A method, device, equipment and medium for predicting soil stratigraphic distribution

Abstract

The application provides a soil stratum distribution prediction method, a prediction device, equipment and a medium. The method comprises the following steps: for each target position point in a soil area to be predicted, a plurality of adjacent drill hole points closest to the target position point are determined from a drill hole set; stratum category information corresponding to each adjacent drill hole point is obtained, and a feature matrix corresponding to the target position point is generated by using a multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drill hole point; the feature matrix is input into a pre-trained stratum distribution prediction model to determine a stratum distribution prediction result corresponding to the target position point, and a target stratum category to which the target position point belongs is determined based on the stratum distribution prediction result. Through the method and the device, the prediction accuracy, the boundary recognition ability and the recognition ability of a small number of strata are improved, manual intervention is reduced, and the surveying precision and efficiency are improved.

Description

Technical Field

[0001] The present application relates to the technical field of stratum distribution prediction, and in particular to a soil stratum distribution prediction method, prediction device, equipment and medium. Background Art

[0002] In geotechnical engineering, an accurate understanding of the distribution of subsurface strata is crucial for engineering design and construction. Traditional geological modeling methods, such as spline interpolation, inverse distance weighting, and kriging interpolation, typically model the expected profile through two-dimensional stratigraphic interpolation. While these methods are effective in homogeneous regions, their high data volume and interpolation function requirements limit their predictive accuracy and applicability in complex three-dimensional stratigraphic conditions, making them difficult to automate. In recent years, with the rise of data-driven approaches, the application of machine learning and deep learning in geological modeling has rapidly developed. In addition to traditional interpolation methods and probabilistic models, methods such as random forests and deep neural networks (DNNs) have demonstrated greater adaptability and predictive accuracy in handling complex stratigraphic features. These methods extract knowledge directly from existing data without relying on prior assumptions, demonstrating high robustness in handling non-uniform borehole distribution or highly heterogeneous stratigraphic conditions.

[0003] Despite this, current stratigraphic modeling still faces numerous challenges. Many existing models rely too heavily on limited borehole data, resulting in insufficient generalization. Some regions have a limited number of stratigraphic categories, making it difficult for models to effectively identify a few. Therefore, it is crucial to develop a more refined soil stratigraphic distribution prediction method that is applicable to a wider range of stratigraphic categories. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a soil stratum distribution prediction method, prediction device, equipment and medium, which only requires multiple adjacent drilling points to quickly predict the stratum distribution of any unknown point within the range, thereby improving the prediction accuracy, boundary recognition ability and recognition ability of a few types of strata, improving the automation level of stratum modeling, reducing manual intervention, and improving exploration accuracy and efficiency.

[0005] In a first aspect, an embodiment of the present application provides a method for predicting soil stratum distribution, the method comprising:

[0006] For each target location point in the soil area to be predicted, determine a plurality of adjacent drill hole points closest to the target location point from the drill hole set;

[0007] Obtaining stratum category information corresponding to each adjacent drilling point, and using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point to generate a feature matrix corresponding to the target location point;

[0008] The feature matrix is ​​input into a pre-trained stratum distribution prediction model to determine the stratum distribution prediction result corresponding to the target location point, and the target stratum category to which the target location point belongs is determined based on the stratum distribution prediction result; wherein the stratum prediction distribution result represents the probability that the target location point belongs to each preset stratum category, and the training data used by the stratum distribution prediction model in the training process is obtained by expanding the SMOTE oversampling method.

[0009] Furthermore, the step of determining a plurality of adjacent drilling points closest to the target location point from the drilling set includes:

[0010] For each drill hole point to be screened in the drill hole set, determining the spatial position coordinates of the drill hole point to be screened relative to the target position point, and calculating the Euclidean distance between the drill hole point to be screened and the target position point using the spatial position coordinates;

[0011] The multiple drilling points to be screened are sorted from small to large distance using the Euclidean distance corresponding to each drilling point to be screened, and a preset number of drilling points to be screened are selected from the sorting as the adjacent drilling points.

[0012] Furthermore, the method of generating a feature matrix corresponding to the target location point using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point includes:

[0013] Gridding the stratum distribution of the soil area to be predicted to form a high-resolution stratum distribution matrix;

[0014] For each adjacent drilling point, based on the stratum category information corresponding to the adjacent drilling point, an initial feature code corresponding to the adjacent drilling point is determined from the multidimensional vectors corresponding to multiple preset stratum categories, and depth information corresponding to the adjacent drilling point is determined from the stratum distribution matrix;

[0015] The spatial coordinates of the neighboring drilling point relative to the target location point, the Euclidean distance between the drilling point to be screened and the target location point, the depth information corresponding to the neighboring drilling point, and the initial feature code corresponding to the neighboring drilling point are spliced ​​to generate a target feature code corresponding to the neighboring drilling point;

[0016] The target feature codes corresponding to multiple adjacent drilling points are used to generate a feature matrix corresponding to the target position point.

[0017] Furthermore, the stratum distribution prediction model is trained by the following steps:

[0018] Determine a test set from the original drilling data set, and use the remaining data in the original drilling data set as a training set and a validation set;

[0019] Using a five-fold cross-validation method to divide the remaining data in the original drilling data set into a training set and a validation set;

[0020] Perform SMOTE oversampling on the minority class samples in the training set to obtain an expanded target training set;

[0021] The target training set and the validation set are used to train and validate the original prediction model of stratum distribution respectively until convergence;

[0022] The converged original stratum distribution prediction model is tested using the test set, and the converged original stratum distribution prediction model that passes the test is used as the stratum distribution prediction model.

[0023] Furthermore, the SMOTE oversampling is performed on the minority class samples in the training set to obtain an expanded target training set, including:

[0024] Obtaining minority class samples in the training set, and determining a plurality of neighbor samples corresponding to each minority class sample;

[0025] For each minority class sample, a target neighbor sample is determined from a plurality of neighbor samples corresponding to the minority class sample, and linear interpolation is performed between the minority class sample and the target neighbor sample to generate an expanded sample to obtain the target training set.

[0026] In a second aspect, an embodiment of the present application further provides a device for predicting soil stratum distribution, the device comprising:

[0027] A drilling point determination module is used to determine, for each target location point in the soil area to be predicted, a plurality of adjacent drilling points that are closest to the target location point from the drilling set;

[0028] A feature matrix generation module is used to obtain the stratum category information corresponding to each adjacent drilling point, and use the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point to generate a feature matrix corresponding to the target location point;

[0029] The stratum category prediction module is used to input the feature matrix into a pre-trained stratum distribution prediction model, determine the stratum distribution prediction result corresponding to the target location point, and determine the target stratum category to which the target location point belongs based on the stratum distribution prediction result; wherein the stratum prediction distribution result represents the probability that the target location point belongs to each preset stratum category, and the training data used by the stratum distribution prediction model in the training process is obtained after expansion through the SMOTE oversampling method.

[0030] Furthermore, when the drilling point determination module is used to determine a plurality of adjacent drilling points closest to the target location point from the drilling set, the drilling point determination module is further used to:

[0031] For each drill hole point to be screened in the drill hole set, determining the spatial position coordinates of the drill hole point to be screened relative to the target position point, and calculating the Euclidean distance between the drill hole point to be screened and the target position point using the spatial position coordinates;

[0032] The multiple drilling points to be screened are sorted from small to large distance using the Euclidean distance corresponding to each drilling point to be screened, and a preset number of drilling points to be screened are selected from the sorting as the adjacent drilling points.

[0033] Furthermore, when the feature matrix generation module is used to generate the feature matrix corresponding to the target location point using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point, the feature matrix generation module is further used to:

[0034] Gridding the stratum distribution of the soil area to be predicted to form a high-resolution stratum distribution matrix;

[0035] For each adjacent drilling point, based on the stratum category information corresponding to the adjacent drilling point, an initial feature code corresponding to the adjacent drilling point is determined from the multidimensional vectors corresponding to multiple preset stratum categories, and depth information corresponding to the adjacent drilling point is determined from the stratum distribution matrix;

[0036] The spatial coordinates of the neighboring drilling point relative to the target location point, the Euclidean distance between the drilling point to be screened and the target location point, the depth information corresponding to the neighboring drilling point, and the initial feature code corresponding to the neighboring drilling point are spliced ​​to generate a target feature code corresponding to the neighboring drilling point;

[0037] The target feature codes corresponding to multiple adjacent drilling points are used to generate a feature matrix corresponding to the target position point.

[0038] In a third aspect, an embodiment of the present application further provides an electronic device comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the soil stratum distribution prediction method as described above are performed.

[0039] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, executes the steps of the soil stratum distribution prediction method as described above.

[0040] An embodiment of the present application provides a soil stratum distribution prediction method, prediction device, equipment and medium. First, for each target location point in the soil area to be predicted, multiple adjacent borehole points closest to the target location point are determined from a borehole set; then, the stratum category information corresponding to each adjacent borehole point is obtained, and a feature matrix corresponding to the target location point is generated using a multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent borehole point; finally, the feature matrix is ​​input into a pre-trained stratum distribution prediction model to determine the stratum distribution prediction result corresponding to the target location point, and the target stratum category to which the target location point belongs is determined based on the stratum distribution prediction result; wherein the stratum prediction distribution result represents the probability that the target location point belongs to each preset stratum category, and the training data used by the stratum distribution prediction model in the training process is obtained by expanding the data using the SMOTE oversampling method.

[0041] This application introduces feature extraction of neighboring borehole data, which is applicable to sites containing multiple stratigraphic categories. By combining multiple technologies of model design and data processing, efficient prediction of shallow complex strata is achieved. By selecting data from neighboring borehole points and combining the spatial neighborhood strategy, multiple borehole points closest to the target point are selected as input features, so that the model can effectively capture the local geological characteristics of the target location and can flexibly adapt to unevenly distributed borehole data. Encoding multiple stratigraphic categories and mapping them into multidimensional vectors can ensure that the model can better identify and distinguish complex multi-classified stratigraphic types. For rare stratigraphic categories in the model training data set, the synthetic minority oversampling technique (SMOTE) is used to expand the minority sample data. This application is applicable to complex site conditions containing multiple stratigraphic categories. Only multiple neighboring borehole points are needed to quickly predict the stratigraphic distribution of any unknown point within the range, providing technical support for refined geological modeling. It has improved its prediction accuracy, boundary recognition ability and the ability to identify a few types of strata, increased the automation level of stratum modeling, reduced manual intervention, and improved exploration accuracy and efficiency.

[0042] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 A flow chart of a method for predicting soil stratum distribution provided in an embodiment of the present application;

[0045] Figure 2 A schematic diagram of the spatial distribution of drilling points and neighborhood selection strategy provided in an embodiment of the present application;

[0046] Figure 3 A schematic diagram of a 9-dimensional vector conversion of a stratum based on one-hot encoding provided in an embodiment of the present application;

[0047] Figure 4 A schematic diagram of a characteristic matrix data format and a stratum distribution prediction model architecture provided in an embodiment of the present application;

[0048] Figure 5A schematic diagram of the loss curve and evaluation index curve of a CNN model in the cross-validation process provided in an embodiment of the present application;

[0049] Figure 6 A schematic diagram of the structure of a device for predicting soil stratum distribution provided in an embodiment of the present application;

[0050] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for which protection is claimed, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, each other embodiment obtained by those skilled in the art without making creative work falls within the scope of protection of the present application.

[0052] First, the application scenarios to which this application is applicable are introduced. This application can be applied in the field of stratum distribution prediction technology.

[0053] In geotechnical engineering, an accurate understanding of the distribution of subsurface strata is crucial for engineering design and construction. Traditional geological modeling methods, such as spline interpolation, inverse distance weighting, and kriging interpolation, typically model the expected profile through two-dimensional stratigraphic interpolation. While these methods are effective in homogeneous regions, their high data volume and interpolation function requirements limit their predictive accuracy and applicability in complex three-dimensional stratigraphic conditions, making them difficult to automate. In recent years, with the rise of data-driven approaches, the application of machine learning and deep learning in geological modeling has rapidly developed. In addition to traditional interpolation methods and probabilistic models, methods such as random forests and deep neural networks (DNNs) have demonstrated greater adaptability and predictive accuracy in handling complex stratigraphic features. These methods extract knowledge directly from existing data without relying on prior assumptions, demonstrating high robustness in handling non-uniform borehole distribution or highly heterogeneous stratigraphic conditions.

[0054] Despite this, current stratigraphic modeling still faces numerous challenges. Many existing models rely too heavily on limited borehole data, resulting in insufficient generalization. Some regions have a limited number of stratigraphic categories, making it difficult for models to effectively identify a few. Therefore, it is crucial to develop a more refined soil stratigraphic distribution prediction method that is applicable to a wider range of stratigraphic categories.

[0055] Based on this, the embodiment of the present application provides a method for predicting soil stratum distribution, which only requires multiple adjacent drilling points to quickly predict the stratum distribution of any unknown point within the range, thereby improving the prediction accuracy, boundary recognition ability and the ability to recognize a few types of strata, improving the automation level of stratum modeling, reducing manual intervention, and improving exploration accuracy and efficiency.

[0056] See also Figure 1 , Figure 1 This is a flow chart of a method for predicting soil stratum distribution provided in an embodiment of the present application. Figure 1 As shown in , the prediction method provided in the embodiment of the present application includes:

[0057] S101 , for each target location point in the soil area to be predicted, determining a plurality of adjacent drill hole points closest to the target location point from a drill hole set.

[0058] Regarding step S101 above, during the specific implementation, for each target location within the soil area to be predicted, a spatial neighborhood strategy is employed during the feature extraction process for each target location, selecting the data of multiple neighboring boreholes in the borehole set that are closest to the target location as feature input. In this way, the present application selects the boreholes closest to the target location, improving the subsequent model's ability to capture local geological features, adapting to unevenly distributed borehole data, and enhancing the model's adaptability and generalization capabilities in complex geological environments.

[0059] As an optional embodiment, with respect to the above step S101, determining a plurality of adjacent drilling points closest to the target location point from the drilling set includes:

[0060] Step 1011: for each drilling point to be screened in the drilling set, determine the spatial position coordinates of the drilling point to be screened relative to the target position point, and calculate the Euclidean distance between the drilling point to be screened and the target position point using the spatial position coordinates.

[0061] Regarding the above step 1011, in the specific implementation, for each drilling point to be screened in the drilling set, the spatial position coordinates of the drilling point to be screened relative to the target position point are determined, and the Euclidean distance between the drilling point to be screened and the target position point is calculated using the spatial position coordinates. Specifically, a target position point P is set, and the drilling set It is expressed by the following formula:

[0062] ={B1(x1,y1),B2(x2,t2),…,B n (x n ,y n )}

[0063] Among them, (x i ,y i ) represents a drilling set The spatial position coordinates of the lth drilling point to be screened relative to the target position point P.

[0064] Drilling point B to be screened i The Euclidean distance from the target position point P is calculated using the following formula:

[0065]

[0066] Among them, d(P,B i ) represents a drilling set The Euclidean distance between the i-th drilling point to be screened and the target position point P, (x0, y0) represents the position coordinates of the target position point P.

[0067] Step 1012: sort the multiple drilling points to be screened from smallest to largest distance using the Euclidean distance corresponding to each drilling point to be screened, and select a preset number of drilling points to be screened from the sorted order as the adjacent drilling points.

[0068] Regarding step 1012, in a specific implementation, the multiple drilling points to be screened are sorted in ascending order of distance using the Euclidean distance corresponding to each drilling point to be screened. A preset number of drilling points to be screened are then selected from the sorted order as neighboring drilling points. Here, according to the example provided in this application, the preset number is 5, meaning that the five drilling points to be screened that are closest to the target location are selected as neighboring drilling points. Specifically, the five neighboring drilling points sorted by distance are represented by the following formula:

[0069]

[0070] Among them, k is a numbered item, The five adjacent drill points are sorted by distance.

[0071] See also Figure 2 , Figure 2 This is a schematic diagram of the spatial distribution of drilling points and neighborhood selection strategy provided in an embodiment of the present application. Figure 2The figure shows the spatial distribution of boreholes within the predicted soil area and the distance distribution characteristics of neighboring boreholes. Black dots represent all target locations, hollow circles represent the boreholes in the set, and gray lines connect the five nearest neighbors of the target location. The inset shows a histogram of the distance distribution of neighboring boreholes. The distance distribution is primarily concentrated in the 20-60m range, with the longest distance reaching 89m.

[0072] S102, obtaining stratum category information corresponding to each adjacent drilling point, and generating a feature matrix corresponding to the target position point using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point.

[0073] Here, according to the embodiment provided in this application, there are nine preset stratum categories, namely, plain fill, silt, silty fine sand, silty soil, silty clay, medium sand, residual soil silty clay, strongly weathered muddy siltstone, and moderately weathered muddy siltstone. Each preset stratum category is mapped to a 9-dimensional vector. Figure 3 , Figure 3 A schematic diagram of a 9-dimensional vector conversion of a stratum based on One-hot encoding is provided in an embodiment of the present application. Figure 3 As shown, each stratum corresponds to a set of vector data, which is one-hot encoded according to the stratum category. One-hot encoding is used to convert the category information of different strata into a standardized matrix form. The nine strata are mapped to 9-dimensional vectors. Optionally, the nine stratum categories can be defined differently according to the actual scenario, as long as different strata are defined as unique 9-dimensional vectors. One-hot encoding is performed on multiple stratum categories, and the category information is mapped to multidimensional vectors to ensure the model's ability to distinguish different stratum types.

[0074] Regarding the above-mentioned step S102, in the specific implementation, after multiple adjacent drilling points are determined in step S101, the stratigraphic category information corresponding to each adjacent drilling point is obtained, and then the multidimensional vector corresponding to each preset stratigraphic category and the stratigraphic category information corresponding to each adjacent drilling point are used to generate a feature matrix corresponding to the target position point.

[0075] As an optional embodiment, with respect to the above step S102, the step of generating a feature matrix corresponding to the target location point using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point includes:

[0076] Step 1021 : gridding the stratum distribution of the soil area to be predicted to form a high-resolution stratum distribution matrix.

[0077] For the above-mentioned step 1021, when it is specifically implemented, the stratigraphic distribution of the soil area to be predicted is gridded using high-precision resolution to form a high-resolution stratigraphic distribution matrix. Here, according to the example provided in this application, the stratigraphic distribution of each borehole is gridded using a depth interval of 0.1m, and the grid interval representing the depth of high resolution is 0.1m. The specific method is to resample the soil layer information from the surface of the soil area to be predicted to the maximum drilling depth into equally spaced depth data points (divided by 0.1m) to form a high-resolution stratigraphic distribution matrix. In this way, the depth distribution of the stratum is refined using high-resolution depth data. This high-resolution data enables the model to more accurately capture the changing characteristics of the stratum in the depth direction. High resolution not only improves the model's ability to capture subtle stratigraphic changes and improves the model's ability to recognize thin layer structures and boundary areas, but also makes its predictions in boundary areas more accurate, reducing the misclassification of adjacent strata. Especially in the transition area, the model can more carefully depict the continuity and gradual characteristics of the stratum, reducing the misclassification of adjacent strata. The results show that the developed model has significant advantages in extracting features of most types of formations and learning their spatial distribution, and can provide reliable support for the automated prediction of complex formations.

[0078] Step 1022, for each adjacent drilling point, based on the stratum category information corresponding to the adjacent drilling point, determine the initial feature code corresponding to the adjacent drilling point from the multidimensional vectors corresponding to multiple preset stratum categories, and determine the depth information corresponding to the adjacent drilling point from the stratum distribution matrix.

[0079] Regarding step 1022, in a specific implementation, for each adjacent borehole point, based on the stratum type information corresponding to the adjacent borehole point, the initial feature code corresponding to the adjacent borehole point is determined from the multidimensional vectors corresponding to the multiple preset stratum types. Here, as an example, when the stratum type information corresponding to the adjacent borehole point is plain fill, the multidimensional vector corresponding to the plain fill is the initial feature code corresponding to the adjacent borehole point. Then, the depth information z corresponding to the adjacent borehole point is determined from the stratum distribution matrix. i .

[0080] Step 1023: The spatial position coordinates of the neighboring drilling point relative to the target position point, the Euclidean distance between the drilling point to be screened and the target position point, the depth information corresponding to the neighboring drilling point, and the initial feature code corresponding to the neighboring drilling point are spliced ​​to generate the target feature code corresponding to the neighboring drilling point.

[0081] Regarding the above step 1023, in the specific implementation, the spatial position coordinates (x i ,y i), the Euclidean distance d between the drilling point to be screened and the target position point (P, B i ), the depth information z corresponding to the adjacent drilling point i The initial feature codes corresponding to the adjacent drilling points are spliced ​​together to generate target feature codes for splicing the adjacent drilling points in rows.

[0082] Step 1024 : Generate a feature matrix corresponding to the target location point using the target feature codes corresponding to the multiple adjacent drilling points.

[0083] Regarding the above step 1024, in specific implementation, after the target feature code corresponding to each adjacent drilling point is constructed, the target feature code corresponding to multiple adjacent drilling points is used to generate the feature matrix corresponding to the target position point. Figure 4 , Figure 4 This is a schematic diagram of a characteristic matrix data format and a stratum distribution prediction model architecture provided in an embodiment of the present application. Figure 4 As shown, continuing with the above embodiment, when the multidimensional vector corresponding to the preset stratigraphic category is a 9-dimensional vector, a 5×13 feature matrix can be constructed by selecting data from five adjacent borehole points as the feature input of the stratigraphic distribution prediction model. In this way, the present application constructs a multi-dimensional, multi-level stratigraphic feature matrix by combining the spatial position of adjacent boreholes, stratigraphic depth, and one-hot coding features of stratigraphic categories. This feature matrix can capture the spatial correlation of geological sequences and improve the classification accuracy of subsequent models.

[0084] S103: Input the feature matrix into a pre-trained stratum distribution prediction model to determine a stratum distribution prediction result corresponding to the target location point, and determine the target stratum category to which the target location point belongs based on the stratum distribution prediction result.

[0085] For the above-mentioned step S103, during the specific implementation, the feature matrix is ​​input into the pre-trained stratum distribution prediction model to determine the stratum distribution prediction result corresponding to the target location point. The stratum prediction distribution result represents the probability that the target location point belongs to each preset stratum category, and the target stratum category to which the target location point belongs is determined based on the stratum distribution prediction result, that is, the preset stratum category with the highest probability in the stratum prediction distribution result is used as the target stratum category to which the target location point belongs.

[0086] According to the embodiment provided in this application, the stratum distribution prediction model in this application is a CNN model structure for shallow geological stratum prediction. Figure 4 The stratigraphic distribution prediction model architecture shown in the figure has a basic structure of CNN model, which includes input layer, convolution layer, pooling layer, fully connected layer and output layer. The specific design is as follows:

[0087] The input layer receives the preprocessed 5×13 feature matrix as input; the input feature matrix is ​​in the form of:

[0088]

[0089] Among them, X is the input feature matrix, and C, H, and W are the sizes of the matrix.

[0090] The stratum distribution prediction model provided in this application is designed with a total of 5 convolution layers. Each layer extracts features through a 3×3 convolution kernel and uses a Leaky ReLU (ReLU with a negative slope) activation function, which can retain small-scale negative information and help avoid the gradient vanishing problem. The output of each convolution layer can be expressed as follows:

[0091]

[0092] Here, Y represents the output features of each convolutional layer. It is the result of convolution, bias addition, and activation function. X represents the input feature map of each layer, which is the output of the previous layer or the initial input data. c represents the index of the input channel. k represents the index of the output channel. For example, if the convolutional layer has 16 convolution kernels, 16 output channels are generated accordingly. b represents the bias value, which is used to adjust the base value of the convolution result and enhance the expressiveness of the model. f is the activation function (Leaky ReLU activation function), which is used to introduce nonlinear mapping, enabling the model to learn complex features.

[0093] The Leaky ReLU activation function is shown below:

[0094]

[0095] The high-dimensional features output by the convolutional layer are further reduced in dimension through the global average pooling (GAP) layer, compressing the feature values ​​of each channel into a scalar. Global pooling is used to reduce the number of parameters while retaining the most significant global features. The output of the pooling layer can be expressed as follows:

[0096]

[0097] Finally, the fully connected layer is designed as two layers, mapping the extracted features to the target output. The first layer compresses the features to 32 dimensions, and the second layer maps them to the final nine stratigraphic categories. The category with the largest predictive factor in the final output is used as the model's stratigraphic prediction result.

[0098] Furthermore, according to the embodiment provided by the present application, the training data used by the stratum distribution prediction model in the training process is obtained after expansion by the SMOTE oversampling method. In the stratum prediction task, the uneven distribution of stratum categories is one of the main challenges of model training. The data samples of certain strata (such as silty fine sand, medium sand, etc.) are relatively scarce, while the data samples of other strata (such as silt) account for too large a proportion, which may cause the model to perform well for most types of strata during training, but have poor prediction capabilities for a few types of strata, thereby affecting the overall performance of the model. To solve this problem, the present application introduces the SMOTE technology, which balances the data distribution of each stratum category by generating adjacent synthetic samples of real stratum data in the feature space.

[0099] According to the embodiment provided in this application, the stratum distribution prediction model is trained by the following steps:

[0100] A: A test set is determined from the original drilling data set, and the remaining data in the original drilling data set is used as a training set and a validation set.

[0101] In the specific implementation of step A above, the samples in the original drillhole dataset are partitioned to determine a test set. The remaining data in the original drillhole dataset serves as the training and validation sets. During the data partitioning process, to ensure model generalization and training fairness, 20% of the total drillhole data is selected as an independent test set. This test set is used to evaluate the generalization ability of the final model and does not participate in the model training and validation steps below. This test set is used to evaluate the generalization ability of the final model.

[0102] B: Use the five-fold cross-validation method to divide the remaining data in the original drilling data set to obtain a training set and a validation set.

[0103] Here, 5-fold cross-validation is a commonly used model evaluation method. Its core idea is to more reliably evaluate the generalization performance of the model through multiple data partitioning and model validation.

[0104] For step B above, in the specific implementation, the remaining data is divided into 5 subsets using the five-fold cross-validation method. The remaining data is evenly divided into 5 subsets for cross-validation. During the cross-validation process, one of the subsets is selected as the validation set each time, and the remaining 4 subsets are used as training sets. This method not only maximizes the utilization of the data, but also effectively detects the stability and effectiveness of the model in different data partitions. During the data partitioning process, special attention is paid to ensuring the uniformity of the spatial distribution of each data set. Avoid a data set that only contains data from a certain concentrated area, so as to ensure that the stratigraphic data in training and testing can fully represent the geological distribution of the entire study area.

[0105] C: Perform SMOTE oversampling on the minority class samples in the training set to obtain an expanded target training set.

[0106] Here, the basic principle of SMOTE oversampling is to generate new samples by interpolating the features of minority class samples, thereby increasing the number of minority class samples and achieving the purpose of balancing class distribution.

[0107] Regarding step C above, during the specific implementation, in each cross-validation, SMOTE oversampling is performed on the minority class samples in the training set to obtain an expanded target training set. In this way, by introducing SMOTE technology, this application can generate synthetic minority class samples, effectively balancing the sample distribution of various strata, thereby avoiding training bias caused by data imbalance and improving the model's prediction accuracy for minority classes.

[0108] Specifically, with respect to the above step C, performing SMOTE oversampling on the minority class samples in the training set to obtain an expanded target training set includes:

[0109] a: Obtain minority class samples in the training set, and determine multiple neighbor samples corresponding to each minority class sample.

[0110] b: For each minority class sample, a target neighbor sample is determined from multiple neighbor samples corresponding to the minority class sample, and linear interpolation is performed between the minority class sample and the target neighbor sample to generate an expanded sample to obtain the target training set.

[0111] In the specific implementation of steps a-b above, first identify the minority class samples, determine the minority class samples in the training set, and determine the k neighbor samples of each minority class sample in the training set. Then, for each minority class sample, randomly select one of the k neighbor samples corresponding to the minority class sample as the target neighbor sample. The extended sample is generated by linear interpolation between the minority class sample and the target neighbor sample, as shown in the following formula:

[0112] p new =pi +λ·(p neighbor -p i )

[0113] Among them, p i is a minority class sample; p neighbor is its target neighbor sample; λ is a random number in [0,1].

[0114] In this way, the minority class samples in the training set are expanded to obtain the expanded target training set. The synthetic minority class oversampling technique (SMOTE) is used to generate synthetic samples of scarce classes (such as silt and highly weathered muddy siltstone) to balance the stratigraphic dataset. This addresses the class imbalance issue, improves the model's predictive ability for minority classes, and reduces training bias.

[0115] D: respectively using the target training set and the validation set to train and validate the original prediction model of stratum distribution until convergence.

[0116] E: Using the test set to test the converged original stratum distribution prediction model, the converged original stratum distribution prediction model that passes the test is used as the stratum distribution prediction model.

[0117] Here, as an example, during the model optimization process, we conducted multiple rounds of pre-training experiments to observe the impact of various parameters on model performance, ultimately determining the following parameter configuration: the learning rate was set to 0.0001; the batch size was set to 4096; the number of training rounds was set to 200; the convolution kernel size of each neural network layer was set to 3×3, and 16 to 256 convolution kernels were used in each layer.

[0118] For the above steps D-E, in the specific implementation, the training set is divided into five equal parts, one part is selected as the validation set each time, and the remaining four parts are used as the training set. The training and verification process is repeated for a total of five rounds of training, and the performance index is finally calculated. By comparing the results of cross-validation, the final converged stratigraphic distribution original prediction model is determined. The converged stratigraphic distribution original prediction model is tested in the test set, and the stratigraphic distribution original prediction model that has passed the test is used as the stratigraphic distribution prediction model. Here, for the classification problem of multiple types of non-uniform samples, this application uses three indicators, namely precision, recall and F1 score, to comprehensively evaluate the performance of the model. The formula of the evaluation index is as follows:

[0119]

[0120] See also Figure 5 , Figure 5Schematic diagram of the loss curve and evaluation index curve of a CNN model in the cross-validation process provided by the embodiment of the present application. Figure 5 As shown in Figure 2, the loss function curve and the changes of each evaluation index with the number of training iterations. Figure 5 It can be seen that the difference between the loss of the training set and the validation set is always kept within a certain range, indicating that the model has not experienced overfitting or underfitting during the training process. In addition, the steady decline in the loss of the validation set further proves that the model structure and training strategy can effectively adapt to the data characteristics and have high stability under different data segmentation conditions. In addition, the evaluation indicators also quickly stabilized. In the five-fold cross-validation, the values ​​of the three evaluation indicators were all stable at around 0.7-0.75, and the precision and recall rates were almost the same. This shows that the developed model has achieved a good balance in the ability to correctly identify strata and reduce false alarms, and has good predictive ability for the multi-classification complex sample problems in this application.

[0121] The soil stratum distribution prediction method provided in the embodiment of the present application first determines, for each target position point in the soil area to be predicted, a plurality of adjacent borehole points that are closest to the target position point from the borehole set; then, obtains the stratum category information corresponding to each adjacent borehole point, and uses the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent borehole point to generate a feature matrix corresponding to the target position point; finally, inputs the feature matrix into a pre-trained stratum distribution prediction model to determine the stratum distribution prediction result corresponding to the target position point, and determines the target stratum category to which the target position point belongs based on the stratum distribution prediction result; wherein the stratum prediction distribution result represents the probability that the target position point belongs to each preset stratum category, and the training data used by the stratum distribution prediction model in the training process is obtained by expanding the data using the SMOTE oversampling method.

[0122] This application introduces feature extraction of neighboring borehole data, which is applicable to sites containing multiple stratigraphic categories. By combining multiple technologies of model design and data processing, efficient prediction of shallow complex strata is achieved. By selecting data from neighboring borehole points and combining the spatial neighborhood strategy, multiple borehole points closest to the target point are selected as input features, so that the model can effectively capture the local geological characteristics of the target location and can flexibly adapt to unevenly distributed borehole data. Encoding multiple stratigraphic categories and mapping them into multidimensional vectors can ensure that the model can better identify and distinguish complex multi-classified stratigraphic types. For rare stratigraphic categories in the model training data set, the synthetic minority oversampling technique (SMOTE) is used to expand the minority sample data. This application is applicable to complex site conditions containing multiple stratigraphic categories. Only multiple neighboring borehole points are needed to quickly predict the stratigraphic distribution of any unknown point within the range, providing technical support for refined geological modeling. It has improved its prediction accuracy, boundary recognition ability and the ability to identify a few types of strata, increased the automation level of stratum modeling, reduced manual intervention, and improved exploration accuracy and efficiency.

[0123] See also Figure 6 , Figure 6 This is a schematic diagram of the structure of a soil stratum distribution prediction device provided in an embodiment of the present application. Figure 6 As shown in , the prediction device 600 includes:

[0124] A drilling point determination module 601 is configured to determine, for each target location point in the soil area to be predicted, a plurality of adjacent drilling points that are closest to the target location point from the drilling set;

[0125] The feature matrix generation module 602 is used to obtain the stratum category information corresponding to each adjacent drilling point, and generate the feature matrix corresponding to the target location point using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point;

[0126] The stratum category prediction module 603 is used to input the feature matrix into a pre-trained stratum distribution prediction model, determine the stratum distribution prediction result corresponding to the target location point, and determine the target stratum category to which the target location point belongs based on the stratum distribution prediction result; wherein the stratum prediction distribution result represents the probability that the target location point belongs to each preset stratum category, and the training data used by the stratum distribution prediction model in the training process is obtained after expansion through the SMOTE oversampling method.

[0127] Furthermore, when the drilling point determination module 601 is used to determine a plurality of adjacent drilling points closest to the target location point from the drilling set, the drilling point determination module 601 is further used to:

[0128] For each drill hole point to be screened in the drill hole set, determining the spatial position coordinates of the drill hole point to be screened relative to the target position point, and calculating the Euclidean distance between the drill hole point to be screened and the target position point using the spatial position coordinates;

[0129] The multiple drilling points to be screened are sorted from small to large distance using the Euclidean distance corresponding to each drilling point to be screened, and a preset number of drilling points to be screened are selected from the sorting as the adjacent drilling points.

[0130] Furthermore, when the feature matrix generation module 602 is used to generate the feature matrix corresponding to the target location point using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point, the feature matrix generation module 602 is further used to:

[0131] Gridding the stratum distribution of the soil area to be predicted to form a high-resolution stratum distribution matrix;

[0132] For each adjacent drilling point, based on the stratum category information corresponding to the adjacent drilling point, an initial feature code corresponding to the adjacent drilling point is determined from the multidimensional vectors corresponding to multiple preset stratum categories, and depth information corresponding to the adjacent drilling point is determined from the stratum distribution matrix;

[0133] The spatial coordinates of the neighboring drilling point relative to the target location point, the Euclidean distance between the drilling point to be screened and the target location point, the depth information corresponding to the neighboring drilling point, and the initial feature code corresponding to the neighboring drilling point are spliced ​​to generate a target feature code corresponding to the neighboring drilling point;

[0134] The target feature codes corresponding to multiple adjacent drilling points are used to generate a feature matrix corresponding to the target position point.

[0135] Furthermore, the prediction device 600 further includes a model training module, which is used to train the stratum distribution prediction model through the following steps:

[0136] Determine a test set from the original drilling data set, and use the remaining data in the original drilling data set as a training set and a validation set;

[0137] Using a five-fold cross-validation method to divide the remaining data in the original drilling data set into a training set and a validation set;

[0138] SMOTE oversampling on the minority class samples in the training set to obtain an expanded target training set;

[0139] training and verifying the stratigraphic distribution original prediction model by using the target training set and the verification set respectively until convergence is achieved;

[0140] testing the converged stratigraphic distribution original prediction model by using the test set, and taking the converged stratigraphic distribution original prediction model that passes the test as the stratigraphic distribution prediction model.

[0141] Further, when the model training module is used to perform SMOTE oversampling on the minority class samples in the training set to obtain an expanded target training set, the model training module is further used to:

[0142] obtain the minority class samples in the training set and determine a plurality of neighbor samples corresponding to each minority class sample;

[0143] For each minority class sample, determine a target neighbor sample from the plurality of neighbor samples corresponding to the minority class sample, and generate an expanded sample by performing linear interpolation between the minority class sample and the target neighbor sample to obtain the target training set.

[0144] Please refer to Figure 7 , Figure 7 A structural schematic diagram of an electronic device provided by an embodiment of the present application. As shown in Figure 7 , the electronic device 700 includes a processor 710, a memory 720, and a bus 730.

[0145] The memory 720 stores machine readable instructions executable by the processor 710. When the electronic device 700 is running, the processor 710 and the memory 720 communicate with each other through the bus 730. The machine readable instructions, when executed by the processor 710, can perform the steps of the prediction method of soil stratigraphic distribution in the method embodiment as described above Figure 1 . For specific implementation manners, refer to the method embodiment, which will not be described here.

[0146] The present application also provides a computer readable storage medium having a computer program stored thereon. When the computer program is run by a processor, the steps of the prediction method of soil stratigraphic distribution in the method embodiment as described above Figure 1 . For specific implementation manners, refer to the method embodiment, which will not be described here.

[0147] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0148] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. There may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some communication interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0149] The units described as separate components may or may not be physically separate, and 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0150] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0151] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0152] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The scope of protection of the present application is not limited thereto. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-mentioned embodiments within the technical scope disclosed in the present application, or perform equivalent replacements for some of the technical features thereof. These modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for predicting soil stratum distribution, characterized in that: The prediction method comprises: For each target location point in the soil area to be predicted, determine a plurality of adjacent drill hole points closest to the target location point from the drill hole set; Obtaining stratum category information corresponding to each adjacent drilling point, and using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point to generate a feature matrix corresponding to the target location point; Inputting the feature matrix into a pre-trained stratigraphic distribution prediction model to determine a stratigraphic distribution prediction result corresponding to the target location point, and determining the target stratigraphic category to which the target location point belongs based on the stratigraphic distribution prediction result; wherein the stratigraphic distribution prediction result represents the probability that the target location point belongs to each preset stratigraphic category, and the training data used in the stratigraphic distribution prediction model during training is obtained by expanding the model using the SMOTE oversampling method; The step of determining a plurality of adjacent drilling points closest to the target location point from the drilling set includes: For each drill hole point to be screened in the drill hole set, determining the spatial position coordinates of the drill hole point to be screened relative to the target position point, and calculating the Euclidean distance between the drill hole point to be screened and the target position point using the spatial position coordinates; sorting the plurality of drilling points to be screened in ascending order of distance using the Euclidean distance corresponding to each drilling point to be screened, and selecting a preset number of drilling points to be screened from the sorting as the adjacent drilling points; The method of generating a feature matrix corresponding to the target location point by using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point includes: Gridding the stratum distribution of the soil area to be predicted to form a high-resolution stratum distribution matrix; For each adjacent drilling point, based on the stratum category information corresponding to the adjacent drilling point, an initial feature code corresponding to the adjacent drilling point is determined from the multidimensional vectors corresponding to multiple preset stratum categories, and depth information corresponding to the adjacent drilling point is determined from the stratum distribution matrix; The spatial coordinates of the neighboring drilling point relative to the target location point, the Euclidean distance between the drilling point to be screened and the target location point, the depth information corresponding to the neighboring drilling point, and the initial feature code corresponding to the neighboring drilling point are spliced ​​to generate a target feature code corresponding to the neighboring drilling point; The target feature codes corresponding to multiple adjacent drilling points are used to generate a feature matrix corresponding to the target position point.

2. The prediction method according to claim 1, characterized in that The stratum distribution prediction model is trained by the following steps: Determine a test set from the original drilling data set, and use the remaining data in the original drilling data set as a training set and a validation set; Using a five-fold cross-validation method to divide the remaining data in the original drilling data set into a training set and a validation set; Perform SMOTE oversampling on the minority class samples in the training set to obtain an expanded target training set; The target training set and the validation set are used to train and validate the original prediction model of stratum distribution respectively until convergence; The converged original stratum distribution prediction model is tested using the test set, and the converged original stratum distribution prediction model that passes the test is used as the stratum distribution prediction model.

3. The prediction method according to claim 2, characterized in that The SMOTE oversampling of the minority class samples in the training set to obtain an expanded target training set includes: Obtaining minority class samples in the training set, and determining a plurality of neighbor samples corresponding to each minority class sample; For each minority class sample, a target neighbor sample is determined from a plurality of neighbor samples corresponding to the minority class sample, and linear interpolation is performed between the minority class sample and the target neighbor sample to generate an expanded sample to obtain the target training set.

4. A device for predicting soil stratum distribution, characterized in that: The prediction device comprises: A drilling point determination module is used to determine, for each target location point in the soil area to be predicted, a plurality of adjacent drilling points that are closest to the target location point from the drilling set; A feature matrix generation module is used to obtain the stratum category information corresponding to each adjacent drilling point, and use the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point to generate a feature matrix corresponding to the target location point; a stratum category prediction module, configured to input the feature matrix into a pre-trained stratum distribution prediction model, determine a stratum distribution prediction result corresponding to the target location point, and determine the target stratum category to which the target location point belongs based on the stratum distribution prediction result; wherein the stratum distribution prediction result represents the probability that the target location point belongs to each preset stratum category, and the training data used in the training process of the stratum distribution prediction model is obtained by expanding the stratum distribution prediction model using the SMOTE oversampling method; When the drilling point determination module is used to determine a plurality of adjacent drilling points closest to the target location point from the drilling set, the drilling point determination module is further used to: For each drill hole point to be screened in the drill hole set, determining the spatial position coordinates of the drill hole point to be screened relative to the target position point, and calculating the Euclidean distance between the drill hole point to be screened and the target position point using the spatial position coordinates; sorting the plurality of drilling points to be screened in ascending order of distance using the Euclidean distance corresponding to each drilling point to be screened, and selecting a preset number of drilling points to be screened from the sorting as the adjacent drilling points; When the feature matrix generation module is used to generate the feature matrix corresponding to the target location point using the multidimensional vector corresponding to each preset stratum category and the stratum category information corresponding to each adjacent drilling point, the feature matrix generation module is further used to: Gridding the stratum distribution of the soil area to be predicted to form a high-resolution stratum distribution matrix; For each adjacent drilling point, based on the stratum category information corresponding to the adjacent drilling point, an initial feature code corresponding to the adjacent drilling point is determined from the multidimensional vectors corresponding to multiple preset stratum categories, and depth information corresponding to the adjacent drilling point is determined from the stratum distribution matrix; The spatial coordinates of the neighboring drilling point relative to the target location point, the Euclidean distance between the drilling point to be screened and the target location point, the depth information corresponding to the neighboring drilling point, and the initial feature code corresponding to the neighboring drilling point are spliced ​​to generate a target feature code corresponding to the neighboring drilling point; The target feature codes corresponding to multiple adjacent drilling points are used to generate a feature matrix corresponding to the target position point.

5. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus. When the processor is running, the machine-readable instructions execute the steps of the method for predicting soil stratum distribution as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for predicting soil stratum distribution according to any one of claims 1 to 3 are executed.

Images