Method and device for determining ground subsidence parameters under a probability integral model

Abstract

The present application relates to the technical field of data processing, and provide a kind of probability integral model under surface subsidence parameter determination method and device, comprising: obtaining the engineering data of the area to be predicted;Respectively from the engineering data, the feature data required for each type of surface subsidence parameter under probability integral model is extracted;Wherein, the feature data required for each type of surface subsidence parameter is obtained by correlation analysis and recursive feature elimination screening;The feature data required for each type of surface subsidence parameter is respectively input into the corresponding prediction model, and the prediction value corresponding to each type of surface subsidence parameter is obtained.The present application can realize the accurate, efficient prediction of various surface subsidence parameters.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for determining surface subsidence parameters under a probability integral model. Background Technology

[0002] Surface subsidence prediction is a crucial technical issue in coal mining, directly impacting the safety of surface structures and environmental protection. The probabilistic integral method, a widely adopted approach for surface subsidence prediction, treats the impact of coal seam mining on overlying strata as the superposition of countless tiny units, deriving the laws governing surface subsidence, horizontal movement, and deformation through integral calculations. The probabilistic integral method is a parameter-driven mathematical model. While its prediction formula is fixed, the method requires determining seven parameters: a subsidence coefficient, a horizontal movement coefficient, the tangent of the main influence angle, and four inflection point offset parameters. These seven parameters are variables, and their values ​​directly affect the reliability of the prediction results.

[0003] Currently, parameter determination using probabilistic integral methods often relies excessively on human experience and subjective judgment. Different engineers may arrive at different parameter values ​​for the same geological conditions, lacking objectivity and consistency, which seriously affects prediction reliability. Secondly, there is a lack of systematic data support, failing to fully utilize historical engineering experience. Traditional methods rely on small-scale, scattered data and lack large-scale standardized databases, leading to wasted engineering experience and limited accuracy. Finally, the low level of intelligence prevents the use of modern machine learning, artificial intelligence, and other advanced technologies to improve the accuracy and efficiency of parameter selection, hindering the overall development of the industry's technological level. Summary of the Invention

[0004] This invention provides a method and apparatus for determining surface subsidence parameters under a probability integral model, in order to solve the technical problems of excessive reliance on subjective experience, low efficiency, lack of systematic data support, and low level of intelligence in the methods for determining surface subsidence parameters under a probability integral model.

[0005] This invention provides a method for determining surface subsidence parameters under a probability integral model, comprising the following steps: Acquire engineering data for the area to be predicted; the engineering parameters include geological data and mining data. The feature data required for each type of surface subsidence parameter under the probability integral model are extracted from the engineering data; wherein, the feature data required for each type of surface subsidence parameter is obtained through correlation analysis and recursive feature elimination screening; The characteristic data required for each type of surface subsidence parameter are input into the corresponding prediction model to obtain the predicted value for each type of surface subsidence parameter. The prediction model corresponding to each type of surface subsidence parameter is obtained by supervised training of the initial model matching each type of surface subsidence parameter using the feature data corresponding to each type of surface subsidence parameter and the corresponding measured values ​​of historical surface subsidence parameters from historical engineering data under multiple historical projects.

[0006] According to the method for determining surface subsidence parameters under a probability integral model provided by the present invention, after inputting the feature data required for each type of surface subsidence parameter into the corresponding prediction model to obtain the predicted value corresponding to each type of surface subsidence parameter, the method further includes: Determine the confidence level of the predicted values ​​for each type of surface subsidence parameter; For each type of surface subsidence parameter, if the confidence level is greater than the first threshold, the predicted value is determined to be the target value corresponding to the surface subsidence parameter. If the confidence level is greater than the second threshold but not greater than the first threshold, the target value corresponding to the surface subsidence parameter is determined based on the predicted value and the historical experience value corresponding to the surface subsidence parameter. If the confidence level is not greater than the second threshold, the target value corresponding to the surface subsidence parameter is determined based on the historical experience value corresponding to the surface subsidence parameter.

[0007] According to the present invention, a method for determining surface subsidence parameters under a probability integral model is provided. The prediction model includes a first prediction model corresponding to the subsidence coefficient under the probability integral model, and the first prediction model is constructed in the following manner: Data related to the subsidence coefficient are selected from the feature data as the first training data, and the corresponding measured values ​​of the subsidence coefficient in the measured values ​​of the historical surface subsidence parameters are used as the first training labels. The random forest model is trained based on the first training data and the first training label to obtain the trained first prediction model.

[0008] According to the present invention, a method for determining surface subsidence parameters under a probability integral model is provided, wherein the prediction model includes a second prediction model corresponding to the horizontal movement coefficient under the probability integral model, and the second prediction model is constructed in the following manner: Data related to the horizontal movement coefficient are selected from the feature data as the second training data, and the corresponding measured values ​​of the horizontal movement coefficient in the measured values ​​of the historical surface subsidence parameters are used as the second training labels. Based on the second training data and the second training labels, the support vector regression model is trained to obtain the trained second prediction model.

[0009] According to the method for determining surface subsidence parameters under a probability integral model provided by the present invention, the prediction model includes a third prediction model corresponding to the tangent value of the main influence angle under the probability integral model, and the third prediction model is constructed in the following manner: Data related to the tangent of the main influence angle are selected from the feature data as the third training data, and the corresponding measured values ​​of the tangent of the main influence angle in the measured values ​​of the historical surface subsidence parameters are used as the third training labels. The gradient boosting tree model is trained based on the third training data and the third training labels to obtain the trained third prediction model.

[0010] According to the present invention, a method for determining surface subsidence parameters under a probability integral model is provided. The prediction model includes a fourth prediction model corresponding to the inflection point offset parameters under the probability integral model, and the fourth prediction model is obtained in the following manner: Data related to the inflection point offset parameter are selected from the feature data as the fourth training data, and the corresponding measured values ​​of the inflection point offset parameter in the measured values ​​of the historical surface subsidence parameters are used as the fourth training labels. The neural network model is trained based on the fourth training data and the fourth training labels to obtain the trained fourth prediction model.

[0011] According to the method for determining surface subsidence parameters under a probability integral model provided by the present invention, the step of extracting the feature data required for each type of surface subsidence parameter under the probability integral model from the engineering data includes: Based on the mining data and geological data in the engineering data, basic engineering features are constructed; the basic engineering features include at least one of the following: average mining depth, working face area, and total rock stratum thickness. Based on the aforementioned basic engineering characteristics, comprehensive engineering characteristics are constructed; the comprehensive engineering characteristics include at least one of the following: mining depth-to-thickness ratio, working face length-to-width ratio, rock stratum thickness variation coefficient, and comprehensive geological index. Z-score standardization and KNN interpolation are performed on the basic engineering features and the comprehensive engineering features to obtain the preprocessed target basic engineering features and target comprehensive engineering features. The feature data required for each type of surface subsidence parameter under the probability integral model are extracted from the target basic engineering features and / or the target comprehensive engineering features, respectively.

[0012] The present invention also provides a device for determining surface subsidence parameters under a probability integral model, comprising: The data acquisition module is used to acquire engineering data for the area to be predicted; the engineering parameters include geological data and mining data. The feature extraction module is used to extract the feature data required for each type of surface subsidence parameter under the probability integral model from the engineering data; wherein, the feature data required for each type of surface subsidence parameter is obtained through correlation analysis and recursive feature elimination screening; The parameter prediction module is used to input the feature data required for each type of surface subsidence parameter into the corresponding prediction model to obtain the predicted value for each type of surface subsidence parameter. The prediction model corresponding to each type of surface subsidence parameter is obtained by supervised training of the initial model matching each type of surface subsidence parameter using the feature data corresponding to each type of surface subsidence parameter and the corresponding measured values ​​of historical surface subsidence parameters from historical engineering data under multiple historical projects.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the method for determining surface subsidence parameters under the probability integral model as described above or the cooling method described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for determining surface subsidence parameters under the probability integral model as described above or the cooling method described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the method for determining surface subsidence parameters under any of the above-described probability integral models or the cooling method.

[0016] The method and apparatus for determining surface subsidence parameters under the probability integral model provided by this invention acquires engineering data including geological and mining data of the area to be predicted, and accurately extracts the feature data required for various surface subsidence parameters under the probability integral model based on correlation analysis and recursive feature elimination. Then, the various feature data are input into a dedicated prediction model that has been trained under the supervision of corresponding feature data from multiple historical projects and measured values ​​of historical surface subsidence parameters. This enables accurate and efficient prediction of various surface subsidence parameters, overcomes the technical problems of high subjectivity, over-reliance on expert experience, and poor consistency in traditional manual experience-based value determination, and effectively improves the pertinence and reliability of surface subsidence parameter prediction. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the method for determining surface subsidence parameters under the probability integral model provided by the present invention.

[0019] Figure 2 This is a schematic diagram of the layered model architecture provided by the present invention.

[0020] Figure 3 This is a schematic diagram of the device for determining surface subsidence parameters under the probability integral model provided by the present invention.

[0021] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] It should be noted that in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0024] The terms "first," "second," etc., used in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0025] This invention provides a method for determining surface subsidence parameters under a probability integral model. Figure 1 This is a flowchart illustrating the method for determining surface subsidence parameters under the probability integral model provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps 110, 120 and 130.

[0026] Step 110: Obtain engineering data for the area to be predicted; the engineering parameters include geological data and mining data.

[0027] In this step, when it is necessary to predict surface subsidence in a specific area, i.e., the area to be predicted, it is first necessary to collect detailed engineering data for that area. Here, the engineering data is a comprehensive description of the geological and mining conditions related to surface subsidence, and it contains at least two types of data: Geological data: This type of data describes the characteristics of the rock and soil mass in the area to be predicted; specifically, it includes, but is not limited to: the thickness of the loose layer covering the surface, the groundwater level, the thickness distribution of each rock layer in the overlying rock column, the lithology of each rock layer, and the physical and mechanical properties of each rock layer. Mining data: This type of data describes the design parameters for mining activities in the area to be predicted; specifically, it includes, but is not limited to: coal seam thickness, coal seam dip angle, average mining depth, working face geometry, stope advance speed, and specific mining methods.

[0028] In this embodiment, these data can be obtained by consulting existing materials such as mine geological reports, borehole columnar sections, and working face design specifications, or by obtaining them through on-site measurement methods, and there are no restrictions on this.

[0029] Step 120: Extract the feature data required for each type of surface subsidence parameter under the probability integral model from the engineering data; wherein, the feature data required for each type of surface subsidence parameter is obtained through correlation analysis and recursive feature elimination screening.

[0030] After acquiring the raw engineering data, this data is transformed into feature data that can be directly used by machine learning models. To improve the specificity and accuracy of predictions, this embodiment employs a decoupled prediction strategy for different parameters in the probability integral model, that is, constructing a separate feature set that best suits each type of parameter.

[0031] Here, each type of surface subsidence parameter in the probability integral model refers to the key parameters required by the prediction method. In this field, there are usually seven, such as the subsidence coefficient q, the horizontal movement coefficient b, the tangent of the main influence angle tanβ, and four inflection point offset parameters s1, s2, s3, and s4.

[0032] In this embodiment, for each type of surface subsidence parameter, during the training process of the prediction model corresponding to that type of surface subsidence parameter, correlation analysis and recursive feature elimination are performed on all input historical engineering data to screen out the feature data required for that type of surface subsidence parameter. Specifically, firstly, the correlation coefficient between each original engineering data and the surface subsidence parameter to be predicted is calculated, such as the Pearson correlation coefficient or the Spearman rank correlation coefficient; by setting a correlation threshold, redundant data with extremely low correlation to the target parameter is initially eliminated, and data with high correlation are retained as candidate features; secondly, based on the correlation analysis, a recursive feature elimination strategy is further adopted. In each round of model training, features with the smallest contribution are eliminated according to the feature weight coefficient or importance score, and then the remaining features are used to repeat this process until the preset number of features is reached or the optimal model performance is obtained. The screened feature data is then used as the feature data required for that type of surface subsidence parameter.

[0033] It should be understood that by combining correlation analysis with recursive feature elimination, this embodiment can accurately select the feature data that has the greatest impact on each surface subsidence parameter from massive amounts of engineering data. This not only reduces the complexity of model calculations but also eliminates the interference of multicollinearity on the model, thereby improving the generalization ability and prediction accuracy of subsequent models.

[0034] Step 130: Input the feature data required for each type of surface subsidence parameter into the corresponding prediction model to obtain the predicted value for each type of surface subsidence parameter.

[0035] After feature extraction and filtering are completed, the specific prediction task is executed. Specifically, in this embodiment, based on the multi-parameter decoupled prediction method, the probability integral model typically includes 7 key parameters: subsidence coefficient q, horizontal movement coefficient b, tangent of the main influence angle tanβ, and 4 inflection point offset parameters s1, s2, s3, and s4. This embodiment does not use a single large model to output all parameters at the same time, but instead constructs a corresponding prediction model for each type of parameter, and uses the corresponding prediction model to obtain the predicted value corresponding to each type of surface subsidence parameter.

[0036] For example, refer to Figure 2 As shown, the feature data selected for the sinking coefficient q is input into the prediction model for the sinking coefficient q, and the model will output the predicted value of the sinking coefficient q. Similarly, the feature data selected for the horizontal movement coefficient b is input into the prediction model for the horizontal movement coefficient b to obtain the predicted value of the horizontal movement coefficient b; the feature data selected for the tangent value of the main influence angle tanβ is input into the prediction model for the tangent value of the main influence angle tanβ to obtain the predicted value of the tangent value of the main influence angle tanβ; and the feature data selected for the inflection point offset parameters s1-s4 is input into the prediction model for the inflection point offset parameters s1-s4, and the model will output the predicted values ​​of the inflection point offset parameters s1-s4.

[0037] In this embodiment, the prediction model corresponding to each type of surface subsidence parameter is obtained by supervised training of the initial model matching each type of surface subsidence parameter using the feature data corresponding to each type of surface subsidence parameter in the historical engineering data of multiple historical projects and the corresponding measured values ​​of historical surface subsidence parameters.

[0038] Specifically, a standardized historical database needs to be collected and organized in advance, containing complete records of multiple historical engineering projects. For each historical record, it includes both the historical engineering data and the measured values ​​of historical surface subsidence parameters obtained through on-site measurement and inversion.

[0039] During the training phase, feature data required for each type of surface subsidence parameter is selected from historical engineering data through correlation analysis and recursive feature elimination. These feature data, along with the corresponding measured parameter values, form a training set to train the initial machine learning model. By continuously adjusting the model parameters to minimize prediction error, a prediction model with generalization capabilities is finally obtained.

[0040] The method for determining surface subsidence parameters under the probability integral model in this embodiment obtains engineering data, including geological and mining data, of the area to be predicted. Based on correlation analysis and recursive feature elimination, it accurately extracts the feature data required for various surface subsidence parameters under the probability integral model. Then, it inputs the various feature data into a dedicated prediction model that has been trained with supervision from the corresponding feature data of multiple historical projects and the measured values ​​of historical surface subsidence parameters. This method can achieve accurate and efficient prediction of various surface subsidence parameters, overcomes the technical problems of traditional manual experience-based value determination, such as high subjectivity, over-reliance on expert experience, and poor consistency, and effectively improves the pertinence and reliability of surface subsidence parameter prediction.

[0041] It should be noted that each implementation method of this application can be freely combined, rearranged, or executed individually, and does not need to rely on or depend on a fixed execution order.

[0042] In some embodiments, the feature data required to extract each type of surface subsidence parameter under the probability integral model from the engineering data includes: Based on the mining data and geological data in the engineering data, basic engineering features are constructed; the basic engineering features include at least one of the following: average mining depth, working face area, and total rock stratum thickness. Based on the aforementioned basic engineering characteristics, comprehensive engineering characteristics are constructed; the comprehensive engineering characteristics include at least one of the following: mining depth-to-thickness ratio, working face length-to-width ratio, rock stratum thickness variation coefficient, and comprehensive geological index. Z-score standardization and KNN interpolation are performed on the basic engineering features and the comprehensive engineering features to obtain the preprocessed target basic engineering features and target comprehensive engineering features. The feature data required for each type of surface subsidence parameter under the probability integral model are extracted from the target basic engineering features and / or the target comprehensive engineering features, respectively.

[0043] In this embodiment, based on mining data and geological data from the engineering data, basic features reflecting the fundamental geometric and geological characteristics of the project are constructed, including but not limited to: Geometric characteristics: such as the average value of the uphill and downhill mining depths, mining dimensions, etc.

[0044] Area characteristics: For example, the area of ​​the working face calculated by multiplying the working face's direction length and its dip length.

[0045] Geological characteristics: For example, the total thickness of the rock strata obtained by summing the thicknesses of each geological unit in a rock strata columnar diagram.

[0046] To delve deeper into the engineering patterns hidden within the data, this step, in addition to the basic engineering characteristics, constructs a series of comprehensive engineering features that better reflect the inherent laws and geological characteristics of the engineering, including but not limited to: Mining depth to thickness ratio: reflects the relative intensity of mining disturbance; The aspect ratio of the working surface reflects its geometric shape. Rock layer thickness variation coefficient: reflects the complexity and heterogeneity of geological conditions; Comprehensive geological index: This is a composite index that takes into account multiple parameters such as rock hardness and rock layer structure, and is used to quantify the overall mechanical properties of overburden.

[0047] Because the aforementioned feature data have different dimensions, directly inputting them into the model would lead to weight bias. Therefore, in this embodiment, Z-score standardization is applied to the basic engineering features and comprehensive engineering features, transforming all features to a standard normal distribution with a mean of 0 and a variance of 1, thus eliminating the influence of dimensions. Furthermore, for missing values ​​commonly found in geological history data, KNN (K-Nearest Neighbor) interpolation is used. This involves finding the K most similar historical samples and using their values ​​to weighted imputation of missing values, ensuring data integrity.

[0048] After the above processing, the preprocessed target basic engineering features and target comprehensive engineering features are obtained. Finally, using correlation analysis and recursive feature elimination methods, the feature data required for each type of surface subsidence parameter can be extracted from these target features.

[0049] The method for determining surface subsidence parameters under the probability integral model in this embodiment transforms the original engineering data into standardized model inputs, thereby ensuring the accuracy of the final trained prediction model from the data source.

[0050] In some embodiments, after inputting the feature data required for each type of surface subsidence parameter into the corresponding prediction model to obtain the predicted value corresponding to each type of surface subsidence parameter, the method further includes: Determine the confidence level of the predicted values ​​for each type of surface subsidence parameter; For each type of surface subsidence parameter, if the confidence level is greater than the first threshold, the predicted value is determined to be the target value corresponding to the surface subsidence parameter. If the confidence level is greater than the second threshold but not greater than the first threshold, the target value corresponding to the surface subsidence parameter is determined based on the predicted value and the historical experience value corresponding to the surface subsidence parameter. If the confidence level is not greater than the second threshold, the target value corresponding to the surface subsidence parameter is determined based on the historical experience value corresponding to the surface subsidence parameter.

[0051] Confidence score is a metric that measures the reliability of a model's predictions. In this embodiment, the confidence score can be determined based on the model's performance metrics on the validation set, such as the coefficient of determination R0. 2 The root mean square error (RMSE) or the variance of the probability distribution of the model output.

[0052] Specifically, using the coefficient of determination R 2 For example, during the training phase, the system records the model's fit to different feature spaces. When new feature data is input, the system calculates the current confidence level based on the historical goodness of fit of the feature space region where the data is located.

[0053] After determining the confidence level of the predicted value, this embodiment introduces a hierarchical decision-making mechanism, which is implemented through two preset first thresholds and a second threshold. Here, both the first and second thresholds are values ​​that can be adjusted according to the actual engineering accuracy requirements, and the first threshold is greater than the second threshold. For example, the first threshold can be set to 0.8, and the second threshold can be set to 0.6.

[0054] After determining the confidence level of the predicted value and the first and second thresholds, when the confidence level of the predicted value is greater than the first threshold, for example, R... 2 A confidence level greater than 0.8 indicates a high degree of consistency between the distribution of the input data and the model training data, signifying excellent model prediction performance. In this case, the predicted value is directly determined to be the target value corresponding to the surface subsidence parameter. When the confidence level of the predicted value is greater than the second threshold but not greater than the first threshold, for example, 0.6... <R 2 When the confidence level is ≤ 0.8, it indicates that the model prediction has reference value but also contains a certain degree of uncertainty. In this case, the target value for the surface subsidence parameter is determined based on the predicted value and the corresponding historical empirical values. A weighted average can be used, and the historical empirical values ​​can be taken from the statistical average of historical engineering projects under similar geological conditions. When the confidence level is not greater than the second threshold, for example, R0... 2 When the value is ≤ 0.6, the intelligent rollback mechanism is triggered, and the target value corresponding to the surface subsidence parameter is determined directly based on the historical experience value corresponding to the surface subsidence parameter. That is, it rolls back to the traditional empirical formula method or theoretical default value as the system's default target value.

[0055] The method for determining surface subsidence parameters under the probability integral model in this embodiment ensures the reliability of the prediction results and the system's adaptability by introducing a three-level decision-making mechanism based on confidence level.

[0056] In some embodiments, the prediction model includes a first prediction model corresponding to the sinking coefficient under a probability integral model, the first prediction model being constructed in the following manner: Data related to the subsidence coefficient are selected from the feature data as the first training data, and the corresponding measured values ​​of the subsidence coefficient in the measured values ​​of the historical surface subsidence parameters are used as the first training labels. The random forest model is trained based on the first training data and the first training label to obtain the trained first prediction model.

[0057] It should be noted that, since the subsidence coefficient is affected by the nonlinear comprehensive influence of various geological and mining factors, such as mining thickness, mining depth, and lithological combination, the random forest model is selected as the initial model corresponding to the subsidence coefficient in this embodiment. As an ensemble learning algorithm, the random forest model consists of multiple decision trees, which is less prone to overfitting when processing high-dimensional data and nonlinear relationships, and the model has strong robustness.

[0058] Specifically, in this embodiment, data preparation is performed from the historical engineering database for the specific parameter of subsidence coefficient, and historical feature data related to the subsidence coefficient is selected as the first training data. These feature data are the factors that have the most significant impact on the subsidence coefficient, selected based on methods such as correlation analysis and recursive feature elimination, such as mining thickness, average mining depth, mining depth-to-thickness ratio, and overburden lithology ratio. Simultaneously, the measured subsidence coefficient values ​​in the historical engineering database that correspond one-to-one with these feature data are used as the first training labels.

[0059] The first training data and the first training labels are input into an initial random forest regression model. Through a supervised training process, the internal parameters of the model are adjusted, such as the number of decision trees, the depth of the trees, and the node splitting strategy, so that the model can learn and fit the complex mapping relationship from the input features to the sinking coefficient value, and finally obtain a trained first prediction model for predicting the sinking coefficient.

[0060] The method for determining surface subsidence parameters under the probability integral model in this embodiment selects a random forest model for supervised training based on the physical characteristics of the subsidence coefficient. Compared with the general model, the random forest model can better capture the complex nonlinear mapping relationship between mining depth, mining thickness and subsidence coefficient, thereby improving the prediction accuracy of the subsidence coefficient.

[0061] In some embodiments, the prediction model includes a second prediction model corresponding to the horizontal movement coefficient under a probability integral model, the second prediction model being constructed in the following manner: Data related to the horizontal movement coefficient are selected from the feature data as the second training data, and the corresponding measured values ​​of the horizontal movement coefficient in the measured values ​​of the historical surface subsidence parameters are used as the second training labels. Based on the second training data and the second training labels, the support vector regression model is trained to obtain the trained second prediction model.

[0062] It should be noted that, since the horizontal movement coefficient is mainly affected by factors such as the dip angle of the working face and the geological conditions of the overlying strata, the actual measured data is relatively small compared to the subsidence coefficient. Therefore, in this embodiment, the support vector regression model is selected as the initial model corresponding to the horizontal movement coefficient. The support vector regression model is based on statistical learning theory and minimizes the prediction error by finding a hyperplane. It performs well in the case of small samples and high dimensionality, and has strong generalization ability, which can effectively avoid the overfitting problem when the sample size is limited.

[0063] Specifically, in this embodiment, data preparation is performed from the historical engineering database for the specific parameter of horizontal movement coefficient, and historical feature data related to the horizontal movement coefficient is selected as the second training data. These feature data are the factors that have the most significant impact on the subsidence coefficient, selected by methods such as correlation analysis and recursive feature elimination, such as the working face dip angle and the geological conditions of the overlying strata. At the same time, the measured values ​​of the horizontal movement coefficient corresponding one-to-one with these feature data in the historical engineering database are used as the second training labels.

[0064] The second training data and the second training labels are input into an initial support vector regression model. Through a supervised training process, a suitable kernel function, such as a linear kernel, a polynomial kernel, or a radial basis function kernel, is selected and relevant parameters, such as the penalty coefficient and kernel function parameters, are optimized. Finally, a trained second prediction model for predicting the horizontal movement coefficient is obtained.

[0065] The method for determining surface subsidence parameters under the probability integral model in this embodiment selects a support vector regression model for supervised training based on the sample characteristics of the horizontal movement coefficient. Compared with the general model, the support vector regression model can better and more effectively capture its key influencing factors, thereby improving the prediction accuracy of the horizontal movement coefficient.

[0066] In some embodiments, the prediction model includes a third prediction model corresponding to the tangent of the principal influence angle under a probability integral model, the third prediction model being constructed in the following manner: Data related to the tangent of the main influence angle are selected from the feature data as the third training data, and the corresponding measured values ​​of the tangent of the main influence angle in the measured values ​​of the historical surface subsidence parameters are used as the third training labels. The gradient boosting tree model is trained based on the third training data and the third training labels to obtain the trained third prediction model.

[0067] It should be noted that the tangent of the main influence angle mainly reflects the movement range of the overlying strata and is closely related to the properties of the strata. Its value range is relatively concentrated. Based on this, the gradient boosting tree model is selected as the initial model corresponding to the tangent of the main influence angle in this embodiment. The gradient boosting tree model is an iterative decision tree algorithm with extremely high prediction accuracy. It can continuously approach the true value through residual iteration and capture the subtle influence of changes in strata properties on the tangent of the main influence angle.

[0068] Specifically, in this embodiment, data preparation is performed from the historical engineering database for the specific parameter of the tangent of the main influence angle, and historical feature data related to the tangent of the main influence angle are selected as the third training data. These feature data are the factors that have the most significant impact on the tangent of the main influence angle, selected by methods such as correlation analysis and recursive feature elimination, such as the comprehensive strength of the overburden, the thickness of the rock strata and their combination, and other property parameters. At the same time, the measured values ​​of the tangent of the main influence angle corresponding one-to-one with these feature data in the historical engineering database are used as the third training labels.

[0069] The aforementioned third training data and third training labels are input into an initial gradient boosting tree model. By adjusting hyperparameters such as the learning rate, the number of trees, and the depth of the trees, the model achieves a good balance between bias and variance, ultimately resulting in a well-trained third prediction model for predicting the tangent value of the main influencing angle.

[0070] The method for determining surface subsidence parameters under the probability integral model in this embodiment achieves accurate prediction of the tangent value of the main influencing angle by using a gradient boosting tree model and leveraging its powerful residual fitting capability.

[0071] In some embodiments, the prediction model includes a fourth prediction model corresponding to the inflection point offset parameter under the probability integral model, and the fourth prediction model is obtained in the following manner: Data related to the inflection point offset parameter are selected from the feature data as the fourth training data, and the corresponding measured values ​​of the inflection point offset parameter in the measured values ​​of the historical surface subsidence parameters are used as the fourth training labels. The neural network model is trained based on the fourth training data and the fourth training labels to obtain the trained fourth prediction model.

[0072] It should be noted that the inflection point offset parameter is not only a geological issue, but also a geometric spatial issue. There is a complex topological mapping relationship between geometric dimensions and offset distance. Based on this, a neural network model is selected as the initial model corresponding to the inflection point offset parameter in this embodiment. The neural network can simulate this complex spatial geometric transformation relationship through the connection of multiple layers of neurons.

[0073] Specifically, in this embodiment, data preparation is performed from the historical engineering database for the specific parameter of inflection point offset, and historical feature data related to the inflection point offset parameter is selected as the fourth training data. These feature data are the factors that have the most significant impact on the inflection point offset parameter, selected by methods such as correlation analysis and recursive feature elimination, such as geometric parameters like working face strike length, dip length, mining depth, and mining thickness. Simultaneously, the measured values ​​of the inflection point offset parameter in the historical engineering database that correspond one-to-one with these feature data are used as the fourth training labels.

[0074] The aforementioned fourth training data and fourth training labels are input into an initial neural network model. The weights and biases in the network are continuously adjusted through optimization methods such as backpropagation and gradient descent until the model's performance on the validation set reaches its optimal level. Finally, a trained fourth prediction model for predicting inflection point offset parameters is obtained.

[0075] The method for determining surface subsidence parameters under the probability integral model in this embodiment utilizes the nonlinear fitting capability of the neural network model to handle the complex mapping relationship between the inflection point offset parameters and the geometric features of the working face, thereby achieving accurate prediction of the inflection point offset parameters.

[0076] The following describes the device for determining surface subsidence parameters under the probability integral model provided by the present invention. The device for determining surface subsidence parameters under the probability integral model described below and the method for determining surface subsidence parameters under the probability integral model described above can be referred to in correspondence.

[0077] The device for determining surface subsidence parameters under the probability integral model in this invention embodiment, such as... Figure 3 As shown, it includes the following modules: Data acquisition module 310 is used to acquire engineering data of the area to be predicted; the engineering parameters include geological data and mining data; The feature extraction module 320 is used to extract the feature data required for each type of surface subsidence parameter under the probability integral model from the engineering data; wherein, the feature data required for each type of surface subsidence parameter is obtained through correlation analysis and recursive feature elimination screening; The parameter prediction module 330 is used to input the feature data required for each type of surface subsidence parameter into the corresponding prediction model to obtain the predicted value corresponding to each type of surface subsidence parameter. The prediction model corresponding to each type of surface subsidence parameter is obtained by supervised training of the initial model matching each type of surface subsidence parameter using the feature data corresponding to each type of surface subsidence parameter and the corresponding measured values ​​of historical surface subsidence parameters from historical engineering data under multiple historical projects.

[0078] The surface subsidence parameter determination device under the probability integral model in this embodiment acquires engineering data including geological and mining data of the area to be predicted. Based on correlation analysis and recursive feature elimination, it accurately extracts the feature data required for various surface subsidence parameters under the probability integral model. Then, it inputs the various feature data into a dedicated prediction model that has been trained with supervision from the corresponding feature data of multiple historical projects and the measured values ​​of historical surface subsidence parameters. This enables accurate and efficient prediction of various surface subsidence parameters, overcoming the technical problems of high subjectivity, over-reliance on expert experience, and poor consistency in traditional manual experience-based value determination. This effectively improves the pertinence and reliability of surface subsidence parameter prediction.

[0079] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 330, and a communication bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a method for determining surface subsidence parameters under a probability integral model, the method including: Acquire engineering data for the area to be predicted; the engineering parameters include geological data and mining data. The feature data required for each type of surface subsidence parameter under the probability integral model are extracted from the engineering data; wherein, the feature data required for each type of surface subsidence parameter is obtained through correlation analysis and recursive feature elimination screening; The characteristic data required for each type of surface subsidence parameter are input into the corresponding prediction model to obtain the predicted value for each type of surface subsidence parameter. The prediction model corresponding to each type of surface subsidence parameter is obtained by supervised training of the initial model matching each type of surface subsidence parameter using the feature data corresponding to each type of surface subsidence parameter and the corresponding measured values ​​of historical surface subsidence parameters from historical engineering data under multiple historical projects.

[0080] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., each of which can store program code.

[0081] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is capable of executing the method for determining surface subsidence parameters under the probability integral model provided by each of the above methods, the method comprising: Acquire engineering data for the area to be predicted; the engineering parameters include geological data and mining data. The feature data required for each type of surface subsidence parameter under the probability integral model are extracted from the engineering data; wherein, the feature data required for each type of surface subsidence parameter is obtained through correlation analysis and recursive feature elimination screening; The characteristic data required for each type of surface subsidence parameter are input into the corresponding prediction model to obtain the predicted value for each type of surface subsidence parameter. The prediction model corresponding to each type of surface subsidence parameter is obtained by supervised training of the initial model matching each type of surface subsidence parameter using the feature data corresponding to each type of surface subsidence parameter and the corresponding measured values ​​of historical surface subsidence parameters from historical engineering data under multiple historical projects.

[0082] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for determining surface subsidence parameters under a probability integral model provided by each of the above methods, the method comprising: Acquire engineering data for the area to be predicted; the engineering parameters include geological data and mining data. The feature data required for each type of surface subsidence parameter under the probability integral model are extracted from the engineering data; wherein, the feature data required for each type of surface subsidence parameter is obtained through correlation analysis and recursive feature elimination screening; The characteristic data required for each type of surface subsidence parameter are input into the corresponding prediction model to obtain the predicted value for each type of surface subsidence parameter. The prediction model corresponding to each type of surface subsidence parameter is obtained by supervised training of the initial model matching each type of surface subsidence parameter using the feature data corresponding to each type of surface subsidence parameter and the corresponding measured values ​​of historical surface subsidence parameters from historical engineering data under multiple historical projects.

[0083] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0084] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in each of the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A method for determining surface subsidence parameters under a probabilistic integral model, characterized in that, include: Acquire engineering data for the area to be predicted; the engineering parameters include geological data and mining data. The feature data required for each type of surface subsidence parameter under the probability integral model are extracted from the engineering data; wherein, the feature data required for each type of surface subsidence parameter is obtained through correlation analysis and recursive feature elimination screening; The characteristic data required for each type of surface subsidence parameter are input into the corresponding prediction model to obtain the predicted value for each type of surface subsidence parameter. The prediction model corresponding to each type of surface subsidence parameter is obtained by supervised training of the initial model matching each type of surface subsidence parameter using the feature data corresponding to each type of surface subsidence parameter and the corresponding measured values ​​of historical surface subsidence parameters from historical engineering data under multiple historical projects.

2. The method for determining surface subsidence parameters under the probability integral model according to claim 1, characterized in that, After inputting the required feature data for each type of surface subsidence parameter into the corresponding prediction model to obtain the predicted value for each type of surface subsidence parameter, the method further includes: Determine the confidence level of the predicted values ​​for each type of surface subsidence parameter; For each type of surface subsidence parameter, if the confidence level is greater than the first threshold, the predicted value is determined to be the target value corresponding to the surface subsidence parameter. If the confidence level is greater than the second threshold but not greater than the first threshold, the target value corresponding to the surface subsidence parameter is determined based on the predicted value and the historical experience value corresponding to the surface subsidence parameter. If the confidence level is not greater than the second threshold, the target value corresponding to the surface subsidence parameter is determined based on the historical experience value corresponding to the surface subsidence parameter.

3. The method for determining surface subsidence parameters under the probability integral model according to claim 1, characterized in that, The prediction model includes a first prediction model corresponding to the sinking coefficient under the probability integral model, which is constructed in the following manner: Data related to the subsidence coefficient are selected from the feature data as the first training data, and the corresponding measured values ​​of the subsidence coefficient in the measured values ​​of the historical surface subsidence parameters are used as the first training labels. The random forest model is trained based on the first training data and the first training label to obtain the trained first prediction model.

4. The method for determining surface subsidence parameters under the probability integral model according to claim 1, characterized in that, The prediction model includes a second prediction model corresponding to the horizontal shift coefficient under the probability integral model, and the second prediction model is constructed in the following manner: Data related to the horizontal movement coefficient are selected from the feature data as the second training data, and the corresponding measured values ​​of the horizontal movement coefficient in the measured values ​​of the historical surface subsidence parameters are used as the second training labels. Based on the second training data and the second training labels, the support vector regression model is trained to obtain the trained second prediction model.

5. The method for determining surface subsidence parameters under the probability integral model according to claim 1, characterized in that, The prediction model includes a third prediction model corresponding to the tangent of the main influence angle under the probability integral model, and the third prediction model is constructed in the following way: Data related to the tangent of the main influence angle are selected from the feature data as the third training data, and the corresponding measured values ​​of the tangent of the main influence angle in the measured values ​​of the historical surface subsidence parameters are used as the third training labels. The gradient boosting tree model is trained based on the third training data and the third training labels to obtain the trained third prediction model.

6. The method for determining surface subsidence parameters under the probability integral model according to claim 1, characterized in that, The prediction model includes a fourth prediction model corresponding to the inflection point offset parameter under the probability integral model, which is obtained in the following way: Data related to the inflection point offset parameter are selected from the feature data as the fourth training data, and the corresponding measured values ​​of the inflection point offset parameter in the measured values ​​of the historical surface subsidence parameters are used as the fourth training labels. The neural network model is trained based on the fourth training data and the fourth training labels to obtain the trained fourth prediction model.

7. The method for determining surface subsidence parameters under the probability integral model according to claim 1, characterized in that, The feature data required to extract each type of surface subsidence parameter under the probability integral model from the engineering data includes: Based on the mining data and geological data in the engineering data, basic engineering features are constructed; the basic engineering features include at least one of the following: average mining depth, working face area, and total rock stratum thickness. Based on the aforementioned basic engineering characteristics, comprehensive engineering characteristics are constructed; the comprehensive engineering characteristics include at least one of the following: mining depth-to-thickness ratio, working face length-to-width ratio, rock stratum thickness variation coefficient, and comprehensive geological index. The basic engineering features and the comprehensive engineering features are subjected to Z-score standardization and KNN interpolation to obtain the preprocessed target basic engineering features and target comprehensive engineering features. The feature data required for each type of surface subsidence parameter under the probability integral model are extracted from the target basic engineering features and / or the target comprehensive engineering features, respectively.

8. A device for determining surface subsidence parameters under a probability integral model, characterized in that, include: The data acquisition module is used to acquire engineering data for the area to be predicted; the engineering parameters include geological data and mining data. The feature extraction module is used to extract the feature data required for each type of surface subsidence parameter under the probability integral model from the engineering data; wherein, the feature data required for each type of surface subsidence parameter is obtained through correlation analysis and recursive feature elimination screening; The parameter prediction module is used to input the feature data required for each type of surface subsidence parameter into the corresponding prediction model to obtain the predicted value for each type of surface subsidence parameter. The prediction model corresponding to each type of surface subsidence parameter is obtained by supervised training of the initial model matching each type of surface subsidence parameter using the feature data corresponding to each type of surface subsidence parameter and the corresponding measured values ​​of historical surface subsidence parameters from historical engineering data under multiple historical projects.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the method for determining surface subsidence parameters under the probability integral model as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for determining surface subsidence parameters under the probability integral model as described in any one of claims 1 to 7.

Images