Underground engineering surrounding rock excavation-support parameter dynamic design method and system

Abstract

The invention provides an underground engineering surrounding rock excavation-support parameter dynamic design method and system, and the method comprises the steps: obtaining the geological features and mechanical parameters of an underground engineering construction region, and building a three-dimensional numerical model according to the geological conditions; performing a numerical simulation experiment according to the three-dimensional numerical model on the basis of an excavation mode and support parameters of the underground engineering construction area, and generating a surrounding rock excavation dynamic response database and a surrounding rock support dynamic response database by taking surrounding rock deformation as an index; an underground engineering excavation scheme design model based on a GA-BP algorithm is constructed, training is conducted through the surrounding rock excavation dynamic response database, and an optimal excavation mode is determined; and constructing an underground engineering support design scheme model based on a PSO-DNN algorithm, and training by using the surrounding rock support dynamic response database to obtain a support scheme under the optimal excavation mode. The dynamic response characteristics of the surrounding rock can be effectively predicted, an excavation support design scheme is optimized, and the safety and economical efficiency of underground engineering are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of underground engineering, and particularly relates to a method and system for dynamic design of excavation-support parameters of surrounding rock of underground engineering. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute the prior art.

[0003] With the growth of the construction scale of underground engineering (such as underground caverns of hydropower stations, traffic tunnels, mine roadways, etc.), the geological conditions faced by the construction thereof are becoming increasingly complex, such as fault fracture zones, high ground stress, high permeation pressure, etc., resulting in a significant increase in disaster risks such as large deformation of surrounding rock, collapse and failure of support structures during excavation. The excavation and support design of underground engineering is the core to ensure construction safety and engineering stability, and the key lies in dynamically adjusting the excavation and support parameters according to the geological conditions to balance safety and economy. Therefore, how to design the excavation and support parameters to adapt to complex geological conditions and ensure the safety and economy of underground engineering has become a key problem to be solved.

[0004] At present, the excavation and support design methods of underground engineering mainly include engineering analogy method, empirical formula method, theoretical calculation method and numerical simulation method, which have certain guiding significance for the safe construction of underground engineering. However, with the increasingly complex geological conditions faced by underground engineering, the limitations of traditional excavation and support design methods are increasingly highlighted, which are specifically manifested in the following aspects: The engineering analogy method highly depends on similar engineering cases and the experience of designers, but the actual engineering geological conditions have significant differences, resulting in low reliability of the analogy excavation and support scheme, which is difficult to meet the design requirements of future deep underground engineering; the empirical formula method is difficult to quantitatively consider the dynamic relationship between surrounding rock and support structure, and is insufficient in considering the coupling of multiple factors, resulting in certain limitations; the theoretical analysis model is usually based on a large number of simplifying assumptions, and it is difficult to fully consider the complex geological effects and dynamic change process of multiple physical fields in actual engineering, resulting in that the excavation design scheme is difficult to meet the engineering requirements; in recent years, numerical simulation methods have been widely used, which can simulate the dynamic response characteristics of surrounding rock under complex geological conditions and provide important theoretical support for excavation and support design. However, the numerical simulation method usually has high calculation cost and long time consumption, and is highly dependent on model parameters and boundary conditions, which is difficult to meet the demand of rapid response to engineering site. SUMMARY

[0005] In order to overcome the deficiencies of the prior art, the present application provides a method and system for dynamic design of excavation-support parameters of surrounding rock of underground engineering, which realizes the intelligentization of the excavation and support design of underground engineering, improves the design precision and efficiency of excavation and support, and ensures the safe and efficient construction of underground engineering through the steps of obtaining geological information, establishing a numerical simulation database, training and optimizing a deep learning model, etc.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: A first aspect of the present invention provides a method for dynamic design of surrounding rock excavation-support parameters in underground engineering; A dynamic design method for surrounding rock excavation-support parameters of underground engineering, comprising: Obtain the geological characteristics and mechanical parameters of the underground engineering construction area and establish a three-dimensional numerical model based on the geological conditions; Based on the excavation mode and support parameters of the underground engineering construction area, numerical simulation experiments are carried out according to the three-dimensional numerical model. The surrounding rock deformation is used as an indicator to generate the underground engineering surrounding rock excavation dynamic response database and the surrounding rock support dynamic response database. Constructing an underground engineering excavation scheme design model based on the GA-BP algorithm, training the model using the surrounding rock excavation dynamic response database, and determining the optimal excavation method; A support design scheme model for underground engineering based on the PSO-DNN algorithm is constructed, and the model is trained using the surrounding rock support dynamic response database to obtain the optimal support scheme under the optimal excavation method.

[0007] As a further technical solution, the process of establishing the three-dimensional numerical model includes: Establish a three-dimensional numerical calculation model based on on-site geological data and project scale characteristics; Set the model boundary conditions; the bottom boundary of the numerical calculation model is a fixed boundary, the surrounding boundaries are displacement constraint boundaries, and the top is a free boundary; The numerical calculation model is assigned parameters based on the rock mechanics parameters obtained from indoor experiments, and initial geostress calculation is carried out to ensure that the initial geostress field of the numerical calculation model is consistent with the measured geostress value in the underground engineering construction area.

[0008] As a further technical solution, the process of generating a database of dynamic responses to surrounding rock excavation in underground engineering is as follows: Investigate the excavation method of the underground engineering construction area, and combine the excavation method with the excavation distance to obtain a numerical simulation scheme for underground engineering excavation; Based on the obtained numerical simulation scheme, a three-dimensional numerical calculation model is used to carry out numerical simulation experiments to record the dynamic response characteristics of the surrounding rock of the underground engineering after changing the influencing factors; The extracted influencing factors and the dynamic response characteristics of the surrounding rock are used as labels to combine and obtain a data set, forming a database of dynamic response characteristics of surrounding rock excavation in underground engineering.

[0009] As a further technical solution, the process of generating the surrounding rock support dynamic response database is as follows: Based on the obtained support parameter range, random functions are used to perform random combinations to obtain support simulation schemes; Based on the obtained support simulation scheme, numerical simulation experiments were carried out using a three-dimensional numerical calculation model; the dynamic response characteristics of the surrounding rock of the underground project after changing the influencing factors were recorded; The extracted influencing factors and the dynamic impact characteristics of the surrounding rock are used as labels to combine and obtain the data set, forming a dynamic response characteristic database of the surrounding rock support.

[0010] As a further technical solution, a design model for underground engineering excavation scheme based on GA-BP algorithm was constructed, including: Set the number of layers, number of neurons and activation function of the BP neural network; Genetic algorithm is used to optimize the initial weights and biases of the neural network, and the root mean square error is used as the fitness evaluation criterion; Gradient descent training is performed through the back-propagation algorithm to further improve model accuracy and reduce the risk of overfitting.

[0011] As a further technical solution, a model for underground engineering support design based on the PSO-DNN algorithm is constructed, including: Set the number of layers, number of neurons and activation function of the neural network; The particle swarm optimization algorithm is used to optimize the initial weights and biases of the neural network, and the root mean square error or classification cross entropy is used as the objective function of PSO optimization; Fine-tune the model through the back-propagation algorithm to reduce the risk of overfitting.

[0012] As a further technical solution, the method for dynamic design of surrounding rock excavation and support parameters for underground engineering further includes: The practical application effects of the underground engineering excavation scheme design model based on the GA-BP algorithm and the underground engineering support design scheme model based on the PSO-DNN algorithm were evaluated. The dynamic response data of the surrounding rock in actual projects were collected and compared with the model prediction results. Based on the evaluation results, the model is fed back and adjusted.

[0013] A second aspect of the present invention provides a dynamic design system for surrounding rock excavation-support parameters of underground engineering.

[0014] A dynamic design system for surrounding rock excavation and support parameters of underground engineering, comprising: The three-dimensional numerical model building module is configured to: obtain geological characteristics and mechanical parameters of the underground engineering construction area and build a three-dimensional numerical model based on the geological conditions; The database construction module is configured to: conduct numerical simulation experiments based on the three-dimensional numerical model based on the excavation method and support parameters of the underground engineering construction area, and use the surrounding rock deformation as an indicator to generate the underground engineering surrounding rock excavation dynamic response database and the surrounding rock support dynamic response database; The optimal excavation method determination module is configured to: construct an underground engineering excavation scheme design model based on the GA-BP algorithm, train the model using the surrounding rock excavation dynamic response database, and determine the optimal excavation method; The support scheme determination module is configured to: construct an underground engineering support design scheme model based on the PSO-DNN algorithm, train the model using the surrounding rock support dynamic response database, and obtain the support scheme under the optimal excavation method.

[0015] The third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a dynamic design method for surrounding rock excavation and support parameters of an underground engineering as described in the first aspect of the present invention.

[0016] The fourth aspect of the present invention provides an electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for dynamic design of surrounding rock excavation and support parameters for underground engineering as described in the first aspect of the present invention are implemented.

[0017] One or more of the above technical solutions have the following beneficial effects: (1) This invention overcomes the subjectivity of the traditional engineering analogy method and the simplification defects of the empirical formula method by integrating numerical simulation and deep learning technology; it adopts GA-BP and PSO-DNN intelligent algorithms to accurately predict the deformation response characteristics of surrounding rock under complex geological conditions, and solves the design bottlenecks of traditional underground engineering excavation and support design, such as strong dependence on engineering experience, excessive model simplification construction, and slow numerical simulation speed.

[0018] (2) The present invention significantly improves design efficiency. Traditional underground engineering excavation support scheme design uses a method that first designs based on manual experience and then verifies it using numerical simulation. This method is time-consuming (2 to 3 months) and requires many iterations (3 to 5 rounds). The present invention completely breaks away from the inefficient cycle of "manual trial and error-simulation verification" of the traditional method by combining deep learning algorithms with numerical models, significantly improving the efficiency of underground engineering support scheme design.

[0019] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0021] Figure 1 This is a flow chart of the method of the first embodiment.

[0022] Figure 2 This is a diagram of the three-dimensional numerical model constructed in the first embodiment.

[0023] Figure 3 This is the layout diagram of the underground engineering monitoring solution of the first embodiment.

[0024] Figure 4 This is a comparison curve between the predicted value and the simulated value of the concentrated surrounding rock excavation scheme verified by the GA-BP model in the first embodiment.

[0025] Figure 5 This is a comparison curve between the predicted value and the simulated value of the centralized surrounding rock support scheme verified by the PSO-DNN model in the first embodiment.

[0026] Figure 6 In the first embodiment, the GA-BP model and the PSO-DNN model are used to obtain a comparison curve between the support scheme and the surrounding rock deformation monitored on site.

[0027] Figure 7 This is a system structure diagram of the second embodiment. DETAILED DESCRIPTION

[0028] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0029] It should be noted that the terms used herein are for describing particular embodiments only and are not intended to limit the exemplary embodiments according to the present invention.

[0030] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0031] In the present invention, the geological characteristics and mechanical parameters of the underground engineering construction area are first obtained, and a three-dimensional numerical model is established according to the geological conditions; secondly, different underground engineering excavation methods are investigated, a large number of numerical simulation tests are carried out, and the surrounding rock deformation is selected as an indicator to generate a dynamic response database of surrounding rock excavation for underground engineering. A surrounding rock excavation design model is established based on a deep learning model to determine the optimal excavation method; then, a support scheme combination is adopted for underground engineering, including anchor rods, length, annular spacing, radial spacing, spacing, and quantity, and numerical experiments are carried out under different support parameter combinations to generate a surrounding rock support dynamic response database, and a surrounding rock support design model is established based on a deep learning model; finally, through a deep learning algorithm, the optimal excavation support parameter combination that meets the safety design requirements is intelligently selected, thereby improving the intelligence level of surrounding rock excavation and support design for underground engineering.

[0032] Example 1 This embodiment discloses a method for dynamic design of surrounding rock excavation-support parameters for underground engineering; like Figure 1 As shown, a dynamic design method for surrounding rock excavation-support parameters of underground engineering includes: Step S1: Obtain the geological characteristics and mechanical parameters of the underground engineering construction area and establish a three-dimensional numerical model based on the geological conditions.

[0033] Through on-site investigation of the proposed underground construction area's three-dimensional geological information, underground engineering excavation methods, underground engineering support types, and laboratory-measured engineering rock mass mechanical parameters within the underground engineering construction area, the three-dimensional geological information includes burial depth, stratum lithology, geological structure, and geostress; underground engineering excavation methods include full-section method, bench method, circular excavation with core soil retention method, middle septum method, cross-middle septum method, and double-side wall pilot method; underground engineering support types include anchor length range, circumferential spacing range, radial spacing range, prestress range, and anchor number range; and laboratory-measured rock mass mechanical parameters within the underground engineering construction area include density, Young's modulus, Poisson's ratio, cohesion, internal friction angle, and tensile strength.

[0034] Based on the geological characteristics and mechanical parameters obtained from the above investigation, a three-dimensional numerical model was established according to the geological conditions, and the boundary conditions of the model were set. The boundary conditions include the bottom boundary of the numerical calculation model being a fixed boundary, the surrounding boundaries being displacement-constrained boundaries, and the top being a free boundary; Finally, the rock mechanics parameters were assigned based on the indoor experiments, and the initial geostress calculation was carried out, so that the initial geostress field of the numerical calculation model was consistent with the measured geostress value in the underground engineering construction area.

[0035] Furthermore, the surrounding rock of the underground engineering construction area is also monitored. Specifically, monitoring sections are arranged within 0-20m behind the heading face, with a spacing of 2-4m between the monitoring sections; monitoring lines are arranged at the arch crown, left arch spandrel, right arch spandrel, left arch waist, right arch waist, left arch foot, and right arch foot of the monitoring section, with a spacing of 0.5-1m between the measuring points on the measuring line, to monitor the deformation of the surrounding rock during the excavation process.

[0036] Step S2: Based on the excavation method and support parameters of the underground engineering construction area, a numerical simulation experiment is performed according to the three-dimensional numerical model, and the surrounding rock deformation is used as an indicator to generate a dynamic response database of surrounding rock excavation and a dynamic response database of surrounding rock support for underground engineering; Specifically, the process of constructing a database of dynamic response characteristics of surrounding rock excavation is as follows: based on the excavation method of the underground engineering construction area, the excavation method and excavation distance are combined to obtain a numerical simulation scheme, and numerical simulation experiments are carried out using a three-dimensional numerical calculation model to record the dynamic response characteristics of the surrounding rock of the underground engineering after changing the influencing factors; the extracted influencing factors and the dynamic response characteristics of the surrounding rock are used as labels, and are combined to obtain a data set to form a database of dynamic response characteristics of surrounding rock excavation.

[0037] The process for constructing a dynamic response database for surrounding rock support involves randomly combining anchor length, circumferential spacing, radial spacing, prestress, and number of anchors using MATLAB random functions based on support parameters obtained from engineering surveys to generate support simulation schemes. Numerical simulation experiments are then conducted using a three-dimensional numerical calculation model to record the dynamic response characteristics of the surrounding rock in underground projects after changing the influencing factors. The extracted influencing factors and the dynamic impact characteristics of the surrounding rock are used as labels to generate a dataset, forming a database of dynamic response characteristics for surrounding rock support.

[0038] The surrounding rock excavation dynamic response database and the surrounding rock support dynamic response database are divided into training sets and validation sets according to a preset ratio (8:2 or 7:3) to facilitate subsequent model training.

[0039] Step S3, constructing an underground engineering excavation scheme design model based on the GA-BP algorithm, training the model using the surrounding rock excavation dynamic response database, and determining the optimal excavation method; Step S31, constructing an underground engineering excavation scheme design model based on the GA-BP algorithm, including: Step S311: Set the number of BP deep neural network layers (input layer, hidden layer, output layer), the number of neurons in each layer, and the activation function type (such as ReLU / Sigmoid), and clarify the dimensions of weights and biases.

[0040] Step S312: Optimizing the initial weights and biases of the neural network using a genetic algorithm, with the root mean square error as the fitness evaluation criterion. The process includes: Step S312-1, define the population size, the number of iterations, the crossover probability (Pc), the mutation probability (Pm), the selection strategy (such as roulette / carnival selection), and randomly generate the initial population.

[0041] Step S312-2, take the root mean square error (RMSE) as the fitness evaluation standard, convert the error value into the fitness value (such as fitness = 1 / (1+RMSE)), and measure the performance of the individual corresponding to the DNN.

[0042] Step S312-3, select the parent individual according to the fitness value (such as roulette probability selection), and retain the high fitness individual to maintain the population diversity. The selected parent individual is subjected to gene exchange according to the crossover probability to generate offspring chromosomes. The offspring individual is subjected to gene disturbance according to the mutation probability to enhance the local search ability.

[0043] Step S312-4, combine the parent and offspring populations, retain the optimal individual (elite retention strategy), and generate a new generation of population.

[0044] Step S312-5, if the preset number of iterations is reached or the optimal fitness value meets the threshold (such as RMSE≤1e-4), stop optimization; otherwise, return to step S312-2 and continue evolution.

[0045] Step S313, further improve the model accuracy and reduce the risk of overfitting by gradient descent training through the back propagation algorithm. Among them, the optimal chromosome output by the genetic algorithm is decoded into the initial weights and biases of the DNN, and the gradient descent training is performed by the back propagation algorithm to further improve the model accuracy and reduce the risk of overfitting.

[0046] Step S32, further, the surrounding rock excavation dynamic response database is used to train the model to determine the optimal excavation method.

[0047] Based on the surrounding rock excavation dynamic response database, the model is trained to verify the prediction performance of the model. By analyzing the prediction results in the validation set, the error source of the deep learning model is analyzed, the cross-validation method is used to further optimize the model hyperparameters, and the prediction accuracy and stability of the deep learning model are improved.

[0048] The input and output parameters of the underground engineering excavation scheme design model based on GA-BP algorithm are defined. The input parameters are the excavation method and excavation distance of the underground engineering; the output parameter is the surrounding rock deformation. Then the surrounding rock deformation can be predicted and compared with the safe deformation, and the support type is selected if it meets the requirements, and the excavation method is adjusted until the optimal excavation method and excavation distance within the safe deformation range are obtained.

[0049] Step S4: construct an underground engineering support design scheme model based on the PSO-DNN algorithm, train the model using the surrounding rock support dynamic response database, and obtain a support scheme under the optimal excavation mode.

[0050] Step S41, constructing an underground engineering support design scheme model based on the PSO-DNN algorithm, including: Step S411: Set the number of deep neural network layers (input layer, hidden layer, output layer), the number of neurons in each layer, and the activation function type (such as ReLU / Sigmoid).

[0051] Step S412-1, using a particle swarm optimization algorithm to optimize the initial weights and biases of the neural network, with the root mean square error or classification cross entropy as the objective function of the PSO optimization; specifically, step S412 includes: Step S412-2, define the particle swarm size, number of iterations, learning factor (c1 / c2), inertia weight (ω), speed and position range, and randomly generate the particle swarm matrix of DNN initial weights and biases.

[0052] In step S412-3, the root mean square error (RMSE) or cross-entropy is used as the objective function of PSO optimization to measure the DNN prediction performance.

[0053] Step S412-4, record the historical optimal solution (pbest) of each particle and the current global optimal solution (gbest) of the particle swarm, corresponding to the optimal combination of DNN weights and biases.

[0054] Step S412-5, iterate the particle swarm according to the speed-position update formula:

[0055] Where, , 、 is the learning factor, , is the inertia weight, which controls the influence of the particle's historical velocity on the current velocity and decreases with iteration. Indicates the Particles ( , is the population size); The number space Dimension (such as a weight or bias of a DNN); is the current evolution generation (time step); For the Particles in Speed ​​in dimension; For the Particles in Position on the dimension; 、 is a random number, obeying uniform distribution ; 、 Respectively particles, the entire particle group The optimal position of individual history and global history in the dimension.

[0056] Step S412-6: If the preset number of iterations is reached or the fitness function value is lower than the target threshold (such as RMSE ≤ 1e-4), stop the optimization; otherwise, return to step S412-5 to continue updating.

[0057] Step S413: Fine-tune the model using the back-propagation algorithm to reduce the risk of overfitting. Substitute the optimal weights and biases output by the PSO into the DNN and perform fine-tuning training using the back-propagation algorithm to achieve high-precision, low-overfitting deep network fitting.

[0058] Step S42: training the model using the surrounding rock support dynamic response database to obtain a support scheme under the optimal excavation mode.

[0059] The model was trained based on a database of surrounding rock support dynamic responses to verify the predictive performance of the support design model. By analyzing the prediction results from the validation set and identifying the sources of error in the deep learning model, a cross-validation method was used to further optimize the model's hyperparameters and improve the deep learning model's predictive accuracy and stability.

[0060] Define the input and output parameters of the PSO-DNN algorithm-based underground engineering support design model. The input parameters are support parameters, including anchor length, circumferential spacing, radial spacing, prestressing force, and number of anchors; the output parameter is surrounding rock deformation.

[0061] Based on the optimal excavation method obtained in step S3, the support design parameters are input, the surrounding rock deformation is predicted and judged with the safe deformation (relevant engineering specifications and empirical values). If it meets the safe deformation (relevant engineering specifications and empirical values), it is applied to the on-site engineering. If it does not meet the requirements, the support scheme is adjusted until a support method that meets the safe deformation range is obtained.

[0062] In addition, the method adopted in this implementation also includes: evaluating the actual application effects of the underground engineering excavation scheme design model based on the GA-BP algorithm and the underground engineering support design scheme model based on the PSO-DNN algorithm, collecting the dynamic response data of the surrounding rock in the actual project, and comparing and analyzing them with the model prediction results; based on the evaluation results, providing feedback and adjustment to the model to further improve the prediction accuracy and applicability of the model.

[0063] Furthermore, in order to further verify the effectiveness of the method of the present invention, an underground engineering excavation and support scheme design is selected as an example to illustrate.

[0064] 1. Through on-site investigation of an underground project with a depth of 200m and surrounding rock types of III1, III2, and IV, the measured ground stress results are maximum principal stress σ1 = 6.68MPa, intermediate principal stress σ2 = 5.25MPa, and minimum principal stress σ3 = 4.82MPa. The density, Young's modulus, Poisson's ratio, cohesion, internal friction angle, and tensile strength parameters were measured by on-site sampling and laboratory testing as shown in Table 1: Table 1 Rock mass mechanical parameters

[0065] 2. Numerical simulation database generation 2.1 Establishing a three-dimensional numerical model Based on the geological conditions obtained in the first step of investigation, a three-dimensional numerical calculation model is established, such as Figure 2 As shown in Figure 1, the model dimensions are 400 m long, 400 m wide, and 320 m high. The bottom boundary of the three-dimensional numerical model is a fixed boundary, the surrounding boundaries are displacement constraint boundaries, and the upper part is a free boundary. According to the indoor experimental parameters in Table 1, the initial equilibrium calculation of the ground stress is carried out to make the ground stress field of the numerical model consistent with the measured value.

[0066] 2.2 Establishing a monitoring plan for surrounding rock of underground projects The monitoring section is arranged within 15m behind the tunnel face, with a radial spacing of 2m. The measuring lines S arch top, S left arch spandrel, S right arch spandrel, S left arch spandrel, S right arch spandrel, S left arch spandrel, S right arch spandrel, S left arch spandrel, S right arch spandrel, S right arch spandrel, S right arch spandrel, S right arch spandrel, S right arch spandrel, S right arch spandrel, S right arch spandrel, S right arch spandrel, S right arch spandrel, S right arch spandrel, S right arch spandrel, S left ... left arch spandrel, S right arch spandrel, S left arch spandrel, S left arch spandrel, S right arch spandrel, S left arch spandrel, S left arch spandrel, S right arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch spandrel, S left arch span Figure 3 shown.

[0067] 2.3 Construct a database of dynamic response of surrounding rock excavation and support for underground engineering.

[0068] 2.3.1 Based on the excavation methods (full-section method, bench method, circular excavation with core soil retention method, center bulkhead method, cross-center bulkhead method, and double-sidewall pilot method) and excavation distances obtained through research, a large-scale simulation scheme was generated using MATLAB random functions. Numerous simulations were conducted within a three-dimensional numerical model, and the dynamic response characteristics of the surrounding rock mass of the underground project were recorded after varying the influencing factors. The extracted influencing factors and the surrounding rock mass dynamic response characteristics were used as labels and combined to generate a dataset, forming a database of surrounding rock excavation dynamic response characteristics.

[0069] Table 2 Database of excavation dynamic response characteristics of surrounding rock in underground engineering under different excavation methods

[0070] 2.3.2 After the excavation of the surrounding rock of the underground project was completed, a large-scale simulation scheme was generated using MATLAB random functions based on the range of investigated support parameters (anchor length, circumferential spacing, radial spacing, prestressing force, and number of anchors). Numerous simulation tests were conducted within the three-dimensional numerical calculation model, and the dynamic response characteristics of the surrounding rock of the underground project were recorded after varying the influencing factors. The extracted influencing factors and the surrounding rock dynamic response characteristics were used as labels and combined to generate a dataset, forming a database of the dynamic response characteristics of the surrounding rock excavation.

[0071] Table 3 Database of dynamic response characteristics of underground engineering surrounding rock under different support parameters

[0072] 2.3.3 Using the MATLAB random function, the surrounding rock excavation dynamic response database and the surrounding rock support dynamic response database were divided into training and validation sets according to a preset ratio (8:2). The training set and test set for the surrounding rock excavation dynamic response data contained 80 and 20 data sets, respectively; the training set and test set for the surrounding rock support dynamic response database contained 84 and 22 data sets, respectively.

[0073] 3. Construct GA-BP and PSO-DNN deep learning algorithms.

[0074] 3.1 Building a GA-BP Deep Learning Algorithm 3.1.1 The deep neural network is configured with five input layers (full-section method, step method, circular excavation with core soil retention method, intermediate partition wall method, and excavation distance); two hidden layers, with 32 neurons in hidden layer 1 and 16 neurons in hidden layer 2, using the RELU activation function.

[0075] 3.1.2 Define the population size as 30, the maximum number of iterations as 50, the crossover probability as 0.8, the mutation probability as 0.1, the elite retention ratio as 0.1, and adopt a tournament strategy, selecting the best from three random individuals each time.

[0076] 3.1.3 Using the root mean square error (RMSE) as the fitness evaluation criterion, the error value is converted into a fitness value (fitness = 1 / (1 + RMSE)) to measure the DNN performance corresponding to the individual.

[0077] 3.1.4 A tournament strategy was used to select parent individuals. Three individuals were randomly selected each time, and the one with the highest fitness was retained as the parent. Gene mixing was performed on the selected parent with probability Pc=0.8, and Gaussian noise (μ=0, σ=0.05) was added to each gene of the offspring with probability Pm=0.1.

[0078] 3.1.5 Merge the parent and offspring populations and retain the best individuals (elite retention strategy). Keep the top 10% of elite individuals and select the rest using a roulette wheel.

[0079] 3.1.6 The number of prediction iterations has been reached, and optimization is stopped 3.1.7 Convert the encoding of the optimal individual into a DNN weight matrix, use the Adam optimizer, and add early stopping to prevent overfitting.

[0080] 3.2 Building a PSO-DNN Deep Learning Algorithm 3.2.1 The input layer is composed of six dimensions: anchor length, circumferential spacing, radial spacing, prestress, number, and surrounding rock deformation. The hidden layer has two layers, with 16 and 8 neurons in hidden layers 1 and 2, respectively, for a total of 257 parameters.

[0081] 3.2.2 The particle swarm size is 30, the maximum number of iterations is 100, the learning factor c1=c2=20, and the initial and final values ​​of the inertia weight are 0.9 and 0.4 respectively.

[0082] 3.3.3 Use the root mean square error (RMSE) or cross-entropy as the objective function of PSO optimization to measure the DNN prediction performance.

[0083] 3.2.4 Record the historical optimal solution (pbest) of each particle and the current global optimal solution (gbest) of the particle swarm, corresponding to the optimal combination of DNN weights and biases.

[0084] 3.2.5 Iterate the particle swarm according to the velocity-position update formula:

[0085] Where, , 、 is the learning factor, , is the inertia weight, which controls the influence of the particle's historical velocity on the current velocity and decreases with iteration. Indicates the Particles ( , is the population size); The number space Dimension (such as a weight or bias of a DNN); is the current evolution generation (time step); For the Particles in Speed ​​in dimension; For the Particles in Position on the dimension; 、 is a random number, obeying uniform distribution ; 、 Respectively particles, the entire particle group The optimal position of individual history and global history in the dimension.

[0086] 3.2.6 Stop optimization when the preset number of iterations is reached.

[0087] 3.2.7 Substitute the optimal weights and biases output by PSO into the DNN and use the backpropagation algorithm for fine-tuning training to achieve high-precision, low-overfitting deep network fitting.

[0088] 4. Construct a design model for underground engineering excavation support scheme based on GA-BP and PSO-DNN deep learning algorithms.

[0089] 4.1 Define the input and output parameters of the deep learning model for underground engineering excavation design based on the GA-BP deep learning algorithm. The input parameters are: excavation type (full section method, bench method, circular excavation with core soil retention method, and partition wall method) and excavation distance; the output parameter is the surrounding rock deformation.

[0090] 4.2 Define the input and output parameters of the underground engineering support design optimization model based on the PSO-DNN deep learning algorithm. The input parameters are: support design parameters (anchor / cable length, spacing, prestressing force, quantity, lining thickness); the output parameter is the surrounding rock deformation.

[0091] 4.3 Based on the generated test set data of the surrounding rock excavation dynamic response database and the surrounding rock support dynamic response database, the GA-BP and PSO-DNN deep learning algorithms were used for model training, respectively. By adjusting the network weights and biases, the model was able to accurately predict the dynamic response characteristics of the surrounding rock.

[0092] 5. Optimize the underground engineering excavation and support design model based on GA-BP and PSO-DNN algorithms Based on the generated surrounding rock excavation dynamic response database and surrounding rock support dynamic response database test set data, the prediction performance of the underground engineering excavation and support design model based on GA-BP and PSO-DNN algorithms was verified. By analyzing the centralized prediction results and the error sources of the deep learning model, the cross-validation method was used to further optimize the model hyperparameters and improve the prediction accuracy and stability of the deep learning model. The comparison curve between the predicted value and the simulated value of the centralized surrounding rock excavation scheme verified by the GA-BP model is shown in the figure below. Figure 4As shown in the figure, the comparison curve between the predicted value and the simulated value of the centralized surrounding rock support scheme verified by the PSO-DNN model is shown in the figure. Figure 5 shown.

[0093] 6. Engineering application of underground engineering excavation and support design models 6.1 The trained GA-BP and PSO-DNN models were applied to the design of excavation and support schemes for actual underground projects. First, the proposed excavation method and distance for the underground project were input. The surrounding rock deformation was predicted and compared with the safe deformation range (relevant engineering specifications and empirical values). If the predictions met the requirements, a support method was selected. If not, the excavation method was adjusted until the optimal excavation method and distance within the safe deformation range were obtained. Then, based on the optimal excavation method, support design parameters were input, and the surrounding rock deformation was predicted and compared with the safe deformation range (relevant engineering specifications and empirical values). If the predictions met the safe deformation range (relevant engineering specifications and empirical values), the project was applied on site. If not, the support scheme was adjusted until the optimal support method within the safe deformation range was obtained. Figure 6 The comparison curves of the support scheme obtained by using the GA-BP model and the PSO-DNN model and the surrounding rock deformation monitored on site are shown.

[0094] 6.2 Model Evaluation and Feedback. Evaluate the model's practical application results, collect dynamic response data from actual projects, and compare and analyze them with the model's prediction results. Based on the evaluation results, provide feedback and adjustments to the model to further improve its prediction accuracy and applicability.

[0095] Example 2 This embodiment discloses a system for dynamic design of surrounding rock excavation and support parameters for underground engineering; like Figure 7 As shown, a dynamic design system for surrounding rock excavation and support parameters of underground engineering includes: The three-dimensional numerical model building module is configured to: obtain geological characteristics and mechanical parameters of the underground engineering construction area and build a three-dimensional numerical model based on the geological conditions; The database construction module is configured to: conduct numerical simulation experiments based on the three-dimensional numerical model based on the excavation method and support parameters of the underground engineering construction area, and use the surrounding rock deformation as an indicator to generate the underground engineering surrounding rock excavation dynamic response database and the surrounding rock support dynamic response database; The optimal excavation method determination module is configured to: construct an underground engineering excavation scheme design model based on the GA-BP algorithm, train the model using the surrounding rock excavation dynamic response database, and determine the optimal excavation method; The support scheme determination module is configured to: construct an underground engineering support design scheme model based on the PSO-DNN algorithm, train the model using the surrounding rock support dynamic response database, and obtain the optimal support scheme under the optimal excavation method.

[0096] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0097] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a dynamic design method for surrounding rock excavation and support parameters of an underground engineering as described in Example 1.

[0098] Example 4 The purpose of this embodiment is to provide an electronic device.

[0099] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of a method for dynamic design of surrounding rock excavation and support parameters for underground engineering as described in Example 1 are implemented.

[0100] The steps involved in the apparatuses of Examples 2, 3, and 4 above correspond to those of Method Example 1. For detailed implementations, please refer to the relevant description of Example 1. The term "computer-readable storage medium" should be understood to mean a single medium or multiple media containing one or more instruction sets; it should also be understood to include any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and causing the processor to perform any method of the present invention.

[0101] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0102] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A method for dynamic design of surrounding rock excavation and support parameters for underground engineering, characterized in that: include: Obtain the geological characteristics and mechanical parameters of the underground engineering construction area and establish a three-dimensional numerical model based on the geological conditions; Based on the excavation mode and support parameters of the underground engineering construction area, numerical simulation experiments are carried out according to the three-dimensional numerical model. The surrounding rock deformation is used as an indicator to generate the underground engineering surrounding rock excavation dynamic response database and the surrounding rock support dynamic response database. Constructing an underground engineering excavation scheme design model based on the GA-BP algorithm, training the model using the surrounding rock excavation dynamic response database, and determining the optimal excavation method; A support design scheme model for underground engineering based on the PSO-DNN algorithm is constructed, and the model is trained using the surrounding rock support dynamic response database to obtain the optimal support scheme under the optimal excavation method.

2. A method for dynamic design of surrounding rock excavation and support parameters for underground engineering according to claim 1, characterized in that: The process of establishing the three-dimensional numerical model includes: Establish a three-dimensional numerical calculation model based on on-site geological data and project scale characteristics; Set the model boundary conditions; the bottom boundary of the numerical calculation model is a fixed boundary, the surrounding boundaries are displacement constraint boundaries, and the top is a free boundary; The numerical calculation model is assigned parameters based on the rock mechanics parameters obtained from indoor experiments, and initial geostress calculation is carried out to ensure that the initial geostress field of the numerical calculation model is consistent with the measured geostress value in the underground engineering construction area.

3. The method for dynamic design of surrounding rock excavation and support parameters for underground engineering according to claim 1, characterized in that: The process of generating the underground engineering surrounding rock excavation dynamic response database is as follows: Investigate the excavation method of the underground engineering construction area, and combine the excavation method with the excavation distance to obtain a numerical simulation scheme for underground engineering excavation; Based on the obtained numerical simulation scheme, a three-dimensional numerical calculation model is used to carry out numerical simulation experiments to record the dynamic response characteristics of the surrounding rock of the underground engineering after changing the influencing factors; The extracted influencing factors and the dynamic response characteristics of the surrounding rock are used as labels to combine and obtain a data set, forming a database of dynamic response characteristics of surrounding rock excavation in underground engineering.

4. A method for dynamic design of surrounding rock excavation and support parameters for underground engineering according to claim 1, characterized in that: The process of generating the surrounding rock support dynamic response database is as follows: Based on the obtained support parameter range, random functions are used to perform random combinations to obtain support simulation schemes; Based on the obtained support simulation scheme, numerical simulation experiments are carried out using a three-dimensional numerical calculation model; Record the dynamic response characteristics of the surrounding rock of underground engineering after changing the influencing factors; The extracted influencing factors and the dynamic impact characteristics of the surrounding rock are used as labels to combine and obtain the data set, forming a dynamic response characteristic database of the surrounding rock support.

5. The method for dynamic design of surrounding rock excavation and support parameters for underground engineering according to claim 1, characterized in that: Construct an underground engineering excavation scheme design model based on the GA-BP algorithm, including: Set the number of layers, number of neurons and activation function of the BP neural network; Genetic algorithm is used to optimize the initial weights and biases of the neural network, and the root mean square error is used as the fitness evaluation criterion; Gradient descent training is performed through the back-propagation algorithm to further improve model accuracy and reduce the risk of overfitting.

6. A method for dynamic design of surrounding rock excavation and support parameters for underground engineering according to claim 1, characterized in that: Construct an underground engineering support design model based on the PSO-DNN algorithm, including: Set the number of layers, number of neurons and activation function of the neural network; The particle swarm optimization algorithm is used to optimize the initial weights and biases of the neural network, and the root mean square error or classification cross entropy is used as the objective function of PSO optimization; Fine-tune the model through the back-propagation algorithm to reduce the risk of overfitting.

7. A method for dynamic design of surrounding rock excavation and support parameters for underground engineering according to claim 1, characterized in that: The method for dynamic design of surrounding rock excavation and support parameters for underground engineering further includes: The practical application effects of the underground engineering excavation scheme design model based on the GA-BP algorithm and the underground engineering support design scheme model based on the PSO-DNN algorithm were evaluated. The dynamic response data of the surrounding rock in actual projects were collected and compared with the model prediction results. Based on the evaluation results, the model is fed back and adjusted.

8. A dynamic design system for surrounding rock excavation and support parameters for underground engineering, characterized by: include: The three-dimensional numerical model building module is configured to: obtain geological characteristics and mechanical parameters of the underground engineering construction area and build a three-dimensional numerical model based on the geological conditions; The database construction module is configured to: conduct numerical simulation experiments based on the three-dimensional numerical model based on the excavation method and support parameters of the underground engineering construction area, and use the surrounding rock deformation as an indicator to generate the underground engineering surrounding rock excavation dynamic response database and the surrounding rock support dynamic response database; The optimal excavation method determination module is configured to: construct an underground engineering excavation scheme design model based on the GA-BP algorithm, train the model using the surrounding rock excavation dynamic response database, and determine the optimal excavation method; The support scheme determination module is configured to: construct an underground engineering support design scheme model based on the PSO-DNN algorithm, train the model using the surrounding rock support dynamic response database, and obtain the optimal support scheme under the optimal excavation method.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for dynamic design of surrounding rock excavation-support parameters for underground engineering as described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method for dynamic design of surrounding rock excavation and support parameters for underground engineering as described in any one of claims 1 to 7 are implemented.

Images