Coal seam thickness prediction method, device, equipment and medium

Abstract

The present disclosure proposes a method, device, equipment and medium for predicting the thickness of a coal seam, and relates to the field of artificial intelligence technology. The method includes: using a ground penetrating radar to detect the coal and rock layers to be measured, and obtaining a radar signal to be measured; using a first sub-network in a coal-rock identification model to extract features from the radar signal to be measured, so as to obtain a prediction feature for indicating the spatial distribution of coal seams and rock layers in the coal and rock layers to be measured; using a second sub-network in a coal-rock identification model to predict the thickness of a coal seam based on the prediction feature, so as to obtain a first thickness distribution of the coal seam in the coal and rock layers to be measured on the detection path of the ground penetrating radar. Thus, based on deep learning technology, the thickness of the coal seam at each position point on the detection path of the ground penetrating radar (i.e., the thickness distribution of the coal seam in the coal and rock layers to be measured) is predicted, which can improve the accuracy of the prediction results, and the prediction speed is fast, and real-time prediction can be achieved. And there is no need to measure the relative dielectric constant of the coal seam, and there is no need to calculate parameters such as the propagation velocity, which can reduce the workload of the prediction process.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and medium for predicting coal seam thickness. Background Art

[0002] Coal rock (coal seam and rock stratum) identification is a technology for identifying the lithology of geological bodies. Coal rock identification technology is of great significance for intelligent mining, improving the coal mining environment, and achieving safe and efficient production in coal mines. Among them, identifying the thickness of coal seams in coal rocks can meet the needs of accurate coal rock detection for unmanned intelligent mining in mines.

[0003] In the related technology, ground penetrating radar can be used to detect coal and rock in mines to obtain radar data. After pre-processing the radar data, the coal and rock strata are calculated using the seed layer selection and tracking method (also called layer tracking method, coal and rock layer tracking method). Finally, the coal and rock strata coordinates are calculated using algorithms such as the trace correlation coefficient and the location of the radar antenna, and the thickness information of the coal seam is obtained from the radar data.

[0004] However, the above coal seam location method requires the propagation speed of electromagnetic waves in the measured coal seam or the relative dielectric constant of the measured coal seam to be determined in advance, and the above parameters need to be measured additionally, which increases the workload of coal seam thickness prediction. In addition, the measurement requires sampling measurement, and it is impossible to automatically complete the prediction of coal seam thickness. Summary of the invention

[0005] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.

[0006] The present disclosure proposes a coal seam thickness prediction method, device, equipment and medium to achieve prediction of the coal seam thickness at each position point on the ground penetrating radar detection path (i.e., the thickness distribution of the coal seam in the coal rock layer to be measured) based on deep learning technology. On the one hand, the accuracy of the prediction result can be improved, and on the other hand, the prediction speed is fast, and real-time prediction can be achieved. In addition, compared with the method of using physical formulas to calculate the thickness of the coal seam in the prior art, in the present disclosure, there is no need to measure the relative dielectric constant of the coal seam, and there is no need to calculate parameters such as the propagation speed, which can reduce the workload of the prediction process. In addition, the thickness distribution of the coal seam can be directly predicted based on the radar signal to be measured without measuring the relative dielectric constant, which can achieve automatic prediction of the thickness of the coal seam.

[0007] The first embodiment of the present disclosure provides a method for predicting coal seam thickness, comprising:

[0008] Using ground penetrating radar to detect the coal and rock layers to be tested, so as to obtain radar signals to be tested;

[0009] The first sub-network in the coal-rock identification model is used to extract features of the radar signal to be measured to obtain prediction features; wherein the prediction features are used to indicate the spatial distribution of coal seams and rock seams in the coal-rock seam to be measured;

[0010] The second subnetwork in the coal-rock identification model is used to predict the coal seam thickness based on the prediction feature to obtain a first thickness distribution of the coal seam in the coal-rock layer to be detected on the ground penetrating radar detection path.

[0011] The coal seam thickness prediction method of the embodiment of the present disclosure uses a ground penetrating radar to detect the coal and rock layers to be tested to obtain the radar signal to be tested; the first sub-network in the coal and rock identification model is used to extract the features of the radar signal to be tested to obtain the prediction features; wherein the prediction features are used to indicate the spatial distribution of the coal seams and rock layers in the coal and rock layers to be tested; the second sub-network in the coal and rock identification model is used to predict the coal seam thickness of the prediction features to obtain the first thickness distribution of the coal seam in the coal and rock layers to be tested on the ground penetrating radar detection path. Thus, based on the deep learning technology, the coal seam thickness of each position point on the ground penetrating radar detection path (i.e., the thickness distribution of the coal seam in the coal and rock layers to be tested) is predicted. On the one hand, the accuracy of the prediction result can be improved, and on the other hand, the prediction speed is fast, and real-time prediction can be achieved. Moreover, compared with the method of calculating the thickness of the coal seam using physical formulas in the prior art, in the present disclosure, there is no need to measure the relative dielectric constant of the coal seam, and there is no need to calculate parameters such as the propagation speed, which can reduce the workload of the prediction process. In addition, the thickness distribution of the coal seam can be directly predicted according to the radar signal to be tested without measuring the relative dielectric constant, and the automatic prediction of the coal seam thickness can be achieved.

[0012] The second aspect of the present disclosure provides a coal seam thickness prediction device, comprising:

[0013] A detection module is used to detect the coal and rock layers to be detected by using a ground penetrating radar to obtain a radar signal to be detected;

[0014] An extraction module, used for extracting features of the radar signal to be measured by using the first sub-network in the coal-rock recognition model to obtain prediction features; wherein the prediction features are used to indicate the spatial distribution of coal seams and rock seams in the coal-rock seam to be measured;

[0015] The prediction module is used to use the second sub-network in the coal-rock identification model to predict the coal seam thickness based on the prediction feature to obtain the first thickness distribution of the coal seam in the coal-rock layer to be tested on the ground penetrating radar detection path.

[0016] The coal seam thickness prediction device of the embodiment of the present disclosure detects the coal and rock layers to be tested by using a ground penetrating radar to obtain a radar signal to be tested; the first sub-network in the coal and rock identification model is used to extract features of the radar signal to be tested to obtain prediction features; wherein the prediction features are used to indicate the spatial distribution of coal seams and rock layers in the coal and rock layers to be tested; the second sub-network in the coal and rock identification model is used to predict the coal seam thickness of the prediction features to obtain the first thickness distribution of the coal seam in the coal and rock layers to be tested on the ground penetrating radar detection path. Thus, based on the deep learning technology, the coal seam thickness of each position point on the ground penetrating radar detection path (i.e., the thickness distribution of the coal seam in the coal and rock layers to be tested) is predicted. On the one hand, the accuracy of the prediction result can be improved, and on the other hand, the prediction speed is fast, and real-time prediction can be achieved. Moreover, compared with the method of calculating the thickness of the coal seam using physical formulas in the prior art, in the present disclosure, there is no need to measure the relative dielectric constant of the coal seam, and there is no need to calculate parameters such as the propagation speed, which can reduce the workload of the prediction process. In addition, the thickness distribution of the coal seam can be directly predicted according to the radar signal to be tested without measuring the relative dielectric constant, and the automatic prediction of the coal seam thickness can be achieved.

[0017] The third aspect embodiment of the present disclosure proposes an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the coal seam thickness prediction method proposed in the first aspect embodiment of the present disclosure is implemented.

[0018] The fourth aspect embodiment of the present disclosure proposes a non-temporary computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the coal seam thickness prediction method proposed in the first aspect embodiment of the present disclosure is implemented.

[0019] The fifth aspect embodiment of the present disclosure proposes a computer program product. When the instructions in the computer program product are executed by a processor, the coal seam thickness prediction method proposed in the first aspect embodiment of the present disclosure is executed.

[0020] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0022] Figure 1 A schematic diagram of a flow chart of a coal seam thickness prediction method provided by an embodiment of the present disclosure;

[0023] FIG. 2( a ) is a schematic diagram of ground penetrating radar data acquisition provided by an embodiment of the present disclosure Figure 1 ;

[0024] FIG2( b ) is a second schematic diagram of ground penetrating radar data collection provided by an embodiment of the present disclosure;

[0025] Figure 3 A schematic diagram of a flow chart of another coal seam thickness prediction method provided by an embodiment of the present disclosure;

[0026] Figure 4 The structure of the coal-rock identification model provided in the embodiment of the present disclosure is shown in FIG. Figure 1 ;

[0027] Figure 5 A second structural diagram of the coal-rock identification model provided in the embodiment of the present disclosure;

[0028] Figure 6 A schematic diagram of data involved in an embodiment of the present disclosure;

[0029] Figure 7 A schematic diagram of a training process of a coal-rock identification model provided in an embodiment of the present disclosure;

[0030] Figure 8 A schematic diagram of the training process of another coal-rock identification model provided in an embodiment of the present disclosure;

[0031] Fig. 9 A schematic diagram of the structure of a coal seam thickness prediction device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0032] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0033] At present, the thickness of coal seams in coal rocks can be determined in the following ways:

[0034] The first method is to use ground penetrating radar to detect coal and rock in the mine, obtain radar data, and after pre-processing the radar data, use the seed layer selection and tracking method (also called layer tracking method, coal and rock layer tracking method) to calculate the coal and rock layer. Finally, use algorithms such as the trace correlation coefficient and the location of the radar antenna to calculate the coal and rock layer coordinates and obtain the thickness information of the coal seam from the radar data.

[0035] The second method is to use ground penetrating radar to detect the interface between the coal roof and floor, then correct the relative dielectric constant through drilling data, and finally use radar supporting software to calculate the thickness of the coal seam.

[0036] However, the first method mentioned above requires the pre-determination of the propagation speed of electromagnetic waves in the measured coal seam or the relative dielectric constant of the measured coal seam, and the above parameters need to be additionally measured, which increases the workload of coal seam thickness prediction and fails to realize the automation of coal seam thickness prediction.

[0037] The second method mentioned above does not involve the specific algorithm of the data processing software, and manual participation is required in the calculation process. In addition, when predicting the thickness of the coal seam, the drilling map is needed to obtain the relative dielectric constant of the coal seam, and the prediction of the thickness of the coal seam requires manual participation, and automatic identification and calculation cannot be achieved.

[0038] Therefore, in order to address at least one of the above problems, the present disclosure proposes a method, device, electronic device and storage medium for predicting coal seam thickness.

[0039] The following describes the coal seam thickness prediction method, device, equipment and medium of the embodiment of the present disclosure with reference to the accompanying drawings. Before describing the embodiment of the present disclosure in detail, for the sake of easy understanding, the commonly used technical terms are first introduced:

[0040] Ground Penetrating Radar: A geophysical method that uses antennas to transmit and receive high-frequency electromagnetic waves to detect the material characteristics and distribution patterns inside a medium. That is, ground penetrating radar technology is a geophysical method that detects the material characteristics and distribution patterns inside a medium, and can be used to identify the interface between coal seams and rock strata.

[0041] Deep learning: A machine learning method that can independently build basic rules based on example data during the learning process.

[0042] Figure 1 A schematic flow chart of a method for predicting coal seam thickness provided in an embodiment of the present disclosure.

[0043] In the disclosed embodiments, the coal seam thickness prediction method can be applied to any electronic device so that the electronic device can perform the coal seam thickness prediction function.

[0044] Among them, the electronic device can be any device with computing capabilities, such as a PC (Personal Computer), a mobile terminal, a server, etc. The mobile terminal can be, for example, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, and other hardware devices with various operating systems, touch screens and / or display screens.

[0045] like Figure 1 As shown, the coal seam thickness prediction method may include the following steps:

[0046] Step 101: Use ground penetrating radar to detect the coal stratum to be tested, so as to obtain the radar signal to be tested.

[0047] In the disclosed embodiment, a ground penetrating radar may be used to detect the coal strata to be detected to obtain a radar signal to be detected.

[0048] As an example, the radiation direction of the ground penetrating radar can be vertically upward and toward the coal rock layer (such as the coal rock layer to be tested), the moving direction is horizontal, and the moving mode is uniform movement. For example, the moving mode and moving direction of the ground penetrating radar can be shown in Figure 2(a), and the ground penetrating radar can move horizontally at a constant speed for a certain distance to collect the radar signal to be tested. For example, when the ground penetrating radar moves at a constant speed, a fixed number of frames per second can be collected to ensure that the spacing between each position point in the collected radar signal image is equal.

[0049] As another example, the radiation direction of the ground penetrating radar can be vertically upward and toward the coal rock layer (such as the coal rock layer to be measured), the moving direction is horizontal, and the moving mode can be uniform movement or non-uniform movement. For example, the moving direction of the ground penetrating radar can be as shown in Figure 2(b), and the ground penetrating radar can be connected to a distance measuring device (such as a distance measuring wheel, a laser distance measuring device, a displacement sensor, etc.), wherein the distance measuring device can move at a uniform speed or non-uniform speed, and the present disclosure does not limit this. Whenever the distance measuring device moves a set interval, the ground penetrating radar can be controlled to collect data once.

[0050] Step 102: Use the first subnetwork in the coal-rock identification model to extract features from the radar signal to be tested to obtain prediction features; wherein the prediction features are used to indicate the spatial distribution of coal seams and rock seams in the coal-rock seam to be tested.

[0051] In the disclosed embodiment, the coal-rock identification model is a trained model, which has learned the corresponding relationship between the radar signal and the coal seam thickness.

[0052] In the disclosed embodiment, the first subnetwork (or lithology identification subnetwork) in the coal-rock identification model can be used to extract features of the radar signal to be tested to obtain prediction features; wherein the prediction features are used to indicate the spatial distribution of the coal seam and the rock layer in the coal-rock layer to be tested. That is, the first subnetwork can perform semantic segmentation on the radar signal to be tested to obtain the interface between the coal seam and the rock layer.

[0053] Step 103: Use the second subnetwork in the coal-rock identification model to predict the coal seam thickness based on the predicted features to obtain a first thickness distribution of the coal seam in the coal-rock layer to be detected on the ground penetrating radar detection path.

[0054] In the embodiment of the present disclosure, the second subnetwork (or called the coal seam thickness prediction subnetwork) in the coal rock identification model can be used to predict the coal seam thickness based on the prediction features to obtain the coal seam thickness at each position point on the ground penetrating radar detection path, that is, the thickness distribution of the coal seam in the coal rock layer to be measured, which is recorded as the first thickness distribution in the present disclosure.

[0055] The coal seam thickness prediction method of the embodiment of the present disclosure uses a ground penetrating radar to detect the coal and rock layers to be tested to obtain the radar signal to be tested; the first sub-network in the coal and rock identification model is used to extract the features of the radar signal to be tested to obtain the prediction features; wherein the prediction features are used to indicate the spatial distribution of the coal seams and rock layers in the coal and rock layers to be tested; the second sub-network in the coal and rock identification model is used to predict the coal seam thickness of the prediction features to obtain the first thickness distribution of the coal seam in the coal and rock layers to be tested on the ground penetrating radar detection path. Thus, based on the deep learning technology, the coal seam thickness of each position point on the ground penetrating radar detection path (i.e., the thickness distribution of the coal seam in the coal and rock layers to be tested) is predicted. On the one hand, the accuracy of the prediction result can be improved, and on the other hand, the prediction speed is fast, and real-time prediction can be achieved. Moreover, compared with the method of calculating the thickness of the coal seam using physical formulas in the prior art, in the present disclosure, there is no need to measure the relative dielectric constant of the coal seam, and there is no need to calculate parameters such as the propagation speed, which can reduce the workload of the prediction process. In addition, the thickness distribution of the coal seam can be directly predicted according to the radar signal to be tested without measuring the relative dielectric constant, and the automatic prediction of the coal seam thickness can be achieved.

[0056] In order to clearly explain how the first thickness distribution of the coal seam in the coal and rock layer to be detected on the ground penetrating radar detection path is predicted based on the coal and rock identification model in the above embodiment, the present disclosure also proposes a coal seam thickness prediction method.

[0057] Figure 3 A schematic flow chart of another method for predicting coal seam thickness provided in an embodiment of the present disclosure.

[0058] like Figure 3 As shown, the coal seam thickness prediction method may include the following steps:

[0059] Step 301: Use ground penetrating radar to detect the coal stratum to be tested, so as to obtain the radar signal to be tested.

[0060] For explanation of step 301, reference may be made to the relevant description in any embodiment of the present disclosure, and no further details will be given here.

[0061] Step 302: Use the first feature extraction layer in the first subnetwork in the coal-rock recognition model to extract features from the radar signal to be measured to obtain a first intermediate feature.

[0062] In the disclosed embodiment, the first feature extraction layer in the first subnetwork in the coal-rock identification model may be used to extract features from the radar signal to be measured to obtain a first intermediate feature.

[0063] The structure of the coal-rock identification model is as follows: Figure 4 or Figure 5For example, the first feature extraction layer may include 3 convolution layers and 2 maximum pooling layers, where the convolution layer is used to extract the features of the radar signal image to be tested and reduce the number of parameters. The maximum pooling layer has the following two functions: first, it reduces the amount of calculation and prevents overfitting; second, it increases the receptive field so that the subsequent convolution kernel can learn more global information.

[0064] The radar signal image to be measured can be input into the convolution layer, the maximum pooling layer, the convolution layer, the maximum pooling layer, and the convolution layer of the first feature extraction layer in sequence to obtain the first intermediate feature output by the last convolution layer in the first feature extraction layer.

[0065] Step 303: Use the second feature extraction layer in the first sub-network to extract features from the first intermediate features to obtain second intermediate features, wherein the scale of the second intermediate features is smaller than that of the first intermediate features.

[0066] In the embodiment of the present disclosure, the second feature extraction layer in the first sub-network may also be used to extract features from the first intermediate features to obtain second intermediate features, wherein the scale of the second intermediate features is smaller than that of the first intermediate features.

[0067] The structure of the coal-rock identification model is as follows Figure 4 or Figure 5 For example, the second feature extraction layer may include one maximum pooling layer and one convolution layer, and the first intermediate feature may be sequentially input into the maximum pooling layer and the convolution layer of the second feature extraction layer to obtain a second intermediate feature output by the convolution layer of the second feature extraction layer, wherein the scale of the second intermediate feature is smaller than that of the first intermediate feature. Figure 4 and Figure 5 The first intermediate feature may be a three-dimensional feature of 128 (width W) × 16 (height H) × 64 (channel C or feature dimension), and the second intermediate feature may be a three-dimensional feature of 64 × 8 × 128.

[0068] Step 304: Use the third feature extraction layer in the first sub-network to extract the second intermediate feature to obtain a third intermediate feature.

[0069] In the embodiment of the present disclosure, the third feature extraction layer in the first sub-network can be used to extract features from the second intermediate features to obtain third intermediate features, wherein the scale of the third intermediate features is smaller than that of the second intermediate features.

[0070] The structure of the coal-rock identification model is as follows Figure 4 or Figure 5For example, the third feature extraction layer may include 1 maximum pooling layer and 2 convolution layers, and the second intermediate feature may be sequentially input into the maximum pooling layer, the convolution layer, and the convolution layer of the third feature extraction layer to obtain the third intermediate feature output by the last convolution layer of the third feature extraction layer, wherein the scale of the third intermediate feature is smaller than that of the second intermediate feature. Figure 4 and Figure 5 The second intermediate feature may be a 64×8×128 stereoscopic feature, and the third intermediate feature may be a 32×4×128 stereoscopic feature.

[0071] Step 305: upsample the third intermediate feature to obtain a fourth intermediate feature.

[0072] In the embodiment of the present disclosure, the third intermediate feature may be upsampled to obtain a fourth intermediate feature.

[0073] The structure of the coal-rock identification model is as follows Figure 4 or Figure 5 As an example, the third intermediate feature can be upsampled by the upsampling layer in the first sub-network, and the obtained fourth intermediate feature can be a 64×8×128 stereo feature.

[0074] Among them, the upsampling layer is used to restore the features to a larger size for further calculation.

[0075] Step 306: Generate a prediction feature based on the fourth intermediate feature, wherein the prediction feature is used to indicate the spatial distribution of the coal seams and rock seams in the coal and rock seams to be measured.

[0076] In the disclosed embodiment, a prediction feature may be generated based on the fourth intermediate feature, wherein the prediction feature is used to indicate the spatial distribution of coal seams and rock seams in the coal and rock seams to be measured.

[0077] In a possible implementation of the embodiment of the present disclosure, the fourth intermediate feature can be fused with the second intermediate feature to obtain a first fused feature. For example, the scale of the fourth intermediate feature can match the scale of the second intermediate feature, so that the fourth intermediate feature with matching scales can be spliced ​​with the second intermediate feature to obtain the first fused feature. Thus, fusing two features of the same dimension can prevent the deep network from losing shallow feature information.

[0078] For example, the structure of the coal-rock identification model is as follows Figure 4 As shown in the example, the fourth intermediate feature can be fused with the second intermediate feature through the skip connection layer in the first sub-network to obtain the first fused feature. The skip connection layer is used to introduce feature information at the corresponding scale into the upsampling process to solve the problem of gradient disappearance.

[0079] Afterwards, the fourth feature extraction layer in the first sub-network may be used to perform feature extraction on the first fused feature to obtain a fifth intermediate feature.

[0080] The structure of the coal-rock identification model is as follows Figure 4 As shown in the example, the fourth feature extraction layer may include one convolution layer, and the first fusion feature may be input into the convolution layer of the fourth feature extraction layer to obtain the fifth intermediate feature output by the convolution layer.

[0081] Afterwards, the fifth intermediate feature can be upsampled to obtain the sixth intermediate feature. Figure 4 As shown for example, the fifth intermediate feature can be upsampled by the upsampling layer in the first sub-network, and the obtained sixth intermediate feature can be, for example, a 128×16×64 stereo feature.

[0082] Afterwards, the sixth intermediate feature can be fused with the first intermediate feature to obtain a second fused feature. For example, the scale of the sixth intermediate feature can match the scale of the first intermediate feature, so that the scale-matched sixth intermediate feature can be spliced ​​with the first intermediate feature to obtain the second fused feature. Finally, the fifth feature extraction layer in the first sub-network can be used to extract features from the second fused feature to obtain a prediction feature.

[0083] The structure of the coal-rock identification model is as follows Figure 4 As shown in the example, the fifth feature extraction layer may include one convolution layer, and the second fusion feature may be input into the convolution layer of the fifth feature extraction layer to obtain the prediction feature output by the convolution layer.

[0084] In another possible implementation of the embodiment of the present disclosure, the fourth feature extraction layer in the first sub-network is used to perform feature extraction on the fourth intermediate feature to obtain the seventh intermediate feature.

[0085] For example, the structure of the coal-rock identification model is as follows Figure 5 As shown in the example, the fourth feature extraction layer may include 1 convolution layer, and the fourth intermediate feature may be input into the convolution layer of the fourth feature extraction layer to obtain the seventh intermediate feature output by the convolution layer.

[0086] Afterwards, the seventh intermediate feature can be upsampled to obtain the eighth intermediate feature. Figure 5 As shown for example, the eighth intermediate feature may be, for example, a 128×16×64 three-dimensional feature.

[0087] Finally, the fifth feature extraction layer in the first sub-network may be used to perform feature extraction on the eighth intermediate feature to obtain a prediction feature.

[0088] The structure of the coal-rock identification model is as follows Figure 5 As shown in the example, the fifth feature extraction layer may include one convolution layer, and the eighth intermediate feature may be input into the convolution layer of the fifth feature extraction layer to obtain the predicted feature output by the convolution layer.

[0089] Therefore, it is possible to extract prediction features indicating the spatial distribution of coal seams and rock strata in the coal and rock strata to be tested from the radar signal to be tested through the first sub-network with different structures, which can improve the flexibility and applicability of the method.

[0090] Step 307: Use the second subnetwork in the coal-rock identification model to predict the coal seam thickness based on the predicted features to obtain a first thickness distribution of the coal seam in the coal-rock layer to be detected on the ground penetrating radar detection path.

[0091] For explanation of step 307, reference may be made to the relevant description in any embodiment of the present disclosure, and will not be repeated here.

[0092] In a possible implementation of the embodiment of the present disclosure, the sixth feature extraction layer in the second sub-network is used to extract the predicted feature to obtain a first candidate feature, wherein the scale of the first candidate feature is smaller than the predicted feature.

[0093] The structure of the coal-rock identification model is as follows: Figure 4 or Figure 5 For example, the sixth feature extraction layer may include two maximum pooling layers and one convolution layer, and the predicted feature may be sequentially input into the maximum pooling layer, the convolution layer, and the maximum pooling layer of the sixth feature extraction layer to obtain the first candidate feature output by the last maximum pooling layer of the sixth feature extraction layer, wherein the scale of the first candidate feature is smaller than the scale of the predicted feature. For example, the first candidate feature may be a 32×4×64 stereo feature.

[0094] Afterwards, each element in the first candidate feature can be stretched or flattened to obtain the second candidate feature. In other words, the multi-dimensional first candidate feature can be transformed into one dimension to obtain the second candidate feature. Figure 4 or Figure 5 As an example, the first candidate feature can be flattened by the flattening layer in the second sub-network, and the obtained second candidate feature can be a one-dimensional feature of 8192.

[0095] Finally, the second candidate features can be sequentially input into the multi-layer fully connected layers in the second sub-network to determine the first thickness distribution of the coal seam in the coal rock layer to be tested according to the output of the last fully connected layer in the multi-layer fully connected layers.

[0096] The structure of the coal-rock identification model is as follows Figure 4 or Figure 5As an example, the second subnetwork may include 2 fully connected layers, and the second candidate features may be sequentially input into the 2 fully connected layers in the second subnetwork to determine the first thickness distribution of the coal seam in the coal rock layer to be tested based on the output of the last fully connected layer.

[0097] Among them, the fully connected layer is used to map the learned feature representation to the label space of the sample.

[0098] in, Figure 4 and Figure 5 The downsampling module in the first subnetwork is used to extract features from the radar signal image to obtain lithology information and structural information; the upsampling module is used to spatially reconstruct the extracted lithology information to reflect the lithology spatial distribution characteristics, that is, the spatial distribution of coal seams and rock strata. The second subnetwork is used to extract the coal seam structural characteristics from the lithology spatial distribution characteristics and map them into coal seam thickness distribution.

[0099] As an example, the radar signal image to be measured collected by the ground penetrating radar can be as follows: Figure 6 As shown in (a), the spatial distribution of coal seams and rock layers in the coal and rock layers to be tested indicated by the prediction features output by the first sub-network can be as follows: Figure 6 As shown in (b), the coal seam thickness distribution output by the second subnetwork can be expressed as Figure 6 (c), where Figure 6 (c) Used to indicate the thickness of the coal seam at different locations along the GPR detection path.

[0100] in, Figure 6 (a) Figure 6 (b) and Figure 6 The horizontal axis of (c) represents the spatial position of each point on the GPR detection path. Figure 6 The vertical axis of (a) represents time, for example, 0-10ns. In this example, 10ns is divided into 512 sampling points to correspond to the input size of the coal-rock identification model. Figure 6 The vertical axis of (b) represents height, Figure 6 The vertical axis of (c) represents thickness.

[0101] The coal seam thickness prediction method of the embodiment of the present disclosure can predict the thickness of the coal seam in the coal rock layer to be tested based on deep learning technology, and can improve the accuracy of the prediction results.

[0102] In order to clearly explain how the coal-rock identification model in the above embodiment is trained, the present disclosure also proposes a training method for the coal-rock identification model.

[0103] Figure 7 A schematic diagram of the training process of a coal-rock identification model provided in an embodiment of the present disclosure.

[0104] like Figure 7As shown, in Figure 1 or Figure 3 Based on the embodiment shown, the coal-rock identification model can be trained by the following steps:

[0105] Step 701: Use ground penetrating radar to detect the sample coal and rock layer to obtain a sample radar signal.

[0106] In the embodiments of the present disclosure, the sample coal rock layer and the coal rock layer to be tested may be the same or different, and the present disclosure does not limit this. For example, the sample coal rock layer may be a coal rock layer with a known coal rock layer thickness in a certain area, and the coal rock layer to be tested may be a coal rock layer with an unknown coal rock layer thickness.

[0107] In the disclosed embodiment, a ground penetrating radar may be used to detect the sample coal layer to obtain a sample radar signal.

[0108] As an example, the radiation direction of the ground penetrating radar can be vertically upward and toward the sample coal rock layer, the moving direction is horizontal, and the moving speed is uniform.

[0109] Step 702: measuring the coal seam thickness at each position point on the ground penetrating radar detection path to obtain a second thickness distribution.

[0110] In the disclosed embodiment, a relevant measuring device may be used to measure the thickness of the coal seam at each position point on the ground penetrating radar detection path to obtain a second thickness distribution of the coal seam.

[0111] As an example, some coal and rock layers are exposed. For example, when the mining equipment was mining for the last time, part of the rock layer was cut or naturally fell off, exposing the coal and rock layer in front. At this time, the boundary between the coal seam and the rock layer can be directly observed by eye. Therefore, the thickness of the coal seam can be directly measured with a ruler, thereby obtaining the second thickness distribution of the coal seam.

[0112] As another example, a core drill may be used to obtain cores at set intervals (eg, 0.5 meters) on the ground penetrating radar detection path to measure the thickness of the coal seam to obtain a second thickness distribution of the coal seam.

[0113] Step 703: label the sample radar signal according to the second thickness distribution to obtain a training sample.

[0114] In the embodiment of the present disclosure, the sample radar signal can be annotated according to the second thickness distribution to obtain a training sample, that is, the sample radar signal can be used as the input of the model, and the second thickness distribution can be used as the annotation information of the sample radar signal.

[0115] Step 704: input the training sample into the initial coal-rock identification model to obtain a third thickness distribution output by the initial coal-rock identification model.

[0116] In the disclosed embodiment, the training sample can be input into the initial coal-rock identification model to obtain the third thickness distribution output by the initial coal-rock identification model. That is, the first sub-network in the initial coal-rock identification model can be used to extract features of the sample radar signal to obtain predicted sample features, and the second sub-network in the initial coal-rock model can be used to predict the coal seam thickness of the predicted sample features to obtain the coal seam thickness of each position point on the ground penetrating radar detection path, that is, the third thickness distribution of the coal seam in the sample coal-rock layer.

[0117] Step 705: training the initial coal-rock identification model according to the third thickness distribution and the second thickness distribution to obtain a trained coal-rock identification model.

[0118] In the disclosed embodiment, the initial coal-rock identification model may be trained according to the difference between the third thickness distribution and the second thickness distribution to obtain a trained coal-rock identification model.

[0119] As an example, the calculated value of the loss function (referred to as the loss value in the present disclosure) can be determined based on the difference between the third thickness distribution and the second thickness distribution, where the loss function may include but is not limited to a square loss function, a mean square error loss function, a root mean square error loss function, a smoothed mean absolute error loss function, an absolute value loss function, a mean absolute percentage error loss function, a root mean square logarithmic error loss function, an L1 norm (Huber) loss function, a regression (Log-Cosh) loss function, a quantile (Quantile) loss function, and the like.

[0120] Therefore, in the present disclosure, the model parameters in the initial coal-rock identification model can be adjusted according to the loss value to minimize the loss value and obtain a trained coal-rock identification model.

[0121] It should be noted that the above only uses the termination condition of model training as the minimization of the loss value as an example. In actual application, other termination conditions can also be set, such as the number of training times reaching a set number of times, the training time reaching a set time, the loss value convergence, etc. The present disclosure does not impose any restrictions on this.

[0122] In summary, by pre-training the initial coal-rock identification model to obtain a trained coal-rock identification model, and then using the trained coal-rock identification model to predict the thickness distribution of coal seams, the accuracy of the prediction results can be improved.

[0123] When there are multiple training samples, in order to clearly explain how the initial coal-rock identification model is trained in the above embodiment, the present disclosure also proposes a method for training the coal-rock identification model.

[0124] Figure 8A schematic diagram of the training process of another coal-rock identification model provided in an embodiment of the present disclosure.

[0125] like Figure 8 As shown, in Figure 1 or Figure 3 Based on the embodiment shown, the coal-rock identification model can be trained by the following steps:

[0126] Step 801: Use ground penetrating radar to detect the sample coal and rock layer to obtain a sample radar signal.

[0127] Step 802, measuring the coal seam thickness at each position point on the ground penetrating radar detection path to obtain a second thickness distribution.

[0128] Step 803: labeling the sample radar signal according to the second thickness distribution to obtain a training sample.

[0129] The explanation of steps 801 to 803 can be found in the relevant description of the above embodiment, which will not be repeated here.

[0130] Step 804: input the multiple training samples into the initial coal-rock identification model respectively to obtain a third thickness distribution of the multiple training samples output by the initial coal-rock identification model.

[0131] In the disclosed embodiment, multiple training samples can be respectively input into the initial coal-rock identification model to obtain the third thickness distribution of the multiple training samples output by the initial coal-rock identification model. That is, for any one of the multiple training samples, the training sample can be input into the initial coal-rock identification model to obtain the third thickness distribution of the training sample output by the initial coal-rock identification model. Specifically, the first sub-network in the initial coal-rock identification model can be used to extract features of the sample radar signal in the training sample to obtain predicted sample features, and the second sub-network in the initial coal-rock model can be used to predict the thickness of the coal seam on the predicted sample features to obtain the third thickness distribution of the coal seam in the sample coal rock layer on the ground penetrating radar detection path.

[0132] Step 805 , training the initial coal-rock identification model according to the third thickness distribution marked on the multiple training samples and the second thickness distribution of the multiple training samples to obtain a trained coal-rock identification model.

[0133] In the embodiment of the present disclosure, the initial coal-rock identification model may be trained based on the second thickness distribution marked on multiple training samples and the third thickness distribution of multiple training samples to obtain a trained coal-rock identification model.

[0134] As a possible implementation method, for any training sample among multiple training samples, the difference between the third thickness distribution and the second thickness distribution corresponding to the any training sample can be determined, and based on the difference between the third thickness distribution and the second thickness distribution of the multiple training samples, the calculated value of the loss function (referred to as the loss value in this disclosure) can be determined, wherein the loss function may include but is not limited to a square loss function, a mean square error loss function, an absolute value loss function, etc.

[0135] Therefore, in the present disclosure, the model parameters in the initial coal-rock identification model can be adjusted according to the loss value to minimize the loss value and obtain a trained coal-rock identification model.

[0136] For example, when the loss function is a square loss function, the loss value L can be determined by the following formula:

[0137]

[0138] Among them, y i is the second thickness distribution of the i-th training sample, x i is the i-th training sample, f(x i ) is the third thickness distribution of the i-th training sample output by the initial coal-rock identification model, and n is the number of training samples.

[0139] For another example, when the loss function is a mean square error loss function, the loss value L can be determined by the following formula:

[0140]

[0141] For another example, when the loss function is an absolute value loss function, the loss value L can be determined by the following formula:

[0142]

[0143] It should be noted that the above only uses the termination condition of model training as the minimization of the loss value as an example. In actual application, other termination conditions can also be set, such as the number of training times reaching a set number of times, the training time reaching a set time, the loss value convergence, etc. The present disclosure does not impose any restrictions on this.

[0144] In the present disclosure, the spatial distribution of coal seams and rock strata is directly predicted by deep learning technology, and the thickness of coal seams is calculated, so that the automatic identification and prediction of the boundary between coal seams and rock strata and the thickness of coal seams can be realized. After the model training is completed, the prediction of coal seams is automatically completed by electronic devices with computing power without human participation, which meets the needs of intelligent identification of coal and rock for unmanned intelligent mining in mines. In addition, the coal and rock stratum selection and dielectric constant correction methods are optimized through deep learning methods, and the prediction results are more accurate.

[0145] Compared with the first method in the related art, in the present disclosure, in the prediction stage, there is no need to measure the relative dielectric constant of the coal seam, and there is no need to calculate parameters such as the propagation speed of the electromagnetic wave in the measured coal seam. Instead, deep learning technology is used to extract characteristic information in the radar signal to be measured. The spatial distribution of geological bodies with different relative dielectric constants can be directly predicted based on the above characteristic information, thereby realizing the division or identification of the interface between the coal seam and the rock layer, and the prediction of the thickness of the coal seam.

[0146] Compared with the second method in the related art, in the present disclosure, in the prediction stage, deep learning technology is used to directly predict the thickness of the coal seam without the need for human intervention or obtaining the relative dielectric constant parameters of the drilling map, which can reduce the workload of prediction and the degree of human involvement.

[0147] Specifically, the prediction of coal seam thickness can be achieved through the following steps:

[0148] Step A: collecting a training set for training a coal-rock identification neural network (referred to as a coal-rock identification model in this disclosure).

[0149] Step A1, the radiation direction of the ground penetrating radar can be vertically upward and toward the coal rock layer (such as the sample coal rock layer), the moving direction is horizontal, and the moving mode can be uniform movement or non-uniform movement. As an example, the moving mode and moving direction of the ground penetrating radar can be shown in Figure 2 (a), and the ground penetrating radar can move horizontally at a uniform speed for a certain distance to collect radar signals (recorded as sample radar signals in this disclosure). As another example, the moving direction of the ground penetrating radar can be shown in Figure 2 (b), and the ground penetrating radar can be connected to a ranging device, wherein the ranging device can move at a uniform speed or non-uniform speed. Whenever the ranging device moves a set interval, the ground penetrating radar can be controlled to collect data once.

[0150] Step A2, measuring the thickness of the coal seam at each position point on the ground penetrating radar detection path to obtain the thickness distribution of the coal seam, which is recorded as the second thickness distribution in the present disclosure.

[0151] Step A3, use the second thickness distribution to mark the sample radar signal to obtain training data (referred to as training samples in this disclosure), that is, the second thickness distribution of the coal seam can be used as a prediction target, and the sample radar signal can be used as an input of the coal rock recognition model.

[0152] Step A4, repeat steps A1-A3, collect no less than 1000 training data, and generate a training set based on the collected training data, wherein the amount of training data in the training set is 90%, and the remaining 10% of the training data is used as a validation set.

[0153] Step B: construct a coal-rock identification model.

[0154] A coal-rock identification model consisting of a first subnetwork (or lithology identification subnetwork) and a second subnetwork (or coal seam thickness prediction subnetwork) is constructed, and model parameters (or network parameters) are set.

[0155] As an example, the structure of the coal-rock identification model can be as follows: Figure 4 or Figure 5 As shown. The first subnetwork is used to extract the features of the sample radar signal, wherein the features are used to indicate the spatial distribution of the coal seam and the rock layer, that is, the first subnetwork can perform semantic segmentation on the sample radar signal to obtain the interface between the coal seam and the rock layer. The second subnetwork is composed of a maximum pooling layer, a convolutional layer, a flattening layer, and a fully connected layer, and outputs the coal seam thickness at each position point on the ground penetrating radar acquisition path or detection path (referred to as the third thickness distribution in this disclosure).

[0156] Step C: training the coal-rock identification model.

[0157] The training data in the training set is input into the coal-rock recognition model. The coal-rock recognition model is trained for no less than 100 times using the random gradient descent method, RMSProp (root mean square propagation gradient descent) optimization method, Adam (Adaptive Moment Estimation) optimization method, Ftrl (Follow-the-regularized-Leader) optimization method, etc., to obtain a trained coal-rock recognition model.

[0158] Step D: deploy the coal-rock identification model.

[0159] The trained coal-rock identification model is transmitted to the storage medium of the computer equipment in the mine.

[0160] Step E: collecting radar signals of the coal stratum to be tested.

[0161] Connect the ground penetrating radar to the computer equipment in the mine, and obtain the radar signal to be tested in real time through the computer equipment.

[0162] Step F, predicting and outputting the coal seam thickness.

[0163] The collected radar signal to be tested is input into the trained coal-rock recognition model, and the coal-rock recognition model predicts the thickness distribution of the coal seam at each position point on the ground penetrating radar collection path or detection path (referred to as the first thickness distribution in this disclosure) and outputs it.

[0164] As an example, the structure of the coal-rock identification model is as follows Figure 4 As shown in the example, the coal rock recognition model can be composed of 9 convolutional layers, 6 pooling layers, 2 upsampling layers, 2 skip connection layers, 1 flattening layer, and 2 fully connected layers.

[0165] As another example, the structure of the coal-rock identification model is as follows: Figure 5 As shown in the example, the coal rock recognition model can be composed of 9 convolutional layers, 6 pooling layers, 2 upsampling layers, 1 flattening layer, and 2 fully connected layers.

[0166] Among them, the convolution kernel size of the convolution layer can be 3×3, the step size is all 1, and the padding parameter is to pad and retain the spatial dimension; the size of the maximum pooling layer and the upsampling layer can be 2×2, the maximum pooling layer step size is all 1 and no padding; the sizes of the two fully connected layers are 1024 and 160 (or 1024 and n) respectively, where n is the number of sampling points on the detection path.

[0167] Optionally, a Dropout layer can be set after the flattening layer and after the first fully connected layer, and the dropout rate is set to 0.4. The function of the Dropout layer is to alleviate the occurrence of overfitting and achieve a regularization effect to a certain extent.

[0168] Among them, the last fully connected layer of the coal rock recognition model does not use an activation function, and the remaining convolutional layers and fully connected layers can use the following activation functions: Sigmoid function, hyperbolic tangent function (Tanh function), linear rectification function (ReLU function), Leaky ReLU function, ELU function, etc.

[0169] Optionally, the coal-rock identification model can use mean square error as the loss function and use an SGD (Stochastic Gradient Descent) optimizer to optimize the loss function.

[0170] In summary, a coal-rock identification neural network (referred to as a coal-rock identification model in this disclosure) is constructed using a deep learning method, and based on the coal-rock identification model, the radar signal (such as the radar signal to be measured) collected by the ground-penetrating radar of the coal-rock identification scene is processed or identified to obtain the thickness of the coal seam. The prediction process does not require manual operation, and there is no need to predict or measure the lithology of the coal seam (such as the relative dielectric constant), so that the automatic prediction of the thickness of the coal seam can be achieved. Among them, the first subnetwork of the coal-rock identification model (or the lithology identification subnetwork) is used to predict the spatial distribution of coal seams and rock seams. The first subnetwork can be, for example, a neural network structure similar to the "UNet" structure or a neural network structure similar to the "SegNet" structure. The second subnetwork of the coal-rock identification model (or the coal seam thickness prediction subnetwork) is used to predict the thickness distribution of the coal seam. The second subnetwork can be a sequential structure connected after the first subnetwork (lithology identification subnetwork).

[0171] Compared with the solutions in the related art, the present disclosure has at least the following advantages:

[0172] First, it can realize automatic prediction of coal seam thickness. It uses deep learning technology to directly extract features from the radar signal to predict coal seam thickness, without the need for manual preprocessing and analysis of the radar signal.

[0173] Second, in the prediction stage, only the radar signal to be measured needs to be input, and the model can output the thickness distribution of the coal seam. Compared with the traditional method of using physical formulas to calculate the thickness of the coal seam, which requires measuring the relative dielectric constant of the coal seam, the workload in the prediction process can be reduced. In addition, the thickness distribution of the coal seam (i.e., the first thickness distribution) can be directly predicted based on the radar signal to be measured without measuring the relative dielectric constant, which can realize the automatic prediction of the thickness of the coal seam.

[0174] Third, the prediction calculation speed is fast and real-time prediction can be achieved. The reason is that the coal seam thickness prediction using the trained model requires less computing resources, while traditional methods (such as full waveform inversion) require a large amount of computing.

[0175] It should be noted that, although a dielectric constant tester can be added to the acquisition system based on the first method in the related art to achieve accurate calculation of coal and rock layers, the measurement of dielectric constant requires sampling, and real-time, in-situ measurement cannot be achieved. In addition, the quality of radar signals is affected by many on-site factors, and it is difficult to automatically select coal and rock layers by computer programs alone, which makes application difficult.

[0176] In addition, the dielectric constant distribution image of the tunnel geological body can be predicted based on the neural network, and the spatial distribution of steel bars, water-free cracks and water-containing defects in the tunnel geological body can be predicted based on the dielectric constant distribution image. Although the above method can be extended to the field of coal-rock identification, the interface between the coal seam and the rock layer can be identified based on the neural network, and the thickness of the coal seam can be predicted based on the interface between the coal seam and the rock layer. However, this neural network is very cumbersome in the collection of training data, because its prediction result is the dielectric constant distribution map image of the geological body. Therefore, when making the training set, it is necessary not only to measure the thickness of the coal seam, but also to measure the dielectric constant of the coal seam and the rock layer at different locations. If the dielectric constant is freely defined by computer modeling and the radar signal is obtained by model calculation, although the difficulty of obtaining training data can be reduced, the obtained training data cannot reflect the impact of the real world on the radar signal, and the prediction effect is not good.

[0177] With the above Figures 1 to 8 Corresponding to the method for predicting coal seam thickness provided in the embodiment, the present disclosure also provides a coal seam thickness prediction device. Figures 1 to 8The coal seam thickness prediction method provided in the embodiment corresponds to the embodiment, so the implementation method of the coal seam thickness prediction method is also applicable to the coal seam thickness prediction device provided in the embodiment of the present disclosure, and will not be described in detail in the embodiment of the present disclosure.

[0178] Fig. 9 A schematic diagram of the structure of a coal seam thickness prediction device provided in an embodiment of the present disclosure.

[0179] like Fig. 9 As shown, the coal seam thickness prediction device 900 may include: a first detection module 901 , an extraction module 902 and a prediction module 903 .

[0180] The first detection module 901 is used to detect the coal stratum to be detected by using a ground penetrating radar to obtain a radar signal to be detected.

[0181] The extraction module 902 is used to extract features of the radar signal to be tested using the first sub-network in the coal-rock identification model to obtain prediction features; wherein the prediction features are used to indicate the spatial distribution of coal seams and rock seams in the coal-rock seam to be tested.

[0182] The prediction module 903 is used to predict the coal seam thickness based on the prediction features using the second sub-network in the coal-rock identification model, so as to obtain the first thickness distribution of the coal seam in the coal-rock layer to be detected on the ground penetrating radar detection path.

[0183] In a possible implementation of the embodiment of the present disclosure, the extraction module 902 is specifically used to: use the first feature extraction layer in the first subnetwork to perform feature extraction on the radar signal to be measured to obtain a first intermediate feature; use the second feature extraction layer in the first subnetwork to perform feature extraction on the first intermediate feature to obtain a second intermediate feature, wherein the scale of the second intermediate feature is smaller than that of the first intermediate feature; use the third feature extraction layer in the first subnetwork to perform feature extraction on the second intermediate feature to obtain a third intermediate feature, wherein the scale of the third intermediate feature is smaller than that of the second intermediate feature; upsample the third intermediate feature to obtain a fourth intermediate feature; and generate a prediction feature based on the fourth intermediate feature.

[0184] In a possible implementation of the embodiment of the present disclosure, the extraction module 902 is specifically used to: fuse the fourth intermediate feature with the second intermediate feature to obtain a first fused feature; use the fourth feature extraction layer in the first sub-network to perform feature extraction on the first fused feature to obtain a fifth intermediate feature; upsample the fifth intermediate feature to obtain a sixth intermediate feature, wherein the scale of the sixth intermediate feature matches the scale of the first intermediate feature; fuse the sixth intermediate feature with the first intermediate feature to obtain a second fused feature; and use the fifth feature extraction layer in the first sub-network to perform feature extraction on the second fused feature to obtain a prediction feature.

[0185] In a possible implementation of the embodiment of the present disclosure, the extraction module 902 is specifically used to: use the fourth feature extraction layer in the first sub-network to perform feature extraction on the fourth intermediate feature to obtain the seventh intermediate feature; upsample the seventh intermediate feature to obtain the eighth intermediate feature; use the fifth feature extraction layer in the first sub-network to perform feature extraction on the eighth intermediate feature to obtain a prediction feature.

[0186] In a possible implementation of the embodiment of the present disclosure, the prediction module 903 is specifically used to: use the sixth feature extraction layer in the second sub-network to extract the predicted feature to obtain a first candidate feature, wherein the scale of the first candidate feature is smaller than the predicted feature; stretch each element in the first candidate feature to obtain a second candidate feature; and sequentially input the second candidate feature into the multi-layer fully connected layer in the second sub-network to determine the first thickness distribution of the coal seam in the coal rock layer to be measured based on the output of the last fully connected layer in the multi-layer fully connected layer.

[0187] In a possible implementation of the embodiment of the present disclosure, the coal-rock identification model is obtained by training the following modules:

[0188] The second detection module is used to detect the sample coal and rock layer using a ground penetrating radar to obtain a sample radar signal;

[0189] A measurement module, used for measuring the thickness of the coal seam at each position point on the detection path of the ground penetrating radar to obtain a second thickness distribution;

[0190] A labeling module, used for labeling the sample radar signal according to the second thickness distribution to obtain a training sample;

[0191] An input module, used for inputting the training samples into the initial coal-rock identification model to obtain a third thickness distribution output by the initial coal-rock identification model;

[0192] The training module is used to train the initial coal-rock identification model according to the third thickness distribution and the second thickness distribution to obtain a trained coal-rock identification model.

[0193] In a possible implementation of the embodiment of the present disclosure, there are multiple training samples, and the training module is specifically used to: determine the difference between the third thickness distribution and the second thickness distribution corresponding to any training sample among the multiple training samples; generate a loss value based on the difference between the multiple training samples; and adjust the model parameters in the initial coal-rock identification model based on the loss value to minimize the loss value.

[0194] The coal seam thickness prediction device of the embodiment of the present disclosure detects the coal and rock layers to be tested by using a ground penetrating radar to obtain a radar signal to be tested; the first sub-network in the coal and rock identification model is used to extract features of the radar signal to be tested to obtain prediction features; wherein the prediction features are used to indicate the spatial distribution of coal seams and rock layers in the coal and rock layers to be tested; the second sub-network in the coal and rock identification model is used to predict the coal seam thickness of the prediction features to obtain the first thickness distribution of the coal seam in the coal and rock layers to be tested on the ground penetrating radar detection path. Thus, based on the deep learning technology, the coal seam thickness of each position point on the ground penetrating radar detection path (i.e., the thickness distribution of the coal seam in the coal and rock layers to be tested) is predicted. On the one hand, the accuracy of the prediction result can be improved, and on the other hand, the prediction speed is fast, and real-time prediction can be achieved. Moreover, compared with the method of calculating the thickness of the coal seam using physical formulas in the prior art, in the present disclosure, there is no need to measure the relative dielectric constant of the coal seam, and there is no need to calculate parameters such as the propagation speed, which can reduce the workload of the prediction process. In addition, the thickness distribution of the coal seam can be directly predicted according to the radar signal to be tested without measuring the relative dielectric constant, and the automatic prediction of the coal seam thickness can be achieved.

[0195] In order to implement the above embodiments, the present disclosure also proposes an electronic device, wherein the electronic device can be any device with computing capability, and the electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the program, a coal seam thickness prediction method as proposed in any of the aforementioned embodiments of the present disclosure is implemented.

[0196] In order to implement the above embodiments, the present disclosure also proposes a non-temporary computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the coal seam thickness prediction method proposed in any of the aforementioned embodiments of the present disclosure is implemented.

[0197] In order to implement the above embodiments, the present disclosure also proposes a computer program product. When the instructions in the computer program product are executed by a processor, the coal seam thickness prediction method proposed in any of the above embodiments of the present disclosure is executed.

[0198] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0199] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0200] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present disclosure belong.

[0201] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0202] It should be understood that the various parts of the present disclosure can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0203] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0204] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0205] The storage medium mentioned above may be a read-only memory, a disk or an optical disk, etc. Although the embodiments of the present disclosure have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present disclosure. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present disclosure.

Claims

1. A method for predicting coal seam thickness, characterized in that: The method comprises: Using ground penetrating radar to detect the coal and rock layers to be tested, so as to obtain radar signals to be tested; The first subnetwork in the coal-rock identification model is used to extract features from the radar signal to be measured to obtain prediction features; wherein the prediction features are used to indicate the spatial distribution of coal seams and rock seams in the coal-rock seams to be measured; the first feature extraction layer in the first subnetwork is used to extract features from the radar signal to be measured to obtain a first intermediate feature; the second feature extraction layer in the first subnetwork is used to extract features from the first intermediate feature to obtain a second intermediate feature, wherein the scale of the second intermediate feature is smaller than that of the first intermediate feature; the third feature extraction layer in the first subnetwork is used to extract features from the second intermediate feature to obtain a third intermediate feature, wherein the scale of the third intermediate feature is smaller than that of the second intermediate feature; the third intermediate feature is upsampled to obtain a fourth intermediate feature; and the prediction features are generated according to the fourth intermediate feature; The second subnetwork in the coal-rock identification model is used to predict the coal seam thickness based on the prediction feature to obtain a first thickness distribution of the coal seam in the coal-rock layer to be detected on the ground penetrating radar detection path.

2. The method according to claim 1, characterized in that: Generating the prediction feature according to the fourth intermediate feature includes: fusing the fourth intermediate feature with the second intermediate feature to obtain a first fused feature; Using the fourth feature extraction layer in the first sub-network to perform feature extraction on the first fused feature to obtain a fifth intermediate feature; Upsampling the fifth intermediate feature to obtain a sixth intermediate feature, wherein a scale of the sixth intermediate feature matches a scale of the first intermediate feature; fusing the sixth intermediate feature with the first intermediate feature to obtain a second fused feature; The fifth feature extraction layer in the first sub-network is used to perform feature extraction on the second fused feature to obtain the prediction feature.

3. The method according to claim 1, characterized in that Generating the prediction feature according to the fourth intermediate feature includes: Using the fourth feature extraction layer in the first sub-network to perform feature extraction on the fourth intermediate feature to obtain a seventh intermediate feature; Upsampling the seventh intermediate feature to obtain an eighth intermediate feature; The fifth feature extraction layer in the first sub-network is used to perform feature extraction on the eighth intermediate feature to obtain the prediction feature.

4. The method according to claim 1, characterized in that The method of using the second subnetwork in the coal-rock identification model to predict the coal seam thickness based on the prediction feature to obtain a first thickness distribution of the coal seam in the coal-rock layer to be detected on the ground penetrating radar detection path includes: Using a sixth feature extraction layer in the second sub-network to perform feature extraction on the prediction feature to obtain a first candidate feature, wherein a scale of the first candidate feature is smaller than that of the prediction feature; Stretching each element in the first candidate feature to obtain a second candidate feature; The second candidate features are sequentially input into the multiple fully connected layers in the second sub-network to determine the first thickness distribution of the coal seam in the coal rock layer to be tested according to the output of the last fully connected layer in the multiple fully connected layers.

5. The method according to claim 1, characterized in that The coal-rock identification model is trained by the following steps: Using the ground penetrating radar to detect the sample coal and rock layer to obtain a sample radar signal; Measuring the thickness of the coal seam at each position point on the ground penetrating radar detection path to obtain a second thickness distribution; labeling the sample radar signal according to the second thickness distribution to obtain a training sample; Inputting the training sample into an initial coal-rock identification model to obtain a third thickness distribution output by the initial coal-rock identification model; The initial coal-rock identification model is trained according to the third thickness distribution and the second thickness distribution to obtain the trained coal-rock identification model.

6. The method according to claim 5, characterized in that The number of the training samples is multiple, and the initial coal-rock identification model is trained according to the third thickness distribution and the second thickness distribution, including: For any training sample among the plurality of training samples, determining a difference between a third thickness distribution and a second thickness distribution corresponding to the any training sample; Generating a loss value according to the difference of the plurality of training samples; The model parameters in the initial coal-rock identification model are adjusted according to the loss value to minimize the loss value.

7. A coal seam thickness prediction device, characterized in that: The device comprises: A detection module is used to detect the coal and rock layers to be detected by using a ground penetrating radar to obtain a radar signal to be detected; An extraction module, configured to use the first subnetwork in the coal-rock identification model to perform feature extraction on the radar signal to be measured to obtain a prediction feature; wherein the prediction feature is used to indicate the spatial distribution of coal seams and rock seams in the coal-rock layer to be measured, use the first feature extraction layer in the first subnetwork to perform feature extraction on the radar signal to be measured to obtain a first intermediate feature, use the second feature extraction layer in the first subnetwork to perform feature extraction on the first intermediate feature to obtain a second intermediate feature, wherein the scale of the second intermediate feature is smaller than that of the first intermediate feature, use the third feature extraction layer in the first subnetwork to perform feature extraction on the second intermediate feature to obtain a third intermediate feature, wherein the scale of the third intermediate feature is smaller than that of the second intermediate feature, upsample the third intermediate feature to obtain a fourth intermediate feature, and generate the prediction feature according to the fourth intermediate feature; The prediction module is used to use the second sub-network in the coal-rock identification model to predict the coal seam thickness based on the prediction feature to obtain the first thickness distribution of the coal seam in the coal-rock layer to be tested on the ground penetrating radar detection path.

8. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images