Method for predicting geology of tunnel face based on geophysical prospecting parameters and geologic characteristic parameters

Abstract

The invention mainly relates to the technical field of geological prediction. In order to improve the accuracy and stability of a tunnel face geological prediction result, the invention provides a method for performing tunnel face geological prediction based on geophysical prospecting parameters and geological characteristic parameters, and the core idea of the method is as follows: extracting key geophysical prospecting parameters for performing geological prediction on a tunnel face from geophysical prospecting parameters of an excavated tunnel face; geologic characteristic parameters of the excavated tunnel face are collected according to the formulated geologic characteristic parameter description standard, and key geologic characteristic parameters for conducting geologic prediction on the tunnel face are obtained; a tunnel face geological prediction model is established, and the tunnel face geological prediction model comprehensively considers the characteristics of a geologic body from the two aspects of physical properties reflected by the key geophysical prospecting parameters and the geological background reflected by the key geologic characteristic parameters based on the key geophysical prospecting parameters and the key geologic characteristic parameters; and rock grade categories and geological disasters under complex geological conditions can be predicted more accurately.

Description

Technical Field

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

[0002] In the field of geological survey, especially in infrastructure construction projects of large hydropower stations, such as the construction of underground caverns such as water diversion tunnels and water diversion traffic tunnels, the prediction of geological conditions behind the face is crucial to ensure engineering safety and optimize design. Although traditional geological exploration methods, including geophysical exploration and direct drilling, provide basic geological data, they often face limitations in predicting the geological conditions behind complex faces due to long data processing time, strong subjectivity in interpretation, and insufficient data integration. Especially in the forward prediction of the face, these methods often cannot provide sufficient response speed and accuracy, thus affecting the timeliness and effectiveness of engineering decisions. Summary of the invention

[0003] Technical Problems to be Solved by the Invention

[0004] A method for tunnel face geological prediction based on geophysical parameters and geological characteristic parameters is provided, aiming to improve the accuracy and efficiency of tunnel face geological prediction results.

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

[0006] The method of predicting the geological conditions of the tunnel face based on geophysical parameters and geological characteristic parameters includes:

[0007] Step S1: extracting key geophysical parameters for geological prediction of the tunnel face from the geophysical parameters of the excavated tunnel face;

[0008] Step S2: collecting geological characteristic parameters of the excavated tunnel face according to the established geological characteristic parameter description standard, matching the key geophysical parameters with the collected geological characteristic parameters, and obtaining key geological characteristic parameters for geological prediction of the tunnel face;

[0009] Step S3: Establishing a tunnel face geological prediction model, taking the key geophysical parameters and key geological characteristic parameters as inputs of the tunnel face geological prediction model, taking the rock grade category of the tunnel face and possible geological disasters as outputs, and training the established tunnel face geological prediction model;

[0010] Step S4: Based on the trained tunnel face geological prediction model, geological prediction is performed on the unexcavated tunnel face to obtain the rock grade category and possible geological disaster types of the unexcavated tunnel face.

[0011] Further, step S1 includes:

[0012] S11: Preprocess the collected geophysical parameters and standardize the preprocessed geophysical parameters Where X is the original data of geophysical parameters, μ and σ are the mean and standard deviation of the corresponding geophysical parameters, and Z is the standardized geophysical parameters;

[0013] S12: Data fusion of geophysical parameters from multiple sources: X 融合 =∑w i x i , where x i is the geophysical parameter of the ith source, w i is the weight coefficient of the geophysical parameters of the corresponding source;

[0014] S13: Screening the fused geophysical parameters based on principal component analysis, evaluating the importance of the geophysical parameters screened by principal component analysis based on the XGBoost algorithm, and selecting key geophysical parameters based on the evaluation results;

[0015] Among them, the objective function of the XGBoost algorithm is: is the loss function, y i is the true value of the i-th sample, is the predicted value of the i-th sample, K is the number of trees, and f k is the complexity of the Kth tree, Ω(f k ) is the regularization term, T K is the number of leaf nodes in the tree, ω j is the weight of the leaf node, γ and λ are regularization parameters.

[0016] Furthermore, the formulation of the geological characteristic parameter description standard in step S2 specifically includes:

[0017] Step S21: defining the types of geological characteristic parameters, including: stratum lithology, weathering unloading, structural development, rock mass structure, groundwater conditions, and fracture density and distribution of the face;

[0018] Step S22: setting description parameters for the defined geological characteristic parameters to quantify the geological characteristic parameters;

[0019] Step S23: formulating a unified geological characteristic parameter data entry format based on the set geological characteristic parameter description parameters.

[0020] Furthermore, in step S2, principal component analysis is used to determine the correspondence between key geophysical parameters and the collected geological characteristic parameters, so as to obtain key geological characteristic parameters for geological prediction of the tunnel face.

[0021] Furthermore, in step S3, a tunnel face geological prediction model is established based on a convolutional neural network; the tunnel face geological prediction model includes an input layer, a hidden layer and an output layer:

[0022] The input layer is used to input the key geophysical parameters and key geological characteristic parameters of the tunnel face;

[0023] The hidden layer is used to extract the characteristic vectors of the input key geophysical parameters and key geological characteristic parameters. The hidden layer includes a convolution layer, an activation layer and a pooling layer. The convolution layer is specifically: Where f(x,y) is the pixel in the output feature map, g(x,y) is the input image, ω(i,j) is the weight of the convolution kernel, and a and b define the size of the convolution kernel; the activation layer uses the ReLU activation function h(x)=max(0,x); the pooling layer uses maximum pooling or average pooling to reduce the feature dimension;

[0024] The output layer is the probability of occurrence of rock grade category and geological disasters at the face, P(y|x; θ) = softmax(Wh+b), where P(y|x; θ) is the probability of occurrence of rock grade category or geological disaster y under a given set of key geophysical parameters and key geological characteristic parameters x and model parameters θ, W is the weight matrix, h is the output of the hidden layer, and b is the bias vector.

[0025] Furthermore, in step S4, the cross entropy loss function is used to evaluate the level categories of the tunnel face geological prediction model and the prediction results of possible geological disasters. Among them, L represents the loss value, y o,c is the one-hot encoding of the true label, P O,C is the prediction probability of each geological prediction result of the tunnel face geological parameter prediction model, and M is the total number of geological prediction categories that may be output.

[0026] Furthermore, during the training process of the tunnel face geological prediction model, the Adam optimizer is used to update the parameters of the tunnel face geological prediction model.

[0027] Beneficial effects of the present invention

[0028] (1) The method for performing face geological prediction based on geophysical parameters and geological characteristic parameters provided by the present invention effectively integrates geophysical parameters and geological characteristic parameters based on a machine learning method, obtains key geophysical parameters for geological prediction of the face, and screens out geological characteristic parameters corresponding to the key geophysical parameters as key geological characteristic parameters, establishes a face geological prediction model based on the key geophysical parameters and key geological characteristic parameters to predict geological disasters, and comprehensively considers the characteristics of the geological body from two perspectives, namely, the physical properties reflected by the key geophysical parameters and the geological background reflected by the key geological characteristic parameters, so as to more deeply associate the formation mechanism of geological disasters. The established face geological prediction model can more accurately predict rock grade categories and geological disasters under complex geological conditions;

[0029] (2) By first determining the key geophysical parameters that can characterize the physical properties of the geological area and then conducting a detailed analysis in combination with the geological characteristic parameters, it is possible to avoid conducting a large amount of exploration work in non-critical areas, thereby reducing exploration costs, reducing human intervention, and improving the convenience and automation of geological prediction operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flow chart of the method for predicting tunnel face geology based on geophysical parameters and geological characteristic parameters according to the present invention. DETAILED DESCRIPTION

[0031] like Figure 1 As shown, the method for predicting the geology of the tunnel face based on geophysical parameters and geological characteristic parameters described in the present invention specifically includes the following contents.

[0032] Step S1: Unify the description standards of the geological characteristic parameters of the tunnel face:

[0033] First, define the standards for geological characteristic parameters, including:

[0034] ① Stratigraphic lithology: record in detail the basic physical properties of the rock mass, such as type, color, texture, and hardness.

[0035] ② Weathering unloading: describes the degree of rock weathering, such as the thickness of the surface weathering layer, color changes, degree of structural looseness, etc.

[0036] ③ Structural development: record the type, scale, direction, dip, etc. of structures such as faults and folds.

[0037] ④ Rock structure conditions: including the development of structural surfaces such as bedding and joints of the rock mass, such as spacing, extension length, degree of opening and closing, etc.

[0038] ⑤ Groundwater conditions: describe the type of groundwater, such as still water or flowing water, as well as the water level, flow direction, flow rate, etc.

[0039] ⑥Crack density and distribution: quantify the number of cracks per unit area and their spatial distribution pattern.

[0040] Descriptive parameters are set for the above geological characteristic parameters, and all geological characteristic parameters are quantified and statistically classified to a certain extent. A unified data entry format is also formulated to ensure that information consistency can be maintained between different survey teams and different projects, including the creation of standardized tables, input interfaces, or the use of specific software applications to collect and store geological characteristic parameters. In addition, the classification of geological characteristic parameters and their descriptive parameters need to be regularly reviewed and field-verified to ensure the applicability and accuracy of the standards, and necessary updates should be made based on the latest geological survey conditions and technological advances.

[0041] Step S2: Use geophysical exploration equipment such as HSP, geological radar and transient electromagnetic instrument to collect geophysical data of the excavated tunnel face, and identify and extract key geophysical parameters for geological prediction of the tunnel face from the collected geophysical exploration data, including:

[0042] S21: Preprocess the collected geophysical parameters and standardize the preprocessed geophysical parameters Where X is the original data of geophysical parameters, μ and σ are the mean and standard deviation of the corresponding geophysical parameters, and Z is the standardized geophysical parameters;

[0043] S22: Data fusion of geophysical parameters from multiple sources: X 融合 =∑w i x i , where x i is the geophysical parameter of the ith source, w i is the weight coefficient of the geophysical parameters of the corresponding source;

[0044] S23: Screening the fused geophysical parameters based on principal component analysis, evaluating the importance of the geophysical parameters screened by principal component analysis based on the XGBoost algorithm, and selecting key geophysical parameters based on the evaluation results;

[0045] Among them, the objective function of the XGBoost algorithm is: is the loss function, y i is the true value of the i-th sample, is the predicted value of the i-th sample, K is the number of trees, and f k is the complexity of the Kth tree, Ω(f k ) is the regularization term, T K is the number of leaf nodes in the tree, ω j is the weight of the leaf node, γ and λ are regularization parameters.

[0046] Step S3: Match the key geophysical parameters with the collected geological characteristic parameters to obtain the key geological characteristic parameters for geological prediction of the tunnel face. Specifically including:

[0047] Independent component analysis (ICA) is used to process geological characteristic parameters. ICA separates mixed signal sources by maximizing statistical independence, and its mathematical model can be expressed as: X = AS;

[0048] Where X is the observed mixed signal matrix, A is the mixing matrix, and S is the independent source signal matrix. In this way, ICA helps identify the unique signals of various geological features, such as fracture density and rock physical properties, thereby improving the feature extraction quality and prediction accuracy.

[0049] Formatting and data cleaning of key geophysical parameters and geological characteristic parameters, including removal of outliers and noise, ensures that all collected and integrated key geophysical parameters and geological characteristic parameters follow the established formats and standards to facilitate subsequent machine learning model processing.

[0050] Data cleaning uses the Isolation Forest algorithm to identify and remove outliers. The algorithm is based on the following model:

[0051]

[0052] Where E(h(x)) is the average path length of x from the isolation tree, n is the number of samples, c(n) is the average path length under normal circumstances, and in addition, in order to adapt the model training, target encoding is implemented on the data to handle categorical features.

[0053] Step S4: Match the key geophysical parameters with the collected geological characteristic parameters to obtain the most influential geological characteristic parameters for geological prediction of the tunnel face.

[0054] The present invention adopts the principal component analysis method to determine the correspondence between key geophysical parameters and collected geological characteristic parameters, and obtains the key geological characteristic parameters for geological prediction of the tunnel face.

[0055] Step S5: Establishing a tunnel face geological prediction model, taking key geophysical parameters and key geological characteristic parameters as inputs of the tunnel face geological prediction model, taking the rock grade category of the tunnel face and possible geological disasters as outputs, and training the established tunnel face geological prediction model;

[0056] In the present invention, a tunnel face geological prediction model is established based on a convolutional neural network, and the trained tunnel face geological prediction model includes an input layer, a hidden layer and an output layer;

[0057] The input layer is used to input the key geophysical parameters and key geological characteristic parameters of the tunnel face;

[0058] The main function of the hidden layer is to extract high-level abstract features from key geophysical parameters and key geological characteristic parameters layer by layer. The design of the output layer depends on the nature of the prediction task. For example, the output layer of the classification task will use the softmax activation function, while the regression task may use a linear activation function:

[0059]

[0060] Where θ is the parameter, η is the learning rate, and are the bias-corrected first- and second-order moment estimates, respectively, and ∈ is a small constant to prevent division by zero.

[0061] The hidden layer includes a convolution layer, an activation layer and a pooling layer, and the convolution layer is specifically: Where f(x,y) is the pixel in the output feature map, g(x,y) is the input image, ω(i,j) is the weight of the convolution kernel, and a and b define the size of the convolution kernel; the activation layer uses the ReLU activation function h(x)=max(0,x); the pooling layer uses maximum pooling or average pooling to reduce the feature dimension;

[0062] The output layer is the probability of occurrence of rock grade category and geological disasters at the face, P(y|x; θ) = softmax(Wh+b), where P(y|x; θ) is the probability of occurrence of rock grade category or geological disaster y under a given set of key geophysical parameters and key geological characteristic parameters x and model parameters θ, W is the weight matrix, h is the output of the hidden layer, and b is the bias vector.

[0063] The cross entropy loss function is used to evaluate the level categories and possible geological disaster prediction results of the tunnel face geological prediction model. The formula is:

[0064]

[0065] Among them, L represents the loss value, y o,c is the one-hot encoding of the true label, P O,C is the prediction probability of each geological prediction result of the tunnel face geological parameter prediction model, and M is the total number of geological prediction categories that may be output.

[0066] In the process of training the tunnel face geological prediction model, the Adam optimizer is used to update the parameters of the tunnel face geological prediction model. The update rule includes the first-order moment estimation (mean) and the second-order moment estimation (uncentered variance) of the gradient.

[0067] In the process of training the face geological prediction model, batch processing is performed on the key geophysical parameters and geological characteristic parameters input, and the model is trained through small batches of data to reduce memory consumption. At the same time, the risk of overfitting is reduced through random selection of each batch. At the same time, regularization is performed, and the Dropout technology is used to randomly "discard" a part of the neurons during the training process to avoid overfitting of the model to the training data. The formula is expressed as:

[0068]

[0069] in, is the Bernoulli distribution of the variable, r (l) is a randomly generated (l) A 0-1 vector of the same dimension, p is the probability of retaining the neuron.

[0070] The performance of the geological disaster prediction model is evaluated using indicators such as precision and recall rate to ensure the accuracy and reliability of the model in actual tests.

[0071] Step S6: Based on the trained tunnel face geological prediction model, geological prediction is performed on the unexcavated tunnel face to obtain the rock grade category and possible geological disaster types of the unexcavated tunnel face, and the prediction results are provided to the on-site geological engineers and related project management teams as a reference for actual construction and engineering design.

Claims

1. A method for predicting tunnel face geology based on geophysical parameters and geological characteristic parameters, characterized in that: include: Step S1: extracting key geophysical parameters for geological prediction of the tunnel face from the geophysical parameters of the excavated tunnel face; Step S2: collecting geological characteristic parameters of the excavated tunnel face according to the established geological characteristic parameter description standard, matching the key geophysical parameters with the collected geological characteristic parameters, and obtaining key geological characteristic parameters for geological prediction of the tunnel face; Step S3: Establishing a tunnel face geological prediction model, taking the key geophysical parameters and key geological characteristic parameters as inputs of the tunnel face geological prediction model, taking the rock grade category of the tunnel face and possible geological disasters as outputs, and training the established tunnel face geological prediction model; Step S4: Based on the trained tunnel face geological prediction model, geological prediction is performed on the unexcavated tunnel face to obtain the rock grade category and possible geological disaster types of the unexcavated tunnel face.

2. The method for predicting tunnel face geology based on geophysical parameters and geological characteristic parameters according to claim 1, characterized in that: Step S1 includes: S11: Preprocess the collected geophysical parameters and standardize the preprocessed geophysical parameters Where X is the original data of geophysical parameters, μ and σ are the mean and standard deviation of the corresponding geophysical parameters, and Z is the standardized geophysical parameters; S12: Data fusion of geophysical parameters from multiple sources: X 融合 =∑w i x i , where x i is the geophysical parameter of the ith source, w i is the weight coefficient of the geophysical parameters of the corresponding source; S13: Screening the fused geophysical parameters based on principal component analysis, evaluating the importance of the geophysical parameters screened by principal component analysis based on the XGBoost algorithm, and selecting key geophysical parameters based on the evaluation results; Among them, the objective function of the XGBoost algorithm is: is the loss function, y i is the true value of the i-th sample, is the predicted value of the i-th sample, K is the number of trees, and f k is the complexity of the Kth tree, Ω(f k ) is the regularization term, T K is the number of leaf nodes in the tree, ω j is the weight of the leaf node, γ and λ are regularization parameters.

3. The method for predicting tunnel face geology based on geophysical parameters and geological characteristic parameters according to claim 1, characterized in that: The formulation of the geological characteristic parameter description standard in step S2 specifically includes: Step S21: defining the types of geological characteristic parameters, including: stratum lithology, weathering unloading, structural development, rock mass structure, groundwater conditions, and fracture density and distribution of the face; Step S22: setting description parameters for the defined geological characteristic parameters to quantify the geological characteristic parameters; Step S23: formulating a unified geological characteristic parameter data entry format based on the set geological characteristic parameter description parameters.

4. The method for predicting tunnel face geology based on geophysical parameters and geological characteristic parameters according to claim 1, characterized in that: In step S2, the principal component analysis method is used to determine the correspondence between the key geophysical parameters and the collected geological characteristic parameters, so as to obtain the key geological characteristic parameters for geological prediction of the tunnel face.

5. The method for predicting tunnel face geology based on geophysical parameters and geological characteristic parameters according to claim 1, characterized in that: In step S3, a tunnel face geological prediction model is established based on a convolutional neural network; the tunnel face geological prediction model includes an input layer, a hidden layer and an output layer; The input layer is used to input the key geophysical parameters and key geological characteristic parameters of the tunnel face; The hidden layer is used to extract the characteristic vectors of the input key geophysical parameters and key geological characteristic parameters. The hidden layer includes a convolution layer, an activation layer and a pooling layer. The convolution layer is specifically: Where f(x,y) is the pixel in the output feature map, g(x,y) is the input image, ω(i,j) is the weight of the convolution kernel, and a and b define the size of the convolution kernel; the activation layer uses the ReLU activation function h(x)=max(0,x); the pooling layer uses maximum pooling or average pooling to reduce the feature dimension; The output layer is the probability of occurrence of rock grade category and geological disasters at the face, P(y|x; θ) = softmax(Wh+b), where P(y|x; θ) is the probability of occurrence of rock grade category or geological disaster y under a given set of key geophysical parameters and key geological characteristic parameters x and model parameters θ, W is the weight matrix, h is the output of the hidden layer, and b is the bias vector.

6. The method for predicting tunnel face geology based on geophysical parameters and geological characteristic parameters according to claim 5, characterized in that: In step S4, the cross entropy loss function is used to evaluate the level category of the tunnel face geological prediction model and the prediction results of possible geological disasters. Among them, L represents the loss value, y o,c is the one-hot encoding of the true label, P O,C is the prediction probability of each geological prediction result of the tunnel face geological parameter prediction model, and M is the total number of geological prediction categories that may be output.

7. The method for predicting tunnel face geology based on geophysical parameters and geological characteristic parameters according to claim 5, characterized in that: During the training process of the tunnel face geological prediction model, the Adam optimizer is used to update the parameters of the tunnel face geological prediction model.

Images