Dynamic simulation method of water diversion tunnel construction based on adaptive Kriging surrogate model

Abstract

The invention discloses a method for dynamic simulation of water diversion tunnel construction based on an adaptive Kriging proxy model, comprising the following steps: step 1, establishing a Kriging proxy model sample set; step 2, training the Kriging proxy model sample set to establish an adaptive Kriging proxy model; step 3, using the adaptive Kriging proxy model to perform surrounding rock-support structure analysis during water diversion tunnel construction to obtain surrounding rock top arch point displacement and strain structure analysis results; step 4, based on the surrounding rock top arch point displacement and strain structure analysis results, checking the rationality of water diversion tunnel construction support in the initial design stage, selecting reasonable secondary reinforcement measures, and then optimizing initial support time parameters; step 5, coupling a support time parameter optimization module based on the adaptive Kriging proxy model to the water diversion tunnel progress simulation, and realizing dynamic simulation of the water diversion tunnel based on the optimization of the initial support time parameters; the method realizes efficient analysis of the stability of the surrounding rock of the water diversion tunnel.

Description

Technical Field

[0001] The invention belongs to the field of water conservancy and hydropower engineering water diversion tunnel construction, and in particular relates to a water diversion tunnel construction dynamic simulation method based on an adaptive Kriging proxy model. Background Art

[0002] The water diversion tunnel project is an underground concealed project with a complex construction environment and numerous construction processes. It is also characterized by high strength and high risk. Initial support is one of the important measures to limit the deformation of the surrounding rock and ensure the safety of the water diversion tunnel structure. The scientific and reasonable selection of initial support measures is of great significance to the safety of water diversion tunnel construction. At the same time, initial support is a cycle control process for water diversion tunnel excavation, and its support duration directly affects the construction progress.

[0003] In the existing water diversion tunnel construction progress simulation research, Yu [1] et al. proposed a probabilistic risk analysis method for water diversion tunnel construction simulation. Based on the hierarchical simulation model, it was possible to analyze risk factors at the tunnel construction operation level and quantitatively calculate the probability of risk events in each simulation cycle; Bi Lei [2] et al. proposed a multi-factor sensitivity analysis method for the simulation of underground cavern group construction progress based on uniform test design, analyzed the influence of each simulation parameter on the construction progress, compared and screened out the sensitive factors that have a greater impact on the construction progress, and accurately identified the factors that have a significant impact on the construction progress, providing a scientific basis for the rationality and optimization of parameter values ​​in simulation and actual construction; Wang Xiaoling [3] et al. proposed a dynamic visualization simulation method for water diversion tunnel construction based on computer graphics-aided three-dimensional interactive modeling software CATIA, which improved the efficiency of dynamic visualization simulation of water diversion tunnel construction and provided advanced scientific and technological means for construction management personnel to timely understand the construction situation and make effective decisions in time when adjusting the construction plan; Zou [4] et al. considered the mutual influence of schedule risks, identified and quantified the schedule risk path of submarine pipeline projects, provided specific guidance for schedule risk management, and improved the time performance of projects; Xiao Yao [5] et al. proposed a method for dynamically updating the construction simulation parameters of a water diversion tunnel based on adaptive chaotic differential evolution support vector machine, which better fits the time-varying trend of simulation parameters and can improve the accuracy of dynamic simulation of the construction progress of the water diversion tunnel using drilling and blasting method; Wang Tianxing [6] et al. conducted research on the simulation theory and 4D technology of hydraulic tunnel construction progress based on BIM technology, realizing the query of hydraulic tunnel construction progress and the display of construction information; Lin Hanwen [7]et al. proposed an underground cavern group construction simulation model based on M5P-SVR fault prediction, established a transportation simulation loop to calculate the slag transportation time, improved the simulation accuracy of this key process, and proposed a transportation machinery fault prediction method based on M5P-SVR. The cross-validation results showed that this method effectively improved the prediction accuracy.

[0004] However, in the above-mentioned water diversion tunnel progress simulation research, the time parameters for the initial support process simulation are often directly given based on the surrounding rock type and engineering experience, which are highly subjective. At the same time, the influence of other factors such as burial depth and tunnel size on the stability of the surrounding rock is ignored. Therefore, the scientificity of the value of the initial support simulation time parameters needs to be improved.

[0005] At present, in the construction of water diversion tunnels, unless there are major abnormalities such as large deformation or collapse, in order to ensure the rapid and continuous construction, the safety of the initial support is rarely evaluated by mechanical calculation and analysis. The general practice is to select the initial support measures according to the design surrounding rock category (given by analysis and calculation and engineering analogy in the design period), and to adjust the support measures appropriately in combination with the surrounding rock properties revealed on site, and further manage them through monitoring and measurement and according to the allowable deformation provided by the specifications. [5] If deformation outside the specifications occurs, support reinforcement, i.e. secondary support, shall be carried out.

[0006] Since the surrounding rock classification directly guides the selection of initial support, the accuracy of the surrounding rock classification directly affects the accuracy of the initial support selection. At present, the classification of tunnel surrounding rock is mainly carried out in the survey and design stage. However, due to the complex and changeable geological conditions and the limited geological survey methods, the tunnel surrounding rock classification results in the survey and design stage are often quite different from the actual surrounding rock conditions presented during construction. [6] . When the actual surrounding rock quality is better than the estimated surrounding rock quality in the design stage, the support design will be too strong, resulting in a certain degree of waste; when the actual surrounding rock quality is worse than the estimated surrounding rock quality in the design stage, the support design will be too weak, posing a huge safety hazard. Secondly, at present, during the construction stage of the water diversion tunnel, most people still rely on empirical observation methods and initial exploration data to judge the surrounding rock level. The results obtained are highly subjective and less accurate, which makes the grading indicators fail to follow the specifications and the grading results are prone to errors. [6] At the same time, the selection of initial support parameters should theoretically be determined according to the stability of the surrounding rock. However, the surrounding rock type is only an objective geological condition. The selection of support measures based only on the surrounding rock type characterizes the surrounding rock stability. The influencing factors considered are too single, ignoring the changes in initial ground stress caused by different burial depths and the influence of different water diversion tunnel structure sizes on surrounding rock deformation. In addition, secondary support is carried out only after a large deformation is detected through monitoring and measurement, which cannot effectively ensure construction safety in a timely manner.

[0007] In summary, in the existing water diversion tunnel progress simulation research, the support time parameters are determined only according to the surrounding rock type, ignoring the influence of factors such as initial ground stress caused by burial depth and tunnel structure size; at the same time, there are the following problems in the construction of water diversion tunnels: the subjective judgment of surrounding rock types on site is large, which leads to the subjectivity of the adjustment of initial support measures; the initial support selection considers a single factor, ignoring the influence of factors such as burial depth and structural size, which easily leads to the design of initial support measures that cannot meet the actual surrounding rock stability requirements and requires secondary reinforcement. The finite element model can comprehensively consider the influence of multiple factors such as structural size, surrounding rock conditions, and burial depth, analyze the surrounding rock deformation under different cycles and different influencing factors, and further scientifically guide the selection of initial support parameters. Therefore, this paper considers multiple factors such as structural size, surrounding rock conditions, and burial depth, establishes a surrounding rock-support integrated finite element model, and uses the deformation of finite element simulation results as a quantitative basis to replace the surrounding rock category to select support parameters. However, due to the high calculation time cost of the finite mechanism model, it is difficult to meet the requirements of tunnel dynamic construction simulation. Therefore, this paper proposes to use the Kriging proxy model to replace the finite element mechanism model to quickly predict the deformation of the surrounding rock of the water diversion tunnel, and then reasonably determine the support time parameters. At the same time, the MMSE adaptive point addition criterion optimized by the parallel point addition method is used to improve the proxy model, establish an adaptive proxy model, reduce the number of training samples, and improve modeling efficiency. In summary, this paper conducts a rapid analysis of the stability of the surrounding rock of the water diversion tunnel based on the Kriging proxy model, and uses the deformation of the analysis results as a quantitative basis to select the initial support measures, thereby realizing the optimization of the initial support time parameters of the water diversion tunnel. Summary of the invention

[0008] The purpose of the present invention is to propose a finite element structural analysis method to identify the structural safety situation, and to check the designed initial support based on the result of the structural analysis, and to select secondary reinforcement measures to optimize the initial support time parameters. At the same time, in view of the shortcoming that the existing tunnel structure analysis research directly uses the finite element mechanism model to perform structural safety analysis, resulting in a large calculation time cost, an adaptive Kriging proxy model is introduced to replace the Abaqus finite element calculation mechanism model to achieve efficient structural analysis. In view of the large time cost required for obtaining samples by the proxy model, an improved MMSE adaptive point addition criterion is proposed to reduce the total number of samples required for model establishment. Finally, an optimization module of the initial support time parameters based on the adaptive Kriging proxy model is integrated into the progress simulation CYCLONE cycle to achieve dynamic simulation of the construction progress of the water diversion tunnel.

[0009] The present invention is implemented by the following technical solutions:

[0010] 1. A dynamic simulation method for water diversion tunnel construction based on an adaptive Kriging proxy model, the dynamic simulation method for water diversion tunnel construction comprising the following steps:

[0011] Step 1: Use Abaqus finite element LHS sampling to obtain surrounding rock mechanical parameters, burial depth, and structural size parameters to calculate surrounding rock deformation and establish a Kriging proxy model sample set;

[0012] Step 2, training the Kriging proxy model sample set to establish an adaptive Kriging proxy model;

[0013] Step 3: Use the adaptive Kriging proxy model to analyze the surrounding rock-support structure during the construction period of the water diversion tunnel to obtain the displacement and strain structural analysis results of the surrounding rock top arch point;

[0014] Step 4: Based on the displacement and strain structure analysis results of the surrounding rock top arch point, check the rationality of the initial water diversion tunnel construction support design, select reasonable secondary reinforcement measures, and then optimize the initial support time parameters;

[0015] Step 5: Couple the support time parameter optimization module based on the adaptive Kriging proxy model to the diversion tunnel progress simulation to achieve dynamic simulation of the diversion tunnel based on initial support time parameter optimization.

[0016] Furthermore, the process of establishing the sample set of the adaptive Kriging proxy model in step 1 is as follows:

[0017] (1-1) Establish a local surrounding rock-support integrated finite element model for water diversion tunnel chamber stability analysis, obtain the displacement, stress and deformation of the surrounding rock at the top arch point, and verify the finite element model based on actual on-site monitoring data;

[0018] (1-2) Obtain model input parameters through LHS sampling;

[0019] (1-3) The displacement and stress-strain simulation results of the surrounding rock under different input parameters are obtained through finite element model simulation calculation, which are the output parameters;

[0020] (1-4) The input parameters and output parameters are summarized to establish the Kriging proxy model sample set, which is divided into a training set and a validation set.

[0021] Furthermore, the specific process of establishing the adaptive Kriging proxy model with the training sample set is as follows:

[0022] The training set is trained to establish the initial Kriging proxy model; the maximum prediction variance addition criterion is used to add samples, and the new sample set is iterated cyclically to finally establish the adaptive Kriging proxy model:

[0023] (2-1) Use the Kriging proxy model to train the initial training sample points, and use the improved genetic algorithm to optimize the Kriging model hyperparameters θ and p to establish the initial Kriging proxy model;

[0024] (2-2) Based on the improved maximum prediction variance adding point criterion and combined with the prediction variance result of the Kriging model, the point with the largest variance is obtained, and the improved firefly foraging algorithm is used to accelerate the optimization process. The input parameters of this point are used as new input parameters to substitute into the finite element mechanism model for calculation, and a new set of training samples is obtained and added to the initial training samples to establish a new training set; wherein: the maximum prediction variance mathematical model is:

[0025] find x=[x1,x2,…,x m ] T

[0026]

[0027] stg i (x)≤0 i=1,2,…,l

[0028] h j (x)≤0 j=1,2,…,s

[0029] x L ≤x≤x U

[0030] Among them, x is the optimal design variable, is the prediction variance, g i (x), h j (x) is the constraint function, x L 、x U are the upper and lower bounds of the design variables;

[0031] (2-3) The Kriging proxy model is re-established by training the new training set, and the mean square error index is used to evaluate the global accuracy of the Kriging proxy model. If the accuracy requirements are met, the Kriging proxy model is established. Otherwise, steps (2-2) and (2-3) are repeated to iterate the establishment of the Kriging proxy model until the model accuracy meets the requirements.

[0032] Furthermore, in step 4, reasonable initial support measures and corresponding support time parameters are selected according to the surrounding rock deformation analysis results obtained, specifically:

[0033] (4-1) According to the requirements of the specifications and engineering construction experience, the corresponding relationship between the surrounding rock deformation and the secondary reinforcement parameters is established;

[0034] (4-2) Based on the surrounding rock deformation analysis results obtained in step 3, according to the established correspondence between the surrounding rock deformation and the secondary reinforcement parameters, the corresponding secondary support parameters are selected and combined with the designed initial support as the optimized initial support, and the support time corresponding to the support parameters is quantitatively calculated.

[0035] Furthermore, the step 5 couples the support parameter optimization module to the diversion tunnel progress simulation to realize the diversion tunnel dynamic simulation based on the initial support time parameter optimization. The specific implementation process is as follows:

[0036] (5-1) Dynamically update and read the latest input parameter data in each cycle of progress simulation;

[0037] (5-2) Calling an external subroutine, namely the adaptive Kriging proxy model module, obtains the surrounding rock deformation output parameter through structural safety analysis of the input parameters, and determines the secondary reinforcement measures through the deformation, further obtains the optimized initial support parameters, and then determines the corresponding initial support time parameters;

[0038] (5-3) Feeding back the initial support time parameters to the simulation main program to simulate the current cycle progress;

[0039] (5-4) Repeat the above steps until the simulation ends.

[0040] Beneficial Effects

[0041] Advantages of the present invention:

[0042] 1. The optimization selection of tunnel simulation support time parameters based on structural analysis makes up for the shortcomings of strong subjectivity and incomplete consideration of factors caused by the selection of initial support based on surrounding rock types in the current water diversion tunnel construction simulation;

[0043] 2. The adaptive Kriging proxy model was introduced to replace the Abaqus finite element computer model, realizing efficient analysis of the stability of the surrounding rock of the water diversion tunnel;

[0044] 3. The MMSE adaptive addition rule improved by parallel addition is used to improve the Kriging proxy model, which reduces the total number of samples required to establish the Kriging proxy model, thereby reducing the time cost required to obtain samples;

[0045] 4. The initial support parameter optimization module is coupled to the progress simulation CYCLONE cycle to achieve the optimization of the initial support time parameters in the progress simulation of the water diversion tunnel, and based on this, the dynamic simulation of the water diversion tunnel is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Cross-section of the Abaqus finite element model.

[0047] Figure 2 Longitudinal section of the Abaqus finite element model.

[0048] Figure 3 Adaptive Kriging surrogate model building process.

[0049] Figure 4 A schedule simulation model coupled with initial support time parameter optimization. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and examples. The following embodiments are only descriptive and not restrictive, and the protection scope of the present invention cannot be limited thereto.

[0051] In the existing water diversion tunnel progress simulation, the time parameters of the initial support process are directly given by only the surrounding rock type and engineering experience, ignoring the influencing factors such as burial depth and structural size, lacking scientific rationality. In view of this main problem, the present invention specifically provides a water diversion tunnel construction progress dynamic simulation optimization method based on an adaptive Kriging proxy model; the method is to establish an efficient analysis method for surrounding rock stability based on an adaptive Kriging proxy model; wherein, an improved meta-heuristic algorithm is used to optimize model parameters, and an improved MMSE adaptive point addition criterion is used to optimize the Kriging proxy model, thereby overcoming the problem of long time consumption of the finite element model, realizing efficient analysis of the structural stability during the construction period of the water diversion tunnel, further predicting the long-term deformation of the surrounding rock based on the structural analysis results, checking the rationality of the designed support, and selecting reasonable secondary reinforcement parameters to supplement the designed initial support, realizing the optimization of the initial support parameters during the construction period of the water diversion tunnel, and finally realizing the dynamic simulation construction process of the water diversion tunnel progress.

[0052] The specific implementation of the present invention is further described in detail below with reference to the accompanying drawings.

[0053] 1. According to the specific engineering data and using Abaqus software to establish a finite element calculation model, such as Figure 1 , Figure 2 As shown in the figure, the input parameter samples of the model are obtained through sampling design, and the finite element structure calculation is repeated based on the input samples to obtain the output parameters of the deformation and establish the sample set of the adaptive Kriging proxy model. The specific steps are as follows:

[0054] (1) The initial model parameters such as the cross-sectional type and size of the water diversion tunnel are obtained through the construction design data and actual construction record data. Considering the mechanical boundary conditions of the drilled and blasted water diversion tunnel, an integrated finite element model of the surrounding rock and design support of the drilled and blasted water diversion tunnel is established. On this basis, the accuracy of the model is verified according to the field measured data (input parameters such as surrounding rock parameters obtained by drilling-deformation monitoring data). The basic model parameters are as follows:

[0055] a. Model size: According to relevant specifications and literature, the diameter of the water diversion tunnel is d (unit: m, the same below). Considering the limited influence range of the tunnel face, 2d is taken in front and behind the tunnel face, that is, 4d is taken in the excavation direction; the effect of the surrounding rock mass beyond 6 times the tunnel diameter is very small, and the vertical symmetry axis is 6d up and down, that is, 12d in height, and 6d is taken in the horizontal direction, that is, 12d in width.

[0056] b. Model calculation section: the section that has been excavated at a distance of one cycle from the tunnel face. This section is the section with the least constraint from the tunnel face in the exposed rock mass after one cycle of excavation. Therefore, it is the section with the largest deformation and the most dangerous. Therefore, the safety of this section is used to ensure the safety of each cycle of excavation.

[0057] c. Model calculation point: According to the preliminary mechanical analysis, the top arch point of the same section has the largest deformation, which is the dangerous point, so this point is selected as the calculation point.

[0058] d. Simplification of model mechanical parameters: rock mass integrity, groundwater, etc.

[0059] Groundwater: In the surrounding rock of the same level, the surrounding rock level will be appropriately reduced after encountering water. The extent of the reduction mainly depends on ① the lithology of the surrounding rock and the state of the structural surface; ② the nature, size, circulation conditions of the groundwater and the infiltration status and degree of harm to the surrounding rock.

[0060] Rock mass integrity: reflected through the adjustment of surrounding rock parameters.

[0061] (2) The value ranges of various input parameters are determined according to the specifications and engineering experience. The surrounding rock parameters (elastic modulus, Poisson's ratio, number of surrounding rock volume joints, etc.) and burial depth parameters (converted into initial geostress) (30 groups) are obtained through Latin hypercube sampling and input into the initial finite element model in turn to establish a computable finite element model.

[0062] (3) The simulation results of surrounding rock displacement, stress and strain under different input parameters are obtained through Abaqus finite element simulation, which are the output parameters.

[0063] (4) The input parameters and output parameters are summarized to establish a proxy model sample set, which is divided into a training set and a validation set in a ratio of 7:3.

[0064] 2. Use the improved maximum prediction variance (MMSE) point addition criterion to cyclically add points to update the training sample set, establish an adaptive proxy model, and use the improved meta-heuristic algorithm to cyclically iterate the Kriging proxy model until the accuracy of the Kriging proxy model meets the requirements. Figure 3 As shown, the specific steps are as follows:

[0065] (1) The initial training sample points were trained using the DACE toolbox provided by Matlab, and the improved meta-heuristic algorithm was used to optimize the model hyperparameters θ and p, which accelerated the model building process and improved the model accuracy, and established the initial Kriging proxy model.

[0066] The Kriging model is an interpolation model whose interpolation result is defined as the linear weighting of the known sample function response value, that is,

[0067]

[0068] Therefore, as long as the weighting coefficient ω=[ω (1) ω (2) …ω (n) ], the estimated value of any design solution in the design space can be calculated. In order to obtain the weighting coefficient, the Kriging model introduces a statistical assumption: the unknown function is regarded as a specific implementation of a Gaussian static random process. The static random process is defined as

[0069] Y(x)=β0+Z(x)

[0070] Where: β0 is an unknown constant, also known as the global trend model, representing the mathematical expectation of Y(x); Z(x) is a model with a mean of zero and a variance of There is a certain correlation between these random variables at different locations in the design space, which can be expressed by covariance:

[0071] Cov[Z(x),Z(x')]=σ 2 R(x,x')

[0072] Where: R(x,x') is the "correlation function" (only related to spatial distance), and it satisfies that it is equal to 1 when the distance is zero; it is equal to 0 when the distance is infinite; the correlation decreases as the distance increases.

[0073] Based on the above assumptions, the Kriging model seeks the optimal weighting coefficient ω so that the mean square error

[0074]

[0075] Minimum and meets the following interpolation conditions (or no deviation conditions):

[0076]

[0077] The meta-heuristic algorithm is an optimization algorithm that combines random algorithms with local algorithms under the inspiration of bionics and random phenomena in nature. The meta-heuristic algorithm has random factors, and the results obtained will be different, so there is a certain probability of jumping out of the local optimum and it is more likely to obtain the global optimal solution. Moreover, the meta-heuristic algorithm does not have any special requirements for the objective function (such as differentiable or convex optimization) and is not limited to specific problems. The present invention selects an improved meta-heuristic algorithm in order to obtain the global optimal solution of the proxy model parameters and the MMSE plus point criterion prediction variance more quickly.

[0078] (2) Based on the improved maximum prediction variance (MMSE) point addition criterion and combined with the prediction variance results of the Kriging model, the maximum and sub-maximum points of the variance are obtained, and the input parameters of the two points are used as new input parameter samples to substitute into the finite element mechanism model for mechanism calculation, and two new training samples are obtained to add to the initial training samples, realize parallel point addition, and establish a new training set. The mathematical model of the MMSE point addition criterion is as follows:

[0079] find x=[x1,x2,…,x m ] T

[0080]

[0081] stg i (x)≤0i=1,2,…,l

[0082] h j (x)≤0j=1,2,…,s

[0083] x L ≤x≤x U

[0084] Among them, x is the optimal design variable, is the prediction variance, g i (x), h j (x) is the constraint function, x L 、x U are the upper and lower bounds of the design variables.

[0085] The model is optimized and solved using global optimization algorithms such as improved metaheuristic algorithm.

[0086] (3) The Kriging proxy model is re-established by training the sample set after adding points. The global accuracy of the Kriging proxy model is evaluated using the validation set and the mean square error as an indicator. If the accuracy requirements are met, the Kriging proxy model is established. Otherwise, steps (2) and (3) are repeated to perform an adaptive loop iteration of the Kriging proxy model until the model accuracy meets the requirements.

[0087] 3. Based on the Kriging proxy model established in step 2 and combined with the field measured input data, an efficient structural analysis is performed to obtain the Kriging proxy model output parameters such as surrounding rock deformation. The specific steps are as follows:

[0088] (1) The data obtained every 30-50 m at the project site, including surrounding rock parameters (elastic modulus, Poisson's ratio, rock mass integrity, etc.), burial depth (converted initial geostress), type and size of the diversion tunnel, etc., are used to obtain the factors affecting surrounding rock deformation in each cycle as input parameters of the Kriging proxy model.

[0089] (2) The obtained construction parameters are input into the Kriging proxy model for structural simulation analysis, and the surrounding rock deformation analysis results (displacement, stress, and strain) of each cycle under the input parameters are efficiently obtained.

[0090] 4. Based on the surrounding rock deformation analysis results obtained in step 3, select reasonable secondary reinforcement measures and comprehensively obtain the optimized initial support time parameters. The specific steps are as follows:

[0091] (1) During the design period, the corresponding relationship between surrounding rock deformation (range value) and secondary reinforcement measures is established through engineering analogy method and the experience of engineering construction experts.

[0092] (2) Based on the surrounding rock structure analysis results obtained in step 3, the corresponding secondary support parameters are selected from the corresponding relationship between the surrounding rock deformation and the secondary support parameters, and are combined with the designed initial support as the optimized initial support measures. The support time corresponding to the optimized initial support parameters is quantitatively calculated.

[0093] 5. The water diversion tunnel initial support parameter optimization module based on the adaptive Kriging proxy model is coupled into the water diversion tunnel progress simulation CYCLONE cycle to realize the dynamic simulation of the water diversion tunnel progress during the construction period based on the initial support parameter optimization, such as Figure 4 As shown, the specific implementation process is as follows:

[0094] (1) Progress simulation In each CYCLONE cycle, the latest input parameter data (surrounding rock parameters, burial depth, etc.) of the current cycle is dynamically read and updated according to the pile number.

[0095] (2) Calling an external subroutine, namely the adaptive Kriging proxy model module, performs efficient structural safety analysis based on the latest input parameters of the current cycle, obtains the output parameters of the surrounding rock deformation of the current cycle, and further finds the initial support time parameters corresponding to the current cycle through the deformation.

[0096] (3) Feedback the initial support time parameters to the simulation main program to simulate the current cycle progress.

[0097] (4) Repeat the above steps until the simulation ends.

[0098] The present invention is not limited to the embodiments described above. The above description of the specific embodiments is intended to describe and illustrate the technical solution of the present invention. The above specific embodiments are merely illustrative and not restrictive. Without departing from the scope of the present invention and the scope of protection of the claims, a person of ordinary skill in the art can also make many forms of specific changes under the guidance of the present invention, which all fall within the scope of protection of the present invention.

Claims

1. A dynamic simulation method for water diversion tunnel construction based on an adaptive Kriging proxy model, characterized in that: The water diversion tunnel construction dynamic simulation method comprises the following steps: Step 1: Use LHS sampling to obtain surrounding rock mechanical parameters, burial depth, and structural size parameters, and obtain surrounding rock deformation based on Abaqus finite element calculation to establish a Kriging proxy model sample set; Step 2: Training the Kriging proxy model sample set to establish an adaptive Kriging proxy model; the specific process of step 2: Training the sample set to establish an adaptive Kriging proxy model is as follows: The training set is trained to establish the initial Kriging proxy model; the maximum prediction variance addition criterion is used to add samples, and the new sample set is iterated cyclically to finally establish the adaptive Kriging proxy model: (2-1) Use the Kriging proxy model to train the initial training sample points, and use the improved genetic algorithm to optimize the Kriging model hyperparameters θ and p to establish the initial Kriging proxy model; (2-2) Based on the improved maximum prediction variance adding point criterion and combined with the prediction variance result of the Kriging model, the point with the largest variance is obtained, and the improved firefly foraging algorithm is used to accelerate the optimization process. The input parameters of this point are used as new input parameters to substitute into the finite element mechanism model for calculation, and a new set of training samples is obtained and added to the initial training samples to establish a new training set; wherein: the maximum prediction variance mathematical model is: Among them, x is the optimal design variable, is the prediction variance, g i (x), h j (x) is the constraint function, x L 、x U are the upper and lower bounds of the design variables; (2-3) Re-establish the Kriging proxy model by training the new training set, and use the mean square error index to evaluate the global accuracy of the Kriging proxy model. If the accuracy requirements are met, the Kriging proxy model is established. Otherwise, repeat steps (2-2) and (2-3) to iterate the establishment of the Kriging proxy model until the model accuracy meets the requirements. Step 3: Use the adaptive Kriging proxy model to analyze the surrounding rock-support structure during the construction period of the water diversion tunnel to obtain the displacement and strain structural analysis results of the surrounding rock top arch point; Step 4: Based on the displacement and strain structure analysis results of the surrounding rock top arch point, check the rationality of the initial water diversion tunnel construction support design, select reasonable secondary reinforcement measures, and then optimize the initial support time parameters; Step 5: Couple the support time parameter optimization module based on the adaptive Kriging proxy model to the diversion tunnel progress simulation to achieve dynamic simulation of the diversion tunnel based on initial support time parameter optimization.

2. The method for dynamic simulation of water diversion tunnel construction based on the adaptive Kriging proxy model according to claim 1 is characterized in that: The process of establishing the sample set of the adaptive Kriging proxy model in step 2: (1-1) Establish a local surrounding rock-support integrated finite element model for water diversion tunnel chamber stability analysis, obtain the displacement, stress and deformation of the surrounding rock at the top arch point, and verify the finite element model based on actual on-site monitoring data; (1-2) Obtain model input parameters through LHS sampling; (1-3) The displacement and stress-strain simulation results of the surrounding rock under different input parameters are obtained through finite element model simulation calculation, which are the output parameters; (1-4) The input parameters and output parameters are summarized to establish the Kriging proxy model sample set, which is divided into a training set and a validation set.

3. The method for dynamic simulation of water diversion tunnel construction based on the adaptive Kriging proxy model according to claim 1, characterized in that: The step 4 selects reasonable initial support measures and corresponding support time parameters according to the surrounding rock deformation analysis results, specifically: (4-1) According to the requirements of the specifications and engineering construction experience, the corresponding relationship between the surrounding rock deformation and the secondary reinforcement parameters is established; (4-2) Based on the surrounding rock deformation analysis results obtained in step 3, according to the established correspondence between the surrounding rock deformation and the secondary reinforcement parameters, the corresponding secondary support parameters are selected and combined with the designed initial support as the optimized initial support, and the support time corresponding to the support parameters is quantitatively calculated.

4. The method for dynamic simulation of water diversion tunnel construction based on the adaptive Kriging proxy model according to claim 1 is characterized in that: The step 5 couples the support parameter optimization module to the diversion tunnel progress simulation to realize the diversion tunnel dynamic simulation based on the initial support time parameter optimization. The specific implementation process is as follows: (5-1) Dynamically update and read the latest input parameter data in each cycle of progress simulation; (5-2) Calling an external subroutine, namely the adaptive Kriging proxy model module, obtains the surrounding rock deformation output parameter through structural safety analysis of the input parameters, and determines the secondary reinforcement measures through the deformation, further obtains the optimized initial support parameters, and then determines the corresponding initial support time parameters; (5-3) Feeding back the initial support time parameters to the simulation main program to simulate the current cycle progress; (5-4) Repeat the above steps until the simulation ends.

Images