Method of geological disaster prediction based on geophysical parameters

Abstract

The present invention mainly relates to the field of geological prediction technology. The present invention provides a method for predicting geological disasters based on geophysical parameters, the core idea of ​​which is: classify and screen out geophysical parameters with a high degree of correlation with specific geological condition parameters, determine the type of geological disaster with a high degree of correlation with the specific geological condition parameters, use the geophysical parameters with a high degree of correlation with the specific geological condition parameters as key geophysical parameters corresponding to the type of geological disaster with a high degree of correlation with the corresponding geological condition parameters, establish a geological disaster classification model based on the key geophysical parameters corresponding to the type of geological disaster, the geological disaster classification model outputs the possible type of geological disaster based on the input geophysical parameters, calculates the probability of the corresponding type of geological disaster under the conditions of the corresponding geophysical parameters, and makes a decision on whether the corresponding geological disaster will occur according to the set probability threshold of the corresponding type of geological disaster, thereby improving the accuracy and efficiency of geological disaster prediction based on geophysical parameters.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of geological prediction, and in particular to a method for predicting geological disasters based on geophysical exploration parameters. Background Art

[0002] In the prior art, tunnel construction and geological surveys often use a variety of geophysical testing methods to predict the geological conditions of unexcavated sections and possible geological disasters. These geophysical methods include seismic wave testing, electrical testing, and magnetic testing, which reveal geological structures by measuring the different physical properties of underground rocks or soil layers. These data are important for predicting geological disasters caused by adverse geological conditions that may be encountered during tunnel construction, such as tectonic fracture zones, rock bursts, karst, etc.

[0003] However, the existing technology has some significant limitations and shortcomings in the application process. The first is the subjectivity of data analysis. The existing geophysical parameter analysis often relies on the experience and judgment of geophysical and geological experts, which increases the subjectivity of the prediction results to a certain extent, resulting in different experts may come up with different interpretations and predictions; the second is that the parameter correlation is not clear. Traditional methods often consider various geophysical parameters independently, and do not fully explore the correlation between these geophysical parameters, as well as the relationship between them and specific geological conditions, which limits the accuracy and reliability of the prediction model; the third is the arbitrariness of threshold setting. In traditional applications, the threshold setting of geophysical parameters often lacks a unified standard, and different projects may adopt different threshold judgment criteria, which makes it difficult to unify and standardize the results, reducing the efficiency of engineering decision-making. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method for predicting geological disasters based on geophysical parameters, thereby improving the accuracy and efficiency of geological disaster prediction based on geophysical parameters.

[0005] The technical solution adopted by the present invention to solve the above technical problems is:

[0006] Methods for predicting geological hazards based on geophysical parameters include:

[0007] Step S1: Perform sensitivity analysis on geological condition parameters and geophysical parameters, and classify and select geophysical parameters with high correlation with specific geological condition parameters;

[0008] Step S2: performing correlation analysis between geological condition parameters and specific geological disaster types, and determining geological condition parameters with a high degree of correlation with specific geological disaster types;

[0009] Step S3: taking the geophysical parameters with a high correlation with the specific geological condition parameters in step S1 as the key geophysical parameters corresponding to the specific geological disaster type with a high correlation with the corresponding geological condition parameters in step S2, and establishing a geological disaster classification model based on the key geophysical parameters corresponding to the specific geological disaster type;

[0010] Step S4: The geological hazard classification model outputs the possible geological hazard types based on the input geophysical parameters, calculates the probability of the corresponding geological hazard type under the corresponding geophysical parameter conditions, and makes a decision on whether the corresponding geological hazard will occur based on the set probability threshold of the corresponding geological hazard type.

[0011] Furthermore, step S1 specifically includes:

[0012] Step S11: calculating the correlation between the geophysical parameters and the specific geological condition parameters based on the Pearson and Spearman correlation coefficients, and preliminarily screening out the geophysical parameters whose correlation with the specific geological condition parameters is higher than a preset value;

[0013] Step S12: Perform decision tree screening on the geophysical parameters screened in step S11, and use the geophysical parameters screened by the decision tree as key geophysical parameters, wherein the constructed decision tree is: ,in is the information gain, is impurity, and Represents the parent node and the The data set of child nodes, It is The number of samples of nodes, is the number of samples of the parent node, is the number of child nodes, It is the geophysical exploration parameter.

[0014] Furthermore, step S2 is specifically as follows: based on the support and confidence, the correlation between the specific geological hazard type and the geological condition parameters is evaluated, and the geological condition parameters with support and confidence values ​​higher than the set values ​​are screened out as the geological condition parameters with a high correlation with the specific geological hazard type.

[0015] Furthermore, the step S3 of establishing a geological disaster classification model based on key geophysical parameters corresponding to a specific geological disaster type specifically includes:

[0016] A geological hazard classification model is established based on the SVM model. The key geophysical parameters corresponding to specific geological hazard types are used as the input of the SVM model, and the corresponding geological hazard types are used as the output of the SVM model to train the SVM model.

[0017] In the process of training the SVM model, the kernel function parameters and penalty factors of the SVM model are optimized based on grid search and Hyperopt. The optimal kernel function parameters and penalty factors are substituted into the SVM model. The trained SVM model is cross-validated based on accuracy and recall rate to obtain a trained geological disaster classification model.

[0018] Furthermore, the probability threshold of the geological disaster type in step S4 is: ,in is the standard deviation of the geophysical parameters, is the sample mean of the geophysical parameters.

[0019] Further, in step S4, the probability of occurrence of each type of geological disaster is calculated based on the logistic regression model. ,in and For a given geophysical parameter set The following model parameters.

[0020] Furthermore, the method also includes step S5: matching the geological disaster types with corresponding key geophysical parameters, geological condition parameters and probability thresholds of the occurrence of geological disaster types to establish a geological disaster prediction table.

[0021] Furthermore, step S1 also includes preprocessing the image data in the collected geophysical exploration parameters, specifically including:

[0022] Segment the image data of the collected geophysical parameters based on convolutional neural network;

[0023] Extracting geophysical parameters from the segmented image based on feature descriptors;

[0024] The extracted geophysical parameters are converted into formatted text data and stored according to the categories of the geophysical parameters.

[0025] Furthermore, step S1 also includes performing principal component analysis and regularized linearization processing on the same geophysical parameters from different sources to obtain fused geophysical parameters, and screening out geophysical parameters with a high degree of correlation with specific geological condition parameters from the fused geophysical parameters.

[0026] Furthermore, the method also includes, if new geophysical parameters are added, removing some of the geophysical parameters based on the recursive feature elimination method, and the decision tree reselects geophysical parameters with a high degree of correlation with specific geological condition parameters based on the data set after removing some of the geophysical parameters, and then retraining the geological hazard classification model according to steps S2 and S3.

[0027] The beneficial effects of the present invention are:

[0028] The method for predicting geological disasters based on geophysical parameters described in the present invention screens out geophysical parameters with a high degree of correlation with specific geological condition parameters and geological disaster types with a high degree of correlation with specific geological condition parameters, uses the geological condition parameters as associated variables of the geophysical parameters and the geological disaster types, determines a key geophysical parameter combination set with a high degree of correlation with the geological disaster types as the key geophysical parameters, and establishes a geological disaster prediction model based on the key geophysical parameters to predict geological disasters. It not only takes into account the single geological condition reflected by a single geophysical parameter, but also takes into account the comprehensive geological conditions under the combination of multiple geophysical parameters, thereby improving the accuracy and efficiency of geological disaster prediction, providing scientific guidance for engineering construction, significantly reducing the risk of geological disasters, and ensuring engineering safety and benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a flow chart of the method for predicting geological disasters based on geophysical parameters described in the present invention. DETAILED DESCRIPTION

[0030] like Figure 1 As shown, the method for predicting geological disasters based on geophysical parameters described in the present invention includes the following steps.

[0031] Step S1: Collect various geophysical parameters measured by seismic waves, electrical methods and magnetic methods under different geological conditions, including: collecting geophysical parameters according to the topography and geomorphology corresponding to the pile number and test time of the tunnel test location, stratum lithology (such as sandstone, shale), preliminary surrounding rock evaluation index, geological structure (such as faults, joints) and geophysical test time. In practice, the collected geophysical parameters include a large amount of image data, so before performing sensitivity analysis of geophysical parameters and geological condition parameters, the image data needs to be preprocessed and converted into formatted text data that can be recognized by the machine.

[0032] In this embodiment, the preprocessing step of the image data includes: segmenting the image data using a convolutional neural network. In this embodiment, a U-Net segmentation network is selected: The collected image type geophysical parameters are segmented, where is the pixel in the output feature map, is the input image, is the weight of the convolution kernel, and Defines the size of the convolution kernel.

[0033] Based on the feature descriptor SIFT, the geophysical parameter feature extraction is performed in the segmented image, and the shape, size, and parameter meaning of each geophysical parameter in the image are described in detail. In the SIFT algorithm, the construction of each geophysical parameter feature vector can be obtained by the following steps: ;in, is the Laplace function in scale space, is the Gaussian blur function, is the input image, Represents the convolution operation. The direction of each feature is assigned The calculation of the gradient yields: ; and The images are and Each geophysical parameter feature is encoded as a feature vector, which includes position information and parameter information.

[0034] The feature vector text of the extracted geophysical parameters is converted into formatted text data, and data entries are constructed and stored according to the geophysical parameter type corresponding to the feature vector. In this embodiment, the feature vector is stored in txt format to facilitate data transmission and application reading.

[0035] Step S2: Conduct sensitivity analysis on geological condition parameters and geophysical parameters, and classify and select geophysical parameters with a high degree of correlation with specific geological condition parameters.

[0036] In the present invention, the traditional statistical method is combined with the decision tree algorithm to screen the geophysical parameters; first, the correlation between the geophysical parameters and the specific geological condition parameters is calculated based on the Pearson and Spearman correlation coefficients, and the geophysical parameters with a correlation degree higher than the preset value with the specific geological condition parameters are preliminarily screened out; after the geophysical parameters with a correlation degree higher than the preset value with the specific geological condition parameters are preliminarily screened out based on the Pearson and Spearman correlation coefficients, the geophysical parameters with a higher correlation degree with the specific geological condition parameters are further screened out as key geophysical parameters based on the decision tree, and the decision tree selects the geophysical parameters that can best reduce the prediction uncertainty by evaluating the information gain or the Gini impurity, and the constructed decision tree is: ,in is the information gain, is impurity, and Represents the parent node and the The data set of child nodes, It is The number of samples of nodes, is the number of samples of the parent node, is the number of child nodes, It is the geophysical exploration parameter.

[0037] Since the same geophysical parameter may have data sources of multiple measurement methods, this embodiment uses principal component analysis (PCA) and regularized linear models to comprehensively process data from geological radar, electrical testing and other related geophysical technologies for geophysical test formatting data from different sources. ,in Represents the original data of geophysical parameters from different sources, Represents the weight coefficient of the geophysical parameters of the corresponding source, Indicates the first Geophysical parameters.

[0038] In this embodiment, a visualization tool is also provided to display a heat map or scatter plot of geophysical parameters and geological condition parameters. The user can observe the correlation changes between the geophysical parameters and the geological condition parameters in real time by adjusting the parameter settings, and explore how different geophysical parameters affect the geological condition parameters, so as to better understand the relationship between the data and manually adjust the geological hazard classification model established subsequently. The main formula is: ,in and Represent the geophysical parameter set and specific geological condition parameters respectively. is the heat map function, Represents a geophysical parameter set and geological parameters The correlation coefficient between .

[0039] Step S3: Conduct a correlation analysis between geological condition parameters and specific geological disaster types to determine geological condition parameters that are highly correlated with specific geological disaster types.

[0040] In this embodiment, the correlation between geological condition parameters and geological disaster types is evaluated based on support and confidence. Support: , confidence level: In the present invention, p represents a geological condition parameter, d represents a geological disaster type, It indicates the proportion of data in which a certain type of geological condition parameter and geological disaster type appear simultaneously in all data records. A high support degree means that the geological condition parameter and the corresponding geological disaster type appear simultaneously at a high frequency, indicating that there may be a strong correlation between them. It indicates the proportion of data records containing a certain type of geological disaster among all data records containing geological condition parameters. A high confidence level indicates that when the geological condition parameter appears, the probability of the corresponding type of geological disaster occurring is very high.

[0041] Step S4: Use the geophysical parameters with a high degree of correlation with the specific geological condition parameters in step S2 as the key geophysical parameters corresponding to the specific geological disaster type with a high degree of correlation with the corresponding geological condition parameters in step S3, and establish a geological disaster classification model based on the key geophysical parameters corresponding to the specific geological disaster type.

[0042] In the present invention, a support vector machine model is developed based on the correlation between key geophysical parameters and geological disasters to establish a geological disaster classification model to classify geological disasters. The working principle of the SVM model is to find the optimal classification hyperplane based on maximizing the interval between different categories. The optimization formula is as follows: , the constraints are: ,in, , is the weight vector, is the bias, is the slack variable, are slack variables for different categories of parameters, is the regularization parameter, and are the labels and feature vectors of the training samples respectively.

[0043] The key geophysical parameters corresponding to specific geological hazard types are used as the input of the SVM model, and the corresponding geological hazard types are used as the output of the SVM model to train the SVM model. In the process of training the SVM model, the kernel function parameters and penalty factors of the SVM model are optimized based on grid search and Hyperopt. and The trained SVM model is cross-validated. After passing the validation, the trained geological disaster classification model is obtained, where TP, TN, FP and FN represent the number of true positive, true negative, false positive and false negative samples, respectively.

[0044] As a preferred embodiment, in this embodiment, the decision tree is combined with the recursive feature elimination (RFE) method, and the decision tree model is passed to the RFE as the basic model, and combined with the feedback of the trained geological hazard classification model to adapt to the addition of new geophysical parameters. The importance evaluation of geophysical parameters by RFE is described as: , It is a feature The weight in the decision tree model. When the geophysical parameters are removed one by one, the performance change of the decision tree model is the change in the model performance index containing all features and the model performance index after removing the feature fi. In this way, RFE not only selects geophysical parameters based on the importance of geophysical parameters in the current data set, but also dynamically adjusts the geophysical parameter selection strategy to adapt to the update of new geophysical parameter data, ensuring that the established geological hazard classification model always maintains the best prediction ability.

[0045] Step S5: The geological hazard classification model outputs the possible geological hazard types based on the input geophysical parameters, calculates the probability of the corresponding geological hazard type under the corresponding geophysical parameter conditions, and makes a decision on whether the corresponding geological hazard will occur based on the set probability threshold of the corresponding geological hazard type.

[0046] The present invention uses the coefficient of variation of geophysical parameters to set the threshold for geological disasters , to describe the ratio of the variation degree in the geophysical parameters to the average value of the geophysical parameters, the formula is as follows: ,in is the standard deviation of the geophysical parameter samples, is the mean value of the sample geophysical parameter sample. The geological disaster classification model is based on the logistic regression model to evaluate the probability of disaster occurrence corresponding to each indicator, where the threshold is used as the decision boundary, and the formula is expressed as: ,in Given the geophysical parameters The probability of geological disasters occurring is and It is a model parameter, which makes a decision on whether the corresponding geological disaster will occur based on the set threshold of the corresponding geological disaster type and the probability threshold of the geological disaster calculated by the geological disaster classification model.

[0047] In the present invention, in order to ensure the accuracy and applicability of the threshold value for the occurrence of geological disasters, the threshold value for the occurrence of geological disasters is regularly evaluated according to a set time period, and adjusted and compiled into a geological disaster prediction table to guide and support engineering decision-making.

Claims

1. A method for predicting geological disasters based on geophysical parameters, characterized in that: include: Step S1: Perform sensitivity analysis on geological condition parameters and geophysical parameters, and classify and select geophysical parameters with high correlation with specific geological condition parameters; Step S2: performing correlation analysis between geological condition parameters and specific geological disaster types, and determining geological condition parameters with a high degree of correlation with specific geological disaster types; Step S3: taking the geophysical parameters with a high correlation with the specific geological condition parameters in step S1 as the key geophysical parameters corresponding to the specific geological disaster type with a high correlation with the corresponding geological condition parameters in step S2, and establishing a geological disaster classification model based on the key geophysical parameters corresponding to the specific geological disaster type; Step S4: The geological hazard classification model outputs the possible geological hazard types based on the input geophysical parameters, calculates the probability of the corresponding geological hazard type under the corresponding geophysical parameter conditions, and makes a decision on whether the corresponding geological hazard will occur based on the set probability threshold of the corresponding geological hazard type.

2. The method for predicting geological disasters based on geophysical parameters according to claim 1, characterized in that: Step S1 specifically includes: Step S11: calculating the correlation between the geophysical parameters and the specific geological condition parameters based on the Pearson and Spearman correlation coefficients, and preliminarily screening out the geophysical parameters whose correlation with the specific geological condition parameters is higher than a preset value; Step S12: Perform decision tree screening on the geophysical parameters screened in step S11, and use the geophysical parameters screened by the decision tree as key geophysical parameters, wherein the constructed decision tree is: ,in is the information gain, is impurity, and Represents the parent node and the The data set of child nodes, It is The number of samples of nodes, is the number of samples of the parent node, is the number of child nodes, It is the geophysical exploration parameter.

3. The method for predicting geological disasters based on geophysical parameters according to claim 1, characterized in that: Step S2 is specifically: based on the support and confidence, the correlation between the specific geological hazard type and the geological condition parameters is evaluated, and the geological condition parameters with support and confidence values ​​higher than the set values ​​are selected as the geological condition parameters with a high correlation with the specific geological hazard type.

4. The method for predicting geological disasters based on geophysical parameters according to claim 1, characterized in that: The step S3 of establishing a geological disaster classification model based on key geophysical parameters corresponding to a specific geological disaster type specifically includes: A geological hazard classification model is established based on the SVM model. The key geophysical parameters corresponding to specific geological hazard types are used as the input of the SVM model, and the corresponding geological hazard types are used as the output of the SVM model to train the SVM model. In the process of training the SVM model, the kernel function parameters and penalty factors of the SVM model are optimized based on grid search and Hyperopt. The optimal kernel function parameters and penalty factors are substituted into the SVM model. The trained SVM model is cross-validated based on accuracy and recall rate to obtain a trained geological disaster classification model.

5. The method for predicting geological disasters based on geophysical parameters according to claim 1, characterized in that: The probability threshold of the geological disaster type described in step S4 is: ,in is the standard deviation of the geophysical parameters, is the sample mean of the geophysical parameters.

6. The method for predicting geological disasters based on geophysical parameters according to claim 1, characterized in that: In step S4, the probability of occurrence of each type of geological disaster is calculated based on the logistic regression model. ,in and For a given geophysical parameter set The following model parameters.

7. The method for predicting geological disasters based on geophysical parameters according to claim 1, characterized in that: The method also includes step S5: matching the geological disaster types with corresponding key geophysical parameters, geological condition parameters and probability thresholds of the occurrence of geological disaster types to establish a geological disaster prediction table.

8. The method for predicting geological disasters based on geophysical parameters according to any one of claims 1 to 7, characterized in that: Step S1 also includes preprocessing the image data in the collected geophysical exploration parameters, specifically including: Segment the image data of the collected geophysical parameters based on convolutional neural network; Extracting geophysical parameters from the segmented image based on feature descriptors; The extracted geophysical parameters are converted into formatted text data and stored according to the categories of the geophysical parameters.

9. The method for predicting geological disasters based on geophysical parameters according to claim 8, characterized in that: Step S1 also includes performing principal component analysis and regular linearization processing on the same geophysical parameters from different sources to obtain fused geophysical parameters, and screening out geophysical parameters with a high degree of correlation with specific geological condition parameters from the fused geophysical parameters.

10. The method for predicting geological disasters based on geophysical parameters according to claim 8, characterized in that: The method also includes, if new geophysical parameters are added, removing some of the geophysical parameters based on the recursive feature elimination method, and the decision tree reselects geophysical parameters with a high degree of correlation with specific geological condition parameters based on the data set after removing some of the geophysical parameters, and then retraining the geological disaster classification model according to steps S2 and S3.

Images