signal to output three-dimensional wave velocity data of the area to be detected ; in, is the spatial position vector; is the time variable; The time fractional viscoelastic wave equation is:

[0084] in, is the coal seam medium density; is the displacement vector field; is the displacement vector field in The direction component, ; are the stress tensor components, where represents the normal direction of the stress action plane, represents the direction of stress action, ; for The spatial coordinate component of the direction; is the source term, representing The body force in the direction satisfies:

[0085] is the coordinate of the artificial earthquake source; is the spatial Dirac function; is the time function of the source signal; is the viscoelastic constant, calibrated by laboratory dynamic mechanical testing; is the Kronick function, hour =1, otherwise =0; is the fractional order, ∈[0.3,0.8]; is the displacement divergence; is the strain tensor, ; The formula for solving the fractional derivative term using the short memory fast algorithm is:

[0086] in, is the current time step; The step length is 1 time step; L is the finite memory window length; is the fractional order The weight coefficient of the decision; The wave field forward simulation uses the finite difference method to discretize the wave equation and calculate the forward simulation wave field in the full range of the spatial domain. .

[0087] Step S3, performing reverse time migration imaging on the seismic wave response signal based on the three-dimensional wave velocity data to obtain a three-dimensional coal seam structure mask map, including: S301. Based on seismic wave response signal Generate a virtual source signal, the expression is:

[0088] in, For the Detection points at time seismic wave response signal; is the total recording time of the seismic wave response signal; For the Detection points at time The virtual source model of the monitoring point; Will As a virtual source applied to the position of the corresponding detection point, the wave field is reversely propagated using the same time fractional viscoelastic wave equation based on the three-dimensional wave velocity data to obtain the reverse wave field ; S302. Calculating 3D Imaging Values : ; S303. After performing gradient enhancement and high-pass filtering on the three-dimensional imaging value, set the threshold , by threshold Generate a 3D coal seam structure mask map that identifies the geological boundaries. The expression is: ; Geological boundaries include geological faults and interlayer boundaries.

[0089] Step S4, slicing the 3D velocity data and the 3D coal seam structure mask image along the same spatial dimension and combining them into an input tensor, inputting it into a pre-trained convolutional neural network, and outputting a 2D stress prediction, specifically includes: S401. 3D wave velocity data With 3D coal seam structure mask map Slice along the Y axis at a fixed step size to obtain a wave velocity data slice sequence and coal seam structure mask slice sequence ; in, is the total number of slices, Indicates the slices; S402. Combine the velocity data slice and the corresponding coal seam structure mask slice into an input tensor: ; S403. Input the input tensor into the pre-trained convolutional neural network and output a 2D stress prediction slice. ; The convolutional neural network is a U-Net encoding-decoding structure, which includes an encoder and a decoder. The encoder and decoder use a jump connection, where: The encoder consists of multiple sequentially connected encoding blocks. Each encoding block contains two 3×3 convolutional layers. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function. At the end of each encoding block, a 2×2 maximum pooling layer with a stride of 2 is connected to perform spatial downsampling. The parameter settings of the encoder are shown in Table 1: Table 1 Encoder parameter setting table

[0090] The decoder consists of multiple sequentially connected decoding blocks. Each decoding block contains a 2×2 transposed convolutional layer with a stride of 2 for spatial upsampling. The transposed convolutional layer is followed by a feature concatenation layer, which concatenates the feature maps output by the transposed convolutional layer with the encoder feature maps of the corresponding level in the channel dimension. After concatenation, two 3×3 convolutional layers are connected. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function. The decoder parameter settings are shown in Table 2: Table 2 Decoder parameter setting table

[0091] The skip connection connects the encoder feature map output by each encoding block in the encoder to the feature concatenation layer of the corresponding level in the decoder; The structural boundary attention mechanism is used in the jump connection process. After obtaining the encoder feature map, the mask feature map after downsampling the structural mask map is first fused with the encoder feature map to obtain a fused feature map, and then the fused feature map is input into the feature splicing layer. The specific fusion method of the structural boundary attention mechanism is as follows: For each encoder depth level , perform the calculation:

[0092] Indicates the Feature maps output by the layer encoder; Structure mask image The mask feature map after downsampling; is a trainable guidance factor, with a value range of [0.5, 2.0]; Represents element-wise multiplication operation; Represents the fused feature map; The training samples used by the convolutional neural network come from the three-dimensional wave velocity data obtained by inverting the seismic wave response signal through the time fractional viscoelastic wave equation, and the three-dimensional coal seam structure mask image extracted based on the reverse time migration imaging technology; The convolutional neural network uses measured stress interpolation maps as labels for semi-supervised training and is jointly trained on multiple geological samples. The measured stress interpolation map is a coarse-resolution stress distribution map generated by interpolation based on the measured stress values ​​at each detection point. The loss function of the convolutional neural network is:

[0093] in, For The measured stress interpolation diagram composed of the stress values ​​of all detection points with equal height on the Y axis; 、 、 is the weight coefficient,

[0094] MAE is the mean absolute error; SSIM is the structural similarity index; Edgeloss is the boundary alignment loss, which is the difference between the predicted stress gradient and the measured stress gradient at the geological boundary identified by the 3D coal seam structure mask.

[0095] Step S5: All two-dimensional stress prediction slices are spliced ​​together to construct a three-dimensional stress body, and three-dimensional stress point cloud data is generated based on the three-dimensional stress body; The 3D stress point cloud data is converted into a 3D stress mesh model through a 3D reconstruction algorithm; Delineate hazardous areas in a three-dimensional stress mesh model using the finite element method; After combining the 3D stress grid model and the 3D coal seam structure mask map, stress value-color mapping rendering is performed to generate a structure-stress integrated visual 3D cloud map, and dangerous areas are marked in the visual 3D cloud map. The steps of generating 3D stress point cloud data include: S501. Slice all two-dimensional predictions Splice back to the three-dimensional coordinate axis in the original spatial order to form a complete three-dimensional stress body: ; S502. Perform voxel sampling on the 3D stress volume, extract voxel coordinates and stress values, and generate 3D stress point cloud data:

[0096] in, The first The spatial coordinates of the sampling points; The first The predicted stress value of each sampling point; Displays the lower threshold for stress; The steps for delineating hazardous areas in a 3D stress mesh model using the finite element method include: Model discretization: Divide the three-dimensional stress body mesh model into tetrahedral or hexahedral unit meshes; Constitutive model loading: assigning elastic-plastic constitutive relations to coal and rock masses; Boundary condition setting: applying displacement constraints according to mining conditions; Stress solver: solves the equilibrium equations and outputs the stress tensor of each unit grid; Danger judgment: If the stress long span burst of the unit grid meets the uniaxial compressive strength calibrated by the laboratory, the unit grid will be marked as a dangerous area; During the stress value-color mapping rendering process, the corresponding relationship between stress value and color is:

[0097] The available color schemes include Jet, Viridis, Thermal, or Turbo; Visualization technology selection includes: Use OpenGL / WebGL / Potree / Three.js or VTK to achieve 3D point cloud rendering; Rendering parameters include point size, color transparency, and response speed optimization.

[0098] The following is an embodiment of the coal seam three-dimensional stress prediction and visualization system provided in the embodiments of the present application. The coal seam three-dimensional stress prediction and visualization system and the coal seam three-dimensional stress prediction and visualization methods of the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiments of the coal seam three-dimensional stress prediction and visualization system, please refer to the embodiments of the above-mentioned coal seam three-dimensional stress prediction and visualization method.

[0099] like Figure 2 As shown in the figure, the coal seam 3D stress prediction and visualization system includes: The data acquisition module is used to set an artificial seismic source and multiple detection points in the area to be detected in the coal seam, stimulate the source signal at the artificial seismic source, and synchronously collect the seismic wave response signal and stress value of each detection point; The wavefield forward modeling module is used to perform wavefield forward modeling based on the source signal using the time fractional viscoelastic wave equation to obtain the forward simulated wavefield; Full waveform inversion module, used to perform full waveform inversion based on the forward simulation wave field and seismic wave response signal, and output three-dimensional wave velocity data of the area to be detected; A reverse time migration imaging module is used to perform reverse time migration imaging on the seismic wave response signal based on the three-dimensional wave velocity data to obtain a three-dimensional coal seam structure mask map; The 2D stress prediction module is used to combine the 3D wave velocity data and the 3D coal seam structure mask image into an input tensor after slicing along the same spatial dimension. The input is fed into a pre-trained convolutional neural network and the 2D stress prediction slice is output. 3D visualization module for: All 2D stress prediction slices are spliced ​​together to construct a 3D stress body, and 3D stress point cloud data is generated based on the 3D stress body; The 3D stress point cloud data is converted into a 3D stress mesh model through a 3D reconstruction algorithm; Use mechanical analysis methods to delineate dangerous areas in a three-dimensional stress grid model; The three-dimensional stress grid model and the three-dimensional coal seam structure mask map are combined and then rendered with stress value-color mapping to generate a structure-stress integrated visual three-dimensional cloud map, and the dangerous areas are marked in the visual three-dimensional cloud map.

[0100] The coal seam three-dimensional stress prediction and visualization system of this embodiment is used to implement the coal seam three-dimensional stress prediction and visualization method.

[0101] This application also provides an electronic device for implementing each embodiment of this application. Figure 3 A hardware structure diagram of an electronic device for implementing various embodiments of the present application is shown in FIG. Figure 3 As shown, the electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0102] Those skilled in the art will understand that the electronic device structure involved in the embodiments of the present application does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0103] In the embodiments of the present application, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described and / or claimed herein.

[0104] In the embodiment of the present application, the processor can be implemented by using at least one of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a processor, a controller, a microcontroller, a microprocessor, and an electronic unit designed to perform the functions described herein. In some cases, such an embodiment can be implemented in a controller. For software implementation, an embodiment such as a process or function can be implemented with a separate software module that allows the execution of at least one function or operation. The software code can be implemented by a software application (or program) written in any appropriate programming language, and the software code can be stored in a memory and executed by a controller.

[0105] In addition, the electronic device includes some functional modules not shown, which will not be described here.

[0106] Those skilled in the art will appreciate that various aspects of the electronic device provided herein may be implemented as a system, method, or program product. Therefore, various aspects of the present application may be specifically implemented in the following forms: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be collectively referred to herein as "circuits," "modules," or "systems."

[0107] This application also provides a storage medium storing a program product capable of implementing a method for three-dimensional coal seam stress prediction and visualization. In some possible implementations, various aspects of this application may also be implemented in the form of a program product comprising program code. When the program product is executed on a terminal device, the program code is configured to cause the terminal device to execute the steps described in the "Exemplary Methods" section above according to various exemplary embodiments of this application.

[0108] The storage medium can be any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0109] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting and visualizing three-dimensional stress in coal seams, characterized in that: include: S1. Set up an artificial seismic source and multiple detection points in the coal seam to be detected, stimulate the source signal at the artificial seismic source, and simultaneously collect seismic wave response signals and stress values ​​at each detection point; S2. Based on the source signal, the wave field is forward modeled using the time fractional viscoelastic wave equation to obtain the forward simulated wave field. , in which a short-memory fast algorithm is used to solve the fractional derivative terms of the time fractional-order viscoelastic wave equation; Perform full waveform inversion based on forward simulation wave field and seismic wave response signal to output three-dimensional wave velocity data of the area to be detected ; in, is the spatial position vector; is the time variable; S3. Perform reverse time migration imaging on the seismic wave response signal based on the 3D wave velocity data to obtain a 3D coal seam structure mask map, including: Time-reverse the seismic wave response signals of each detection point to generate virtual source signals; Using the virtual source signal as the excitation source, the wavefield is back-propagated based on the three-dimensional wave velocity data to generate the back-propagation wavefield; Perform time-space cross-correlation calculation on the forward simulation wave field and the reverse propagation wave field to obtain the three-dimensional imaging value; Perform gradient enhancement and threshold segmentation on the three-dimensional imaging values ​​to generate a three-dimensional coal seam structure mask map that identifies the geological boundary; S4. Slice the 3D velocity data and the 3D coal seam structure mask image along the same spatial dimension and combine them into an input tensor. This is then fed into a pre-trained convolutional neural network to output a 2D stress prediction slice. S5. Construct a 3D stress volume by stitching all 2D stress prediction slices together, and generate 3D stress point cloud data based on the 3D stress volume. The 3D stress point cloud data is converted into a 3D stress mesh model through a 3D reconstruction algorithm; Use mechanical analysis methods to delineate dangerous areas in a three-dimensional stress grid model; The three-dimensional stress grid model and the three-dimensional coal seam structure mask map are combined and then rendered with stress value-color mapping to generate a structure-stress integrated visual three-dimensional cloud map, and the dangerous areas are marked in the visual three-dimensional cloud map.

2. The method for three-dimensional stress prediction and visualization of coal seams according to claim 1, wherein: In step S2, the time fractional-order viscoelastic wave equation is: in, is the coal seam medium density; is the displacement vector field; is the displacement vector field in The direction component, ; are the stress tensor components, where represents the normal direction of the stress action plane, represents the direction of stress action, ; for The spatial coordinate component of the direction; is the source term, representing The body force in the direction satisfies: is the coordinate of the artificial earthquake source; is the spatial Dirac function; is the time function of the source signal; is the viscoelastic constant, calibrated by laboratory dynamic mechanical testing; is the Kronick function, hour =1, otherwise =0; is the fractional order, ∈[0.3,0.8]; is the displacement divergence; is the strain tensor, ; The formula for solving the fractional derivative term using the short memory fast algorithm is: in, is the current time step; The step length is 1 time step; L is the finite memory window length; is the fractional order The weight coefficient of the decision; The wave field forward simulation uses the finite difference method to discretize the wave equation and calculate the forward simulation wave field in the full range of the spatial domain. .

3. The method for three-dimensional stress prediction and visualization of coal seams according to claim 1, wherein: Step S3 includes: S301. Based on seismic wave response signal Generate a virtual source signal, the expression is: in, For the Detection points at time seismic wave response signal; is the total recording time of the seismic wave response signal; For the Detection points at time The virtual source model of the monitoring point; Will As a virtual source applied to the position of the corresponding detection point, the wave field is reversely propagated using the same time fractional viscoelastic wave equation based on the three-dimensional wave velocity data to obtain the reverse wave field ; S302. Calculating 3D Imaging Values : ； S303. After performing gradient enhancement and high-pass filtering on the three-dimensional imaging value, set the threshold , by threshold Generate a structural mask map, the expression is: 。 4. The method for three-dimensional stress prediction and visualization of coal seams according to claim 1, wherein: Step S4 includes: S401. 3D wave velocity data With 3D coal seam structure mask map Slice along the Y axis at a fixed step size to obtain a wave velocity data slice sequence and coal seam structure mask slice sequence ; in, is the total number of slices, Indicates the slices; S402. Combine the velocity data slice and the corresponding coal seam structure mask slice into an input tensor: ； S403. Input the input tensor into the pre-trained convolutional neural network and output a 2D stress prediction slice. .

5. The method for three-dimensional stress prediction and visualization of coal seams according to claim 1, wherein: In step S4, the convolutional neural network is a U-Net encoding-decoding structure, including an encoder and a decoder, and a jump connection is used between the encoder and the decoder, where: The encoder consists of multiple sequentially connected encoding blocks. Each encoding block contains two 3×3 convolutional layers. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function. At the end of each encoding block, a 2×2 maximum pooling layer with a stride of 2 is connected to perform spatial downsampling. The decoder consists of multiple sequentially connected decoding blocks. Each decoding block contains a 2×2 transposed convolutional layer with a stride of 2 for spatial upsampling. The transposed convolutional layer is followed by a feature concatenation layer, which concatenates the feature maps output by the transposed convolutional layer with the encoder feature maps of the corresponding level in the channel dimension. After concatenation, two 3×3 convolutional layers are connected. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function. The skip connection connects the encoder feature map output by each encoding block in the encoder to the feature concatenation layer of the corresponding level in the decoder.

6. The method for three-dimensional stress prediction and visualization of coal seams according to claim 5, characterized in that: The training samples used by the convolutional neural network come from the three-dimensional wave velocity data obtained by inverting the seismic wave response signal through the time fractional viscoelastic wave equation, and the three-dimensional coal seam structure mask image extracted based on the reverse time migration imaging technology; The convolutional neural network uses measured stress interpolation maps as labels for semi-supervised training and is jointly trained on multiple geological samples. The measured stress interpolation map is a coarse-resolution stress distribution map generated by interpolation based on the measured stress values ​​at each detection point. The loss function of the convolutional neural network is: in, For The measured stress interpolation diagram composed of the stress values ​​of all detection points with equal height on the Y axis; 、 、 is the weight coefficient, MAE is the mean absolute error; SSIM is the structural similarity index; Edgeloss is the boundary alignment loss, which is the difference between the predicted stress gradient and the measured stress gradient at the geological boundary identified by the 3D coal seam structure mask.

7. The method for three-dimensional stress prediction and visualization of coal seams according to claim 1, wherein: In step S5, the step of generating three-dimensional stress point cloud data includes: S501. Slice all two-dimensional predictions Splice back to the three-dimensional coordinate axis in the original spatial order to form a complete three-dimensional stress body: ； S502. Perform voxel sampling on the 3D stress volume, extract voxel coordinates and stress values, and generate 3D stress point cloud data: in, The first The spatial coordinates of the sampling points; The first The predicted stress value of each sampling point; Displays the lower threshold value for stress.

8. A coal seam three-dimensional stress prediction and visualization system, characterized in that: A method for implementing three-dimensional stress prediction and visualization of coal seams according to any one of claims 1 to 7, comprising: The data acquisition module is used to set an artificial seismic source and multiple detection points in the area to be detected in the coal seam, stimulate the source signal at the artificial seismic source, and synchronously collect the seismic wave response signal and stress value of each detection point; The wavefield forward modeling module is used to perform wavefield forward modeling based on the source signal using the time fractional viscoelastic wave equation to obtain the forward simulated wavefield; Full waveform inversion module, used to perform full waveform inversion based on the forward simulation wave field and seismic wave response signal, and output three-dimensional wave velocity data of the area to be detected; A reverse time migration imaging module is used to perform reverse time migration imaging on the seismic wave response signal based on the three-dimensional wave velocity data to obtain a three-dimensional coal seam structure mask map; The 2D stress prediction module is used to combine the 3D wave velocity data and the 3D coal seam structure mask image into an input tensor after slicing along the same spatial dimension. The input is fed into a pre-trained convolutional neural network and the 2D stress prediction slice is output. 3D visualization module for: All 2D stress prediction slices are spliced ​​together to construct a 3D stress body, and 3D stress point cloud data is generated based on the 3D stress body; The 3D stress point cloud data is converted into a 3D stress mesh model through a 3D reconstruction algorithm; Use mechanical analysis methods to delineate dangerous areas in a three-dimensional stress grid model; The three-dimensional stress grid model and the three-dimensional coal seam structure mask map are combined and then rendered with stress value-color mapping to generate a structure-stress integrated visual three-dimensional cloud map, and the dangerous areas are marked in the visual three-dimensional cloud map.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: The processor is used to implement the steps of the coal seam three-dimensional stress prediction and visualization method as described in any one of claims 1 to 7 when executing the computer program.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the coal seam three-dimensional stress prediction and visualization method as described in any one of claims 1 to 7 are implemented.

Images