Downhole geophysical prospecting sampling data quality monitoring method combined with dynamic cross validation

Abstract

The invention relates to the technical field of data processing, and provides an underground geophysical prospecting sampling data quality monitoring method combined with dynamic cross validation. The method comprises the following steps: performing coal mine geophysical prospecting through a seismic reflection method to obtain a plurality of data sets; analyzing reflection data stability of the data set, setting a cross proportion, and performing cross verification on the data set to obtain multiple pieces of verification information; generating coal thickness distribution according to the verification information; and setting a verification proportion based on the cross proportion, carrying out cross verification on the coal thickness distribution, obtaining accuracy information, and calculating total accuracy information as a sampling data quality monitoring result of the current underground coal mine geophysical prospecting. The technical problem that the sampling data quality is difficult to accurately analyze due to insufficient accuracy of the underground geophysical prospecting sampling data is solved, the cross proportion is automatically adjusted through a dynamic cross validation method, the stability and accuracy of the data are enhanced, and the accuracy of the data is improved. And the coal thickness distribution generation precision of the coal mine and the sampling data quality monitoring reliability are improved.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and in particular to a method for monitoring the quality of downhole geophysical sampling data combined with dynamic cross-validation. Background Art

[0002] Underground geophysical exploration technology is a common detection method in coal mining. Especially in complex geological environments, the use of seismic reflection method to detect underground coal seams can provide important data support for coal mining and safety. However, the current underground geophysical sampling data monitoring method faces multiple technical difficulties. In practical applications, due to the complexity of geological conditions and the uneven distribution of sampling points, the collected data often have large fluctuations, and it is difficult to directly obtain stable coal thickness distribution information through simple analysis. In addition, traditional coal thickness data analysis methods are often difficult to accurately and stably generate coal thickness distribution information in the face of large-scale and diverse data sets, which in turn affects the quality monitoring results of sampling data. Summary of the invention

[0003] The present application provides a method for monitoring the quality of downhole geophysical sampling data combined with dynamic cross-validation, aiming to solve the technical problem that it is difficult to accurately analyze the quality of sampling data due to insufficient accuracy of downhole geophysical sampling data.

[0004] In view of the above problems, the present application provides a method for monitoring the quality of downhole geophysical sampling data combined with dynamic cross-validation.

[0005] The present application provides a method for monitoring the quality of underground geophysical sampling data combined with dynamic cross-validation, the method comprising: performing underground coal exploration at multiple underground monitoring points in a target mining area by means of a seismic reflection method to obtain multiple seismic reflection data sets; analyzing the reflection data stability of the multiple seismic reflection data sets, setting multiple cross-ratios for performing cross-validation analysis of coal thickness on the multiple seismic reflection data sets, performing cross-validation analysis of coal thickness on the multiple seismic reflection data sets, and obtaining multiple verified coal thickness information of the coal mines; generating coal thickness distributions of the coal mines in the target mining area according to the multiple verified coal thickness information of the coal mines, and obtaining multiple coal thickness distributions of the coal mines; setting multiple coal thickness distribution verification ratios based on the multiple cross-ratios, cross-validating the multiple coal thickness distributions of the coal mines, and obtaining cross-validation accuracy information of the multiple coal thickness distributions of the coal mines, and calculating and obtaining total cross-validation accuracy information as the sampling data quality monitoring result of the current underground coal exploration.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The above-mentioned underground geophysical sampling data quality monitoring method combined with dynamic cross-validation, this method performs coal prospecting by seismic reflection method at multiple underground monitoring points in the target mining area, and obtains multiple seismic reflection data sets. Subsequently, the reflection data stability of these data sets is analyzed to determine the fluctuation of each data set. Based on this, multiple cross-validation ratios are set to ensure that the stability of different data sets can be adapted during cross-validation analysis. Through the analysis of different cross-validation ratios, the accurate verification of coal thickness information can be effectively improved, thereby obtaining multiple reliable verified coal thickness information. Afterwards, a coal thickness distribution map in the target mining area is generated based on the verified coal thickness information. This process can more accurately reflect the actual changes in coal thickness. In order to ensure the reliability of the generated coal thickness distribution map, the verification ratio is set based on the cross-ratio, and these distribution maps are further cross-validated to ensure the accuracy of each distribution. In this process, the verification parameters can be dynamically adjusted to make the results closer to the actual situation. Finally, by calculating the total cross-validation accuracy, the quality monitoring results of the current geophysical sampling data are obtained, providing more reliable data support for coal mining, while reducing manual intervention and errors.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of a process of a method for monitoring downhole geophysical exploration sampling data quality in combination with dynamic cross-validation in one embodiment;

[0011] Figure 2 The figure is a flow chart of obtaining multiple seismic reflection data sets in a method for monitoring downhole geophysical sampling data quality in combination with dynamic cross-validation in one embodiment. DETAILED DESCRIPTION

[0012] The embodiment of the present application solves the technical problem that it is difficult to accurately analyze the quality of sampling data due to insufficient accuracy of downhole geophysical sampling data by providing a downhole geophysical sampling data quality monitoring method combined with dynamic cross-validation.

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0014] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0015] Examples, such as Figure 1 As shown, the present application provides a method for monitoring downhole geophysical sampling data quality in combination with dynamic cross-validation, the method comprising:

[0016] At multiple underground monitoring points in the target mining area, underground coal exploration is carried out using the seismic reflection method to obtain multiple seismic reflection data sets.

[0017] In one embodiment, at multiple underground monitoring points in the target mining area, the system terminal uses seismic reflection method for detection. Seismic reflection method is a technology that obtains underground structure information by sending seismic waves and recording reflected waves. Specifically, the source equipment (such as explosive source or vibration truck source) will send vibration signals underground. When these vibration waves encounter different geological layers, they will be partially reflected back. By receiving these reflected waves and recording them, multiple seismic reflection data sets can be formed, which provide a basis for subsequent coal thickness analysis and verification.

[0018] Further, if Figure 2 As shown, the present application provides multiple underground monitoring points in the target mining area, and uses seismic reflection method to conduct underground coal exploration and obtain multiple seismic reflection data sets, including:

[0019] At multiple underground monitoring points in the target mining area, the seismic reflection method is used to conduct underground coal exploration, and the reflection wave data at each moment is received, wherein each reflection wave data includes the reflection time and the reflection wave amplitude; the reflection wave data at multiple moments of the multiple underground monitoring points are integrated to obtain multiple seismic reflection data sets.

[0020] Preferably, at multiple underground monitoring points in the target mining area, the system terminal uses the seismic reflection method to conduct coal prospecting. The specific operation is to send out vibration waves through the source equipment. When these waves propagate underground, they will encounter different geological layers and reflect back. The receiving equipment (such as a seismic detector) at each underground monitoring point will receive these reflected wave data in real time, and record the time of each reflected wave return and the amplitude intensity of the wave. These reflection time and amplitude information can reflect the different characteristics of the underground structure. Subsequently, all the reflected wave data received by multiple monitoring points in different time periods are integrated to generate multiple seismic reflection data sets. Each seismic reflection data set contains detailed reflection information of the monitoring point at different times and locations, providing complete geological and coal seam thickness data for subsequent analysis.

[0021] Analyze the reflection data stability of the multiple seismic reflection data sets, set multiple cross-ratios for cross-validation analysis of coal mine thickness for the multiple seismic reflection data sets, perform cross-validation analysis of coal mine thickness on the multiple seismic reflection data sets, and obtain multiple verified coal mine thickness information respectively.

[0022] In one embodiment, after obtaining multiple seismic reflection data sets, the system terminal analyzes the volatility of each seismic reflection data set, and calculates the fluctuation amplitude of the data at each moment by comparing the amplitude of the reflection wave at each moment. The larger the fluctuation amplitude, the lower the stability of the data, indicating that the inaccuracy of the data is high. Subsequently, the fluctuation coefficient of each seismic reflection data set is obtained based on the fluctuation amplitude of each seismic reflection data set and the maximum seismic reflection wave amplitude in the seismic reflection data set. Based on the multiple fluctuation coefficients calculated again, the system terminal evaluates the stability of the reflection data based on these fluctuation coefficients, determines the stability coefficient of each seismic reflection data set, and uses each stability coefficient as the cross ratio of the corresponding seismic reflection data set for subsequent cross-validation analysis. Afterwards, based on these cross ratios, the system terminal combines the coal thickness predictor of the coal mine to cross-validate multiple seismic reflection data sets, and respectively verifies the coal thickness information of each seismic reflection data set to ensure the accuracy and stability of the data analysis.

[0023] Further, the present application provides a method for analyzing the stability of reflection data of the multiple seismic reflection data sets, and setting multiple crossover ratios for performing coal mine coal thickness cross-validation analysis on the multiple seismic reflection data sets, including:

[0024] In the multiple seismic reflection data sets, the fluctuation amplitude of the seismic reflection data is identified respectively to obtain multiple fluctuation coefficients as the reflection data instability coefficients of the multiple seismic reflection data sets; based on the multiple reflection data instability coefficients, multiple reflection data stability coefficients are calculated and obtained, and are set as multiple cross-ratios for cross-validation analysis of coal mine coal thickness of the multiple seismic reflection data sets.

[0025] Preferably, in each seismic reflection data set, the system terminal identifies and analyzes the fluctuation amplitude of the reflected wave data at each moment, and calculates the fluctuation amplitude of the data at each moment by comparing the amplitude of the reflected wave at each moment. Subsequently, the fluctuation amplitude of each seismic reflection data set is calculated by ratio calculation with the maximum seismic reflection wave amplitude in the data set to obtain the fluctuation coefficient of each seismic reflection data set. The fluctuation coefficient reflects the volatility of the data, and the greater the fluctuation, the more unstable the data. The fluctuation coefficient of each data set is defined as the reflection data instability coefficient, which is used to quantify the fluctuation of the data. Afterwards, based on these instability coefficients, the reflection data stability coefficient of each seismic reflection data set is calculated. The reflection data stability coefficient is obtained by subtracting the reflection data instability coefficient from 1, indicating the stability of the reflection data. According to these reflection data stability coefficients, the system terminal determines the crossover ratio of each seismic reflection data set in the subsequent coal mine coal thickness cross-validation analysis, ensuring that a greater weight is given to data with higher stability during cross-validation, and optimizing the accuracy and reliability of the validation results.

[0026] Further, the present application provides a method for performing a coal mine thickness cross-validation analysis on the multiple seismic reflection data sets, including:

[0027] According to the first cross ratio among the multiple cross ratios, an identification data set of the first cross ratio is randomly selected in the first seismic reflection data set, and other data in the first seismic reflection data set is used as a verification data set; a pre-trained coal mine coal thickness predictor is used to predict the coal thickness of the identification data set and the verification data set, and the mean of the predicted coal thickness is calculated respectively, so as to obtain the first predicted coal thickness information and the first verification coal thickness information.

[0028] Optionally, the system terminal randomly extracts a cross ratio from multiple cross ratios as the first cross ratio, and then uses the seismic reflection data set corresponding to the cross ratio as the first seismic reflection data set. Subsequently, according to the first cross ratio, a corresponding number of data are selected from the first seismic reflection data set as the identification data set, and the remaining data are used as the verification data set. Subsequently, using the pre-trained coal mine coal thickness prediction model, the identification data set and the verification data set are respectively predicted for coal thickness, and the first identification coal thickness prediction data and the first verification coal thickness prediction data are obtained. Afterwards, the means of the first identification coal thickness prediction data and the first verification coal thickness prediction data are respectively calculated to obtain the first set of predicted coal thickness information and the first set of verification coal thickness information. This process helps to effectively verify the data set to improve the accuracy of coal thickness prediction.

[0029] Calculate the cross-validation deviation of the first predicted coal thickness information and the first verification coal thickness information to obtain the verification accuracy; continue to randomly select the identification data set and the verification data set in the first seismic reflection data set according to the first cross-ratio to predict the coal thickness, and calculate the verification accuracy until the convergence times are reached, output the predicted coal thickness information corresponding to the maximum verification accuracy, and obtain the first verification coal mine thickness information; according to other multiple cross-ratios, perform coal mine coal thickness cross-validation analysis on other multiple seismic reflection data sets to obtain multiple verification coal mine thickness information with the maximum verification accuracy.

[0030] Optionally, after calculating the first predicted coal thickness information and the first verified coal thickness information, the system terminal calculates the difference between the first predicted coal thickness information and the first verified coal thickness information as a cross-validation deviation. This cross-validation deviation reflects the predicted difference between the first predicted coal thickness information and the first verified coal thickness information. The larger the difference, the higher the possibility that the identification data set contains stable data, because the difference in coal thickness prediction between stable data and unstable data is usually large. Therefore, the system terminal uses the calculated predicted difference as the verification accuracy of the identification data set. Subsequently, continue to randomly select different identification data sets and verification data sets in the first seismic reflection data set according to the same first cross ratio, perform multiple coal thickness predictions, and calculate the verification accuracy each time. Through multiple iterations, the verification accuracy gradually converges, that is, the verification accuracy tends to be stable, and the system terminal extracts the first predicted coal thickness information with the largest verification accuracy, and outputs it as the final first verified coal mine coal thickness information of this verification. Afterwards, the same cross-validation process is performed on other seismic reflection data sets according to the corresponding cross ratio, and finally outputs multiple predicted coal thickness information with the highest verification accuracy in each seismic reflection data set as the final multiple verified coal mine coal thickness information. This approach ensures adequate identification of different stability data sets and optimizes the accuracy of coal thickness prediction.

[0031] Furthermore, the present application provides a pre-training step of the coal mine coal thickness predictor, comprising:

[0032] A sample seismic reflection data set is collected, and a sample coal mine thickness information set is obtained according to the actual coal mine thickness annotation; the sample seismic reflection data set and the sample coal mine thickness information set are used to train a coal mine thickness predictor based on K-fold cross validation until the training converges.

[0033] Optionally, the system terminal collects a set of sample seismic reflection data containing multiple time and position, and annotates it in combination with the actual measured coal mine thickness information to generate a corresponding set of sample coal mine thickness information. These data sets provide the association between the actual coal seam thickness information and the seismic reflection data. Subsequently, the coal mine thickness predictor is trained based on the K-fold cross validation method using these sample data. Among them, the K-fold cross validation is to divide the sample data into K parts, where K-1 parts of the data are used for training each time, and the remaining part is used for verification. The coal mine thickness predictor can be constructed according to multi-layer perceptron (MLP), support vector machine (SVM), random forest (RF), etc. Taking the multi-layer perceptron as an example, the system terminal first uses the multi-layer perceptron to construct an initial coal mine thickness predictor structure, including an input layer, a hidden layer, an output layer, etc. The input layer receives the seismic reflection data, the hidden layer performs nonlinear transformation through a series of neurons, and finally the output layer generates the predicted coal thickness value. The number of hidden layers and the number of neurons in each layer can be adjusted according to the complexity of the data. Subsequently, the sample seismic reflection data set and the sample coal mine thickness information set are randomly divided into K subsets. In each iteration, K-1 subsets are used as training sets, and the remaining 1 subset is used as a validation set to ensure that each subset has the opportunity to be used for predictor validation to avoid overfitting of the predictor on certain data sets. After that, the training set is input into the initial coal mine thickness predictor for forward propagation, and is passed layer by layer through the input layer, hidden layer, output layer, etc. to calculate the prediction result of coal mine thickness. The mean square error (MSE) loss function is then used to calculate the loss value between the prediction result and the actual sample coal mine thickness information, and the gradient of the loss to the weight of each layer is calculated layer by layer through back propagation. Then, the Adam optimizer is used to optimize the predictor parameters, and the weights and biases of each layer are adjusted to minimize the value of the loss function. The entire training process will be carried out in K iterations, and each iteration will be verified using a different validation set to ensure the generalization ability of the model on different data sets. After the training is completed, the average error on all validation sets is calculated to evaluate the accuracy of the predictor in the coal mine thickness prediction task. If the accuracy of the predictor reaches the expected effect, the currently trained initial coal mine thickness predictor is output as the final coal mine thickness predictor. If it fails to meet expectations, the learning rate, batch size and other hyperparameters are adjusted according to the performance of the validation set, and training is continued to further improve the prediction effect of the coal mine thickness predictor. Through multiple rounds of training and validation cycles, the parameters of the predictor are gradually adjusted to make it more accurate in predicting coal thickness information. The training process continues until the error of the predictor in training and validation converges and achieves the best prediction effect, thereby completing the training of the coal mine thickness predictor.

[0034] According to the plurality of verified coal mine thickness information, the coal mine thickness distribution in the target mining area is generated respectively to obtain a plurality of coal mine thickness distributions.

[0035] In one embodiment, after obtaining multiple verified coal mine thickness information, the system terminal inputs the multiple verified coal mine thickness information into a pre-built coal thickness distribution generator to calculate the coal thickness distribution of each layer of the coal mine, and generates multiple coal mine thickness distributions to more accurately reflect the changes in coal seam thickness in the target mining area, and provide reliable reference data for subsequent coal mine mining planning and decision-making.

[0036] Furthermore, the present application provides generating the coal thickness distribution of the coal mines in the target mining area respectively according to the plurality of verified coal mine thickness information, and obtaining a plurality of coal mine thickness distributions, including:

[0037] According to the coal mine coal thickness detection data records, a sample coal mine coal thickness information set is collected, and the coal mine coal thickness distribution of the mining area is collected, and the sample coal mine coal thickness distribution set is obtained by annotation; based on the generative adversarial network, a coal thickness distribution generator is constructed; the sample coal mine coal thickness information set and the sample coal mine coal thickness distribution set are used to supervise the coal thickness distribution generator until convergence; the multiple verified coal mine coal thickness information are respectively input into the coal thickness distribution generator, and the coal mine coal thickness distribution in the target mining area is generated respectively, so as to obtain multiple coal mine coal thickness distributions.

[0038] Preferably, the system terminal collects a set of sample coal mine coal thickness information from the coal mine coal thickness detection data record, and collects the actual coal mine coal thickness distribution data of the mining area to form a labeled sample coal mine coal thickness distribution set. These data provide the association between the real coal mine coal thickness information and the corresponding distribution, and provide a basis for the training of the coal thickness distribution generator. Subsequently, based on the generative adversarial network (GAN), a coal thickness distribution generator is constructed, including a generation path and a discrimination path. The task of the generation path is to generate coal thickness distribution information based on the input coal mine coal thickness information, while the task of the discrimination path is to distinguish the generated coal thickness distribution information from the real coal thickness distribution information. The generation path usually contains multiple layers of fully connected layers or convolutional layers to map the input coal thickness information to high-dimensional coal thickness distribution information. The discrimination path is composed of convolutional layers or fully connected layers, which are used to judge the authenticity of the input coal thickness distribution information. Afterwards, the sample coal mine coal thickness information set and the sample coal mine coal thickness distribution set are input into the generation path of the coal thickness distribution generator for forward propagation. The generation path transmits the input coal mine coal thickness information layer by layer to generate the coal thickness distribution information of the coal mine. At the same time, the sample coal mine coal thickness distribution information corresponding to the sample coal mine coal thickness distribution set is input into the discrimination path to distinguish the real distribution information from the distribution information generated by the generation path. After the forward propagation is completed, the difference between the coal thickness distribution information generated by the generation path and the real distribution information is calculated, and the loss value is calculated using the cross entropy loss function. The discrimination path calculates its classification error for the real distribution information and the generated distribution information, and outputs a probability value representing the true and false discrimination result. Then, the back propagation algorithm is used to calculate the gradient of the loss function for the weights of each layer in the generation path and the discrimination path layer by layer. Through these gradients, the Adam optimizer is used to update the weights and biases of the coal thickness distribution generator, optimize the generation path, and make the coal thickness distribution information generated by it closer to the real distribution. The discrimination path also gradually improves its ability to distinguish true and false information. The process is repeated through iterations. New coal thickness information is input in each round of training. The generation path gradually generates new coal thickness distribution information, and the discrimination path performs new true and false discrimination. The generation path and the discrimination path are mutually optimized in adversarial training. The generation path constantly deceives the discrimination path, and the discrimination path tries its best to identify the real and generated distribution information. Training continues until the generation path can generate highly realistic coal thickness distribution information, while the discrimination path's ability to distinguish is gradually weakened. After the training is completed, the coal thickness information of sample coal mines that are not used for training and the coal thickness distribution information of sample coal mines are used to verify the effect of the generation path to ensure that the generated coal thickness distribution information has sufficient accuracy. If the generation effect does not meet expectations, the hyperparameters such as learning rate and batch size are adjusted, and training continues until the generator performance meets expectations. Through this training method, the generation path can generate coal thickness distribution information that is highly similar to the real distribution, and the discrimination path gradually enhances its ability to distinguish between the real and generated distribution information, thereby ensuring that the coal thickness distribution information generated by the generation path can provide an effective reference for mining in the mining area.After constructing the coal thickness distribution generator, the system terminal inputs multiple verified coal mine coal thickness information into the trained coal thickness distribution generator to generate corresponding coal mine coal thickness distribution information, providing accurate reference for subsequent coal mining.

[0039] Based on the multiple cross-ratios, multiple coal thickness distribution verification ratios are set, and the coal thickness distributions of the multiple coal mines are cross-validated to obtain cross-validation accuracy information of the coal thickness distributions of the multiple coal mines, and the total cross-validation accuracy information is calculated as the sampling data quality monitoring result of the current underground coal exploration.

[0040] In one embodiment, the system terminal sets multiple coal thickness distribution verification ratios based on the obtained multiple cross-ratios, and cross-validates the coal thickness distributions of different multiple coal mines. Specifically, the system terminal selects coal thickness distribution data of different sizes for verification based on multiple coal thickness distribution verification ratios, compares the performance of each coal mine's coal thickness distribution under other coal mines' coal thickness distributions, obtains the cross-validation accuracy information of each coal mine's coal thickness distribution, and then calculates the mean of all cross-validation accuracy information to obtain the total cross-validation accuracy information. This result reflects the overall quality of the current underground coal prospecting sampling data and provides a basis for the reliability of coal thickness distribution generation.

[0041] Furthermore, the present application provides generating the coal thickness distribution of the coal mines in the target mining area respectively according to the plurality of verified coal mine thickness information, and obtaining a plurality of coal mine thickness distributions, including:

[0042] The multiple cross-ratios are set as multiple coal thickness distribution verification ratios; according to the first coal thickness distribution verification ratio among the multiple coal thickness distribution verification ratios, the coal thickness distribution of the first coal mine and the coal thickness distributions of multiple other coal mines are cross-validated to obtain first cross-validation accuracy information; according to the other multiple coal thickness distribution verification ratios, the coal thickness distributions of multiple other coal mines are cross-validated to obtain multiple cross-validation accuracy information; the mean of the multiple cross-validation accuracy information is calculated to obtain total cross-validation accuracy information.

[0043] Preferably, the system terminal sets multiple cross-ratios as multiple coal thickness distribution verification ratios. The larger the cross-ratio, the higher the accuracy of the coal thickness distribution of the coal mine, and accordingly, the coal thickness distribution verification ratio during verification will also be larger. During verification, a coal thickness distribution verification ratio is randomly extracted from multiple coal thickness distribution verification ratios as the first coal thickness distribution verification ratio, and the corresponding coal thickness distribution of the coal mine is used as the first coal thickness distribution of the coal mine. Subsequently, the corresponding verification area is divided from the first coal thickness distribution of the coal mine and the coal thickness distributions of multiple other coal mines according to the first coal thickness distribution verification ratio, and the first cross-validation accuracy information is calculated by comparing the similarity of the first coal thickness distribution and the coal thickness distribution of other coal mines in these areas. After that, according to other coal thickness distribution verification ratios, the coal thickness distributions of other coal mines are cross-validated in turn, and the accuracy of the coal thickness distribution generated by each monitoring point under the data of other monitoring points is evaluated. The result of each verification reflects the similarity between the coal thickness distribution of the current monitoring point and the data of other points, which serves as the basis for cross-validation accuracy. Then, the cross-validation results of all coal thickness distributions of coal mines are summarized, and the mean of these cross-validation accuracy information is calculated to obtain the total cross-validation accuracy information. This total cross-validation accuracy information is used to evaluate the overall quality of the current underground coal exploration sampling data. If the validation results show that the data is highly accurate, it can be used for future data analysis or operations, such as in-depth analysis of coal mine distribution characteristics or guidance for further mining planning.

[0044] Further, the present application provides a method for cross-validating the first coal mine coal thickness distribution and the other multiple coal mine coal thickness distributions according to the first coal thickness distribution verification ratio within the multiple coal thickness distribution verification ratios, including:

[0045] According to the first coal thickness distribution verification ratio, the coal thickness distribution of the first coal mine and the coal thickness distribution of the coal mines with the first coal thickness distribution verification ratio near the first underground monitoring point in the coal thickness distribution of other multiple coal mines are divided to obtain a first coal thickness distribution to be verified and multiple basic coal thickness distributions; the similarities of the first coal thickness distribution to be verified and the multiple coal thickness distributions of the multiple basic coal thickness distributions are analyzed respectively, and the mean is calculated to obtain the first cross-validation accuracy information.

[0046] Optionally, the system terminal selects the area near the first underground monitoring point from the first coal mine coal thickness distribution and the coal thickness distribution of multiple other coal mines according to the set first coal thickness distribution verification ratio, divides the coal thickness distribution in these areas, and obtains the first coal thickness distribution to be verified and multiple basic coal thickness distributions. These divided areas reflect the changes in coal thickness near the first underground monitoring point in the coal thickness distribution of different coal mines. By cross-validating the first coal thickness distribution to be verified and multiple basic coal thickness distributions, the accuracy of the coal thickness distribution monitoring production of the first underground monitoring point can be obtained. Subsequently, the first coal thickness distribution to be verified is compared with multiple basic coal thickness distributions, and the similarity of the coal thickness distribution between each area is calculated. The similarity is used to evaluate the consistency between the area to be verified and other basic areas, reflecting the accuracy of the coal thickness distribution of the current monitoring point. Taking cosine similarity as an example, for the first coal thickness distribution to be verified and multiple basic coal thickness distributions, the system terminal arranges these coal thickness values ​​into a one-dimensional vector according to the spatial order of the monitoring points, and constructs the first coal thickness distribution vector to be verified and multiple basic coal thickness distribution vectors. Afterwards, the cosine similarity formula is used to calculate the similarity between the first coal thickness distribution vector to be verified and each basic coal thickness distribution vector, and then all similarity results are averaged to obtain the first cross-validation accuracy information, which is used to indicate the overall accuracy of the coal thickness distribution in the area.

[0047] In summary, the embodiments of the present application have at least the following technical effects:

[0048] The embodiment of the present application collects seismic reflection data sets by seismic reflection method at multiple underground monitoring points, and analyzes the stability of the data sets. According to the analysis results, multiple cross-ratios are set, and the seismic reflection data sets are cross-validated for coal thickness, and coal thickness information of multiple coal mines is obtained respectively. Subsequently, the coal thickness information after verification is used to generate the coal thickness distribution generator constructed by the generative adversarial network to generate the coal thickness distribution of the coal mine area. After that, different coal thickness distribution verification ratios are set based on multiple cross-ratios, and the generated coal thickness distribution is cross-validated, and the performance of different coal thickness distributions under other distributions is compared, and the similarity is evaluated, and the accuracy information of the verification is calculated. By continuously verifying the coal thickness distribution of each coal mine with multiple other coal thickness distributions, multiple cross-validation accuracy information is calculated, and based on this, the total cross-validation accuracy information is calculated as the quality monitoring result of the current underground coal prospecting sampling data. This method ensures the reliability and accuracy of the data quality monitoring results by dynamically adjusting the verification ratio and generating the distribution, and provides an important reference basis for coal mining. These technical effects jointly solve the technical problem that it is difficult to accurately analyze the quality of sampling data due to the insufficient accuracy of underground geophysical sampling data. They realize the automatic adjustment of the cross-ratio through dynamic cross-validation method, enhance the stability and accuracy of the data, and improve the accuracy of coal thickness distribution generation in coal mines and the reliability of sampling data quality monitoring.

[0049] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0050] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0051] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for monitoring downhole geophysical sampling data quality combined with dynamic cross-validation, characterized in that: The method comprises: At multiple underground monitoring points in the target mining area, underground coal exploration is carried out using the seismic reflection method to obtain multiple seismic reflection data sets; Analyzing the reflection data stability of the multiple seismic reflection data sets, setting multiple crossover ratios for performing coal mine thickness cross-validation analysis on the multiple seismic reflection data sets, performing coal mine thickness cross-validation analysis on the multiple seismic reflection data sets, and respectively obtaining multiple verified coal mine thickness information; Generating the coal thickness distribution of the coal mines in the target mining area respectively according to the plurality of verified coal mine thickness information, and obtaining a plurality of coal mine thickness distributions; Based on the multiple cross-ratios, multiple coal thickness distribution verification ratios are set, and the coal thickness distributions of the multiple coal mines are cross-validated to obtain cross-validation accuracy information of the coal thickness distributions of the multiple coal mines, and the total cross-validation accuracy information is calculated as the sampling data quality monitoring result of the current underground coal exploration.

2. The method for monitoring downhole geophysical sampling data quality in combination with dynamic cross-validation according to claim 1, characterized in that: At multiple underground monitoring points in the target mining area, underground coal exploration was carried out using the seismic reflection method to obtain multiple seismic reflection data sets, including: At multiple underground monitoring points in the target mining area, the seismic reflection method is used to conduct underground coal exploration, and the reflected wave data at each moment is received, wherein each reflected wave data includes the reflection time and the reflected wave amplitude; The reflected wave data at multiple times of the multiple downhole monitoring points are integrated to obtain multiple seismic reflection data sets.

3. The method for monitoring downhole geophysical sampling data quality in combination with dynamic cross-validation according to claim 1, characterized in that: Analyzing the reflection data stability of the multiple seismic reflection data sets, and setting multiple crossover ratios for performing coal mine coal thickness cross-validation analysis on the multiple seismic reflection data sets, including: In the multiple seismic reflection data sets, respectively, fluctuation amplitude identification of the seismic reflection data is performed to obtain multiple fluctuation coefficients as reflection data instability coefficients of the multiple seismic reflection data sets; According to the multiple reflection data instability coefficients, multiple reflection data stability coefficients are calculated and obtained, and multiple crossover ratios are set as the multiple seismic reflection data sets for cross-validation analysis of coal mine coal thickness.

4. The method for monitoring downhole geophysical sampling data quality in combination with dynamic cross-validation according to claim 3, characterized in that: A coal thickness cross-validation analysis is performed on the multiple seismic reflection data sets, including: According to a first cross ratio among the multiple cross ratios, randomly selecting an identification data set of the first cross ratio in the first seismic reflection data set, and using other data in the first seismic reflection data set as a verification data set; Using a pre-trained coal mine coal thickness predictor, predicting the coal thickness of the identification data set and the verification data set and respectively calculating the mean of the predicted coal thickness to obtain first predicted coal thickness information and first verified coal thickness information; Calculating a cross-validation deviation between the first predicted coal thickness information and the first verified coal thickness information to obtain a verification accuracy; Continue to randomly select an identification data set and a verification data set in the first seismic reflection data set according to the first cross ratio to predict coal thickness, and calculate the verification accuracy until the number of convergence times is reached, output the predicted coal thickness information corresponding to the maximum verification accuracy, and obtain the first verification coal mine coal thickness information; According to other multiple cross-ratios, a cross-validation analysis of the coal mine thickness is performed on other multiple seismic reflection data sets to obtain multiple verified coal mine thickness information with the highest verification accuracy.

5. The method for monitoring downhole geophysical sampling data quality in combination with dynamic cross-validation according to claim 4, characterized in that: The pre-training step of the coal mine coal thickness predictor comprises: Collecting a sample seismic reflection data set, and obtaining a sample coal mine thickness information set based on the actual coal mine thickness annotation; The sample seismic reflection data set and the sample coal mine thickness information set are used to train the coal mine thickness predictor based on K-fold cross validation until the training converges.

6. The method for monitoring downhole geophysical sampling data quality in combination with dynamic cross-validation according to claim 1, characterized in that: According to the plurality of verified coal mine thickness information, the coal mine thickness distribution in the target mining area is generated respectively to obtain a plurality of coal mine thickness distributions, including: According to the coal mine thickness detection data records, collect the coal mine thickness information set of the sample coal mine, collect the coal mine thickness distribution of the mining area, and mark and obtain the coal mine thickness distribution set of the sample coal mine; Based on the generative adversarial network, a coal thickness distribution generator is constructed; Using the sample coal mine coal thickness information set and the sample coal mine coal thickness distribution set, supervised training is performed on the coal thickness distribution generator until convergence; The plurality of verified coal mine thickness information are respectively input into the coal thickness distribution generator, and the coal thickness distribution of the coal mines in the target mining area is generated respectively to obtain a plurality of coal mine thickness distributions.

7. The method for monitoring downhole geophysical sampling data quality in combination with dynamic cross-validation according to claim 1, characterized in that: The method further comprises: setting a plurality of coal thickness distribution verification ratios based on the plurality of cross ratios, cross-validating the plurality of coal thickness distributions of the coal mines, obtaining cross-validation accuracy information of the plurality of coal thickness distributions of the coal mines, and calculating and obtaining total cross-validation accuracy information, including: Setting the multiple cross ratios as multiple coal thickness distribution verification ratios; According to a first coal thickness distribution verification ratio among the multiple coal thickness distribution verification ratios, cross-validate the first coal mine coal thickness distribution and the other multiple coal mine coal thickness distributions to obtain first cross-validation accuracy information; Continue to cross-validate the coal thickness distributions of other multiple coal mines according to other multiple coal thickness distribution validation ratios to obtain multiple cross-validation accuracy information; The mean of the multiple cross-validation accuracy information is calculated to obtain the total cross-validation accuracy information.

8. The method for monitoring downhole geophysical sampling data quality in combination with dynamic cross-validation according to claim 7, characterized in that: According to the first coal thickness distribution verification ratio among the multiple coal thickness distribution verification ratios, cross-verification is performed on the first coal mine coal thickness distribution and the other multiple coal mine coal thickness distributions, including: According to the first coal thickness distribution verification ratio, the first coal thickness distribution and the coal thickness distributions of the first coal thickness distribution verification ratio near the first underground monitoring point in the other multiple coal thickness distributions are divided to obtain a first coal thickness distribution to be verified and multiple basic coal thickness distributions; The similarities between the first coal thickness distribution to be verified and the multiple coal thickness distributions of the multiple basic coal thickness distributions are analyzed respectively, and the mean is calculated to obtain the first cross-validation accuracy information.

Images