Mine water inrush prediction method, device, equipment, medium and product

Abstract

The invention discloses a mine water inrush prediction method, device, equipment, medium and product, and relates to the field of mine water disaster prediction.The method comprises the steps that to-be-predicted data of a mine is received; inputting to-be-predicted data into the mine water disaster prediction model, and outputting a prediction result of whether water inrush occurs in the mine; the mine water disaster prediction model is trained by the following steps: acquiring a plurality of groups of training samples and the influence weight of each influence factor in each group of training samples on water inrush; obtaining the score of each group of training samples according to the influence weight of each influence factor of each group of training samples; obtaining a water inrush average value and a water inrush variance of the water inrush sample; obtaining a training average value and a training variance of all training samples; when the water inrush average value, the training average value, the water inrush variance and the training variance meet preset conditions, all the training samples are determined as target training samples; and training the deep learning model according to the target training sample. The accuracy of the water inrush prediction result can be improved.

Description

Technical Field

[0001] The present application relates to the field of mine water hazard prediction, and in particular to a mine water inrush prediction method, device, equipment, medium and product. Background Art

[0002] The hydrogeological conditions of many coalfields in my country are very complex. Coal mines are threatened by various water disasters during production, which seriously affects mine safety production.

[0003] Since deep learning technology has demonstrated powerful capabilities in prediction, classification, clustering, etc., it is also used in the field of mine water hazard prediction. However, deep learning requires a large number of training samples to train the model. In the field of mine water hazard prevention and control, due to limited historical water inrush accident data, the number of training samples for deep learning models is naturally limited during training, resulting in the prediction results of the deep learning model obtained after training being inaccurate. Summary of the invention

[0004] The purpose of this application is to provide a mine water inrush prediction method, device, equipment, medium and product, which can improve the accuracy of the prediction results.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for predicting water inrush in a mine, comprising:

[0007] receiving data to be predicted from the mine;

[0008] Inputting the data to be predicted into a mine water hazard prediction model, and outputting a prediction result of whether water inrush occurs in the mine;

[0009] The mine water hazard prediction model is trained in the following way:

[0010] Acquire multiple groups of training samples, wherein each group of the training samples includes multiple influencing factors, the influencing factors include geological data and geophysical data, and the multiple groups of the training samples include water inrush samples;

[0011] Obtaining the influence weight of each influencing factor on water inrush in each group of training samples;

[0012] Obtaining the score of each group of training samples according to the influence weight of each influencing factor of each group of training samples and the sample value corresponding to each influencing factor;

[0013] Obtaining the average water inrush value and the variance of the water inrush corresponding to the water inrush sample according to the score corresponding to the water inrush sample;

[0014] Obtaining the training mean and training variance corresponding to all the training samples according to the scores corresponding to all the training samples;

[0015] When the water inrush average value and the training average value meet a first preset condition, and the water inrush variance and the training variance meet a second preset condition, determining all the training samples as target training samples;

[0016] The deep learning model is trained according to the target training samples to obtain the mine water hazard prediction model.

[0017] In a second aspect, the present application provides a mine water inrush prediction device, the mine water inrush prediction device comprising:

[0018] A receiving module, used for receiving the data to be predicted of the mine;

[0019] A processing module, used for inputting the data to be predicted into the mine water hazard prediction model, and outputting a prediction result of whether water inrush occurs in the mine;

[0020] The processing module is also used to train the mine water hazard prediction model in the following manner:

[0021] Acquire multiple groups of training samples, wherein each group of the training samples includes multiple influencing factors, the influencing factors include geological data and geophysical data, and the multiple groups of the training samples include water inrush samples;

[0022] Obtaining the influence weight of each influencing factor on water inrush in each group of training samples;

[0023] Obtaining the score of each group of training samples according to the influence weight of each influencing factor of each group of training samples and the sample value corresponding to each influencing factor;

[0024] Obtaining the average water inrush value and the variance of the water inrush corresponding to the water inrush sample according to the score corresponding to the water inrush sample;

[0025] Obtaining the training mean and training variance corresponding to all the training samples according to the scores corresponding to all the training samples;

[0026] When the water inrush average value and the training average value meet a first preset condition, and the water inrush variance and the training variance meet a second preset condition, determining all the training samples as target training samples;

[0027] The deep learning model is trained according to the target training samples to obtain the mine water hazard prediction model.

[0028] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the above-described methods for predicting water inrush in the mine.

[0029] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-mentioned methods for predicting water inrush in the mine.

[0030] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of any of the above-mentioned methods for predicting water inrush in the mine.

[0031] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0032] The present application provides a method, device, equipment, medium and product for predicting water inrush in a mine, comprising: receiving data to be predicted of a mine; inputting the data to be predicted into a mine water hazard prediction model, and outputting a prediction result of whether water inrush occurs in the mine; the mine water hazard prediction model is trained in the following manner: obtaining multiple groups of training samples, wherein each group of training samples includes multiple influencing factors, the influencing factors include geological data and geophysical data, and the multiple groups of training samples include water inrush samples; obtaining the influence weight of each influencing factor on water inrush in each group of training samples; obtaining the score of each group of training samples according to the influence weight of each influencing factor of each group of training samples and the sample value corresponding to each influencing factor; obtaining the average water inrush value and the variance of water inrush corresponding to the water inrush sample according to the score corresponding to the water inrush sample; obtaining the training average value and the training variance corresponding to all training samples according to the score corresponding to all training samples; when the average water inrush value and the training average value meet the first preset condition, and the variance of water inrush and the training variance meet the second preset condition, determining all training samples as target training samples; training a deep learning model according to the target training samples to obtain a mine water hazard prediction model. Among them, by calculating the mean and variance of the water inrush sample and all training samples, and then comparing the similarity between the water inrush sample and all training samples through the mean and variance, if the mean and variance of the water inrush sample are close to the mean and variance of all training samples respectively, it is considered that all training samples are similar to the water inrush sample, and all training samples are marked as water inrush samples. In this way, a large number of labeled water inrush samples can be obtained for subsequent deep model training, thereby improving the accuracy of model training, thereby improving the accuracy of water inrush prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 is a flow chart of a method for predicting water inrush in a mine according to an exemplary embodiment;

[0035] Figure 2 is a schematic diagram of a hierarchical structure according to an exemplary embodiment;

[0036] Figure 3 is a functional module schematic diagram of a mine water inrush prediction device according to an exemplary embodiment;

[0037] Figure 4 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0038] 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, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0039] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0040] The hydrogeological conditions of many coalfields in my country are very complex. Coal mines are threatened by various water bodies during production, which seriously affects the safe production of mines. According to statistics, from 2000 to 2022, there were 1,206 water disasters in my country's coal mines, killing 5,018 people, second only to gas accidents in threatening coal mine safety and miners' lives. The causes of mine water disasters are complex and are related to multiple factors such as structure, aquiclude conditions, and mining methods. It is a hydrogeological and mining composite dynamic phenomenon that is controlled by multiple factors and has very complex nonlinear dynamic characteristics.

[0041] In terms of mine water disaster prediction, the traditional method is mainly the water inrush coefficient method, where the water inrush coefficient = water pressure / thickness of aquiclude. Compared with the critical water inrush coefficient (the critical water inrush coefficient comes from previous data statistics, generally 0.1MP / m in normal areas and 0.06MP / m in structural areas), the concept of the water inrush coefficient is clear, and the formula is simple and practical. Although only two simple factors, water pressure (P) and thickness of aquiclude (M), appear in the expression, it reflects the comprehensive effect of water inrush factors, including the role of mine pressure in destroying the floor or promoting the reactivation of fractures. It has played a positive role in coal mine production, so it has been used to this day. However, there are still several problems in the actual application of water inrush coefficient: 1. The critical water inrush coefficient value is mainly obtained based on the statistical analysis of the previous water inrush data of the mining area, but the water inrush data are mostly distributed near the fault. The critical water inrush coefficient obtained by using these data is conservative when used to judge whether a normal block has water inrush; 2. The critical water inrush coefficient increases with the increase of mining depth, and its increment also increases with the increase of mining depth, that is, the critical water inrush coefficient is linearly related to the depth in the shallow part, but in the deep part, this linear relationship no longer exists. And the shallow water inrush data is used for deep prediction, and the result is conservative; 3. When the mining area lacks water inrush data, the critical water inrush coefficient is often difficult to determine; 4. The water inrush coefficient does not take into account the water-richness of the bottom aquifer and its hydrodynamic characteristics, the water-blocking performance of the aquiclude, the lithology and other factors, and the obtained results lack persuasiveness. In addition, there are the "lower three zones" theory and the key layer theory. These methods all attempt to study the mechanism of water inrush. These theories involve complex knowledge such as rock mechanics, and the heterogeneity of deep strata is more prominent, making it difficult to obtain the required data. Regardless of the practicality of the method or the difficulty for on-site technicians to master it, it is difficult to promote it.

[0042] In recent years, deep learning technology has flourished, showing powerful prediction, classification, clustering and other capabilities, and has achieved very good application results in many industries. In the field of mine water hazard prevention and control, due to the limited number of water hazard samples, there have been about 1,206 water hazard accidents in my country in the past 20 years. Considering my country's vast territory, the geological conditions of several major coal-producing areas vary greatly, and the types of water hazard accidents are diverse (roof, floor, fault, separation, etc.). There are even fewer cases of different water hazard accidents under different geological conditions. The number of samples that can be used to train the model is seriously insufficient, resulting in unsatisfactory application results of most mainstream deep learning models in the field of mine water hazard prevention and control.

[0043] In order to solve the above technical problems, the present invention provides a method and device for predicting water inrush in a mine.

[0044] Figure 1 is a flow chart of a method for predicting water inrush in a mine according to an exemplary embodiment. Figure 1 As shown, the mine water inrush prediction method includes the following steps S101-S109:

[0045] In step S101, data to be predicted of a mine is received.

[0046] In step S102, the data to be predicted is input into a mine water hazard prediction model, and a prediction result of whether water inrush occurs in the mine is output.

[0047] The mine water hazard prediction model is obtained by training through the following steps S103-S109:

[0048] In step S103, multiple groups of training samples are obtained, wherein each group of training samples includes multiple influencing factors, the influencing factors include geological data and geophysical data, and the multiple groups of training samples include water inrush samples.

[0049] Among them, the water inrush samples are geological data and geophysical data corresponding to historical water inrush accidents.

[0050] The geological data include at least one of the following: water pressure of aquifer, water yield, thickness and structure of aquiclude.

[0051] Geophysical data include at least one of the following: transient electromagnetic, audio, radio wave tunnel perspective, slot wave and three-dimensional seismic.

[0052] In step S104, the influence weight of each influencing factor on water inrush in each group of training samples is obtained.

[0053] In one achievable manner, the weight of each influencing factor on water inrush in each group of training samples can be obtained by using a hierarchical analysis method.

[0054] Specific: The analytic hierarchy process (AHP) is a hierarchical, systematic, multi-objective, multi-criteria decision analysis method that combines qualitative and quantitative methods. By applying the idea of ​​system analysis, complex decision problems with multiple objectives and multiple criteria are transformed into simple quantitative decision problems. It is widely used in resource allocation, program evaluation, evaluation prediction, and system analysis. The principles and steps of implementing the analytic hierarchy process are as follows:

[0055] According to the characteristics of each influencing factor, a hierarchical analysis structure model is constructed, which generally includes the target layer A, the criterion layer B and the decision-making layer C.

[0056] According to the categories of influencing factors and the different structural levels of their effects on water inrush, the structural model is analyzed. For example, the water pressure of the aquifer, the amount of water inflow, the thickness of the aquiclude, the structure, the cracks, the transient electromagnetic, and the slot wave are considered. Figure 2 The hierarchical structure shown, from top to bottom, consists of three levels: target layer A, criterion layer B, and decision layer C.

[0057] The importance of each influencing factor in the same level in the hierarchical analysis structure model relative to the corresponding influencing factor in the previous level is compared in pairs, and a judgment matrix is ​​constructed, and the 1-9 scale method is used for scoring, as shown in Table 1 and Table 2:

[0058] Table 1

[0059] Bi than Bj same Slightly stronger powerful Very strong Absolutely strong same Slightly weak weak Very weak Absolutely weak <![CDATA[a ij ]]> 1 3 5 7 9 1 1 / 3 1 / 5 1 / 7 1 / 9

[0060] Table 2

[0061] Ci is better than Cj same Slightly stronger powerful Very strong Absolutely strong same Slightly weak weak Very weak Absolutely weak <![CDATA[b ij ]]> 1 3 5 7 9 1 1 / 3 1 / 5 1 / 7 1 / 9

[0062] Among them, a in Table 1 ij Indicates the importance of Bi to A compared with Bj. Bi and Bj represent two influencing factors at the same level. Because a certain level often involves many influencing factors, it is too subjective to compare the importance of multiple influencing factors at the same time. Here, the influencing factors at the same level are compared two by two. We only need to consider which of the two influencing factors is more important and how much more important they are, that is, a scale of 1 to 9. ij Indicates the importance of Ci to B compared to Cj.

[0063] The same level refers to the same target layer, the same criterion layer and the same decision-making layer. Specifically, in the above example, B1~B4 are compared in importance to A, C2 and C3 are compared in importance to B2, C4 and C5 are compared in importance to B3, and C6 and C7 are compared in importance to B4.

[0064] In the following embodiments, taking the level A to B as an example, the final goal is to construct a judgment matrix such as the first 5 columns of Table 3 below. At the same time, in order to further eliminate personal subjective factors when comparing two by two, a consistency test is performed, so it is not a random comparison. In addition to A to B, B1 to B1, B2 to (C2, C3) all need to construct a judgment matrix and perform a consistency test. A = (a ij ) n×n The judgment matrix constructed by comparing the importance of B1 to B4 to A is shown in Table 3:

[0065] Table 3

[0066]

[0067] After the judgment matrix A is constructed (the first 5 columns of Table 3 above), as shown in Table 3, after the judgment matrix A is constructed, the maximum eigenvalue λ of the judgment matrix A is obtained. max , and use its corresponding characteristic equation to solve the corresponding eigenvector W, which is normalized to be the importance weight of each influencing factor at the same level to a certain influencing factor at the previous level.

[0068] At the same time, in order to eliminate subjective factors when constructing the judgment matrix, it is also necessary to calculate the consistency index M according to the following formula:

[0069]

[0070] Among them, λ max represents the maximum eigenvalue of the judgment matrix, and n represents the order of the judgment matrix.

[0071] When M=0, the judgment matrix is ​​consistent. The larger M is, the more serious the inconsistency of the judgment matrix is. Therefore, the random consistency index RI is introduced. The value of RI is shown in Table 4:

[0072] Table 4

[0073] n 3 4 5 6 7 8 9 10 11 RI 0.58 0.9 1.12 1.24 1.32 1.41 1.45 1.49 1.51

[0074] CR is calculated according to CR=M / RI. When CR is less than 0.1, the inconsistency of the judgment matrix is ​​still acceptable, otherwise the judgment matrix must be adjusted.

[0075] Taking all levels into consideration, the weight of each influencing factor in the decision layer (C layer) on the target layer (A layer) is obtained, and the total ranking consistency ratio (CR z ) test, when CR z When it is less than 0.1, the total sorting consistency is considered acceptable, where CR z Obtained according to the following formula:

[0076]

[0077] Among them, CR z Indicates the total order consistency ratio; CI i Indicates (B i ~C) Judgment matrix consistency index; RI i Indicates (B i ~C) Average random consistency index of the judgment matrix; W i Influencing factors B of the criterion layer i The weight of the target layer A. Finally, the total ranking weight value that passes the consistency check is obtained.

[0078] This mainly involves A~B i , B i ~C i There are two levels. The weight of each level relative to the previous level has been calculated before. If the weight of C2 to B2 is 0.3, and the weight of C3 to B2 is 0.7, and the weight of B2 to A in Table 3 is 0.0591, then the weight of C2 to A is 0.0591*0.3=0.1773, and the weight of C3 to A is 0.0591*0.7=0.04137. In this way, the weight of each C to A in the decision-making layer is calculated in turn.

[0079] In step S105, the score of each group of training samples is obtained according to the influence weight of each influencing factor of each group of training samples and the sample value corresponding to each influencing factor.

[0080] Multiply each influencing factor index by the weight and sum them up to get the score of each group of training samples.

[0081] That is, the weight of each decision layer C to A has been calculated before. Multiply each influencing factor in each decision layer C by the corresponding weight and sum them up to get the "score" of each group of samples.

[0082] In step S106, the average water inrush value and the variance of the water inrush value corresponding to the water inrush sample are obtained according to the score corresponding to the water inrush sample.

[0083] In step S107, the training mean values ​​and training variances corresponding to all the training samples are obtained according to the scores corresponding to all the training samples.

[0084] If there are 1000 samples (including 100 water inrush samples), add up all the sample scores and divide by 1000 to get the average sample score; add up the squares of the difference between each sample score and the average and divide by the number of samples 1000 to get the variance.

[0085] The mean and variance were also calculated for the 100 water inrush samples.

[0086] In step S108, when the water inrush average value and the training average value meet the first preset condition, and the water inrush variance and the training variance meet the second preset condition, all training samples are determined to be target training samples.

[0087] The average water inrush value and the training average value meet the first preset condition, including the following sub-steps A1-A3:

[0088] A1. Obtain the first difference between the average value of the water inrush and the average value of the training.

[0089] A2. Obtain a first proportion of the first difference value in the average value of the water inrush.

[0090] A3. When the first proportion is greater than a first threshold, it is determined that the average value of the water inrush current and the average value of the training satisfy a first preset condition.

[0091] Wherein, the water inrush variance and the training variance meet the second preset condition, including the following sub-steps B1-B3:

[0092] B1. Obtain a second difference between the water inrush variance and the training variance.

[0093] B2. Obtain a second proportion of the second difference in the water inrush variance.

[0094] B3. When the second proportion is greater than the second threshold, determining that the water inrush variance and the training variance meet a second preset condition.

[0095] In one possible implementation, the mine water hazard prediction method includes:

[0096] When the average value of the water inrush and the average value of the training do not satisfy the first preset condition, and / or the variance of the water inrush and the variance of the training do not satisfy the second preset condition, deleting the training sample with the lowest score among the training samples;

[0097] The steps of obtaining the training mean and the training variance, and determining whether the training sample is a target training sample are repeated for the remaining training samples until the target training sample is obtained.

[0098] Specifically, the difference between the mean value of known water inrush samples and the mean value of the total score of all training samples is compared, and the difference between the variance of known water inrush samples and the variance of the total score of all training samples is compared. If both differences are less than a certain value (such as 5%), all training samples are marked as water inrush, and the sample labeling is completed. At this time, all training samples can be used as target samples to train the model; if both differences are greater than a certain value (such as 5%), the last group of training samples (i.e., the group of training samples with the lowest total score) is deleted and used as new training samples, and the steps of obtaining the training mean value and training variance are repeated until the total score of the training samples and the known water inrush samples meet the above conditions, and the water inrush sample labeling is completed.

[0099] Among them, after obtaining the scores of all training samples, they can also be sorted from high to low according to the scores. All training samples are arranged from high to low according to the scores for the purpose of subsequent sample marking. In theory, the score of the water inrush sample should be the highest and ranked in front. The closer the training sample score is to the water inrush sample, the more similar it is to the water inrush sample, and the greater the probability of water inrush.

[0100] Here, the first threshold and the second threshold are both taken as 5% as an example, (average value of all samples - average value of water inrush samples) / average value of water inrush samples * 100%, and the variance is calculated in the same way. If the calculated value is greater than 5%, it is considered that the total training samples are not similar to the water inrush samples (that is, there are too many non-water inrush samples mixed in the total samples). In this way, the last group of total samples is deleted (the samples with the lowest scores have the worst similarity with the water inrush samples).

[0101] Then, the mean and variance of the water inrush sample and all samples are recalculated, and the similarity between the water inrush sample and all training samples is compared until it reaches below the required threshold. All training samples are considered similar to the water inrush sample and are marked as water inrush samples. In this way, a large number of labeled water inrush samples can be obtained for subsequent deep model training.

[0102] In step S109, the deep learning model is trained according to the target training samples to obtain a mine water hazard prediction model.

[0103] Specifically, step S109 includes the following sub-steps C1-C4:

[0104] C1. Divide the target training samples into training set, test set and validation set.

[0105] Specifically, in the process of processing and preparing the target training samples, it is first necessary to reasonably distribute these data into different subsets to facilitate subsequent model training and evaluation. Generally, the target training samples need to be divided into three main parts: training set, test set and validation set. Training Set: used to train the model, accounting for about 60%-80%. Test Set: used to adjust model parameters and prevent overfitting, accounting for about 10%-20%. Validation Set: used to evaluate the final performance of the model, accounting for about 10%-20%.

[0106] As you can understand, the training set is used to train the model. Through the training set, the model can learn the mapping relationship from input to output, so as to make predictions for new input data. The training set contains a large amount of labeled sample data, which will be used to guide the model to learn and identify different features and patterns. Through repeated iterations and optimization on the training set, the model can gradually grasp the regularity in the data and establish preliminary prediction capabilities. The test set is used to evaluate the generalization ability of the model, that is, the model's prediction ability for new data. After the model training is completed, the model is tested on the test set to verify the generalization ability of the model on unseen data. The data in the test set is the data that the model has never encountered during the training process, so the model can be adjusted through the test set. The role of the validation set is to monitor performance and adjust parameters during the model training process. The data in the validation set also comes from the original annotated corpus, but they are neither involved in the training of the model nor used for the final performance test, but serve as a reference standard for model tuning. The evaluation of the validation set can more objectively reflect the true performance of the model.

[0107] C2. Input the training set into the deep learning model for training to obtain the trained deep learning model.

[0108] C3. Adjust the trained deep learning model based on the test set to obtain an adjusted deep learning model.

[0109] Specifically, after the deep learning model is trained, in order to make the model output more accurate, the model needs to be further adjusted and optimized. By fine-tuning the model parameters, it is ensured that the model can more accurately predict whether water inrush will occur in the mine based on the input mine data.

[0110] Specifically, C3 includes the following sub-steps C31-C34:

[0111] C31. Input the test set into the trained deep learning model and output the test prediction results.

[0112] In the present disclosure, the preset test results include whether water inrush will occur in the current test set.

[0113] C32. Calculate the loss function based on the test prediction results and the expected output results corresponding to the test set.

[0114] Among them, the expected output result corresponding to the test set is the preset water inrush result actually corresponding to the test set.

[0115] C33. Update the model parameters based on the loss function.

[0116] C34. Iteratively train the deep learning model based on the updated model parameters to obtain an adjusted deep learning model.

[0117] C4. Evaluate the adjusted deep learning model based on the validation set; when the evaluation result meets the preset requirements, the corresponding adjusted deep learning model is the mine water hazard prediction model.

[0118] Specifically, the deficiencies of the model identified during the evaluation process may involve careful review and necessary adjustments to the training data set, or manual optimization and adjustment of the model's parameter settings. In addition, in order to ensure that the performance of the model in actual applications can continue to improve, the validation set needs to be updated regularly to capture the latest data trends and changes. By continuously repeating the above evaluation process, it can be ensured that the performance of the model in the prediction task is continuously improved and optimized. Through continuous optimization, it is ensured that the model is always in the best working condition, providing technical support for the prediction of mine water hazards.

[0119] Specifically, C4 includes the following sub-steps C41-C45:

[0120] C41. Input the validation set into the adjusted deep learning model to obtain the model judgment result.

[0121] Among them, the model judgment result includes whether water inrush will occur in the current verification set.

[0122] C42. Compare the model judgment result with the annotation result corresponding to the validation set. If the model judgment result is completely consistent with the annotation result corresponding to the validation set, the model judgment is recorded as correct.

[0123] C43. Count the total number of correct model judgments and the total number of model judgments, and calculate the accuracy of mine water inrush judgment based on the total number of correct model judgments and the total number of model judgments.

[0124] C44. Evaluate the adjusted deep learning model based on the accuracy of mine water inrush judgment.

[0125] C45. When it is determined that the accuracy of mine water inrush judgment meets the preset accuracy, the corresponding adjusted deep learning model is the mine water hazard prediction model.

[0126] The present invention can generate a large number of water inrush samples based on a small number of existing water inrush samples for deep learning model training, which is more in line with the characteristics of mine water hazards; and it takes into account the influencing factors of water inrush comprehensively, and can flexibly increase or decrease the influencing factors. For predictions of different geological conditions and different types of water hazards, it only needs to adjust the hierarchical analysis structure model, which is suitable for water hazard prevention and control work in most parts of the country; furthermore, it can be combined with various exploration and monitoring technologies in the current field of mine water hazard prevention and control, and the real-time data obtained by various sensors can be input into the trained model for calculation, so as to realize real-time automatic prediction of mine water hazards.

[0127] Based on the same inventive concept, the embodiment of the present application also provides a mine water inrush prediction device for implementing the mine water inrush prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more mine water inrush prediction device embodiments provided below can refer to the limitations of the mine water inrush prediction method above, and will not be repeated here.

[0128] In an exemplary embodiment, Figure 3 As shown, a mine water inrush prediction device is provided, comprising:

[0129] A receiving module 11 is used to receive the mine data to be predicted;

[0130] The processing module 12 is used to input the data to be predicted into the mine water hazard prediction model and output a prediction result of whether water inrush occurs in the mine;

[0131] The processing module 12 is also used to train the mine water hazard prediction model in the following manner:

[0132] Acquire multiple groups of training samples, wherein each group of the training samples includes multiple influencing factors, the influencing factors include geological data and geophysical data, and the multiple groups of the training samples include water inrush samples;

[0133] Obtaining the influence weight of each influencing factor on water inrush in each group of training samples;

[0134] Obtaining the score of each group of training samples according to the influence weight of each influencing factor of each group of training samples and the sample value corresponding to each influencing factor;

[0135] Obtaining the average water inrush value and the variance of the water inrush corresponding to the water inrush sample according to the score corresponding to the water inrush sample;

[0136] Obtaining the training mean and training variance corresponding to all the training samples according to the scores corresponding to all the training samples;

[0137] When the water inrush average value and the training average value meet a first preset condition, and the water inrush variance and the training variance meet a second preset condition, determining all the training samples as target training samples;

[0138] The deep learning model is trained according to the target training samples to obtain a mine water hazard prediction model.

[0139] As an optional implementation manner, the processing module is further used to:

[0140] When the water inrush average value and the training average value do not satisfy the first preset condition, and / or the water inrush variance and the training variance do not satisfy the second preset condition, deleting the training sample with the lowest score among the training samples;

[0141] The steps of obtaining the training mean and the training variance, and determining whether the remaining training samples are the target training samples are executed again until the target training samples are obtained.

[0142] As an optional implementation manner, in determining whether the average water inrush value and the training average value satisfy a first preset condition, the processing module is specifically configured to:

[0143] Obtaining a first difference between the average water inrush value and the training average value;

[0144] Obtaining a first proportion of the first difference value in the average value of the water inrush;

[0145] When the first proportion is greater than a first threshold, it is determined that the average water inrush value and the training average value satisfy the first preset condition.

[0146] As an optional implementation manner, in determining whether the water inrush variance and the training variance satisfy a second preset condition, the processing module is specifically configured to:

[0147] Obtaining a second difference between the water inrush variance and the training variance;

[0148] Obtaining a second proportion of the second difference in the water inrush variance;

[0149] When the second proportion is greater than a second threshold, it is determined that the water inrush variance and the training variance satisfy the second preset condition.

[0150] As an optional implementation, in terms of obtaining the weight of the influence of each influencing factor on the water inrush in each group of training samples, the processing module is specifically used to:

[0151] The weight of each influencing factor on water inrush in each group of training samples is obtained through the hierarchical analysis method.

[0152] As an optional implementation,

[0153] The geological data include at least one of the following: water pressure of aquifer, water yield, thickness and structure of aquiclude;

[0154] The geophysical exploration data includes at least one of the following data: transient electromagnetic, audio, radio wave tunnel perspective, slot wave and three-dimensional seismic.

[0155] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data required for mine water inrush prediction. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting mine water inrush is implemented.

[0156] Those skilled in the art will understand that Figure 4The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0157] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0158] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0159] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0160] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0161] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0162] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0163] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0164] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for predicting water inrush in a mine, characterized in that: The mine water inrush prediction method comprises: receiving data to be predicted from the mine; Inputting the data to be predicted into a mine water hazard prediction model, and outputting a prediction result of whether water inrush occurs in the mine; The mine water hazard prediction model is trained in the following way: Acquire multiple groups of training samples, wherein each group of the training samples includes multiple influencing factors, the influencing factors include geological data and geophysical data, and the multiple groups of the training samples include water inrush samples; Obtaining the influence weight of each influencing factor on water inrush in each group of training samples; Obtaining the score of each group of training samples according to the influence weight of each influencing factor of each group of training samples and the sample value corresponding to each influencing factor; Obtaining the average water inrush value and the variance of the water inrush corresponding to the water inrush sample according to the score corresponding to the water inrush sample; Obtaining the training mean and training variance corresponding to all the training samples according to the scores corresponding to all the training samples; When the water inrush average value and the training average value meet a first preset condition, and the water inrush variance and the training variance meet a second preset condition, determining all the training samples as target training samples; The deep learning model is trained according to the target training samples to obtain the mine water hazard prediction model.

2. The mine water inrush prediction method according to claim 1, characterized in that: The mine water hazard prediction method comprises: When the water inrush average value and the training average value do not satisfy the first preset condition, and / or the water inrush variance and the training variance do not satisfy the second preset condition, deleting the training sample with the lowest score among the training samples; The steps of obtaining the training mean and the training variance, and determining whether the remaining training samples are the target training samples are executed again until the target training samples are obtained.

3. The mine water inrush prediction method according to claim 1, characterized in that: The average water inrush value and the training average value satisfy a first preset condition, including: Obtaining a first difference between the average water inrush value and the training average value; Obtaining a first proportion of the first difference value in the average value of the water inrush; When the first proportion is greater than a first threshold, it is determined that the average water inrush value and the training average value satisfy the first preset condition.

4. The mine water inrush prediction method according to claim 1, characterized in that: The water inrush variance and the training variance satisfy a second preset condition, including: Obtaining a second difference between the water inrush variance and the training variance; Obtaining a second proportion of the second difference in the water inrush variance; When the second proportion is greater than a second threshold, it is determined that the water inrush variance and the training variance satisfy the second preset condition.

5. The mine water inrush prediction method according to claim 1, characterized in that: The obtaining of the weight of each influencing factor on the water inrush in each group of the training samples comprises: The weight of each influencing factor on water inrush in each group of training samples is obtained by using the hierarchical analysis method.

6. The mine water inrush prediction method according to claim 1, characterized in that: The geological data include at least one of the following: water pressure of aquifer, water yield, thickness and structure of aquiclude; The geophysical exploration data includes at least one of the following data: transient electromagnetic, audio, radio wave tunnel perspective, slot wave and three-dimensional seismic.

7. A mine water inrush prediction device, characterized in that: The mine water inrush prediction device comprises: A receiving module, used for receiving the data to be predicted of the mine; A processing module, used for inputting the data to be predicted into the mine water hazard prediction model, and outputting a prediction result of whether water inrush occurs in the mine; The processing module is also used to train the mine water hazard prediction model in the following manner: Acquire multiple groups of training samples, wherein each group of the training samples includes multiple influencing factors, the influencing factors include geological data and geophysical data, and the multiple groups of the training samples include water inrush samples; Obtaining the influence weight of each influencing factor on water inrush in each group of training samples; Obtaining the score of each group of training samples according to the influence weight of each influencing factor of each group of training samples and the sample value corresponding to each influencing factor; Obtaining the average water inrush value and the variance of the water inrush corresponding to the water inrush sample according to the score corresponding to the water inrush sample; Obtaining the training mean and training variance corresponding to all the training samples according to the scores corresponding to all the training samples; When the water inrush average value and the training average value meet a first preset condition, and the water inrush variance and the training variance meet a second preset condition, determining all the training samples as target training samples; The deep learning model is trained according to the target training samples to obtain the mine water hazard prediction model.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the mine water inrush prediction method described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the mine water inrush prediction method described in any one of claims 1-6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the mine water inrush prediction method described in any one of claims 1-6 are implemented.

Images