Hidden fracture prediction method based on multi-source detection data fusion and related device

Abstract

The invention discloses a blind fracture prediction method based on multi-source detection data fusion and a related device, and the method comprises the steps: carrying out the detection processing in a designated region, and obtaining gravity detection data and magnetic detection data; carrying out fusion processing on the gravity detection data and the magnetic detection data; performing enhancement processing on the multi-source fusion detection data; performing data set construction processing on the enhanced multi-source fusion detection data, and performing labeling processing on the constructed data set to form a training sample data set; inputting the training sample data set into a preset machine learning model for training processing to obtain a training convergent machine learning model; and applying the training convergence machine learning model to an unknown fracture target area to carry out blind fracture prediction processing, and generating a blind fracture prediction distribution map covering the unknown fracture target area. According to the embodiment of the invention, the dependence on high-end equipment and professionals is reduced, and the recognition capability of the hidden fracture is greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a hidden fracture prediction method based on multi-source detection data fusion and a related device. Background Art

[0002] Traditional geological survey methods face many challenges in this area, including complex terrain, thick overburden, and surface vegetation cover, which make fault identification difficult and inefficient. With the development of remote sensing technology and geophysical methods, especially the widespread application of gravity and magnetic data in geological research, new approaches have been provided for fault identification and the study of regional geological characteristics.

[0003] The study of regional geological characteristics and deep geological structures is a basic and highly applied geological survey work, which is of great significance to social and economic development. This research plays a key role in various aspects such as lithology identification and classification, tectonic evolution, mineral prediction and development, and exploration of deep geological conditions. Gravity and magnetic data reflect different aspects of the geological characteristics of a region. By superimposing gravity and magnetic data through a comprehensive method, the geological characteristic information is effectively highlighted, providing a basis for regional comprehensive geological analysis. In the past, the interpretation of gravity and magnetic data mainly adopted the method of analyzing Bouguer gravity anomalies and aeromagnetic anomalies separately, and then combined with geological conditions for manual interpretation, and the results of the analysis were multi-solution. Or a variety of geophysical data such as gravity and magnetism were used for joint inversion. The traditional analysis method was used to conduct relevant research on gravity anomalies and magnetic anomalies, and there was a lack of methods for comprehensive analysis of gravity and magnetic interpretation with geological data, resulting in the failure to effectively mine some geological characteristics, laws, and genetic mechanisms. The development of big data analysis capabilities and artificial intelligence technology and the enhancement of computing power have provided new technical means for intelligent mining of geological information from gravity and magnetic data, and provided new opportunities for intelligent interpretation of gravity and magnetic data.

[0004] At present, machine learning technology allows computers to recognize the rules contained in the data from the training samples (i.e. generate models) and identify or predict new sample inputs. It has good processing capabilities for complex and nonlinear problems and is an effective means of data classification and regression prediction. In the era of big data, machine learning provides a data thinking that "emphasizes correlation and neglects causality" for data, opening up a new space for scientific research. Machine learning starts from data and analyzes the correlation of multi-source and heterogeneous data, thereby mining the hidden relationships behind complex data, which is of great significance for mining the hidden value of geological data. In recent years, machine learning has been active in many fields, and some studies have also used machine learning technology to solve problems in the field of earth science. Such as earthquake prediction, lithology classification and mapping, remote sensing information extraction, quantitative prediction of mineral resources, identification and extraction of geochemical anomalies and other earth science fields.

[0005] At present, machine methods have become a hot topic in the field of geology. How to use intelligent technologies and methods to process geophysical comprehensive exploration data, analyze and predict regional geological characteristics and deep geological structures, and improve the efficiency and accuracy of gravity and magnetic anomaly interpretation methods is one of the hot topics in current research. The key scientific issues are: (1) How to process gravity and magnetic data in view of the weak geophysical signals in the coverage area, and perform data enhancement and diversity characterization on gravity and magnetic information; (2) How to design hyperparameters for machine learning to achieve the best model effect; (3) The interpretation of faults is not the more features the better, how to select the best features from the data. (4) How to express small faults in a refined manner. Based on this, gravity and magnetic data were collected on the basis of previous work; after data enhancement, the data were integrated to establish a regional spatial geographic information database. The known faults are grouped and the existing data sets are classified and labeled. The training set and the test set are divided according to a certain ratio. The random forest (RF) and artificial neural network (ANNs) algorithms are used for training, and the hyperparameters of the RF and ANNs parameters are optimized. The selected model is applied to the coverage area to predict the unknown faults in the coverage area and obtain the results of the fault interpretation in the coverage area. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the prior art. The present invention provides a hidden fracture prediction method and related devices based on multi-source detection data fusion, which reduces the dependence on high-end equipment and professionals and greatly improves the ability to identify hidden fractures.

[0007] In order to solve the above technical problems, an embodiment of the present invention provides a hidden fracture prediction method based on multi-source detection data fusion, the method comprising: Based on the gravity instrument and the magnetic instrument, detection and processing are performed in a designated area to obtain gravity detection data and magnetic detection data, wherein the designated area is a known fracture area, the gravity detection data is Bouguer gravity anomaly data, and the magnetic detection data is aeromagnetic anomaly data; Fusing the gravity detection data and the magnetic detection data to form multi-source fused detection data; Performing enhancement processing on the multi-source fusion detection data based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data; The enhanced multi-source fusion detection data is processed into a data set, and the constructed data set is annotated to form a training sample data set; Inputting the training sample data set into a preset machine learning model for training processing to obtain a machine learning model with training convergence, wherein the machine learning model includes a random forest model and an artificial neural network model; The machine learning model that has converged in training is applied to the unknown fracture target area for hidden fracture prediction processing, the prediction results are output in the form of spatial coordinates, and a hidden fracture prediction distribution map covering the unknown fracture target area is generated.

[0008] Optionally, fusing the gravity detection data and the magnetic detection data to form multi-source fused detection data includes: Performing abnormal data filtering processing on the gravity detection data and the magnetic detection data to obtain filtered gravity detection data and filtered magnetic detection data; The filtered gravity detection data and the filtered magnetic detection data are smoothed and compensated in sequence to obtain compensated gravity detection data and compensated magnetic detection data; The compensated gravity detection data and the compensated magnetic detection data are fused based on the data fusion algorithm to obtain multi-source fused detection data.

[0009] Optionally, the performing enhancement processing on the multi-source fusion detection data based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data includes: Use a variety of data enhancement algorithms to enhance the geological structure in the multi-source fusion detection data to obtain enhanced multi-source fusion detection data; The multiple data enhancement algorithms include total horizontal derivative method, vertical derivative method, vertical second-order derivative, tilt angle method, tilt angle horizontal derivative method, analytical signal method, Theat method, total horizontal derivative tilt angle method, horizontal tilt angle method, enhanced analytical signal method, and enhanced tilt angle method.

[0010] Optionally, the data set construction process is performed on the enhanced multi-source fusion detection data, and the constructed data set is annotated to form a training sample data set, including: Performing sample synthesis processing on the enhanced multi-source fusion detection data based on a synthetic minority class oversampling algorithm, and performing data set construction processing based on the synthesized sample data to obtain a constructed data set; The existing fracture distribution map and geological data of the designated area are obtained, and the constructed data set is annotated based on the existing fracture distribution map and geological data to form a training sample data set.

[0011] Optionally, the sample synthesis processing of the enhanced multi-source fusion detection data based on the synthetic minority class oversampling algorithm includes: For each minority class sample data X in the enhanced multi-source fusion detection data, the minority class sample X and its nearest neighboring sample data are calculated according to the required oversampling amount. The difference between is used to synthesize new samples, which can be expressed as: ; in, represents a random function; Represents synthetic sample data.

[0012] Optionally, the step of inputting the training sample data set into a preset machine learning model for training processing to obtain a machine learning model with converged training includes: When the training sample data set is input into a preset machine learning model for training processing, the hyperparameters in the preset machine learning model are adjusted by a grid search algorithm or a genetic optimization algorithm, and the best parameter combination is selected for training processing to obtain a machine learning model with training convergence; When the preset machine learning model is a random forest model, the hyperparameters are the number of decision trees in the random forest model, the maximum depth of the decision tree, the minimum number of samples contained in a leaf node, and the minimum number of samples that can be divided into nodes; when the preset machine learning model is an artificial neural network model, the hyperparameters are the number of layers and nodes of the artificial neural network model.

[0013] Optionally, the prediction results are output in the form of spatial coordinates, and a hidden fracture prediction distribution map covering the unknown fracture target area is generated, including: The prediction results are output in the form of spatial coordinates, and the prediction results output in the form of spatial coordinates are visualized and analyzed in combination with geographic information system software to generate a hidden fault prediction distribution map covering the unknown fault target area.

[0014] In addition, an embodiment of the present invention further provides a hidden fracture prediction device based on multi-source detection data fusion, the device comprising: Detection module: used to perform detection processing in a specified area based on gravity instruments and magnetic instruments to obtain gravity detection data and magnetic detection data. The specified area is a known fracture area, the gravity detection data is Bouguer gravity anomaly data, and the magnetic detection data is aeromagnetic anomaly data; Data fusion module: used for fusing the gravity detection data and the magnetic detection data to form multi-source fusion detection data; Data enhancement module: used to enhance the multi-source fusion detection data based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data; Data construction and annotation module: used to construct and process the data set of the enhanced multi-source fusion detection data, and to annotate the constructed data set to form a training sample data set; Model training module: used to input the training sample data set into a preset machine learning model for training processing to obtain a machine learning model with training convergence, wherein the machine learning model includes a random forest model and an artificial neural network model; Prediction generation module: used to apply the trained converged machine learning model to the unknown fracture target area for hidden fracture prediction processing, output the prediction results in the form of spatial coordinates, and generate a hidden fracture prediction distribution map covering the unknown fracture target area.

[0015] In addition, an embodiment of the present invention further provides an electronic device, including a processor and a memory, wherein the processor runs a computer program or code stored in the memory to implement a hidden fracture prediction method as described in any one of the above.

[0016] In addition, an embodiment of the present invention further provides a computer-readable storage medium for storing a computer program or code. When the computer program or code is executed by a processor, the hidden fracture prediction method as described in any one of the above is implemented.

[0017] In the embodiment of the present invention, by integrating multi-source detection data such as gravity and magnetism, the underground geological structure can be reflected more comprehensively and deeply; the superposition and comprehensive analysis of multi-source data make the fault characteristics more obvious, greatly improving the ability to identify hidden faults; and through intelligent algorithms and automated analysis technology, the operation steps are simplified, the dependence on high-end equipment and professionals is reduced, and the overall cost is significantly reduced; at the same time, by optimizing the analysis process, a large amount of data can be processed quickly, thereby improving the work efficiency of fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a flow chart of a hidden fracture prediction method based on multi-source detection data fusion in an embodiment of the present invention; Figure 2 is a flow chart of a hidden fracture prediction method based on multi-source detection data fusion in another embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of a hidden fracture prediction device based on multi-source detection data fusion in an embodiment of the present invention; Figure 4 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] For example, see Figure 1 , Figure 1 It is a flow chart of a hidden fracture prediction method based on multi-source detection data fusion in an embodiment of the present invention.

[0022] like Figure 1 As shown, a hidden fault prediction method based on multi-source detection data fusion, the method comprises: S101: performing detection processing in a designated area based on gravity instruments and magnetic instruments to obtain gravity detection data and magnetic detection data, wherein the designated area is a known fracture area, the gravity detection data is Bouguer gravity anomaly data, and the magnetic detection data is aeromagnetic anomaly data; In the specific implementation process of the present invention, it is necessary to designate a designated area, which is a known fracture area; then gravity instruments and magnetic instruments are used to perform detection and processing in the designated area, so as to obtain gravity detection data and magnetic detection data; the gravimeter provides Bouguer gravity anomaly data by measuring the slight changes in the gravity field on the earth's surface; the magnetometer records the magnetic field anomalies on the earth's surface and generates aeromagnetic anomaly data; the data is transmitted to the storage device through the data acquisition workstation for subsequent processing.

[0023] S102: fusing the gravity detection data and the magnetic detection data to form multi-source fused detection data; In the specific implementation process of the present invention, the gravity detection data and the magnetic detection data are fused to form multi-source fused detection data, including: filtering the gravity detection data and the magnetic detection data for abnormal data to obtain filtered gravity detection data and filtered magnetic detection data; smoothing and compensating the filtered gravity detection data and the filtered magnetic detection data in turn to obtain compensated gravity detection data and compensated magnetic detection data; fusing the compensated gravity detection data and the compensated magnetic detection data based on a data fusion algorithm to obtain multi-source fused detection data.

[0024] Specifically, the gravity detection data and magnetic detection data obtained by detection first need to be preprocessed, and the preprocessing here includes data cleaning, filtering, smoothing and compensation processing, etc.; that is, the gravity detection data and the magnetic detection data are subjected to abnormal data filtering processing to obtain filtered gravity detection data and filtered magnetic detection data; then the filtered gravity detection data and the filtered magnetic detection data are smoothed and compensated in turn to obtain compensated gravity detection data and compensated magnetic detection data; finally, the compensated gravity detection data and the compensated magnetic detection data are fused through a data fusion algorithm to obtain multi-source fused detection data.

[0025] S103: performing enhancement processing on the multi-source fusion detection data based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data; In the specific implementation process of the present invention, the multi-source fusion detection data is enhanced based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data, including: using multiple data enhancement algorithms to enhance the geological structure in the multi-source fusion detection data to obtain enhanced multi-source fusion detection data; wherein the multiple data enhancement algorithms include total horizontal derivative method, vertical derivative method, vertical second-order derivative, inclination angle method, inclination angle horizontal derivative method, analytical signal method, Theat method, total horizontal derivative inclination angle method, horizontal inclination angle method, enhanced analytical signal method, and enhanced inclination angle method.

[0026] Specifically, after obtaining the multi-source fusion detection data, it is necessary to enhance the geological structure in the multi-source fusion detection data through a variety of data enhancement algorithms to obtain enhanced multi-source fusion detection data; wherein the multiple data enhancement algorithms include a combination of at least two of the total horizontal derivative method, vertical derivative method, vertical second-order derivative, inclination angle method, inclination angle horizontal derivative method, analytical signal method, Theat method, total horizontal derivative inclination angle method, horizontal inclination angle method, enhanced analytical signal method, and enhanced inclination angle method.

[0027] Among them, the total horizontal derivative method first uses the horizontal gradient of gravity anomalies to identify boundaries. Later, this method was widely used in gravity and magnetic anomalies. Its principle is to assume that the anomaly is a vertical body boundary, and the horizontal gradient modulus of the anomaly takes a maximum value at the boundary, thereby determining the location of the boundary. However, with the increase of the buried depth of the geological body and the change of the inclination angle, the offset of the maximum position from the boundary will also increase, but this offset can be ignored for the anomalies in the study area; the calculation formula of the total horizontal derivative method can be expressed as: ; The calculation formula of the vertical derivative method is: ; The vertical derivative method is different from the total horizontal derivative method. The vertical derivative method determines the boundary by finding the zero-value position of the gravity and magnetic anomaly, and its offset to the true position will become more obvious as the burial depth increases. Usually the vertical derivative is implemented in the frequency domain, so for the stability of the result, low-pass filtering methods such as upward extension and Winer filtering are generally used to suppress high-frequency interference. At the same time, the vertical second-order derivative method can also be used, but increasing the order of the derivative may cause the identified boundary to be closer to the center position; as follows: ; The inclination angle method is also called the oblique derivative method, and its calculation formula is: ; in, , There are vertical derivative method and total horizontal derivative method to obtain respectively; the inclination angle method is a boundary enhancement method with better balance; the ratio of the abnormal vertical derivative and the total horizontal derivative is used, which is used to balance the high-amplitude anomalies and the low-amplitude anomalies well, and appropriately stretch the geological bodies of different burial depths. The first two methods perform poorly in identifying deep source anomalies, while the inclination angle method is insensitive to the depth of the field source due to its balance, and can identify the boundary well; but the inclination angle method also has defects, and the "analytic singularity" phenomenon may affect the results; the inclination angle is further used to obtain the total horizontal derivative, which not only eliminates the "analytic singularity", but also uses the maximum position to determine the boundary to improve the lateral resolution of the identification; the improved calculation formula can be expressed as: ; The analytical signal method is also called the total gradient modulus method. Initially, analytical signals were only used for two-dimensional space processing, and then the three-dimensional analytical signal method was developed. The analytical signal method has its own advantages in processing magnetic anomalies. It is not affected by the magnetization direction and magnetic anomaly components or has little influence. However, its lateral resolution is very low compared with other methods, so improving its lateral resolution is also the focus of future research. The calculation formula of the analytical signal method is: ; in, , Obtained from the above formulas respectively.

[0028] The calculation formula of Theta graph method is: ; Among them, THDR and ASM are respectively derived from the above.

[0029] Theta diagram identifies the boundary by taking the maximum value position, and also balances the anomaly amplitude values ​​from different depths well, enhancing shallow source anomalies while also enhancing deep source anomalies; it is believed that this method is affected by the magnetic anomaly component and magnetization direction and is comparable to the vertical derivative method and the inclination angle method; however, the Theta diagram also has the same "analytical singularity" problem as the inclination angle method, which can be improved by introducing a regularization factor.

[0030] The calculation formula of the horizontal tilt angle method is: ; Among them, THDR and VDR are given above respectively; contrary to the normalized vertical derivative amplitude of the tilt angle method, the amplitude of the normalized horizontal derivative of the horizontal tilt angle method shows equally good results for deep and shallow anomalies in boundary recognition, and is a very useful method for feature enhancement in a given direction.

[0031] The enhanced analytical signal tilt angle method is an equalization filter formed based on the tilt angle method and the analytical signal method. Its purpose is to handle the situation where positive anomalies and negative anomalies exist at the same time. Its calculation formula is: ; where ASM is given by the above formula.

[0032] At the same time, a method to increase the resolution is proposed, which uses the vertical first-order derivative of the amplitude of the analytical signal to define the filter: ; Where VAS is the vertical first-order derivative of ASM, that is: ; It is generally believed that the TAAS and TAVAS methods are more accurate than traditional methods in identifying boundaries, and when both positive and negative anomalies are included, no redundant erroneous boundary information will be generated. However, due to the calculation of high-order derivatives, TAAS and TAVAS are easily affected by external noise, so in order to obtain accurate fracture information, filtering operations need to be performed before preprocessing.

[0033] The calculation formula of the enhanced tilt angle method is: ; The enhanced tilt angle method is similar to the tilt angle method. The zero value provides the potential field boundary, and it also achieves balance by normalizing the anomalies of different field source depths. Its greatest advantage is that it can improve the accuracy of identifying small-scale linear structures from superimposed anomalies, and in practical applications it can highlight the linear characteristics of weak changes. Similarly, TAVDR will also be affected by noise, so filtering should also be performed before enhancement.

[0034] In this embodiment, according to the abnormal discontinuity of gravity and magnetism at the geological boundary, the larger change rate of gravity and magnetic anomalies at the boundary is used to reasonably extract the boundary, providing a basis for the subsequent identification of faults; common methods for gravity and magnetic boundary extraction can be roughly divided into mathematical statistics methods (such as small subdomain filtering method, normalized standard deviation method, etc.), numerical calculation methods (various order derivatives of potential field data method) and other types (such as directional filtering method, etc.); the most commonly used numerical calculation methods for boundary detection are used, including 12 methods including the first-order derivative and higher-order derivative of the gravity and magnetic field, and the enhanced results will be used as feature data for fault identification.

[0035] S104: constructing a data set for the enhanced multi-source fusion detection data, and annotating the constructed data set to form a training sample data set; In the specific implementation process of the present invention, the enhanced multi-source fusion detection data is subjected to data set construction processing, and the constructed data set is labeled to form a training sample data set, including: based on a synthetic minority class oversampling algorithm, the enhanced multi-source fusion detection data is subjected to sample synthesis processing, and based on the synthesized sample data, a data set is constructed to obtain a constructed data set; the existing fracture distribution map and geological data of the specified area are obtained, and based on the existing fracture distribution map and geological data, the constructed data set is labeled to form a training sample data set.

[0036] Furthermore, the sample synthesis processing of the enhanced multi-source fusion detection data based on the synthetic minority class oversampling algorithm includes: For each minority class sample data X in the enhanced multi-source fusion detection data, the minority class sample X and its nearest neighboring sample data are calculated according to the required oversampling amount. The difference between is used to synthesize new samples, which can be expressed as: ; in, represents a random function; Represents synthetic sample data.

[0037] Specifically, for the enhanced multi-source fusion detection data, this embodiment uses the synthetic minority over-sampling technique (SMOTE) to deal with the problem of sample imbalance, that is, to increase the amount of data of minority class samples, so as to achieve the purpose of sample balance. The principle of the SMOTE method is to create synthetic samples instead of randomly replacing over-sampling; in this embodiment, for each minority class sample X, according to the required over-sampling number, each group of breakpoints is used as the sampling basis, and the minority sample and its nearest neighbor sample data are calculated. The difference between is used to synthesize new samples, which can be expressed as: ; in, represents a random function; Represents synthetic sample data; the SMOTE method can not only fully learn the characteristics of a few samples and improve the accuracy of the classifier, but also avoid the overfitting phenomenon caused by the random oversampling method.

[0038] S105: Inputting the training sample data set into a preset machine learning model for training processing to obtain a machine learning model with training convergence, wherein the machine learning model includes a random forest model and an artificial neural network model; In the specific implementation process of the present invention, the training sample data set is input into the preset machine learning model for training processing to obtain a machine learning model with training convergence, including: when the training sample data set is input into the preset machine learning model for training processing, the hyperparameters in the preset machine learning model are adjusted by a grid search algorithm or a genetic optimization algorithm, and the optimal parameter combination is selected for training processing to obtain a machine learning model with training convergence; when the preset machine learning model is a random forest model, the hyperparameters are the number of decision trees in the random forest model, the maximum depth of the decision tree, the minimum number of samples contained in a leaf node, and the minimum number of samples that can be divided into a node; when the preset machine learning model is an artificial neural network model, the hyperparameters are the number of layers and the number of nodes of the artificial neural network model.

[0039] Specifically, the random forest model and the artificial neural network model are selected as machine learning models; the random forest model can effectively avoid the overfitting problem by integrating multiple decision trees; the artificial neural network model has the ability to handle complex nonlinear problems.

[0040] When the training sample data set is input into the preset machine learning model for training, the hyperparameters in the preset machine learning model are adjusted through the grid search algorithm or the genetic optimization algorithm, and the best parameter combination is selected for training to obtain a machine learning model with convergence in training; that is, the hyperparameters of the model (the number of decision trees, the maximum depth of the decision tree, the minimum number of samples contained in a leaf node, the minimum number of samples that a node can be divided into, the number of layers and nodes of the neural network, etc.) are adjusted through methods such as grid search or genetic optimization algorithms to select the best parameter combination.

[0041] The artificial neural network model is a supervised feedforward neural network, which maps a set of input vectors to output vectors through hidden layers. It consists of three node layers: input layer, one or more hidden layers and output layer. The function of the input layer is to receive the prediction variable and pass it to the hidden layer. The hidden layer is composed of one or more network layers, in which each neuron is connected to the neurons in the previous and next layers. The number of neurons in the output layer is the same as the number of categories in the actual classification problem. It is essentially a multi-category logistic regression process. The hidden layer and the output layer calculate the output signal of the layer through the activation function based on the received signal of the previous layer. The calculation process of the artificial neural network is summarized as follows. First, the output signal is calculated: ; in, represents the activation function; represents the weighted value; Represents input; b represents bias term; in order to obtain the most appropriate weight and bias term, the back propagation algorithm is used for iterative calculation; for this purpose, the loss function is defined as: ; in, represents the true label (the probability that sample i belongs to the jth class); The output of the model (the predicted probability that sample i belongs to the jth class) is calculated layer by layer from back to front to calculate the gradient of the loss function with respect to the weight and bias terms, that is: ; ; in, and Respectively indicate l The weight and bias of the nth neuron in the layer; Represents the learning rate; the neural network randomly initializes the weights and bias values, then updates the weights and bias items of each layer according to the above formula, and continuously iterates the forward propagation and back propagation processes until the stopping criterion, that is, the set number of cycles or error threshold, is met, and finally the solution is achieved.

[0042] The random forest model is an ensemble learning algorithm based on decision trees. It generates a large number of decision trees (weak classifiers), learns and predicts independently, and then votes on the results of these decision trees to form a strong classifier. A decision tree is a tree structure in which each internal node represents an attribute judgment, each branch represents a judgment result, and each leaf node represents a classification result. Common decision tree algorithms include C4.5, ID3, and CART, among which random forests usually use CART decision trees. CART decision (Classification and Regression Tree) can be used for classification and regression tasks. If the data set contains continuous variables, the decision tree is used as a regression tree to make predictions using the mean of the leaf nodes; if the dependent variable is a discrete value, the decision tree is used as a classifier to solve the classification problem; the decision tree in the random forest traverses all possible split points of the feature variables and finds the split point with the smallest Gini impurity coefficient to divide the data into two subsets; this process is repeated until the predefined stop condition is met, thereby completing the purification of a single decision tree. Random forest generates decision trees through the following two random processes: Randomly generate training sets: Use bootstrap to randomly sample with replacement to generate training subset data for each tree. This process avoids high correlation between trees, and uses bagging to replace the original data for random sampling.

[0043] Randomly select feature attributes: randomly select features without replacement when splitting the nodes of the tree, generate feature subsets, select the best features suitable for splitting, and continue feature selection and tree splitting.

[0044] These two random processes ensure the randomness of the entire algorithm and the independence of each decision tree without any pruning, which improves the classification accuracy. The calculation process of the random forest algorithm is briefly summarized as follows: In the training data X, a subsample subset is generated by random sampling. And select the corresponding feature subset , where n is the number of decision trees; sample and the randomly selected features Training a decision tree ; Decision Tree Voting is performed for each category, and the voting result is used as the final classification value; the voting classification result of a sample can be expressed as: ; in, represents the lithology category predicted by the i-th decision tree; The random forest model can evaluate the importance of predictive variables, and the corresponding calculation formula is calibrated as follows: ; Among them, t, k and m are the number of nodes in each tree, the number of decision trees and the number of predictor variables respectively; The reduction value of the Gini coefficient for the r-th predictor variable at the j-th node of the i-th tree.

[0045] Typically, random forests generate hundreds to thousands of decision trees and return the final result through voting; the goal of random forests is not to pursue the optimal decision tree, but to aggregate a set of predictors that benefit from the broad search of all tree predictors.

[0046] After the training is completed, the performance indicators are used to evaluate whether the performance of the trained machine learning model meets the requirements; in the embodiment of the present application, five indicators are used: accuracy, precision, recall, F1 score, and AUC value; usually, in order to describe these indicators, it is necessary to introduce the following confusion matrix of binary classification data, as shown in Table 1: Table 1 Confusion matrix of binary classification data

[0047] Among them, TP is the correctly classified positive sample, FP is the incorrectly classified negative sample, FN is the incorrectly classified positive sample, and TN is the correctly classified negative sample.

[0048] Accuracy: It measures the classification accuracy of the test set as a whole, indicating the ratio of the number of samples that meet the label in the prediction results to the total number of samples. The calculation formula is: ; Precision: refers to the proportion of samples predicted to be correctly classified to all samples predicted to be positive. The calculation formula is: ; Recall: refers to the proportion of samples that are predicted to be correctly classified to all truly positive samples. The calculation formula is: ; F1 score: It is a combination of precision and recall. Since there is a contradictory situation of high precision and low recall in actual situations, the F1 score is introduced to reconcile the evaluation of the two. The calculation formula is: ; AUC value: The performance of the random forest algorithm is poor when dealing with unbalanced data sets, and it tends to give a higher weight to the majority class; therefore, it is necessary to introduce the receiver operating characteristic curve (ROC), which is a useful tool for evaluating classification performance; first, it is necessary to calculate the negative sample misclassification probability (FPR) and the positive sample split probability (TPR), both of which range from 0 to 1, and their calculation formulas are as follows: ; ; The ROC curve is established with FPR as the horizontal axis and TPR as the vertical axis. The closer the curve is to the (0,1) point, that is, TPR approaches 1 and FPR approaches 0, the more perfect the classifier is. On the contrary, if the curve is closer to the (1,0) point, the performance of the classifier is worse. The area under the curve is the AUC value, which usually ranges from 0.5 to 1. The closer the AUC value is to 1, the better the classifier performance is.

[0049] S106: Apply the trained converged machine learning model to the unknown fracture target area to perform hidden fracture prediction processing, output the prediction results in the form of spatial coordinates, and generate a hidden fracture prediction distribution map covering the unknown fracture target area.

[0050] In the specific implementation process of the present invention, the prediction results are output in the form of spatial coordinates, and a hidden fracture prediction distribution map covering the unknown fracture target area is generated, including: outputting the prediction results in the form of spatial coordinates, and combining the prediction results output in the form of spatial coordinates with geographic information system software to perform visual analysis, and generate a hidden fracture prediction distribution map covering the unknown fracture target area.

[0051] Specifically, the machine learning model that has been trained and converged is applied to the unknown study area to predict hidden faults in the unknown fault target area; the output prediction results are output in the form of spatial coordinates, and the prediction results are visualized and analyzed in combination with the geographic information system (GIS) software to generate a fault distribution map of the coverage area.

[0052] In the embodiment of the present invention, by integrating multi-source detection data such as gravity and magnetism, the underground geological structure can be reflected more comprehensively and deeply; the superposition and comprehensive analysis of multi-source data make the fault characteristics more obvious, greatly improving the ability to identify hidden faults; and through intelligent algorithms and automated analysis technology, the operation steps are simplified, the dependence on high-end equipment and professionals is reduced, and the overall cost is significantly reduced; at the same time, by optimizing the analysis process, a large amount of data can be processed quickly, thereby improving the work efficiency of fault prediction.

[0053] For example 2, please refer to Figure 2 , Figure 2It is a flow chart of a hidden fracture prediction method based on multi-source detection data fusion in another embodiment of the present invention.

[0054] like Figure 2 As shown, a hidden fault prediction method based on multi-source detection data fusion, the method comprises: S201 performs detection processing in a designated area based on gravity instruments and magnetic instruments to obtain gravity detection data and magnetic detection data, wherein the designated area is a known fracture area, the gravity detection data is Bouguer gravity anomaly data, and the magnetic detection data is aeromagnetic anomaly data; S202: fusing the gravity detection data and the magnetic detection data to form multi-source fused detection data; S203: performing enhancement processing on the multi-source fusion detection data based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data; S204: performing sample synthesis processing on the enhanced multi-source fusion detection data based on a synthetic minority class oversampling algorithm, and performing data set construction processing based on the synthesized sample data to obtain a constructed data set; S205: obtaining an existing fault distribution map and geological data of the designated area, and annotating the constructed data set based on the existing fault distribution map and geological data to form a training sample data set; S206: Inputting the training sample data set into a preset machine learning model for training processing to obtain a machine learning model with training convergence, wherein the machine learning model includes a random forest model and an artificial neural network model; S207: Applying the trained converged machine learning model to the unknown fracture target area to perform hidden fracture prediction processing, outputting the prediction results in the form of spatial coordinates, and generating a hidden fracture prediction distribution map covering the unknown fracture target area.

[0055] Specifically, the specific implementation of the second embodiment can refer to the first embodiment, which will not be described in detail here.

[0056] Example 3, please refer to Figure 3 , Figure 3 It is a schematic diagram of the structural composition of a hidden fracture prediction device based on multi-source detection data fusion in an embodiment of the present invention.

[0057] like Figure 3 As shown, a hidden fracture prediction device based on multi-source detection data fusion, the device comprises: Detection module 301: used to perform detection processing in a specified area based on gravity instruments and magnetic instruments to obtain gravity detection data and magnetic detection data, wherein the specified area is a known fracture area, the gravity detection data is Bouguer gravity anomaly data, and the magnetic detection data is aeromagnetic anomaly data; In the specific implementation process of the present invention, it is necessary to designate a designated area, which is a known fracture area; then gravity instruments and magnetic instruments are used to perform detection and processing in the designated area, so as to obtain gravity detection data and magnetic detection data; the gravimeter provides Bouguer gravity anomaly data by measuring the slight changes in the gravity field on the earth's surface; the magnetometer records the magnetic field anomalies on the earth's surface and generates aeromagnetic anomaly data; the data is transmitted to the storage device through the data acquisition workstation for subsequent processing.

[0058] Data fusion module 302: used for fusing the gravity detection data and the magnetic detection data to form multi-source fusion detection data; In the specific implementation process of the present invention, the gravity detection data and the magnetic detection data are fused to form multi-source fused detection data, including: filtering the gravity detection data and the magnetic detection data for abnormal data to obtain filtered gravity detection data and filtered magnetic detection data; smoothing and compensating the filtered gravity detection data and the filtered magnetic detection data in turn to obtain compensated gravity detection data and compensated magnetic detection data; fusing the compensated gravity detection data and the compensated magnetic detection data based on a data fusion algorithm to obtain multi-source fused detection data.

[0059] Specifically, the gravity detection data and magnetic detection data obtained by detection first need to be preprocessed, and the preprocessing here includes data cleaning, filtering, smoothing and compensation processing, etc.; that is, the gravity detection data and the magnetic detection data are subjected to abnormal data filtering processing to obtain filtered gravity detection data and filtered magnetic detection data; then the filtered gravity detection data and the filtered magnetic detection data are smoothed and compensated in turn to obtain compensated gravity detection data and compensated magnetic detection data; finally, the compensated gravity detection data and the compensated magnetic detection data are fused through a data fusion algorithm to obtain multi-source fused detection data.

[0060] Data enhancement module 303: used to enhance the multi-source fusion detection data based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data; In the specific implementation process of the present invention, the multi-source fusion detection data is enhanced based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data, including: using multiple data enhancement algorithms to enhance the geological structure in the multi-source fusion detection data to obtain enhanced multi-source fusion detection data; wherein the multiple data enhancement algorithms include total horizontal derivative method, vertical derivative method, vertical second-order derivative, inclination angle method, inclination angle horizontal derivative method, analytical signal method, Theat method, total horizontal derivative inclination angle method, horizontal inclination angle method, enhanced analytical signal method, and enhanced inclination angle method.

[0061] Specifically, after obtaining the multi-source fusion detection data, it is necessary to enhance the geological structure in the multi-source fusion detection data through a variety of data enhancement algorithms to obtain enhanced multi-source fusion detection data; wherein the multiple data enhancement algorithms include a combination of at least two of the total horizontal derivative method, vertical derivative method, vertical second-order derivative, inclination angle method, inclination angle horizontal derivative method, analytical signal method, Theat method, total horizontal derivative inclination angle method, horizontal inclination angle method, enhanced analytical signal method, and enhanced inclination angle method.

[0062] Among them, the total horizontal derivative method first uses the horizontal gradient of gravity anomalies to identify boundaries. Later, this method was widely used in gravity and magnetic anomalies. Its principle is to assume that the anomaly is a vertical body boundary, and the horizontal gradient modulus of the anomaly takes a maximum value at the boundary, thereby determining the location of the boundary. However, with the increase of the buried depth of the geological body and the change of the inclination angle, the offset of the maximum position from the boundary will also increase, but this offset can be ignored for the anomalies in the study area; the calculation formula of the total horizontal derivative method can be expressed as: ; The calculation formula of the vertical derivative method is: ; The vertical derivative method is different from the total horizontal derivative method. The vertical derivative method determines the boundary by finding the zero-value position of the gravity and magnetic anomaly, and its offset to the true position will become more obvious as the burial depth increases. Usually the vertical derivative is implemented in the frequency domain, so for the stability of the result, low-pass filtering methods such as upward extension and Winer filtering are generally used to suppress high-frequency interference. At the same time, the vertical second-order derivative method can also be used, but increasing the order of the derivative may cause the identified boundary to be closer to the center position; as follows: ; The inclination angle method is also called the oblique derivative method, and its calculation formula is: ; in, , There are vertical derivative method and total horizontal derivative method to obtain respectively; the inclination angle method is a boundary enhancement method with better balance; the ratio of the abnormal vertical derivative and the total horizontal derivative is used, which is used to balance the high-amplitude anomalies and the low-amplitude anomalies well, and appropriately stretch the geological bodies of different burial depths. The first two methods perform poorly in identifying deep source anomalies, while the inclination angle method is insensitive to the depth of the field source due to its balance, and can identify the boundary well; but the inclination angle method also has defects, and the "analytic singularity" phenomenon may affect the results; the inclination angle is further used to obtain the total horizontal derivative, which not only eliminates the "analytic singularity", but also uses the maximum position to determine the boundary to improve the lateral resolution of the identification; the improved calculation formula can be expressed as: ; The analytical signal method is also called the total gradient modulus method. Initially, analytical signals were only used for two-dimensional space processing, and then the three-dimensional analytical signal method was developed. The analytical signal method has its own advantages in processing magnetic anomalies. It is not affected by the magnetization direction and magnetic anomaly components or has little influence. However, its lateral resolution is very low compared with other methods, so improving its lateral resolution is also the focus of future research. The calculation formula of the analytical signal method is: ; in, , Obtained from the above formulas respectively.

[0063] The calculation formula of Theta graph method is: ; Among them, THDR and ASM are respectively derived from the above.

[0064] Theta diagram identifies the boundary by taking the maximum value position, and also balances the anomaly amplitude values ​​from different depths well, enhancing shallow source anomalies while also enhancing deep source anomalies; it is believed that this method is affected by the magnetic anomaly component and magnetization direction and is comparable to the vertical derivative method and the inclination angle method; however, the Theta diagram also has the same "analytical singularity" problem as the inclination angle method, which can be improved by introducing a regularization factor.

[0065] The calculation formula of the horizontal tilt angle method is: ; Among them, THDR and VDR are given above respectively; contrary to the normalized vertical derivative amplitude of the tilt angle method, the amplitude of the normalized horizontal derivative of the horizontal tilt angle method shows equally good results for deep and shallow anomalies in boundary recognition, and is a very useful method for feature enhancement in a given direction.

[0066] The enhanced analytical signal tilt angle method is an equalization filter formed based on the tilt angle method and the analytical signal method. Its purpose is to handle the situation where positive anomalies and negative anomalies exist at the same time. Its calculation formula is: ; where ASM is given by the above formula.

[0067] At the same time, a method to increase the resolution is proposed, which uses the vertical first-order derivative of the amplitude of the analytical signal to define the filter: ; Where VAS is the vertical first-order derivative of ASM, that is: ; It is generally believed that the TAAS and TAVAS methods are more accurate than traditional methods in identifying boundaries, and when both positive and negative anomalies are included, no redundant erroneous boundary information will be generated. However, due to the calculation of high-order derivatives, TAAS and TAVAS are easily affected by external noise, so in order to obtain accurate fracture information, filtering operations need to be performed before preprocessing.

[0068] The calculation formula of the enhanced tilt angle method is: ; The enhanced tilt angle method is similar to the tilt angle method. The zero value provides the potential field boundary, and it also achieves balance by normalizing the anomalies of different field source depths. Its greatest advantage is that it can improve the accuracy of identifying small-scale linear structures from superimposed anomalies, and in practical applications it can highlight the linear characteristics of weak changes. Similarly, TAVDR will also be affected by noise, so filtering should also be performed before enhancement.

[0069] In this embodiment, according to the abnormal discontinuity of gravity and magnetism at the geological boundary, the larger change rate of gravity and magnetic anomalies at the boundary is used to reasonably extract the boundary, providing a basis for the subsequent identification of faults; common methods for gravity and magnetic boundary extraction can be roughly divided into mathematical statistics methods (such as small subdomain filtering method, normalized standard deviation method, etc.), numerical calculation methods (various order derivatives of potential field data method) and other types (such as directional filtering method, etc.); the most commonly used numerical calculation methods for boundary detection are used, including 12 methods including the first-order derivative and higher-order derivative of the gravity and magnetic field, and the enhanced results will be used as feature data for fault identification.

[0070] Data construction and annotation module 304: used to construct a data set for the enhanced multi-source fusion detection data, and to annotate the constructed data set to form a training sample data set; In the specific implementation process of the present invention, the enhanced multi-source fusion detection data is subjected to data set construction processing, and the constructed data set is labeled to form a training sample data set, including: based on a synthetic minority class oversampling algorithm, the enhanced multi-source fusion detection data is subjected to sample synthesis processing, and based on the synthesized sample data, a data set is constructed to obtain a constructed data set; the existing fracture distribution map and geological data of the specified area are obtained, and based on the existing fracture distribution map and geological data, the constructed data set is labeled to form a training sample data set.

[0071] Furthermore, the sample synthesis processing of the enhanced multi-source fusion detection data based on the synthetic minority class oversampling algorithm includes: For each minority class sample data X in the enhanced multi-source fusion detection data, the minority class sample X and its nearest neighboring sample data are calculated according to the required oversampling amount. The difference between is used to synthesize new samples, which can be expressed as: ; in, represents a random function; Represents synthetic sample data.

[0072] Specifically, for the enhanced multi-source fusion detection data, this embodiment uses the synthetic minority over-sampling technique (SMOTE) to deal with the problem of sample imbalance, that is, to increase the amount of data of minority class samples, so as to achieve the purpose of sample balance. The principle of the SMOTE method is to create synthetic samples instead of randomly replacing over-sampling; in this embodiment, for each minority class sample X, according to the required over-sampling number, each group of breakpoints is used as the sampling basis, and the minority sample and its nearest neighbor sample data are calculated. The difference between is used to synthesize new samples, which can be expressed as: ; in, represents a random function; Represents synthetic sample data; the SMOTE method can not only fully learn the characteristics of a few samples and improve the accuracy of the classifier, but also avoid the overfitting phenomenon caused by the random oversampling method.

[0073] Model training module 305: used to input the training sample data set into a preset machine learning model for training processing to obtain a machine learning model with training convergence, wherein the machine learning model includes a random forest model and an artificial neural network model; In the specific implementation process of the present invention, the training sample data set is input into the preset machine learning model for training processing to obtain a machine learning model with training convergence, including: when the training sample data set is input into the preset machine learning model for training processing, the hyperparameters in the preset machine learning model are adjusted by a grid search algorithm or a genetic optimization algorithm, and the optimal parameter combination is selected for training processing to obtain a machine learning model with training convergence; when the preset machine learning model is a random forest model, the hyperparameters are the number of decision trees in the random forest model, the maximum depth of the decision tree, the minimum number of samples contained in a leaf node, and the minimum number of samples that can be divided into a node; when the preset machine learning model is an artificial neural network model, the hyperparameters are the number of layers and the number of nodes of the artificial neural network model.

[0074] Specifically, the random forest model and the artificial neural network model are selected as machine learning models; the random forest model can effectively avoid the overfitting problem by integrating multiple decision trees; the artificial neural network model has the ability to handle complex nonlinear problems.

[0075] When the training sample data set is input into the preset machine learning model for training, the hyperparameters in the preset machine learning model are adjusted through the grid search algorithm or the genetic optimization algorithm, and the best parameter combination is selected for training to obtain a machine learning model with convergence in training; that is, the hyperparameters of the model (the number of decision trees, the maximum depth of the decision tree, the minimum number of samples contained in a leaf node, the minimum number of samples that a node can be divided into, the number of layers and nodes of the neural network, etc.) are adjusted through methods such as grid search or genetic optimization algorithms to select the best parameter combination.

[0076] The artificial neural network model is a supervised feedforward neural network, which maps a set of input vectors to output vectors through hidden layers. It consists of three node layers: input layer, one or more hidden layers and output layer. The function of the input layer is to receive the prediction variable and pass it to the hidden layer. The hidden layer is composed of one or more network layers, in which each neuron is connected to the neurons in the previous and next layers. The number of neurons in the output layer is the same as the number of categories in the actual classification problem. It is essentially a multi-category logistic regression process. The hidden layer and the output layer calculate the output signal of the layer through the activation function based on the received signal of the previous layer. The calculation process of the artificial neural network is summarized as follows. First, the output signal is calculated: ; in, represents the activation function; represents the weighted value; Represents input; b represents bias term; in order to obtain the most appropriate weight and bias term, the back propagation algorithm is used for iterative calculation; for this purpose, the loss function is defined as: ; in, represents the true label (the probability that sample i belongs to the jth class); The output of the model (the predicted probability that sample i belongs to the jth class) is calculated layer by layer from back to front to calculate the gradient of the loss function with respect to the weight and bias terms, that is: ; ; in, and Respectively indicate l The weight and bias of the nth neuron in the layer; Represents the learning rate; the neural network randomly initializes the weights and bias values, then updates the weights and bias items of each layer according to the above formula, and continuously iterates the forward propagation and back propagation processes until the stopping criterion, that is, the set number of cycles or error threshold, is met, and finally the solution is achieved.

[0077] The random forest model is an ensemble learning algorithm based on decision trees. It generates a large number of decision trees (weak classifiers), learns and predicts independently, and then votes on the results of these decision trees to form a strong classifier. A decision tree is a tree structure in which each internal node represents an attribute judgment, each branch represents a judgment result, and each leaf node represents a classification result. Common decision tree algorithms include C4.5, ID3, and CART, among which random forests usually use CART decision trees. CART decision (Classification and Regression Tree) can be used for classification and regression tasks. If the data set contains continuous variables, the decision tree is used as a regression tree to make predictions using the mean of the leaf nodes; if the dependent variable is a discrete value, the decision tree is used as a classifier to solve the classification problem; the decision tree in the random forest traverses all possible split points of the feature variables and finds the split point with the smallest Gini impurity coefficient to divide the data into two subsets; this process is repeated until the predefined stop condition is met, thereby completing the purification of a single decision tree. Random forest generates decision trees through the following two random processes: Randomly generate training sets: Use bootstrap to randomly sample with replacement to generate training subset data for each tree. This process avoids high correlation between trees, and uses bagging to replace the original data for random sampling.

[0078] Randomly select feature attributes: randomly select features without replacement when splitting the nodes of the tree, generate feature subsets, select the best features suitable for splitting, and continue feature selection and tree splitting.

[0079] These two random processes ensure the randomness of the entire algorithm and the independence of each decision tree without any pruning, which improves the classification accuracy. The calculation process of the random forest algorithm is briefly summarized as follows: In the training data X, a subsample subset is generated by random sampling , and select the corresponding feature subset , where n is the number of decision trees; sample and the randomly selected features Training a decision tree ; Decision Tree Voting is performed for each category, and the voting result is used as the final classification value; the voting classification result of a sample can be expressed as: ; in, represents the lithology category predicted by the i-th decision tree; The random forest model can evaluate the importance of predictive variables, and the corresponding calculation formula is calibrated as follows: ; Among them, t, k and m are the number of nodes in each tree, the number of decision trees and the number of predictor variables respectively; The reduction value of the Gini coefficient for the r-th predictor variable at the j-th node of the i-th tree.

[0080] Typically, random forests generate hundreds to thousands of decision trees and return the final result through voting; the goal of random forests is not to pursue the optimal decision tree, but to aggregate a set of predictors that benefit from the broad search of all tree predictors.

[0081] After the training is completed, the performance indicators are used to evaluate whether the performance of the trained machine learning model meets the requirements; in the embodiment of the present application, five indicators are used: accuracy, precision, recall, F1 score, and AUC value; usually, in order to describe these indicators, it is necessary to introduce the following confusion matrix of binary classification data, as shown in Table 1: Table 1 Confusion matrix of binary classification data

[0082] Among them, TP is the correctly classified positive sample, FP is the incorrectly classified negative sample, FN is the incorrectly classified positive sample, and TN is the correctly classified negative sample.

[0083] Accuracy: It measures the classification accuracy of the test set as a whole, indicating the ratio of the number of samples that meet the label in the prediction results to the total number of samples. The calculation formula is: ; Precision: refers to the proportion of samples predicted to be correctly classified to all samples predicted to be positive. The calculation formula is: ; Recall: refers to the proportion of samples that are predicted to be correctly classified to all truly positive samples. The calculation formula is: ; F1 score: It is a combination of precision and recall. Since there is a contradictory situation of high precision and low recall in actual situations, the F1 score is introduced to reconcile the evaluation of the two. The calculation formula is: ; AUC value: The performance of the random forest algorithm is poor when dealing with unbalanced data sets, and it tends to give a higher weight to the majority class; therefore, it is necessary to introduce the receiver operating characteristic curve (ROC), which is a useful tool for evaluating classification performance; first, it is necessary to calculate the negative sample misclassification probability (FPR) and the positive sample split probability (TPR), both of which range from 0 to 1, and their calculation formulas are as follows: ; ; The ROC curve is established with FPR as the horizontal axis and TPR as the vertical axis. The closer the curve is to the (0,1) point, that is, TPR approaches 1 and FPR approaches 0, the more perfect the classifier is. On the contrary, if the curve is closer to the (1,0) point, the performance of the classifier is worse. The area under the curve is the AUC value, which usually ranges from 0.5 to 1. The closer the AUC value is to 1, the better the classifier performance is.

[0084] Prediction generation module 306: used to apply the trained converged machine learning model to the unknown fracture target area to perform hidden fracture prediction processing, output the prediction results in the form of spatial coordinates, and generate a hidden fracture prediction distribution map covering the unknown fracture target area.

[0085] In the specific implementation process of the present invention, the prediction results are output in the form of spatial coordinates, and a hidden fracture prediction distribution map covering the unknown fracture target area is generated, including: outputting the prediction results in the form of spatial coordinates, and combining the prediction results output in the form of spatial coordinates with geographic information system software to perform visual analysis, and generate a hidden fracture prediction distribution map covering the unknown fracture target area.

[0086] Specifically, the machine learning model that has been trained and converged is applied to the unknown study area to predict hidden faults in the unknown fault target area; the output prediction results are output in the form of spatial coordinates, and the prediction results are visualized and analyzed in combination with the geographic information system (GIS) software to generate a fault distribution map of the coverage area.

[0087] In the embodiment of the present invention, by integrating multi-source detection data such as gravity and magnetism, the underground geological structure can be reflected more comprehensively and deeply; the superposition and comprehensive analysis of multi-source data make the fault characteristics more obvious, greatly improving the ability to identify hidden faults; and through intelligent algorithms and automated analysis technology, the operation steps are simplified, the dependence on high-end equipment and professionals is reduced, and the overall cost is significantly reduced; at the same time, by optimizing the analysis process, a large amount of data can be processed quickly, thereby improving the work efficiency of fault prediction.

[0088] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, a hidden fracture prediction method of any one of the above embodiments is implemented. The computer-readable storage medium includes, but is not limited to, any type of disk (including floppy disk, hard disk, optical disk, CD-ROM, and magneto-optical disk), ROM (Read-Only Memory), RAM (Random Access Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic card or optical card. That is, the storage device includes any medium that can store or transmit information in a readable form by a device (for example, a computer, a mobile phone), which can be a read-only memory, a disk or an optical disk, etc.

[0089] An embodiment of the present invention further provides a computer application program, which runs on a computer and is used to execute the hidden fracture prediction method of any one of the above embodiments.

[0090] also, Figure 4 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention.

[0091] The embodiment of the present invention further provides an electronic device, such as Figure 4 The electronic device includes a processor 402, a memory 403, an input unit 404, a display unit 405 and other devices. Those skilled in the art will understand that Figure 4The structural components of the electronic device shown do not constitute a limitation on all devices, and may include more or fewer components than shown, or combine certain components. The memory 403 can be used to store the application 401 and various functional modules, and the processor 402 runs the application 401 stored in the memory 403, thereby executing various functional applications and data processing of the device. The memory can be an internal memory or an external memory, or include both internal and external memories. The internal memory may include a read-only memory (ROM), a programmable ROM (PROM), an electrically programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a flash memory, or a random access memory. The external memory may include a hard disk, a floppy disk, a ZIP disk, a U disk, a magnetic tape, etc. The memory disclosed in the present invention includes but is not limited to these types of memories. The memory disclosed in the present invention is only used as an example and not as a limitation.

[0092] The input unit 404 is used to receive the input of signals and keywords input by the user. The input unit 404 may include a touch panel and other input devices. The touch panel can collect the user's touch operations on or near it (such as the user's operation on or near the touch panel using any suitable object or accessory such as a finger, a stylus, etc.), and drive the corresponding connection device according to a pre-set program; other input devices may include but are not limited to one or more of a physical keyboard, a function key (such as a play control key, a switch key, etc.), a trackball, a mouse, a joystick, etc. The display unit 405 can be used to display the information input by the user or the information provided to the user and various menus of the terminal device. The display unit 405 can be in the form of a liquid crystal display, an organic light emitting diode, etc. The processor 402 is the control center of the terminal device, which uses various interfaces and lines to connect various parts of the entire device, and executes various functions and processes data by running or executing software programs and / or modules stored in the memory 403, and calling data stored in the memory.

[0093] As an embodiment, the electronic device includes: one or more processors 402, a memory 403, and one or more applications 401, wherein the one or more applications 401 are stored in the memory 403 and are configured to be executed by the one or more processors 402, and the one or more applications 401 are configured to execute the hidden fracture prediction method in any one of the above-mentioned embodiments.

[0094] In the embodiment of the present invention, by integrating multi-source detection data such as gravity and magnetism, the underground geological structure can be reflected more comprehensively and deeply; the superposition and comprehensive analysis of multi-source data make the fault characteristics more obvious, greatly improving the ability to identify hidden faults; and through intelligent algorithms and automated analysis technology, the operation steps are simplified, the dependence on high-end equipment and professionals is reduced, and the overall cost is significantly reduced; at the same time, by optimizing the analysis process, a large amount of data can be processed quickly, thereby improving the work efficiency of fault prediction.

[0095] In addition, the above is a detailed introduction to a hidden fracture prediction method based on multi-source detection data fusion and related devices provided by an embodiment of the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A hidden fault prediction method based on multi-source detection data fusion, characterized in that: The method comprises: Based on the gravity instrument and the magnetic instrument, detection and processing are performed in a designated area to obtain gravity detection data and magnetic detection data, wherein the designated area is a known fracture area, the gravity detection data is Bouguer gravity anomaly data, and the magnetic detection data is aeromagnetic anomaly data; Fusing the gravity detection data and the magnetic detection data to form multi-source fused detection data; Performing enhancement processing on the multi-source fusion detection data based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data; The enhanced multi-source fusion detection data is processed into a data set, and the constructed data set is annotated to form a training sample data set; Inputting the training sample data set into a preset machine learning model for training processing to obtain a machine learning model with training convergence, wherein the machine learning model includes a random forest model and an artificial neural network model; The machine learning model that has converged in training is applied to the unknown fracture target area for hidden fracture prediction processing, the prediction results are output in the form of spatial coordinates, and a hidden fracture prediction distribution map covering the unknown fracture target area is generated.

2. The method for predicting hidden fractures according to claim 1, characterized in that: The fusing the gravity detection data and the magnetic detection data to form multi-source fused detection data includes: Performing abnormal data filtering processing on the gravity detection data and the magnetic detection data to obtain filtered gravity detection data and filtered magnetic detection data; The filtered gravity detection data and the filtered magnetic detection data are smoothed and compensated in sequence to obtain compensated gravity detection data and compensated magnetic detection data; The compensated gravity detection data and the compensated magnetic detection data are fused based on the data fusion algorithm to obtain multi-source fused detection data.

3. The method for predicting hidden fractures according to claim 1, characterized in that: The step of performing enhancement processing on the multi-source fusion detection data based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data includes: Use a variety of data enhancement algorithms to enhance the geological structure in the multi-source fusion detection data to obtain enhanced multi-source fusion detection data; The multiple data enhancement algorithms include total horizontal derivative method, vertical derivative method, vertical second-order derivative, tilt angle method, tilt angle horizontal derivative method, analytical signal method, Theat method, total horizontal derivative tilt angle method, horizontal tilt angle method, enhanced analytical signal method, and enhanced tilt angle method.

4. The method for predicting hidden fractures according to claim 1, characterized in that: The step of constructing a data set for the enhanced multi-source fusion detection data and labeling the constructed data set to form a training sample data set includes: Performing sample synthesis processing on the enhanced multi-source fusion detection data based on a synthetic minority class oversampling algorithm, and performing data set construction processing based on the synthesized sample data to obtain a constructed data set; The existing fracture distribution map and geological data of the designated area are obtained, and the constructed data set is annotated based on the existing fracture distribution map and geological data to form a training sample data set.

5. The method for predicting hidden fractures according to claim 4, characterized in that: The sample synthesis processing of the enhanced multi-source fusion detection data based on the synthetic minority class oversampling algorithm includes: For each minority class sample data X in the enhanced multi-source fusion detection data, the minority class sample X and its nearest neighboring sample data are calculated according to the required oversampling amount. The difference is used to synthesize new samples, which can be specifically expressed as: ； in, represents a random function; Represents synthetic sample data.

6. The method for predicting hidden fractures according to claim 1, characterized in that: The step of inputting the training sample data set into a preset machine learning model for training processing to obtain a machine learning model with converged training includes: When the training sample data set is input into a preset machine learning model for training processing, the hyperparameters in the preset machine learning model are adjusted by a grid search algorithm or a genetic optimization algorithm, and the best parameter combination is selected for training processing to obtain a machine learning model with training convergence; When the preset machine learning model is a random forest model, the hyperparameters are the number of decision trees in the random forest model, the maximum depth of the decision tree, the minimum number of samples contained in a leaf node, and the minimum number of samples that can be divided into nodes; when the preset machine learning model is an artificial neural network model, the hyperparameters are the number of layers and nodes of the artificial neural network model.

7. The method for predicting hidden fractures according to claim 1, characterized in that: The prediction results are output in the form of spatial coordinates, and a hidden fracture prediction distribution map covering the unknown fracture target area is generated, including: The prediction results are output in the form of spatial coordinates, and the prediction results output in the form of spatial coordinates are visualized and analyzed in combination with geographic information system software to generate a hidden fault prediction distribution map covering the unknown fault target area.

8. A hidden fracture prediction device based on multi-source detection data fusion, characterized in that: The device comprises: Detection module: used to perform detection processing in a specified area based on gravity instruments and magnetic instruments to obtain gravity detection data and magnetic detection data. The specified area is a known fracture area, the gravity detection data is Bouguer gravity anomaly data, and the magnetic detection data is aeromagnetic anomaly data; Data fusion module: used for fusing the gravity detection data and the magnetic detection data to form multi-source fusion detection data; Data enhancement module: used to enhance the multi-source fusion detection data based on a preset data enhancement algorithm to obtain enhanced multi-source fusion detection data; Data construction and annotation module: used to construct and process the data set of the enhanced multi-source fusion detection data, and to annotate the constructed data set to form a training sample data set; Model training module: used to input the training sample data set into a preset machine learning model for training processing to obtain a machine learning model with training convergence, wherein the machine learning model includes a random forest model and an artificial neural network model; Prediction generation module: used to apply the trained converged machine learning model to the unknown fracture target area for hidden fracture prediction processing, output the prediction results in the form of spatial coordinates, and generate a hidden fracture prediction distribution map covering the unknown fracture target area.

9. An electronic device comprising a processor and a memory, characterized in that: The processor runs a computer program or code stored in the memory to implement the hidden fracture prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium for storing a computer program or code, characterized in that: When the computer program or code is executed by a processor, the hidden fracture prediction method according to any one of claims 1 to 7 is implemented.

Images