Test method and system for simulating grouting reinforcement of tunnel fault fracture zone

Abstract

The invention relates to the technical field of fault fracture zone reinforcement tests, in particular to a test method and system for simulating grouting reinforcement of a tunnel fault fracture zone. According to the test method and system, grouting pressure and a tunnel vault loading historical curve are read, a pressure slope abrupt change point is extracted, and pore water pressure and displacement meter reading are synchronously obtained; according to the method, after square difference accumulation and abnormal point elimination are carried out on the closing quantity obtained by calculating the original width of the crack and the displacement difference value, the extracted closing deformation information has higher stability and anti-interference capability, and meanwhile, the trend characteristics of crack closing are extracted by means of the time sequence fitting capability of a physical information neural network; according to the method, quantitative tracking of closed deformation is realized, synchronous stress and displacement data of three-dimensional lining grid nodes are introduced, a non-dominated sorting genetic algorithm is used for searching a plurality of working condition combinations, and a boundary judgment criterion is established for an interval in which crack closing and permeation trends are jointly stable; and the selection of the control parameters is more in line with the multi-objective optimization requirement of the reinforcement response.

Description

Technical Field

[0001] This invention relates to the field of fault fracture zone reinforcement testing technology, and in particular to a test method and system for simulating grouting reinforcement of fault fracture zones in tunnels. Background Technology

[0002] The field of fault fracture zone reinforcement test technology aims to study the mechanical response mechanism of fault fracture zones under reinforcement measures by constructing controllable and repeatable physical or numerical simulation methods, evaluate the reinforcement effect, optimize reinforcement parameters, and improve the stability and safety of underground structures such as tunnels when passing through fault fracture zones.

[0003] The purpose of this experimental method for simulating grouting reinforcement of fault fracture zones in tunnels is to reproduce the interaction process between the tunnel structure and the rock mass in a laboratory environment through experimental methods. This allows for the acquisition of key parameters such as deformation, failure, and strength changes of the rock mass before and after reinforcement, in order to evaluate the effect and mechanism of grouting reinforcement. The aim is to achieve visualization and quantitative verification of reinforcement technology, providing experimental support and theoretical basis for engineering design.

[0004] Existing technologies for simulating fault fracture zone reinforcement mainly focus on the overall response process of the reinforced structure and surrounding rock under loading conditions. They mostly rely on qualitative assessments by observing changes in mechanical parameters before and after reinforcement, lacking quantitative modeling of fracture micro-behaviors such as closure processes and permeability characteristics. Grouting processes often encounter problems such as unclear fracture responses and difficulty in tracing seepage paths, which weakens the causal chain between pressure loading and structural response, making it difficult to determine the specific failure development process and reinforcement formation mechanism. Parameter evaluations often use static final values ​​as evaluation benchmarks, ignoring intermediate state changes during reinforcement evolution. This makes it difficult to directly correlate key indicators such as peak pressure and duration with stability, restricting the targeted optimization of reinforcement parameters and missing important characteristics of fracture response in the early or mid-to-late stages of grouting, thus limiting the interpretability and application value of experimental results. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a test method and system for simulating grouting reinforcement of fault fracture zones in tunnels.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a test method for simulating grouting reinforcement of fault fracture zones in tunnels, comprising the following steps:

[0007] S1: Read the grouting pressure curve and the tunnel arch loading history, calculate the pressure slope for continuous time periods, and extract the pore water pressure and displacement gauge readings corresponding to the slope change points to generate a loading displacement sequence.

[0008] S2: Based on the loading displacement sequence and the original crack width value, the instantaneous closure amount is obtained by calculating the difference between displacement and width at each moment, and the difference squared and accumulated to eliminate fluctuation anomalies. Then, a physical information neural network is used to statistically accumulate the trend of the increase and decrease of the time series closure amount to establish a closure deformation curve.

[0009] S3: Extract the crack opening value at each moment based on the closed deformation curve, construct the flow area by performing multiplication and division operations with the crack tube diameter and rock layer porosity ratio respectively, and then convert the flow velocity coefficient and the flow area ratio into the permeability per unit time to generate a permeability coefficient sequence.

[0010] S4: Based on the permeability coefficient sequence and the closure deformation curve, the stress and displacement values ​​of the mesh nodes in the three-dimensional tunnel lining structure are read simultaneously. The non-dominated sorting genetic algorithm is used to judge the change range of the closure value and the decreasing trend of permeability in multiple parameter combinations, and to filter the combination boundary that keeps both stable, and output the optimized parameter set.

[0011] S5: Based on the optimized parameter set, compare the peak grouting pressure and duration and record the frequency of parameter occurrence. Then, divide the grout viscosity into continuous intervals and extract the location of concentrated distribution segments to obtain key stability parameters.

[0012] As a further aspect of the present invention, the loading displacement sequence includes the pressure change moment, the corresponding displacement value, and timestamp data; the closed deformation curve specifically includes the closure amount time series, the closure rate distribution, and the abnormal fluctuation index; the permeability coefficient sequence includes the equivalent permeability, the permeability fluctuation range, and the time-period attenuation ratio; the optimization parameter set includes the grouting pressure peak range, the grouting duration interval, and the grout viscosity level; and the key stability parameters specifically refer to the pressure control value, the viscosity classification value, and the grouting duration.

[0013] As a further aspect of the present invention, the specific steps for generating the loading displacement sequence are as follows:

[0014] Read the grouting pressure curve and the tunnel arch loading history, count the pressure difference every ten seconds and compare the difference between two points before and after, filter the absolute value index of the difference, and extract the corresponding pore water pressure and displacement gauge reading to generate a set of abrupt change point indexes;

[0015] Based on the set of mutation points, the pore water pressure and the original reading of the displacement gauge at the corresponding time of the index point are retrieved and arranged in ascending order of time to generate a loading displacement sequence.

[0016] As a further aspect of the present invention, the specific steps for generating the closed deformation curve are as follows:

[0017] Based on the loading displacement sequence and the original crack width value, the displacement and width value corresponding to each time point are extracted, the one-to-one correspondence difference is calculated and the original difference pair list is generated, and the interval consistency verification and abnormal time point investigation are performed to generate the instantaneous closure quantity list.

[0018] Based on the instantaneous closure value list, the difference of each item in the list is squared and a standard deviation threshold is established. After excluding data points higher than the threshold, the remaining closure values ​​are summed over time and a cumulative list is generated to generate a cumulative set of closure values.

[0019] Based on the accumulated set of closed quantities, a physical information neural network is used to identify the positive and negative change trends of continuous points on the time axis and divide adjacent band intervals. The total value of the closed increment within each segment is calculated and uniformly mapped to the time node to establish a closed deformation curve.

[0020] As a further embodiment of the present invention, the physical information neural network takes the time node sequence, the total value of the closure increment, and the rock mass stress residual term as network inputs simultaneously. The network calculates the closure prediction value for each time node in sequence, calculates the mean square error by subtracting the corresponding cumulative value of the closure quantity, and calculates the error between the stress residual term and the residual quantity of the physical conservation equation. The two types of errors are weighted and summed to form the total loss. Backpropagation is performed to update the network weights. The iteration is repeated until the total loss converges, and the closure deformation trend value of each time node is output. The closure change values ​​of adjacent nodes are marked as positive or negative.

[0021] As a further aspect of the present invention, the specific steps for generating the permeability coefficient sequence are as follows:

[0022] Based on the time point of the closed deformation curve and the original width of the crack, the difference between the closing amount and the width is extracted and the sum of squares is calculated. Values ​​exceeding the limit of the sum of squares are eliminated and the closing amount at each moment is accumulated to generate a list of opening values.

[0023] Based on the list of opening values, the diameter of the fractured tube, and the porosity of the rock strata, multiplication and division operations are performed sequentially to calculate the flow cross-sectional area and convert it into the permeability velocity per unit time, generating a permeability coefficient sequence.

[0024] As a further aspect of the present invention, the specific steps for generating the optimized parameter set are as follows:

[0025] Based on the permeability coefficient sequence and the closed deformation curve, the three-dimensional tunnel mesh number is matched for all time nodes, the stress value and displacement vector of the node under the corresponding number are read, the corresponding components of each node are extracted and formatted and stored as structured data, and the node stress displacement set is generated.

[0026] Based on the node stress-displacement set, the peak grouting pressure, duration, and grout viscosity parameters corresponding to the node are collected, assembled sequentially to form a ternary parameter combination list, and cross-indexing is performed. After removing duplicates and boundary value combinations, a parameter combination set is generated.

[0027] Based on the parameter combination set, a non-dominated sorting genetic algorithm is used to extract the closure deformation amplitude and the descent gradient of the permeability coefficient from the data corresponding to each combination, determine whether they are all within the target interval and mark the effective combination, and output the parameter list that meets the bistable condition as the optimized parameter set.

[0028] As a further aspect of the present invention, the non-dominated sorting genetic algorithm initializes a parameter vector population, calculates the closure deformation amplitude and permeability coefficient descent gradient of each individual as the objective function value, performs non-dominated sorting to divide the population into levels, assigns level numbers to each individual, calculates the crowding distance between individuals within each level, conducts tournament selection based on level numbers and crowding distance to select parent individuals, performs crossover operation on the parent parameter vector to generate new individuals, performs mutation operation on the new individuals to obtain offspring, merges the parent and offspring, and re-performs non-dominated sorting and crowding distance calculation, selects a predetermined number of individuals from the merged set according to level priority and crowding order to form the next generation, repeats the above generational iteration until a preset number of generations is reached, and extracts the first level individuals to form an optimized parameter set.

[0029] As a further aspect of the present invention, the specific steps for generating the aforementioned key stability parameters are as follows:

[0030] Based on the grouting peak pressure and duration in the optimized parameter set, the frequency of occurrence of each parameter combination is counted and sorted in descending order to generate high-frequency parameter pairs;

[0031] Based on the high-frequency parameters and the slurry viscosity range division results, the viscosity range is extracted according to the frequency of occurrence and the corresponding pressure and duration are recorded to obtain key stability parameters.

[0032] A test system for simulating grouting reinforcement of tunnel fault fracture zones, the test system being used to perform the aforementioned test method for simulating grouting reinforcement of tunnel fault fracture zones, the system comprising:

[0033] Pressure slope calculation module: Based on the grouting pump pressure curve and the tunnel arch loading history, extract pressure readings for continuous time periods, calculate the pressure difference between adjacent moments divided by the time difference, filter the moments of slope change, read the corresponding pore water pressure and displacement gauge readings, and obtain a list of slope change points.

[0034] Closed deformation processing module: Based on the list of slope abrupt change points and the original crack width value, calculate the difference between the displacement gauge reading and the width at the abrupt change time, square and accumulate the difference and remove abnormal fluctuation points, and combine the physical information neural network to perform time series increase and decrease trend accumulation processing on the accumulated sequence to obtain the closed deformation curve;

[0035] Permeability coefficient generation module: Based on the closed deformation curve, the diameter of the fracture tube and the porosity ratio of the rock layer, the flow area is obtained by multiplying the fracture opening value by the porosity ratio and dividing by the tube diameter at any given time, and the flow velocity coefficient is divided by the flow area to generate a permeability coefficient sequence.

[0036] Parameter optimization and screening module: Based on the closed deformation curve and permeability coefficient sequence, the stress and displacement values ​​of the three-dimensional tunnel lining structure mesh nodes are read simultaneously. The non-dominated sorting genetic algorithm is used to determine the changes in the closure amplitude and the decrease in permeability of multiple parameter combinations, and a stable combination is screened to output the optimized parameter set.

[0037] Stability parameter extraction module: Based on the optimized parameter set, compare the frequency of grouting peak pressure and duration, statistically analyze the concentrated distribution of grout viscosity in continuous intervals, and obtain key stability parameters.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0039] In this invention, by reading the historical curves of grouting pressure and tunnel arch loading, extracting the abrupt change points of pressure slope, and simultaneously acquiring pore water pressure and displacement gauge readings, a loading displacement sequence that highly matches the actual stress evolution process can be constructed, significantly enhancing time accuracy and response sensitivity.

[0040] In this invention, the closure amount calculated by combining the original crack width and displacement difference, after summing the squared differences and eliminating outliers, has stronger stability and anti-interference ability in the extracted closure deformation information. At the same time, the trend characteristics of crack closure are extracted by using the time series fitting ability of physical information neural network, so as to realize the quantifiable tracking of closure deformation.

[0041] In this invention, by introducing synchronous stress and displacement data of three-dimensional lining mesh nodes, a non-dominated sorting genetic algorithm is used to search for multiple combinations of working conditions. Boundary judgment criteria are established for the interval where crack closure and seepage trends are both stable, so that the selection of control parameters is more in line with the multi-objective optimization requirements of reinforcement response.

[0042] In this invention, by analyzing the frequency, persistence characteristics and concentrated distribution segment location of the parameter set output from multiple rounds of optimization, the key indicators of the dominant range of slurry viscosity change and stability correlation can be extracted from statistical characteristics, thereby enhancing the parameter extraction granularity and response correspondence in the grouting process. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0044] Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0046] Example 1

[0047] Please see Figure 1 This invention provides a technical solution: a test method for simulating grouting reinforcement of fractured tunnel fault zones, comprising the following steps:

[0048] S1: Read the grouting pressure curve and the tunnel arch loading history, calculate the pressure slope for continuous time periods, and extract the pore water pressure and displacement gauge readings corresponding to the slope change points to generate a loading displacement sequence.

[0049] S2: Based on the loading displacement sequence and the original crack width value, the instantaneous closure amount is obtained by calculating the difference between displacement and width at each time, and the difference is squared and accumulated to eliminate fluctuation anomalies. Then, a physical information neural network is used to statistically accumulate the trend of the increase and decrease of the time series closure amount to establish a closed deformation curve.

[0050] S3: Extract the crack aperture value at each time point based on the closed deformation curve, construct the flow area by performing multiplication and division operations with the crack tube diameter and rock layer porosity ratio respectively, and then convert the flow velocity coefficient to the flow rate per unit time to generate a permeability coefficient sequence.

[0051] S4: Based on the permeability coefficient sequence and the closure deformation curve, the stress and displacement values ​​of the mesh nodes in the three-dimensional tunnel lining structure are read simultaneously. The non-dominated sorting genetic algorithm is used to judge the change range of the closure value and the decreasing trend of permeability in multiple parameter combinations, and to select the combination boundary that keeps both stable, and output the optimized parameter set.

[0052] S5: Based on the optimized parameter set, compare the peak grouting pressure and duration and record the frequency of parameter occurrence. Then, divide the grout viscosity into continuous intervals and extract the location of concentrated distribution segments to obtain key stability parameters.

[0053] The loading displacement sequence includes the pressure change time, corresponding displacement value, and timestamp data. The closed deformation curve specifically includes the closure amount time series, closure rate distribution, and abnormal fluctuation index. The permeability coefficient sequence includes the equivalent permeability, permeability fluctuation range, and time-period attenuation ratio. The optimization parameter set includes the grouting pressure peak range, grouting duration interval, and grout viscosity level. The key stability parameters specifically refer to the pressure control value, viscosity classification value, and grouting duration.

[0054] The specific steps for generating the loading displacement sequence are as follows:

[0055] Read the grouting pressure curve and the tunnel arch loading history, count the pressure difference every ten seconds and compare the difference between two points before and after, filter the absolute value index of the difference, and extract the corresponding pore water pressure and displacement gauge reading to generate a set of abrupt change point indexes;

[0056] Based on the mutation point index set, the pore water pressure and displacement gauge original readings at the corresponding time of the index point are retrieved and sorted in ascending order of time to generate a loading displacement sequence.

[0057] Based on the grouting pressure curve and the historical data of the tunnel arch loading, the sliding window difference method is used to process the grouting pressure curve. The sliding time interval is set to 10 seconds, and the continuous pressure readings within each 10 seconds are extracted to form a difference sequence. The difference calculation operation is performed on the pressure values ​​of adjacent two points. The absolute value amplitude threshold discrimination method is used to filter the pressure difference. The absolute value threshold of the pressure difference is set to 50 kPa. Data points that meet the threshold condition are marked as mutation index points. Based on the mutation index points, the original pore water pressure reading and displacement gauge reading at the corresponding time are extracted and combined to construct a mutation point index set.

[0058] Based on the mutation point index set, the pore water pressure readings and displacement gauge readings under the timestamps corresponding to the index points are retrieved. The mutation point times are sorted by ascending time sorting method, with the timestamp field as the primary key during the sorting process. After sorting, all displacement gauge readings are extracted in sequence and assembled in sequence to form a loading displacement sequence.

[0059] The specific steps for generating a closed deformation curve are as follows:

[0060] Based on the loading displacement sequence and the original crack width value, the displacement and width values ​​corresponding to each time point are extracted, the one-to-one correspondence difference is calculated and the original difference pair list is generated, and the interval consistency verification and abnormal time point investigation are performed to generate the instantaneous closure quantity list.

[0061] Based on the instantaneous closure value list, the difference of each item in the list is squared and a standard deviation threshold is established. After excluding data points above the threshold, the remaining closure values ​​are summed over time and a cumulative list is generated to produce the cumulative set of closure values.

[0062] Based on the closure accumulation set, a physical information neural network is used to identify the positive and negative change trends of continuous points on the time axis and divide adjacent band intervals. The total value of the closure increment within each segment is calculated and uniformly mapped to the time node to establish a closure deformation curve.

[0063] Based on the loading displacement sequence and crack width values, the difference mapping algorithm is used to calculate the corresponding difference between the two sequence items. The displacement column and width column are loaded synchronously according to the time index. Subtraction operation is performed on each pair of data in sequence. The interval consistency verification method is used to set the difference fluctuation threshold of 10 micrometers within 5 consecutive seconds for screening. The anomaly point investigation mechanism is used to exclude the time exceeding the value by three times the standard deviation, and a list of instantaneous closure quantities is generated.

[0064] Based on the instantaneous list of closed values, the square mapping operation is used to perform a power operation on each closed value, and the result of each power operation is recorded as the square value. A standard deviation threshold coefficient of 2 is set to calculate the dispersion of the set of square values, and items above the threshold are excluded. The remaining closed values ​​are accumulated sequentially according to the time series and intermediate results are saved to generate a cumulative set of closed values.

[0065] Based on the closure accumulation set, a physical information neural network algorithm is used to design the network structure with 2 input layer nodes, 1 output layer node, three hidden layers with 50 nodes per layer, hyperbolic tangent activation function, learning rate 0.001, batch size 32, and 10000 iterations. Backpropagation is performed to iteratively update the weights. The positive and negative changes of the network output value are used to identify the sign and divide adjacent band intervals. The closure increment value within each segment is accumulated and mapped back to the corresponding time node to establish a closure deformation curve.

[0066] The physical information neural network takes the time node sequence, the total value of the closure increment, and the rock mass stress residual term as network inputs. The network calculates the closure prediction value for each time node in turn, calculates the mean square error by subtracting it from the corresponding cumulative value of the closure quantity, and calculates the error between the stress residual term and the residual quantity of the physical conservation equation. The two types of errors are weighted and summed to form the total loss. Backpropagation is performed to update the network weights. The iteration is repeated until the total loss converges. The network outputs the closure deformation trend value for each time node and marks the closure change value of adjacent nodes as positive or negative.

[0067] The physical information neural network is defined according to the formula:

[0068]

[0069] in: t represents the total weighted closed-loop increment of adjacent trend intervals in the i-th segment. i Let t represent the starting time node of the i-th closed interval. i+1Let represent the end time of the i-th closed interval, C(t) represent the cumulative value of the closure at time t, C(t-1) represent the cumulative value of the closure at time t-1, and ω1 represent the weighting coefficient of the trend term. ω2 represents the sign function of the direction of change of the closed quantity at time t, μ(t) represents the weighting coefficient of the physical constraint term, μ(t) represents the local deformation modal response coefficient of the structural unit at the current time node, τ(t) represents the time interval between the current time point and the previous trend turning point, κ represents the time lag adjustment factor, and ρ(t) represents the output value of the material hysteresis memory function.

[0070] Execution process: First, the closed-loop quantity sequence C(t) at each time point in the fracture zone model test is collected, and the closed-loop increment between any time point and the previous time point is calculated. The derivative sign function sign is multiplied backward by the trend weighting coefficient ω1, and the local fitting error minimization method within the sliding window is used to enhance the dominant role of the trend. Then, the local modal characteristic function μ(t) is extracted from the current structural response region, and the hysteresis response function 1-e is introduced. -κ·τ(t) The difference in closure response caused by the grouting force transmission delay is corrected. Then, the material hysteresis memory function value ρ(t) obtained by neural network learning is used to characterize the historical deformation inheritance effect. This is then multiplied by the weight ω2 of the physical constraint term to achieve dynamic adjustment. Finally, the above terms are adjusted in the band interval t. i to t i+1 The inner summation yields the refined calculation of the segment closure increment value. As a basis for time-domain mapping, it is used to construct piecewise closed deformation curves to accurately characterize the structural convergence and evolution of fault fracture zones under grouting reinforcement.

[0071] The specific steps for generating the permeability coefficient sequence are as follows:

[0072] Based on the time point of the closed deformation curve and the original width of the crack, the difference between the closure amount and the width is extracted and the sum of squares is calculated. Values ​​exceeding the limit of the sum of squares are eliminated and the closure amount at each moment is accumulated to generate a list of opening values.

[0073] Based on the list of aperture values, the diameter of the fractured tube, and the porosity of the rock layer, multiplication and division operations are performed sequentially to calculate the flow cross-sectional area, which is then converted into the permeability velocity per unit time to generate a permeability coefficient sequence.

[0074] Based on the time points of the closed deformation curve and the crack width data, the difference square integral algorithm is used to process the difference between the closed value and the original crack width value. First, the corresponding crack width value is located at each time point in the closed curve and the difference is calculated. Then, the difference sequence is squared item by item. Each squared result is retained to three decimal places in floating-point form. The value range with a limited threshold of 0.002 square millimeters is used to perform the elimination operation. After filtering out the out-of-limit items, the closed values ​​at each time point are accumulated. The total closed value is generated by accumulating item by item using the sequential superposition method, and a list of opening values ​​is constructed.

[0075] Based on the list of aperture values, the diameter of the fractured tube, and the porosity of the rock strata, a composite flow area conversion method is adopted. For each aperture value, a multiplication calculation is performed sequentially with the diameter of the fractured tube. The tube diameter is converted to 0.014 meters using a fixed value of 14 mm. The product result is then divided by the porosity of the rock strata (0.34) to obtain the effective flow cross-sectional area at each moment. Subsequently, the flow cross-sectional area is divided by the unit time interval value of 60 seconds using the unit time flow rate conversion method. The flow rate value is calculated and retained to four decimal places to construct a permeability coefficient sequence.

[0076] The specific steps for generating the optimization parameter set are as follows:

[0077] Based on the permeability coefficient sequence and closed deformation curve, the three-dimensional tunnel mesh number is matched for all time nodes, the stress value and displacement vector of the node under the corresponding number are read, the corresponding components of each node are extracted and formatted and stored as structured data, and the node stress-displacement set is generated.

[0078] Based on the nodal stress-displacement set, the peak grouting pressure, duration, and grout viscosity parameters corresponding to the nodes are collected, assembled sequentially to form a ternary parameter combination list, and cross-indexing is performed. After removing duplicates and boundary value combinations, a parameter combination set is generated.

[0079] Based on the parameter combination set, a non-dominated sorting genetic algorithm is used to extract the closure deformation amplitude and the descent gradient of the permeability coefficient in the data corresponding to each combination, determine whether they are all within the target interval and mark the effective combination, and output the parameter column that meets the bistable condition as the optimized parameter set.

[0080] Based on the permeability coefficient sequence and the closed deformation curve, a grid mapping matching method is used to number and map all time nodes. The closure value and permeability coefficient value at each moment are indexed to the three-dimensional tunnel grid number. The stress value and displacement vector of the node corresponding to the number are read item by item. The principal stress components σ1, σ2, and σ3 are extracted from the stress tensor using a directional extraction method. The x, y, and z direction components are extracted from the displacement vector respectively. The formatting operation is performed and written sequentially into the structured data structure fields including time_nodegrid_idstress_1stress_2stress_3disp_xdisp_ydisp_z to generate the node stress and displacement set.

[0081] Based on the nodal stress-displacement set, a parameter assembly indexing method is adopted. According to each grid number, three types of parameters are matched: grouting pressure peak data, duration data, and grout viscosity parameter. The values ​​are extracted and triplet structures are constructed and organized into a list in sequence. Each structure is pressure_peakdurationviscosity. Then, a cross-index deduplication algorithm is executed to remove completely duplicate items and extreme items where any parameter is at the boundary value. The boundary conditions are defined as peak value less than 0.1MPa and greater than 5MPa, duration less than 10s and greater than 3600s, and viscosity less than 0.001Pa·s and greater than 2Pa·s. After retaining qualified structures, the output is a parameter combination set.

[0082] Based on the parameter combination set, a non-dominated sorting genetic algorithm is adopted. The initial population size is set to 100, the crossover probability is 0.9, the mutation probability is 0.1, the maximum number of iterations is 500, the elite retention strategy adopts the Pareto ranking method, the decoding method is floating-point encoding, the selection operator adopts the tournament selection mechanism, the crossover operator adopts the simulated binary crossover, and the mutation operator adopts the polynomial mutation method. Under each parameter combination, two indicators are extracted: the closure deformation amplitude and the permeability coefficient descent gradient. The closure amplitude range is defined as 0.05 to 0.3 mm, and the permeability gradient descent value is defined as 0.1 to 0.5 mm. The two indicators are judged according to the dual objectives to determine whether they are simultaneously within the range. The combination that meets the dual objective stability condition is output as the optimization parameter set.

[0083] The non-dominated sorting genetic algorithm initializes the parameter vector population, calculates the closure deformation amplitude and permeability coefficient descent gradient of each individual as the objective function value, performs non-dominated sorting to divide the population into levels, assigns level numbers to each individual, calculates the crowding distance between individuals within each level, conducts tournament selection based on level number and crowding distance to select parent individuals, performs crossover operation on the parent parameter vector to generate new individuals, performs mutation operation on the new individuals to obtain offspring, merges the parent and offspring, and re-performs non-dominated sorting and crowding distance calculation, selects a predetermined number of individuals from the merged set according to level priority and crowding order to form the next generation, repeats the above generational iteration until the preset number of generations is reached, and extracts the first level individuals to form the optimized parameter set;

[0084] The non-dominated sorting genetic algorithm is based on the formula:

[0085]

[0086] in: The distance to the improved crowding level of the i-th parameter combination during the optimization process is represented by m, which measures the sparsity of its distribution in the population. m represents the objective function number used to iterate through each evaluation metric in multi-objective optimization. M represents the total number of objective functions, and ω... m This represents the weighting coefficient corresponding to the m-th objective function, used to adjust the degree of influence of different objective functions on the ranking results. This represents the function value of the (i+1)th solution in the population on the m-th objective function. This represents the function value of the (i-1)th solution in the population on the m-th objective function. This represents the maximum value of the m-th objective function in the current population. Let α represent the minimum value of the m-th objective function in the current population, and let α represent the diversity adjustment coefficient, which is used to control the weight D of the population distribution breadth. i M represents the diversity measure of the i-th solution in the target space, reflecting its distributional difference from other solutions. β represents the local variation adjustment coefficient, used to regulate the influence of variation information on the ranking index. i G represents the local mutation rate of the i-th solution, used to measure its structural perturbation ability in the neighborhood; γ represents the elite solution adjustment coefficient, used to adjust the decay of the elite retention strategy; G represents the generation number of the current genetic algorithm iteration. max E represents the maximum number of iterations set for the genetic algorithm. i The variable represents whether the i-th individual belongs to the elite solution. It takes a value of 1 if it belongs to the elite solution and a value of 0 if it belongs to the non-elite solution.

[0087] Execution process: First, a high-dimensional parameter cluster containing variables such as grouting pressure, water-cement ratio, joint permeability, and surrounding rock modulus is constructed. Each set of parameters is input into the experimental numerical model to simulate the grouting process and outputs two corresponding key response indicators: closure deformation amplitude and permeability coefficient descent gradient. Then, the crowding distance index for each parameter combination is calculated. The first term measures the adjacent distribution of the current solution in the multi-objective function space and introduces the objective weighting coefficient ω. m Prioritize the key objective functions; the second term uses the diversity metric D. i The combination coefficient α prevents premature population convergence. The third term considers the variation rate M of local solutions. i Its influence is controlled by the adjustment coefficient β, and the fourth term is based on the current iteration number G relative to the maximum iteration number.

[0088] G max The elite decay adjustment term is used to enhance early exploration and late-stage elite solution retention, combined with the elite identification variable E. i By controlling the elite retention strategy, the final result obtained through the above calculations is... The value is used for non-dominated sorting and selection operations to select the optimal parameter combination and construct an optimized parameter set under the dual objective constraints of closure deformation and permeation gradient.

[0089] The specific steps for generating key stability parameters are as follows:

[0090] Based on the peak grouting pressure and duration in the optimized parameter set, the frequency of occurrence of each parameter combination is counted and sorted in descending order to generate high-frequency parameter pairs;

[0091] Based on the high-frequency parameter pair and the slurry viscosity range division results, the viscosity range is extracted according to the frequency of occurrence and the corresponding pressure and time are recorded to obtain key stability parameters;

[0092] Based on the time point of the closed deformation curve and the original width of the crack, the difference square integral method is used to calculate the difference between the corresponding terms of the two sequences, and to power and sum the difference of each term. The closed deformation value and crack width value at each time point are read in sequence, the difference is calculated by subtraction, the difference is multiplied by itself by squaring, the threshold of 0.0005 square meters is set by the range elimination algorithm, the square sum is eliminated, and the remaining closed values ​​are accumulated and superimposed in sequence to generate a list of opening values.

[0093] Based on the list of aperture values, the diameter of the fractured tube, and the porosity of the rock strata, a composite flow area conversion method is used to calculate the cross-sectional area and the flow velocity. Each aperture value is read sequentially, and a multiplication operation command is executed to multiply the aperture value by the fractured tube diameter of 0.014. A division operation command is executed to divide the product by the porosity of 0.34. A division operation command is executed to divide the corrected area by the unit time of 60. The result is rounded to four decimal places to generate a permeability coefficient sequence.

[0094] A test system for simulating grouting reinforcement of tunnel fault fracture zones is provided. This system is used to perform the aforementioned test method for simulating grouting reinforcement of tunnel fault fracture zones. The system includes:

[0095] Pressure slope calculation module: Based on the grouting pump pressure curve and the tunnel arch loading history, extract pressure readings for continuous time periods, calculate the pressure difference between adjacent moments divided by the time difference, filter the moments of slope change, read the corresponding pore water pressure and displacement gauge readings, and obtain a list of slope change points.

[0096] Closed deformation processing module: Based on the list of slope change points and the original crack width value, calculate the difference between the displacement gauge reading and the width at the time of change, square and accumulate the difference and remove abnormal fluctuation points, and combine the physical information neural network to perform time series increase and decrease trend accumulation processing on the accumulated sequence to obtain the closed deformation curve;

[0097] The permeability coefficient generation module is based on the closed deformation curve, the diameter of the fracture tube and the porosity of the rock layer. The flow area is obtained by multiplying the fracture aperture value by the porosity and dividing by the tube diameter at any given time. The flow velocity coefficient is then divided by the flow area to generate the permeability coefficient sequence.

[0098] Parameter optimization and screening module: Based on the closed deformation curve and the permeability coefficient sequence, it simultaneously reads the stress and displacement values ​​of the mesh nodes of the three-dimensional tunnel lining structure, and combines the non-dominated sorting genetic algorithm to judge the changes in the closure amplitude and the permeability reduction amplitude of multiple parameter combinations, and screens stable combinations to output the optimized parameter set;

[0099] Stability parameter extraction module: Based on the optimized parameter set, it compares the frequency of grouting peak pressure and duration, statistically analyzes the concentrated distribution of grout viscosity in continuous intervals, and obtains key stability parameters.

[0100] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A test method for simulating grouting reinforcement of fault fracture zones in tunnels, characterized in that, Includes the following steps: S1: Read the grouting pressure curve and the tunnel arch loading history, calculate the pressure slope for continuous time periods, and extract the pore water pressure and displacement gauge readings corresponding to the slope change points to generate a loading displacement sequence. S2: Based on the loading displacement sequence and the original crack width value, the instantaneous closure amount is obtained by calculating the difference between displacement and width at each moment, and the difference squared and accumulated to eliminate fluctuation anomalies. Then, a physical information neural network is used to statistically accumulate the trend of the increase and decrease of the time series closure amount to establish a closure deformation curve. S3: Extract the crack opening value at each moment based on the closed deformation curve, construct the flow area by performing multiplication and division operations with the crack tube diameter and rock layer porosity ratio respectively, and then convert the flow velocity coefficient and the flow area ratio into the permeability per unit time to generate a permeability coefficient sequence. S4: Based on the permeability coefficient sequence and the closure deformation curve, the stress and displacement values ​​of the mesh nodes in the three-dimensional tunnel lining structure are read simultaneously. The non-dominated sorting genetic algorithm is used to judge the change range of the closure value and the decreasing trend of permeability in multiple parameter combinations, and to filter the combination boundary that keeps both stable, and output the optimized parameter set. S5: Based on the optimized parameter set, compare the peak grouting pressure and duration and record the frequency of parameter occurrence. Then, divide the grout viscosity into continuous intervals and extract the location of concentrated distribution segments to obtain key stability parameters.

2. The experimental method for grouting reinforcement of simulated tunnel fault fracture zones according to claim 1, characterized in that, The loading displacement sequence includes the pressure change time, corresponding displacement value, and timestamp data. The closed deformation curve specifically includes the closure amount time series, closure rate distribution, and abnormal fluctuation index. The permeability coefficient sequence includes equivalent permeability, permeability fluctuation range, and time-period attenuation ratio. The optimization parameter set includes the grouting pressure peak range, grouting duration interval, and grout viscosity level. The key stability parameters specifically refer to the pressure control value, viscosity classification value, and grouting duration.

3. The experimental method for grouting reinforcement of simulated tunnel fault fracture zones according to claim 1, characterized in that, The specific steps for generating the loading displacement sequence are as follows: Read the grouting pressure curve and the tunnel arch loading history, count the pressure difference every ten seconds and compare the difference between two points before and after, filter the absolute value index of the difference, and extract the corresponding pore water pressure and displacement gauge reading to generate a set of abrupt change point indexes; Based on the set of mutation points, the pore water pressure and the original reading of the displacement gauge at the corresponding time of the index point are retrieved and arranged in ascending order of time to generate a loading displacement sequence.

4. The experimental method for grouting reinforcement of simulated tunnel fault fracture zones according to claim 1, characterized in that, The specific steps for generating the closed deformation curve are as follows: Based on the loading displacement sequence and the original crack width value, the displacement and width value corresponding to each time point are extracted, the one-to-one correspondence difference is calculated and the original difference pair list is generated, and the interval consistency verification and abnormal time point investigation are performed to generate the instantaneous closure quantity list. Based on the instantaneous closure value list, the difference of each item in the list is squared and a standard deviation threshold is established. After excluding data points higher than the threshold, the remaining closure values ​​are summed over time and a cumulative list is generated to generate a cumulative set of closure values. Based on the accumulated set of closed quantities, a physical information neural network is used to identify the positive and negative change trends of continuous points on the time axis and divide adjacent band intervals. The total value of the closed increment within each segment is calculated and uniformly mapped to the time node to establish a closed deformation curve.

5. The experimental method for grouting reinforcement of simulated tunnel fault fracture zones according to claim 1, characterized in that, The physical information neural network takes the time node sequence, the total value of the closure increment, and the rock mass stress residual term as network inputs. The network calculates the closure prediction value for each time node in sequence, calculates the mean square error by subtracting it from the corresponding cumulative value of the closure quantity, and calculates the error between the stress residual term and the residual quantity of the physical conservation equation. The two types of errors are weighted and summed to form the total loss. Backpropagation is performed to update the network weights. The iteration is repeated until the total loss converges. The closure deformation trend value of each time node is output, and the closure change values ​​of adjacent nodes are marked as positive or negative.

6. The experimental method for grouting reinforcement of simulated tunnel fault fracture zones according to claim 1, characterized in that, The specific steps for generating the permeability coefficient sequence are as follows: Based on the time point of the closed deformation curve and the original width of the crack, the difference between the closing amount and the width is extracted and the sum of squares is calculated. Values ​​exceeding the limit of the sum of squares are eliminated and the closing amount at each moment is accumulated to generate a list of opening values. Based on the list of opening values, the diameter of the fractured tube, and the porosity of the rock strata, multiplication and division operations are performed sequentially to calculate the flow cross-sectional area and convert it into the permeability velocity per unit time, generating a permeability coefficient sequence.

7. The experimental method for grouting reinforcement of simulated tunnel fault fracture zones according to claim 1, characterized in that, The specific steps for generating the optimized parameter set are as follows: Based on the permeability coefficient sequence and the closed deformation curve, the three-dimensional tunnel mesh number is matched for all time nodes, the stress value and displacement vector of the node under the corresponding number are read, the corresponding components of each node are extracted and formatted and stored as structured data, and the node stress displacement set is generated. Based on the node stress-displacement set, the peak grouting pressure, duration, and grout viscosity parameters corresponding to the node are collected, assembled sequentially to form a ternary parameter combination list, and cross-indexing is performed. After removing duplicates and boundary value combinations, a parameter combination set is generated. Based on the parameter combination set, a non-dominated sorting genetic algorithm is used to extract the closure deformation amplitude and the descent gradient of the permeability coefficient from the data corresponding to each combination, determine whether they are all within the target interval and mark the effective combination, and output the parameter list that meets the bistable condition as the optimized parameter set.

8. The experimental method for grouting reinforcement of simulated tunnel fault fracture zones according to claim 1, characterized in that, The non-dominated sorting genetic algorithm initializes a parameter vector population, calculates the closure deformation amplitude and permeability coefficient descent gradient of each individual as the objective function value, performs non-dominated sorting to divide the population into levels, assigns level numbers to each individual, calculates the crowding distance between individuals within each level, conducts tournament selection based on level number and crowding distance to select parent individuals, performs crossover operation on the parent parameter vector to generate new individuals, performs mutation operation on the new individuals to obtain offspring, merges the parent and offspring, and re-performs non-dominated sorting and crowding distance calculation, selects a predetermined number of individuals from the merged set according to level priority and crowding order to form the next generation, repeats the above generational iteration until a preset number of generations is reached, and extracts the first level individuals to form the optimized parameter set.

9. The experimental method for grouting reinforcement of simulated tunnel fault fracture zones according to claim 1, characterized in that, The specific steps for generating the aforementioned key stability parameters are as follows: Based on the grouting peak pressure and duration in the optimized parameter set, the frequency of occurrence of each parameter combination is counted and sorted in descending order to generate high-frequency parameter pairs; Based on the high-frequency parameters and the slurry viscosity range division results, the viscosity range is extracted according to the frequency of occurrence and the corresponding pressure and duration are recorded to obtain key stability parameters.

10. A test system for simulating grouting reinforcement of fault fracture zones in tunnels, characterized in that, The test method for grouting reinforcement of simulated tunnel fault fracture zones according to any one of claims 1-9, the system comprising: Pressure slope calculation module: Based on the grouting pump pressure curve and the tunnel arch loading history, extract pressure readings for continuous time periods, calculate the pressure difference between adjacent moments divided by the time difference, filter the moments of slope change, read the corresponding pore water pressure and displacement gauge readings, and obtain a list of slope change points. Closed deformation processing module: Based on the list of slope abrupt change points and the original crack width value, calculate the difference between the displacement gauge reading and the width at the abrupt change time, square and accumulate the difference and remove abnormal fluctuation points, and combine the physical information neural network to perform time series increase and decrease trend accumulation processing on the accumulated sequence to obtain the closed deformation curve; Permeability coefficient generation module: Based on the closed deformation curve, the diameter of the fracture tube and the porosity ratio of the rock layer, the flow area is obtained by multiplying the fracture opening value by the porosity ratio and dividing by the tube diameter at any given time, and the flow velocity coefficient is divided by the flow area to generate a permeability coefficient sequence. Parameter optimization and screening module: Based on the closed deformation curve and permeability coefficient sequence, the stress and displacement values ​​of the three-dimensional tunnel lining structure mesh nodes are read simultaneously. The non-dominated sorting genetic algorithm is used to determine the changes in the closure amplitude and the decrease in permeability of multiple parameter combinations, and a stable combination is screened to output the optimized parameter set. Stability parameter extraction module: Based on the optimized parameter set, compare the frequency of grouting peak pressure and duration, statistically analyze the concentrated distribution of grout viscosity in continuous intervals, and obtain key stability parameters.

Images