Tunnel smooth blasting effect prediction method with adaptive blasthole parameters

Abstract

The present invention discloses a method for predicting the smooth blasting effect of a tunnel with self-adaptive blast hole parameters, including: constructing a smooth blasting effect prediction model based on a graph neural network; using the smooth blasting effect prediction model to predict the rock mass displacement data; and calculating the smooth blasting effect of the tunnel in combination with the prediction results of the rock mass displacement data. The present invention constructs a fast prediction model for the smooth blasting effect of self-adaptive blast hole parameters based on a deep graph neural network, uses the graph neural network to comprehensively extract smooth blasting data, improves the accuracy of the calculation method based on deep learning, and has important theoretical significance and application value for tunnel blasting construction.

Description

Technical Field

[0001] The present invention relates to the field of tunnel smooth blasting construction, and is a method for predicting the smooth blasting effect of a tunnel with adaptive hole parameters based on an artificial intelligence algorithm. Background Technique

[0002] The smooth blasting effect is an important factor affecting the quality and safety of blasting engineering. Accurate prediction and optimization can not only improve the efficiency of blasting operations, reduce construction costs, but also effectively reduce the safety hazards caused by poor smooth blasting effects. In blasting operations, the selection of hole parameters (such as hole depth, hole spacing, charge amount, etc.) directly affects the quality and failure mode of the smooth blasting effect.

[0003] Currently, the research on smooth blasting using traditional methods includes the "Intelligent Optimization System for Smooth Blasting of Tunnel Drilling and Blasting Method Based on Digital Twin", which proposes to use information modules to obtain surrounding rock parameters, tunnel parameters, hole layout parameters, and charging parameters; then optimize the construction process parameters, and based on the optimized parameters, construct an explosive state model, a rock mass state model, and an air and water medium state model; collect images of the broken stones and rock walls after blasting, detect the particle size of the broken stones, and obtain the overbreak and underbreak degree and the semi-hole retention degree as reference indicators for evaluating the blasting quality, and study the blasting effect. Traditional methods usually rely on experience or manual adjustment to optimize the hole design. Although they can provide certain guidance for the project, they lack dynamic adaptability when facing complex geological conditions, changing hole parameters, and various non-linear factors.

[0004] Different from traditional rule-based prediction models, deep learning models have strong generalization ability and can handle complex, non-linear data relationships. Through continuous optimization and update of the model, it can adapt to different geological conditions and different types of blasting operations, and self-adjust according to new blasting data. Currently, the research on smooth blasting using artificial intelligence methods includes the "Prediction Method and System for Smooth Blasting of Highway Tunnels", which proposes to establish sample data for the PSO-GA-LSSVM model, train the PSO-GA-LSSVM model through the sample data, and use the trained model to predict the smooth blasting effect of highway tunnels. However, the existing artificial intelligence methods have strict requirements for usage conditions and cannot adapt to changes in hole parameters; moreover, during the construction of sample data, due to inevitable data deviations during the conversion from unstructured data to structured data, it brings unpredictable deviations to the prediction accuracy of the artificial intelligence model. Summary of the Invention

[0005] In order to improve the prediction accuracy and real-time performance of the smooth surface effect in blasting engineering, the present invention provides a prediction method for the smooth surface effect of tunnels with self-adaptive blast hole parameters based on artificial intelligence algorithms, constructs a fast prediction model for the smooth surface effect of self-adaptive blast hole parameters based on a deep graph neural network, uses the graph neural network to comprehensively extract smooth blasting data, improves the accuracy of the calculation method based on deep learning, and has important theoretical significance and application value for tunnel blasting construction.

[0006] The technical solution for achieving the purpose of the present invention is as follows:

[0007] In the first aspect, a prediction method for the smooth surface effect of tunnels with self-adaptive blast hole parameters is provided, including the following steps:

[0008] Based on the graph neural network, construct a smooth surface effect prediction model;

[0009] Use the smooth surface effect prediction model to predict the rock mass displacement data;

[0010] Combined with the prediction results of the rock mass displacement data, calculate the smooth surface effect of the tunnel.

[0011] Preferably, the steps for constructing the smooth surface effect prediction model include:

[0012] Construct a smooth blasting effect graph database including a training data set and a validation data set;

[0013] Based on the graph neural network, construct a neural network model;

[0014] Use the training data set and the validation data set to train and optimize the neural network model to obtain a smooth surface effect prediction model.

[0015] Preferably, the steps for constructing the smooth blasting effect graph database including a training data set and a validation data set include:

[0016] Perform mesh division on the three-dimensional geometric model of the tunnel face to obtain face mesh data;

[0017] Encode the three-dimensional position data of the blast holes, the blast hole depth data, the blast hole blasting time sequence data, and the blast hole blasting charge data into vectors in sequence, and splice them after vector normalization to obtain a blast hole parameter matrix;

[0018] Use the meshes in the face mesh data as graph data nodes, and use the faces between the meshes as the edges between the graph data nodes to generate an adjacency matrix;

[0019] In the graph data nodes, store the mesh position data and the blast hole parameter matrix as graph data samples, and store the corresponding rock mass displacement data as graph data labels;

[0020] Obtain rock mass displacement data under different smooth blasting hole parameters, obtain several different graph data samples and graph data labels, and stack them to form a graph structure data set, where the smooth blasting hole parameters include the smooth blasting scenario of blasting charge, blasting sequence, hole position, and hole depth;

[0021] Divide the graph structure data set, input it into the database system, and complete the construction of the smooth blasting effect graph database.

[0022] Preferably, use a modified finite element simulation model to obtain rock mass displacement data under the smooth blasting scenario of different blasting charges, blasting sequences, hole positions, and hole depths. Among them, the steps for obtaining the modified finite element simulation model include:

[0023] Perform full-coverage scanning of the heading face by three-dimensional laser scanning to obtain the three-dimensional data of the heading face. After preprocessing the three-dimensional data, reconstruct the three-dimensional geometric model of the heading face;

[0024] Set three non-overlapping points in the heading face as reference points, and sequentially measure the distances between the blast holes and the reference points to obtain the three-dimensional position data of the blast holes;

[0025] Use a laser rangefinder to obtain the blast hole depth data of the blast holes;

[0026] Number the blast holes according to the set blasting time sequence information to obtain the blast hole blasting time sequence data;

[0027] Collect the blast hole blasting charge data according to the set blasting equivalent of the blast holes;

[0028] Perform finite element simulation on smooth blasting to obtain smooth blasting simulation data including at least rock mass displacement data and overbreak and underbreak;

[0029] Carry out smooth blasting tests to obtain smooth blasting test data including at least rock mass displacement data and overbreak and underbreak;

[0030] Use the smooth blasting simulation data and smooth blasting test data as smooth effect correction data, and use the finite element model correction method based on SGMD and LWOA-ELM to correct the finite element structure constitutive model to obtain a modified finite element simulation model.

[0031] Preferably, the neural network model includes a feature compression coding layer, an asynchronous feature association layer, and a smooth effect decoding layer;

[0032] In the feature compression coding layer, use a number of stacked graph convolutional networks GCN to extract features from the input data to obtain compressed feature data , where a ReLU activation function is connected after each GCN, and n represents the number of blast holes;

[0033] In the asynchronous feature association layer, n transformer encoders are respectively input for calculation, and the calculation results of each transformer encoder are respectively input into a gated recurrent unit (GRU). The residual connection is used to add to the corresponding GRU output result to obtain the hidden state ; are used as the input of the corresponding GRU. The residual connection is also used to add to the corresponding GRU output result to obtain the hidden state ; and are used as the input of the corresponding GRU. The residual connection is also used to add to the corresponding GRU output result to obtain the hidden state ; to are used as the input of the corresponding GRU. The residual connection is also used to add to the corresponding GRU output result to obtain the hidden state ;

[0034] Smooth surface effect decoding layer: Use stacked GCNs to perform dilation reduction on the hidden state to obtain the rock mass displacement map data , where a ReLU activation function is connected after each GCN.

[0035] Second, a computer-readable storage medium storing one or more programs is also provided. The one or more programs include instructions that, when executed by a computing device, cause the computing device to execute the method as described above.

[0036] Third, an electronic device is also provided, including one or more processors, one or more memories, and one or more programs, where the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors. The one or more programs include instructions for executing the method as described above.

[0037] Compared with the prior art, the significant advantages of the present invention are:

[0038] 1. High computing efficiency: Different from the layer-by-layer reasoning method of traditional numerical simulation, in the tunnel smooth surface effect prediction method, an end-to-end deep learning prediction mode is adopted. Through complete training, the rock mass displacement data can be directly predicted in milliseconds, and then the smooth surface effect data can be quickly calculated;

[0039] 2. Accurate prediction: The graph neural network is used as the prediction model structure, covering all grid data comprehensively, avoiding distortion and deviation during the conversion of grid data;

[0040] 3. Good robustness: By establishing a database of smooth blasting effect diagrams containing multiple blasting scenarios, the graph neural network model is effectively trained, realizing the generalization ability of smooth prediction for different blasting charges, blasting sequences, blast hole positions, and blast hole depths.

[0041] The following further describes the present invention in detail in conjunction with the accompanying drawings and specific embodiments. Description of the Drawings

[0042] Figure 1 is the main flow chart of the tunnel smooth blasting effect prediction method with adaptive blast hole parameters.

[0043] Figure 2 is a schematic diagram of the framework of the smooth blasting effect prediction model. Specific Embodiments

[0044] As Figure 1 shown, a tunnel smooth blasting effect prediction method with adaptive blast hole parameters of the present invention specifically includes the following steps:

[0045] S10. Collection of smooth blasting effect data: Collect experimental data and simulation data, compare the data differences, use the experimental data to correct the simulation model, and then use the simulation model to generate a large amount of smooth blasting effect data and collect the smooth blasting effect data;

[0046] The step of collecting the smooth blasting effect data includes:

[0047] S11. Acquisition of tunnel face geometric data: The face is scanned in a full-coverage manner by three-dimensional laser scanning to obtain the three-dimensional data of the face, and then the three-dimensional data after preprocessing is converted into a three-dimensional geometric model of the face through three-dimensional surface reconstruction technology.

[0048] S12. Acquisition of smooth blasting blast hole parameters: Set three non-overlapping points in the face as reference points, measure the distances from the blast holes to the reference points in sequence, and obtain the three-dimensional position data of the blast holes according to the distances; use a laser rangefinder to obtain the hole depth data of the blast holes; number the blast holes according to the set blasting time sequence information to obtain the blast hole blasting time sequence data; collect the blast hole blasting charge data according to the set blasting equivalent of the blast holes.

[0049] S13. Acquisition of correction data: Conduct finite element simulation on smooth blasting to obtain finite element simulation data, and at the same time carry out smooth blasting experiments to obtain smooth blasting experimental data, and collect the rock mass displacement data and overbreak and underbreak amounts of the simulation and experiments as smooth blasting effect correction data.

[0050] S14. Finite element simulation model correction: Use the finite element model correction method based on SGMD and LWOA-ELM, and correct the finite element structure constitutive model in combination with the correction data.

[0051] S15. Generation of smooth surface effect data: Use Python to call the corrected finite element simulation model to batch generate smooth surface effect data with different blasting hole parameters (i.e., different blasting charges, blasting sequences, hole positions, hole depths), and collect the smooth surface effect data. The smooth surface effect data includes rock mass displacement data and the overbreak and underbreak amounts calculated from the rock mass displacement data, which is the difference between the actually excavated cross-section and the designed excavation contour line.

[0052] S20. Graph database production: Geometrically divide the mesh of the heading face, and then encode the hole parameters; use the OpenFoam source file parsing method to obtain the graph structure data, and then batch store the hole parameter matrix and the rock mass displacement data into the graph structure data respectively to obtain the graph data sample set and the graph data label set, and divide the data set ratio to obtain the graph database.

[0053] The steps of the graph database production include:

[0054] S21. Geometric mesh division of the heading face: Divide the three-dimensional geometric model of the heading face to obtain the heading face mesh data.

[0055] S22. Encoding and processing of hole parameters: Encode each item in the hole parameter data as a vector, perform data normalization processing on the vector, and then splice the normalized vectors to obtain the hole parameter matrix integrating multi-source data.

[0056] S221. Encoding of hole parameter data: Encode the three-dimensional hole position data, hole blasting depth data, hole blasting time sequence data, and hole blasting charge data as vectors in turn.

[0057] S222. Normalization processing of hole parameters: Perform min-max normalization processing on the encoded vectors.

[0058] S223. Splicing of hole parameter vectors: Splice the normalized vectors to obtain the hole parameter matrix integrating multi-source data.

[0059] S23. Generation of graph structure data: Use the idea of taking the mesh as the graph node and the surface between the meshes as the edge between the graph nodes to produce the graph structure data, and specifically implement the conversion from the mesh to the graph structure data by using the OpenFOAM source file parsing method.

[0060] Introduction to the basic principle: The corresponding node and edge information is extracted from the grid data. Nodes represent cells in the grid, while edges represent the connection relationships between these cells. Each node is assigned the corresponding characteristics of the grid it represents, including the coordinates, attributes, or other relevant information of the grid. Secondly, considering the topological characteristics of the graph data structure, if two cells are adjacent in the grid, then in the graph data structure, the nodes corresponding to these two cells will be connected by an edge. The weight of the edge is calculated based on the numerical information in these two cells. For example, the average value or difference of the numerical values in these two cells is taken as the weight of the edge. Finally, in this way, a new graph data structure is obtained, which not only retains the spatial distribution characteristics of the original grid data but also inherits the numerical information in the grid data.

[0061] S231. Parsing of OpenFOAM source files: In the constant / polyMesh directory of the OpenFOAM case, the grid files of the case are stored, and the parsing content includes the information of the points and faces files. The grid data is obtained by parsing these files in sequence.

[0062] S2311. Parsing of the points file: This file records all the nodes of the grid. The coordinate information of each node is arranged in the order of (x y z), that is, the three-dimensional spatial coordinates of the nodes are parsed.

[0063] S2312. Parsing of the faces file: This file records all the faces and the node numbers that make up each face, that is, the order of all the nodes in the points file is parsed.

[0064] S232. Generation of graph structure data: According to the parsed points file and faces file, the node information and edge information are obtained.

[0065] S2321. Generation of graph nodes: Traverse from the lower left corner of the computational domain to the right. When reaching the right boundary, move up one row in the positive direction of the vertical axis, and then loop through the parsing. According to the node information in the parsed points file, in the order of the nodes therein, the coordinate information of the nodes is stored in the array in sequence to obtain the node characteristics of the graph data.

[0066] S2322. Generation of graph edges: According to the parsed faces file, the edge connection relationships of the nodes are obtained. Based on these connection relationships, a graph structure representation of the node edges, that is, an adjacency list, is constructed, and the existing adjacency lists of all nodes are concatenated to obtain the adjacency matrix of the graph structure data.

[0067] For example, for structured grids, each face contains 4 points. The representation of a certain face in the faces file is 4(1 6 31 26). Among them, 4 indicates that this face is composed of four nodes, and the content in the parentheses represents the sequence numbers of the four nodes that make up the first face. According to the node sequence numbers, the edges that make up this face are: [1, 6], [6, 31], [31, 26], [26, 1].

[0068] S24. Graph data sample production: Produce the graph structure data sequence sample input into the prediction model, and store the grid position data and the blast hole parameter matrix in the graph structure data nodes.

[0069] The steps for producing the graph data sample include:

[0070] S241. Graph structure sample generation: Use the grids in the tunnel face grid data as graph data nodes; use the faces between the grids as the edges between the graph data nodes to generate an adjacency matrix.

[0071] S242. Storage of graph structure sample nodes: Assume the number of blast holes is n. Then a single sample input into the prediction model contains n graph structure data, forming a sequence of length n composed of n graph structure data, that is ; In the graph structure data input at the first time step, the grid position data of the grid corresponding to the node and the blast hole parameter matrix of the first blast are stored in each node; in the graph structure data input at the second time step, the grid position data of the grid corresponding to the node and the blast hole parameter matrix of the second blast are stored in each node; and so on to obtain .

[0072] S25. Graph data label production: Produce the label data of the graph structure, and store the smooth blasting effect data in the graph structure data nodes.

[0073] The steps for producing the graph data label include:

[0074] S251. Graph structure label generation: Use the grids in the tunnel face grid data as graph data nodes and use the faces between the grids as the edges between the graph data nodes to generate an adjacency matrix.

[0075] S252. Storage of graph structure label nodes: Correlate the rock mass displacement data with the tunnel face grid data; store the rock mass displacement data in the graph data nodes corresponding to the tunnel face grids to obtain the graph data labels.

[0076] S26. Batch generation of samples and labels: Perform steps S21 to S25 on the smooth blasting data with different blasting charges, blasting sequences, blast hole positions, and blast hole depths in sequence to obtain a large number of different graph data samples and graph data labels.

[0077] S27. Graph dataset production: Stack the results of all smooth blasting scenarios to form a graph-structured data set. Divide all the data into a training graph data set and a validation graph data set in a ratio of 7:3, and then input them into the library system to establish a smooth blasting effect graph database.

[0078] S30. Construction of the smooth blasting effect prediction model framework: Construct a neural network model structure composed of a feature compression and encoding layer, an asynchronous feature association layer, and a smooth blasting effect decoding layer to obtain the smooth blasting effect prediction model framework.

[0079] As Figure 2 shown, the steps for constructing the smooth blasting effect prediction model framework include:

[0080] S31. Feature compression and encoding layer: Use the Graph Convolutional Network (GCN) as the basic module of the feature compression and encoding layer. Three layers of GCN extract features from the input data , and a ReLU activation function is connected after each layer to improve the nonlinearity of the model. After being calculated by the feature compression and encoding layer, the input data obtains compressed feature data .

[0081] S32. Asynchronous feature association layer: Introduce the transformer encoder and the Gated Recurrent Units (GRU) as the basic units of the asynchronous feature association layer; for the compressed feature data after being calculated by the transformer encoder , use GRU to calculate the output of the first time step and add it to the residual connection to calculate the hidden state ; for the compressed feature data after being calculated by the transformer encoder , use GRU to calculate the output of the second time step and add it to the residual connection to calculate the hidden state ; a skip connection is used between each layer of units to prevent the problem of model gradient disappearance during the training phase, and so on to obtain the output hidden state of the last layer .

[0082] S33. Smooth blasting effect decoding layer: Use GCN as the basic module of the smooth blasting effect decoding layer. Three layers of GCN perform dilation and reduction on the input data , and a ReLU activation function is connected after each layer to improve the nonlinearity of the model. After being calculated by the dilation and reduction, the input data obtains the rock mass displacement map data .

[0083] S34. Prediction Model Framework Construction: Based on the neural network model structures of the feature compression and encoding layer, the asynchronous feature correlation layer, and the smooth surface effect decoding layer, a smooth surface effect prediction model framework is obtained.

[0084] S40. Smooth Surface Effect Prediction Model Training: Use the training graph dataset to train the prediction model to obtain a smooth surface effect prediction model, and use the validation graph dataset to verify the generality of the model.

[0085] The steps of training the smooth surface effect prediction model include:

[0086] S41. Model Calculation: After a single sample data is input into the model, the graph data at all time steps passes through the feature compression and encoding layer to obtain compressed feature data ; in the asynchronous feature correlation layer, the graph data at all time steps is input into the basic units of the asynchronous feature correlation layer according to the blasting order, and the graph data after compressing the blasting parameters of the first blast hole is stored as the input of the first time step, and the hidden state is obtained through calculation ; the graph data after compressing the blasting parameters of the second blast hole is stored as the input of the second time step, and combined with the hidden state of the previous time step the hidden state of the second time step is calculated ; the graph data after compressing the blasting parameters of the third blast hole is stored as the input of the third time step, and combined with the hidden states of the previous two time steps and the hidden state of the third time step is calculated ; and so on, the remaining are input in turn, and the hidden state of the last time step is calculated ; the hidden state of the last layer passes through the smooth surface effect decoding layer to calculate the rock mass displacement graph data .

[0087] S42. Parameter Update: Based on the rock mass displacement graph data calculate the value of the mean square error loss (MSE) of the loss function and calculate the gradient of the model parameters with respect to the loss value, and use the backpropagation algorithm to update the model parameters.

[0088] S43. Model Optimization: Input the sample data in all training graph datasets in turn, repeat S41 and S42 to obtain the optimized model.

[0089] S44. Model Verification: Input the data of the validation graph dataset into the optimized model to verify the generality of the model.

[0090] Application of smooth surface effect prediction model: Measure the blasting environment data, divide the target surface into grids, set the blasting hole parameters, and produce the model input data to predict the smooth surface effect;

[0091] The application steps of the smooth surface effect prediction model include:

[0092] Measurement of blasting environment data: Use three-dimensional laser scanning to conduct a full-coverage scan of the target surface, and convert the preprocessed three-dimensional data into a three-dimensional geometric model of the target surface through three-dimensional surface reconstruction technology;

[0093] Grid division of the target surface: Divide the three-dimensional geometric model of the target surface into grids to obtain the target surface grid data;

[0094] Prediction of smooth surface effect: Set the blasting hole parameters and produce the model input data using the steps of S23; Use the smooth surface effect prediction model to predict the displacement data of the rock mass of the bench corresponding to the nodes, and calculate the smooth surface effect data such as overbreak and underbreak amounts and smooth surface areas in combination with the displacement data.

[0095] Based on the same technical solution, the present invention also discloses a computer-readable storage medium storing one or more programs, where the one or more programs include instructions that, when executed by a computing device, cause the computing device to execute the above surface prediction method.

[0096] Based on the same technical solution, the present invention also discloses a computing device, including one or more processors, one or more memories, and one or more programs, where the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing the above prediction method.

[0097] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code.

[0098] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0099] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0100] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0101] The above embodiments are only for illustrating the technical idea of the present invention, and the protection scope of the present invention cannot be limited thereby. Any modification made on the basis of the technical solution according to the technical idea proposed by the present invention falls within the protection scope of the present invention.

Claims

1. A method for predicting the smooth blasting effect of a tunnel with self-adaptive blast hole parameters, characterized in that, It includes the following steps: Based on the graph neural network, construct a smooth surface effect prediction model; Use the smooth surface effect prediction model to predict the rock mass displacement data; Combined with the prediction results of the rock mass displacement data, calculate the smooth surface effect of the tunnel; The steps for constructing the smooth surface effect prediction model include: Construct a smooth blasting effect diagram database including a training data set and a validation data set; Based on the graph neural network, construct a neural network model; the neural network model includes a feature compression encoding layer, an asynchronous feature association layer, and a smooth surface effect decoding layer; In the feature compression and encoding layer, a number of stacked graph convolutional networks (GCNs) are used to extract features from the input data to obtain compressed feature data , where a ReLU activation function is connected after each GCN, and n represents the number of blast holes; In the asynchronous feature association layer, n transformer encoders are respectively input for calculation, and the calculation results of each transformer encoder are respectively input into a gated recurrent unit (GRU). The residual connection is used to add to the corresponding GRU output result to obtain the hidden state ; are taken together as the input of the corresponding GRU. The residual connection is also used to add to the corresponding GRU output result to obtain the hidden state ; and are taken together as the input of the corresponding GRU. The residual connection is also used to add to the corresponding GRU output result to obtain the hidden state ; and so on. to are taken together as the input of the corresponding GRU. The residual connection is also used to add to the corresponding GRU output result to obtain the hidden state ; Smooth surface effect decoding layer: Using stacked GCNs for the hidden state to perform dilation reduction to obtain rock mass displacement map data , where a ReLU activation function is connected after each GCN; Use the training data set and the validation data set to train and optimize the neural network model to obtain a smooth surface effect prediction model.

2. The method according to claim 1, wherein The steps for constructing a smooth blasting effect diagram database including a training data set and a validation data set include: Perform mesh generation on the three-dimensional geometric model of the tunnel face to obtain face mesh data; Encode and process the three-dimensional position data of the blast holes, the blast hole depth data of the blast holes, the blast hole blasting timing data, and the blast hole blasting charge data into vectors in sequence, and splice them after vector normalization to obtain a blast hole parameter matrix; Use the meshes in the face mesh data as graph data nodes, and use the faces between the meshes as the edges between the graph data nodes to generate an adjacency matrix; Within the graph data nodes, store the mesh position data and the blast hole parameter matrix as graph data samples, and store the corresponding rock mass displacement data as graph data labels; Obtain the rock mass displacement data under different blast hole parameters, obtain several different graph data samples and graph data labels, and stack them to form a graph structure data set, where the blast hole parameters include the smooth blasting scenarios of the blasting charge, blasting sequence, blast hole position, and blast hole depth; Divide the graph structure data set, enter it into the library system, and complete the construction of the smooth blasting effect diagram database.

3. The method according to claim 1, wherein Use the modified finite element simulation model to obtain the rock mass displacement data under the smooth blasting scenarios of different blasting charges, blasting sequences, blast hole positions, and blast hole depths. Among them, the steps for obtaining the modified finite element simulation model include: Perform full-coverage scanning of the face by three-dimensional laser scanning to obtain the three-dimensional data of the face. After preprocessing the three-dimensional data, reconstruct the three-dimensional geometric model of the face; Set three non-overlapping points in the face as reference points, and measure the distances between the blast holes and the reference points in sequence to obtain the three-dimensional position data of the blast holes; Use a laser rangefinder to obtain the blast hole depth data of the blast holes; Number the blast holes according to the set blasting timing information to obtain the blast hole blasting timing data; Collect the blast hole blasting charge data according to the set blasting equivalent of the blast holes; Perform finite element simulation on the smooth blasting to obtain smooth blasting simulation data including at least the rock mass displacement data and the overbreak and underbreak; Conduct a smooth blasting test to obtain smooth blasting test data including at least the rock mass displacement data and the overbreak and underbreak; Use the smooth blasting simulation data and the smooth blasting test data as smooth surface effect correction data, and use the finite element model correction method based on SGMD and LWOA-ELM to correct the finite element structural constitutive model to obtain a modified finite element simulation model.

4. A computer-readable storage medium storing one or more programs, the one or more programs including instructions, characterized in that, When executed by a computing device, the instructions cause the computing device to execute the method according to any one of claims 1-3.

5. An electronic device, characterized in that, Comprising one or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing the method according to any one of claims 1-3.

Images