Underground engineering intelligent grouting method based on improved discrete fracture network model

Abstract

The invention discloses an underground engineering intelligent grouting method based on an improved discrete fracture network model. The underground engineering intelligent grouting method is characterized by comprising the following steps: 1) early-stage preparation and rock mass fracture data collection; 2) on the basis of a matrix joint model algorithm, constructing a discrete fracture network model through rock mass fracture data; 3) intelligent optimization of grouting parameters is carried out based on a multi-target particle swarm optimization algorithm; (4) grouting operation is carried out; and 5) effect evaluation and long-term monitoring. According to the underground engineering intelligent grouting method and system based on the improved discrete fracture network model, accurate control and long-term monitoring of the grouting process are achieved by combining the advanced modeling technology, the artificial intelligence algorithm and the Internet of Things technology. According to the method, the grouting effect is improved, resource consumption and environmental influence are remarkably reduced, and an innovative solution is provided for rock mass reinforcement and anti-seepage treatment of important projects such as water conservancy and hydropower, tunnels and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of underground engineering, and in particular to an intelligent grouting method for underground engineering based on an improved discrete fracture network model. Background Art

[0002] In underground engineering construction, rock cracks are one of the main factors affecting engineering safety. The presence of cracks not only reduces the overall strength of the rock mass but also provides channels for water infiltration, increasing the risk of leakage and instability.

[0003] Traditional grouting technology has shortcomings in dealing with these problems, such as low accuracy, low efficiency, and difficulty in monitoring. In response to the problems of traditional grouting technology, improvements have been made in existing technologies. For example, Wang et al. published a study on the application of discrete fracture network models in tunnel engineering in the Journal of Rock Mechanics and Geotechnical Engineering in 2019, and Li et al. published a study on the application of multi-objective optimization algorithms in underground engineering in Computers and Geotechnics in 2020. Although these existing technologies have made some innovations in some aspects, they still have the following shortcomings: (1) Lack of accurate simulation of complex fracture networks, especially ignoring the correlation between fractures and the influence of stress fields; (2) Insufficient accuracy of parameter optimization and dynamic adjustment; (3) Lack of consideration of the chemical reaction between slurry and rock mass during grouting and its influence on the evolution of fracture network; (4) Lack of long-term monitoring and intelligent early warning mechanism.

[0004] Therefore, there is an urgent need for a new technology that can provide accurate fracture network analysis, intelligent grouting parameter optimization, real-time dynamic adjustment, and long-term monitoring to improve grouting efficiency and engineering safety. Summary of the Invention

[0005] In order to overcome the shortcomings of the above-mentioned technologies, the purpose of the present invention is to provide an intelligent grouting method for underground engineering based on an improved discrete fracture network model, solve the problem of low simulation accuracy of complex fracture networks, and provide new technologies for accurate fracture network analysis, intelligent grouting parameter optimization, real-time dynamic adjustment and long-term monitoring to improve grouting efficiency and engineering safety.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] An intelligent grouting method for underground engineering based on an improved discrete fracture network model comprises the following steps:

[0008] 1) Preliminary preparation and rock fracture data collection;

[0009] 2) Based on the matrix joint model algorithm, a discrete fracture network model is constructed using rock mass fracture data;

[0010] 3) Intelligent optimization of grouting parameters based on multi-objective particle swarm optimization algorithm;

[0011] 4) Carry out grouting operations;

[0012] 5) Effect evaluation and long-term monitoring.

[0013] Preferably, the data in step 1) include the spatial distribution, direction, length, width and density of rock fractures.

[0014] Preferably, in step 2), the matrix joint model algorithm is an improved matrix joint model algorithm; the improved matrix joint model algorithm includes introducing a crack correlation coefficient and a stress field influencing factor into the matrix joint model algorithm to adjust the generated cracks.

[0015] Preferably, the calculation formula of the crack correlation coefficient CC is as follows:

[0016]

[0017] Among them, X i and Y i is the characteristic parameter of adjacent cracks, and is the average value of the corresponding parameter;

[0018] The calculation formula of the stress field influencing factor SFIF is as follows:

[0019] SFIF=(σ1-σ3) / (2×τmax);

[0020] Among them, σ1 and σ3 are the maximum and minimum principal stresses respectively, and τmax is the maximum shear stress.

[0021] Preferably, the adjustment includes adjusting various crack parameters until the crack correlation coefficient CC and the stress field influence factor SFIF reach preset values.

[0022] Preferably, in step 2), constructing the discrete fracture network model includes the following steps:

[0023] a. Data preprocessing: Using MATLAB software, we conduct statistical analysis on the collected fracture data based on the matrix joint model algorithm to obtain the probability distribution function of the fracture;

[0024] b. Model parameter setting: set the model boundary and grid size according to the project scale;

[0025] c. Crack generation: Based on the Monte Carlo method, a Python program is used to randomly generate cracks according to the probability distribution function, taking into account CC and SFIF;

[0026] d. Fracture connectivity analysis: Use graph theory algorithms in the NetworkX library to analyze the connectivity of the fracture network generated in step c.

[0027] e. Model verification: Verify the accuracy of the model by comparing it with the on-site drilling data.

[0028] Preferably, in step 3), the multi-objective particle swarm optimization algorithm is an improved multi-objective particle swarm optimization algorithm; the improved multi-objective particle swarm optimization algorithm includes introducing adaptive inertia weight AIW and chaotic perturbation mechanism into the multi-objective particle swarm optimization algorithm.

[0029] Preferably, the calculation formula of the adaptive inertia weight AIW is as follows:

[0030] AIW=(w max -w min )×(itermax-iter) / itermax+w min ;

[0031] Among them, w max and w min are the maximum and minimum inertia weights respectively, itermax is the maximum number of iterations, and iter is the current number of iterations;

[0032] The chaotic perturbation mechanism adopts Logistic mapping, and the formula is as follows:

[0033] x n+1 =μx n (1-x n );

[0034] Among them, x n is the state value of the multi-objective particle swarm optimization algorithm at the nth step; μ is the control parameter, which is usually set to 4 to obtain a completely chaotic state.

[0035] Preferably, the intelligent optimization objectives in step 3) include: maximizing the crack filling rate, minimizing the amount of grouting material used, and minimizing the construction time.

[0036] Preferably, the grouting parameters in step 3) include grouting pressure, grouting flow rate, and water-cement ratio.

[0037] Preferably, the intelligent optimization in step 3) includes the following steps:

[0038] a. Initialize the particle swarm, where each particle represents a set of grouting parameters;

[0039] b. Using the DFN model constructed in step 2) to simulate the grouting effect under each set of parameters;

[0040] c. Evaluate the fitness of each particle and use the weighted sum method to comprehensively score:

[0041] Score = w1 × filling rate + w2 × (1 / material amount) + w3 × (1 / construction time)

[0042] Where w1 is the weight coefficient for fill rate, w2 for material usage, and w3 for construction time, all ranging from [0, 1], with w1 + w2 + w3 = 1. Their relative values ​​determine the weight ratios of the various optimization objectives. Adjusting these three coefficients can balance the relative importance of grouting performance, material usage, and construction efficiency based on actual project requirements.

[0043] d. Update the particle's position and velocity, applying adaptive inertia weights (AIW) and chaotic perturbation mechanisms (CPM).

[0044] e. Repeat steps b to d until the preset number of iterations is reached or the convergence condition is met.

[0045] Preferably, step 4) includes using the parameters intelligently optimized in step 3) to perform grouting operations, and using an anomaly detection algorithm based on a long short-term memory network to analyze the monitoring data during the grouting process in real time, and using a Deep Q-Network reinforcement learning algorithm to dynamically adjust the grouting parameters.

[0046] Preferably, step 5) includes: scanning the grouting treatment area to evaluate the grouting effect; establishing a long-term monitoring system based on the Internet of Things, and using an early warning model based on the random forest algorithm for real-time evaluation and early warning; the long-term monitoring system includes distributed optical fiber sensors and intelligent microseismic monitoring equipment.

[0047] An underground engineering intelligent grouting system for implementing the above method comprises:

[0048] A data acquisition unit for collecting rock fracture data;

[0049] Data processing unit, used for discrete fracture network model construction and parameter optimization;

[0050] Intelligent control unit for real-time optimization and adjustment based on deep reinforcement learning;

[0051] Monitoring unit, used to collect grouting pressure and flow in real time;

[0052] A grouting execution unit, used for executing grouting operations;

[0053] Long-term monitoring unit, used for long-term monitoring of grouting effects;

[0054] Among them, the intelligent control unit uses edge computing devices to achieve real-time optimization and adjustment based on deep reinforcement learning.

[0055] Preferably, the data acquisition unit includes ground penetrating radar, 3D laser scanning, drone aerial photography, and ground three-dimensional seismic exploration; the data processing unit is a high-performance computing server; the intelligent control unit is an edge computing device; the monitoring unit includes a high-precision pressure sensor, a flow sensor, and a data acquisition system; the grouting execution unit includes an intelligent grouting pump and an automatic grouting system; and the long-term monitoring unit includes a distributed fiber optic sensing system and intelligent microseismic monitoring equipment.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The intelligent grouting method and system for underground projects, based on an improved discrete fracture network model, combines advanced modeling techniques, artificial intelligence algorithms, and the Internet of Things (IoT) to achieve precise control and long-term monitoring of the grouting process. This method not only improves grouting effectiveness but also significantly reduces resource consumption and environmental impact, providing an innovative solution for rock reinforcement and anti-seepage control in key projects such as water conservancy, hydropower, and tunnels. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a flow chart of the intelligent grouting method of the present invention;

[0059] Figure 2 Schematic diagram for improving the discrete fracture network model;

[0060] Figure 3 Schematic diagram of the grouting parameter optimization process;

[0061] Figure 4 To monitor and adjust the system architecture diagram in real time;

[0062] Figure 5 Schematic diagram of the long-term monitoring system deployment. DETAILED DESCRIPTION

[0063] In order to better explain the present invention, the main contents of the present invention are further illustrated below in conjunction with specific examples, but the contents of the present invention are not limited to the following examples.

[0064] like Figure 1 As shown, the present invention provides an intelligent grouting method for underground engineering based on an improved discrete fracture network model, comprising the following steps:

[0065] 1) Preliminary preparation and data collection

[0066] Using technologies such as ground penetrating radar (GPR), 3D laser scanning, drone aerial photography, and ground 3D seismic exploration, rock fracture data are collected, including parameters such as the spatial distribution, direction, length, width, and density of the fractures.

[0067] Specifically, a GSSI SIR-4000 GPR system equipped with a 400MHz antenna was used to perform a grid scan of the designated project area with a grid spacing of 0.5 meters. A Leica RTC360 laser scanner was used for 3D scanning, with a scanning accuracy of 2mm at 10m. Aerial photography was performed using a DJI Matrice 300RTK drone equipped with a P1 camera at an altitude of 200 meters, with an 80% front-to-back overlap and a 60% left-to-right overlap.

[0068] 2) Improved discrete fracture network model construction

[0069] A three-dimensional discrete fracture network (DFN) model was constructed using an improved matrix joint model algorithm. This improved algorithm introduced the fracture correlation coefficient (CC) and stress field impact factor (SFIF) to improve the realism of the fracture network.

[0070] The matrix joint model (MJM) is a common method for simulating fracture distribution, particularly suitable for studying fracture networks at various scales. During data preprocessing, this model allows for statistical analysis of fracture parameters such as orientation, spacing, and length, generating corresponding probability distribution functions to more accurately simulate the behavior of fracture networks.

[0071] The calculation formula of crack correlation coefficient (CC) is:

[0072]

[0073] Among them, X i and Y i are the characteristic parameters of adjacent cracks (such as direction, length, etc.), and is the average value of the corresponding parameter.

[0074] Stress field influence factor (SFIF) calculation formula:

[0075] SFIF=(σ1-σ3) / (2×τmax)

[0076] Among them, σ1 and σ3 are the maximum and minimum principal stresses respectively, and τmax is the maximum shear stress.

[0077] like Figure 2As shown in Figure 2, the discrete fracture network model construction process includes:

[0078] a. Data preprocessing: Using MATLAB software, we conduct statistical analysis on the collected fracture data based on the matrix joint model algorithm to obtain the probability distribution function of the fracture;

[0079] b. Model parameter setting: set the model boundary and grid size according to the project scale;

[0080] c. Crack generation: Based on the Monte Carlo method, a self-written Python program is used to randomly generate cracks according to the probability distribution function, taking into account CC and SFIF;

[0081] Among them, the calculation steps of the self-written program are:

[0082] 1. Initialize the fracture network parameters;

[0083] 2. Generate crack parameters based on probability distribution function;

[0084] 3. Consider the crack correlation coefficient (CC);

[0085] 4. Consider stress field influence factor (SFIF);

[0086] 5. Crack generation and simulation iteration;

[0087] 6. Result storage and visualization.

[0088] The core logic of the program is to generate fracture parameters based on a set probability distribution using the Monte Carlo method. Fracture generation is then adjusted based on the calculated results of CC and SFIF until the fracture correlation coefficient CC and stress field influence factor SFIF reach preset values. The resulting fracture network will more closely resemble the fracture distribution in the actual geological environment, achieving a more accurate simulation. The closer the preset values ​​of CC and SFIF are to 1, the more accurate the generated fracture network. In actual operation, the preset values ​​are determined based on field conditions.

[0089] d. Fracture connectivity analysis: Use graph theory algorithms in the NetworkX library to analyze the connectivity of fracture networks;

[0090] e. Model verification: Verify the accuracy of the model by comparing it with the on-site drilling data.

[0091] 3) Intelligent optimization of grouting parameters

[0092] An improved multi-objective particle swarm optimization algorithm (IMOPSO) was used to optimize the grouting parameters. The improvements included the introduction of adaptive inertia weight (AIW) and chaotic perturbation mechanism (CPM).

[0093] The optimization objectives include: maximizing the crack filling rate, minimizing the amount of grouting materials used, and minimizing the construction time.

[0094] Adaptive Inertia Weight (AIW) calculation formula:

[0095] AIW=(w max -w min )×(itermax-iter) / itermax+wmin

[0096] Among them, w max and w min are the maximum and minimum inertia weights respectively, itermax is the maximum number of iterations, and iter is the current number of iterations.

[0097] The Chaotic Perturbation Mechanism (CPM) uses Logistic mapping:

[0098] x n+1 =μx n (1-x n )

[0099] Among them, x n is the state value of the system at the nth step; μ is the control parameter, which is usually taken as 4 to obtain a completely chaotic state.

[0100] like Figure 3 As shown in Figure 2, the intelligent optimization process is as follows:

[0101] a. Initialize the particle swarm, where each particle represents a set of grouting parameters;

[0102] b. Using the DFN model constructed in step 2) to simulate the grouting effect under each set of parameters;

[0103] c. Evaluate the fitness of each particle and use the weighted sum method to comprehensively score:

[0104] Score = w1 × filling rate + w2 × (1 / material amount) + w3 × (1 / construction time)

[0105] Where w1 is the weight coefficient for fill rate, w2 for material usage, and w3 for construction time, all ranging from [0, 1], with w1 + w2 + w3 = 1. Their relative values ​​determine the weight ratios of the various optimization objectives. Adjusting these three coefficients can balance the relative importance of grouting performance, material usage, and construction efficiency based on actual project requirements.

[0106] d. Update the particle's position and velocity, applying adaptive inertia weights (AIW) and chaotic perturbation mechanisms (CPM).

[0107] e. Repeat steps b to d until the preset number of iterations is reached or the convergence condition is met.

[0108] 4) Implement grouting operations

[0109] During the grouting process, high-precision pressure and flow sensors are deployed to monitor key parameters such as grouting pressure and flow in real time. An anomaly detection algorithm based on a long short-term memory network (LSTM) is used to analyze the monitoring data in real time. When an anomaly is detected, a dynamic adjustment mechanism based on Deep Q-Network (DQN) is triggered, using the Deep Q-Network (DQN) reinforcement learning algorithm to dynamically adjust grouting parameters.

[0110] Specifically, during the grouting process, Siemens SITRANS P420 high-precision pressure sensors (accuracy ±0.1%) and Endress+Hauser Promass F300 flow sensors (accuracy ±0.5%) are deployed to monitor key parameters such as grouting pressure and flow in real time.

[0111] Adopt LSTM-based anomaly detection algorithm to analyze monitoring data in real time.

[0112] The LSTM network structure consists of an input layer, two hidden layers (128 neurons each), and an output layer. When an anomaly is detected, such as a sudden pressure increase exceeding 20% ​​of the set value or a flow deviation exceeding 30% from the expected value, a dynamic adjustment mechanism based on the Deep Q-Network (DQN) is triggered.

[0113] like Figure 4 As shown, the steps of the dynamic adjustment mechanism are as follows:

[0114] a. Use the DFN model to quickly simulate the current grouting status;

[0115] b. Based on simulation results and real-time data, the DQN algorithm is used to dynamically adjust grouting parameters:

[0116] State space: current pressure, flow, material usage, etc.;

[0117] Action space: pressure adjustment, flow adjustment, ratio adjustment;

[0118] Reward function: R = α × fill rate increment - β × material usage increment - γ × time increment;

[0119] Wherein, α, β, and γ are weight coefficients, ranging from [0, 1], and α+β+γ=1.

[0120] c. Feedback the adjusted parameters to the grouting equipment in real time to achieve closed-loop control.

[0121] 5) Effect evaluation and long-term monitoring

[0122] After grouting is complete, GPR technology is used to scan the grouting area again to evaluate the grouting effect. A long-term monitoring system based on the Internet of Things is established, including distributed fiber optic sensors and intelligent microseismic monitoring equipment. An early warning model based on the random forest algorithm is used for real-time assessment and early warning.

[0123] Specifically, if Figure 5 As shown in the figure, the long-term monitoring system based on the Internet of Things includes:

[0124] a. Deploy Brillouin optical time domain analysis (BOTDA) distributed fiber optic sensors to monitor rock strain and displacement;

[0125] b. Install ESG's Paladin digital intelligent microseismic monitoring system to capture tiny rock fracture signals in real time;

[0126] c. Develop an early warning model based on the random forest algorithm to conduct real-time assessment and early warning of rock mass stability. The early warning model based on the random forest algorithm is as follows:

[0127] Input features: fiber optic sensing data, microseismic signal characteristics, environmental parameters, etc.

[0128] Output: rock mass stability score (0-100);

[0129] When the score is lower than the preset threshold (such as 60 points), the early warning mechanism is triggered.

[0130] The present invention also provides an underground engineering intelligent grouting system for implementing the above method, which includes a data acquisition unit, a data processing unit, an intelligent control unit, a monitoring unit, a grouting execution unit and a long-term monitoring unit.

[0131] Among them, the data acquisition unit is used to collect rock fracture data, including ground penetrating radar equipment, 3D laser scanner and drone system; the data processing unit is a high-performance computing server, which is used for DFN model construction and parameter optimization; the intelligent control unit is an edge computing device, which realizes real-time optimization and adjustment based on deep reinforcement learning; the monitoring unit is used to collect grouting pressure and flow in real time, including high-precision pressure sensors, flow sensors and data acquisition systems; the grouting execution unit is used to perform grouting operations, including intelligent grouting pumps and automatic grouting systems; the long-term monitoring unit is used for long-term monitoring of grouting effects, including distributed fiber optic sensing systems and intelligent microseismic monitoring equipment.

[0132] Specifically, the above-mentioned intelligent grouting system for underground engineering includes:

[0133] A data acquisition unit for collecting rock fracture data, including a GSSI SIR-4000GPR device, a Leica RTC360 laser scanner, and a DJI Matrice 300RTK drone system;

[0134] Data processing unit for DFN model building and parameter optimization, including Dell PowerEdge R940xa high-performance computing server;

[0135] An intelligent control unit for real-time optimization and adjustment of grouting parameters, including an NVIDIA Jetson AGXXavier edge computing device;

[0136] A monitoring unit, used to collect real-time grouting pressure and flow, includes a Siemens SITRANS P420 high-precision pressure sensor, an Endress+Hauser Promass F300 flow sensor, and a National Instruments CompactDAQ data acquisition system;

[0137] Grouting execution unit, used to perform grouting operations, including the Chemgrout CG-550 / 030MP intelligent grouting pump and the Tecniwell TWM20 automatic grouting system;

[0138] The long-term monitoring unit, used for long-term monitoring of grouting effects, includes the Neubrex NBX-7020 distributed fiber optic sensing system and the ESG Paladin intelligent microseismic monitoring device.

[0139] Next, the present invention will be further described through Examples 1 and 2.

[0140] Example 1: Rock reinforcement project for underground powerhouse of a large hydropower station

[0141] Project Background:

[0142] The underground powerhouse of a large hydropower station is located in an area with complex geological conditions. The rock mass is well-developed with cracks and presents serious leakage problems, impacting project safety and power generation efficiency. Traditional grouting methods are ineffective, necessitating the use of new rock reinforcement technologies.

[0143] Implementation process:

[0144] 1) Preliminary preparation and data collection

[0145] A GSSI SIR-4000 GPR system was used to scan a 100m x 50m x 30m area around the factory building, with a grid spacing of 0.5m. A Leica RTC360 laser scanner was used to perform a 3D scan of the factory interior. A DJI Matrice 300RTK drone was used to capture aerial images of the ground surface and determine the distribution of surface cracks.

[0146] 2) Improved discrete fracture network model construction

[0147] Based on the collected data, the DFN model is constructed using the improved matrix joint model algorithm. The crack correlation coefficient (CC) and stress field influence factor (SFIF) are introduced:

[0148] The calculated average CC value is 0.73, indicating that there is a strong spatial correlation between cracks.

[0149] The SFIF calculation result is 0.62, indicating that the current stress field has a significant impact on the crack distribution.

[0150] Model validation shows that compared with actual drilling data, the fracture density error is less than 8%, the direction error is less than 5°, and the length error is less than 10%.

[0151] Regarding the evaluation of the crack correlation coefficient (CC), the strength of the spatial correlation between cracks can usually be measured based on its value. The CC value is between -1 and 1, and its meaning is as follows:

[0152] 1. CC = 1: It means that there is a complete positive correlation between the cracks, that is, the direction, position and other characteristics of the two cracks are highly consistent or similar.

[0153] 2. CC = 0: It means that there is no correlation between cracks, that is, the distribution of cracks is random and there is no correlation between them.

[0154] 3. CC = -1: Indicates that there is a complete negative correlation between the cracks, that is, the characteristics of one crack are completely opposite to those of another crack.

[0155] The specific evaluation rules are generally as follows:

[0156] -0.7≤CC≤1.0: Indicates a strong positive correlation between cracks, which means that the cracks have a relatively consistent distribution trend in space, and may have similar directions or regular distribution.

[0157] For CC = 0.73, it shows that the spatial distribution of cracks is not completely random, but has strong convergence and regularity.

[0158] The stress field influence factor (SFIF) is used to measure the effect of the stress field on the crack distribution, and its value is usually between 0 and 1. The evaluation rule of SFIF is based on its value, which reflects the degree of control of the stress field on the crack network.

[0159] The specific rules are as follows:

[0160] 1. SFIF = 1: The influence of the stress field on crack distribution reaches its maximum value. This means that the stress field plays a decisive role in the formation and distribution of cracks. The direction, length, and location of the cracks are all determined by the principal stress field.

[0161] 2. 0.7≤SFIF<1: The stress field has a strong control over the distribution of fractures. The direction and distribution of fractures are significantly affected by the stress field, but other geological factors may still interfere.

[0162] 3. 0.5≤SFIF<0.7: The stress field has a moderate influence on the distribution of cracks. The formation of cracks is controlled by the stress field and may also be affected by other factors (such as geological structure, chemical action, historical stress, etc.).

[0163] 4. 0.3≤SFIF<0.5: The influence of stress field on crack distribution is weak, the crack distribution is more random, and there are fewer signs of being controlled by stress field.

[0164] 5. SFIF<0.3: The stress field has little or no effect on the distribution of cracks, and the formation and distribution of cracks are mainly determined by other geological or physical and chemical factors.

[0165] The SFIF value is 0.62, which is considered a moderate to strong influence. This indicates that the stress field has a significant impact on the formation and distribution of cracks, but may also be affected by other geological factors. Therefore, the distribution of cracks generally conforms to the direction of the principal stresses and has a certain regularity, but there is still some variation.

[0166] 3) Intelligent optimization of grouting parameters

[0167] The IMOPSO algorithm was used to optimize the grouting parameters, with weight coefficients w1 = 0.5, w2 = 0.3, and w3 = 0.2. After 1,000 iterations, the optimal grouting parameter combination was obtained: grouting pressure: 6.8 MPa; grouting flow: 75 L / min; and water-cement ratio: 0.8.

[0168] 4) Implement grouting operations

[0169] During the grouting process, the LSTM anomaly detection algorithm identified seven anomalies, triggering the DQN dynamic adjustment mechanism. The most significant adjustment occurred at 30% of the grouting progress. Due to a large crack, the pressure suddenly dropped by 40%. The system automatically increased the grouting pressure to 8.2 MPa, the flow rate to 90 L / min, and the water-cement ratio to 0.7, successfully resolving the anomaly.

[0170] 5) Effect evaluation and long-term monitoring

[0171] After grouting was completed, GPR scans revealed a 95% fracture fill rate, a 35% improvement over traditional methods. The deployed BOTDA fiber optic sensors detected a maximum strain of 0.02% over six months, well below the 0.1% warning level. The Paladin microseismic monitoring system detected no rock failure signals exceeding 50dB during this period.

[0172] Example 2: Treatment of a water-rich section in a tunnel project

[0173] Project Background:

[0174] A highway tunnel passes through a karst-developed area. The water-rich section is about 200 meters long and suffers from severe water gushing. Traditional curtain grouting is not effective, and new technology is needed for precise grouting.

[0175] Implementation process:

[0176] 1) Preliminary preparation and data collection

[0177] A GSSI SIR-4000 GPR system was used to perform a circumferential scan of a 50-meter radius around the tunnel, with sections taken every 5 meters along the tunnel axis. A Leica RTC360 laser scanner was used to perform 3D scans of the tunnel interior. Due to limited surface conditions, drone aerial photography was replaced with ground-based 3D seismic surveys.

[0178] 2) Improved discrete fracture network model construction

[0179] Based on the collected data, a DFN model is constructed. The calculation results are:

[0180] The average CC value is 0.68, indicating that there is a moderate degree of spatial correlation between fractures.

[0181] The SFIF calculation result is 0.75, indicating that the current stress field has a strong influence on the crack distribution.

[0182] Model validation shows that compared with actual drilling data, the fracture density error is less than 10%, the direction error is less than 7°, and the length error is less than 12%.

[0183] 3) Intelligent optimization of grouting parameters

[0184] The IMOPSO algorithm was used to optimize the grouting parameters. Considering the severity of the water inflow problem, the weight coefficients w1 = 0.6, w2 = 0.2, and w3 = 0.2 were set. After 1500 iterations, the optimal grouting parameter combination was obtained: grouting pressure: 5.5 MPa; grouting flow: 60 L / min; water-cement ratio: 0.6.

[0185] 4) Implement grouting operations

[0186] During the grouting process, the LSTM anomaly detection algorithm identified 12 anomalies, triggering the DQN dynamic adjustment mechanism. One key adjustment occurred when a high-pressure water inrush point occurred. The system automatically increased the grouting pressure to 7.0 MPa, reduced the flow rate to 45 L / min, and adjusted the water-cement ratio to 0.5, successfully sealing the inrush point.

[0187] 5) Effect evaluation and long-term monitoring

[0188] After grouting was completed, GPR scanning showed that the crack filling rate reached 92%, which was 28% higher than the traditional method. The water inflow in the tunnel increased from the original 350m 3 / h reduced to 15m 3 / h, a decrease of 95.7%. The deployed BOTDA fiber optic sensors detected a maximum displacement of 2mm over three months, well below the warning value of 10mm. The Paladin microseismic monitoring system detected three minor rock fracture signals during this period, all with intensities below 60dB. Assessments show no impact on tunnel safety.

[0189] Through practical engineering applications, this invention has achieved remarkable results compared with traditional grouting methods:

[0190] 1. Grouting accuracy improved by 30%: The improved DFN model accurately simulates the fracture network, making grouting design more accurate;

[0191] 2. Material usage reduced by 20%: Intelligent parameter optimization significantly reduces material waste;

[0192] 3. Construction period shortened by 15%: Real-time monitoring and dynamic adjustment improve construction efficiency;

[0193] 4. Safety factor increased by 25%: Long-term monitoring system significantly reduces engineering risks;

[0194] 5. Project life is extended by more than 20 years: precise grouting and continuous monitoring ensure the long-term stability of the project;

[0195] 6. Environmental impact is reduced by 40%: Precision grouting reduces the impact on the surrounding groundwater and ecological environment.

[0196] The intelligent grouting method and system for underground engineering projects, based on an improved discrete fracture network model, provided by this invention achieves precise control and long-term monitoring of the grouting process by combining advanced modeling techniques, artificial intelligence algorithms, and Internet of Things technologies. This method not only improves grouting effectiveness but also significantly reduces resource consumption and environmental impact, providing an innovative solution for rock reinforcement and anti-seepage management in important projects such as water conservancy, hydropower, and tunnels. Two specific implementation cases fully demonstrate the applicability and superiority of this invention in different geological conditions and project types, confirming its significant advantages in solving complex engineering problems.

Claims

1. An intelligent grouting method for underground engineering based on an improved discrete fracture network model, characterized in that: The following steps are involved: 1) Preliminary preparation and rock fracture data collection; 2) Based on the matrix joint model algorithm, a discrete fracture network model is constructed through rock mass fracture data; 3) Intelligent optimization of grouting parameters based on multi-objective particle swarm optimization algorithm; 4) Carry out grouting operations; 5) Effect evaluation and long-term monitoring.

2. The intelligent grouting method for underground engineering according to claim 1, characterized in that: The data in step 1) include the spatial distribution, direction, length, width and density of rock mass fractures.

3. The intelligent grouting method for underground engineering according to claim 1, characterized in that: In step 2), the matrix joint model algorithm is an improved matrix joint model algorithm; the improved matrix joint model algorithm includes introducing a crack correlation coefficient and a stress field influencing factor into the matrix joint model algorithm to adjust the generated cracks.

4. The intelligent grouting method for underground engineering according to claim 3, characterized in that: The calculation formula of the crack correlation coefficient CC is as follows: Among them, X i and Y i is the characteristic parameter of adjacent cracks, and is the average value of the corresponding parameter; The calculation formula of the stress field influencing factor SFIF is as follows: SFIF=(σ1-σ3) / (2×τmax); Among them, σ1 and σ3 are the maximum and minimum principal stresses respectively, and τmax is the maximum shear stress.

5. The intelligent grouting method for underground engineering according to claim 1, characterized in that: In step 3), the multi-objective particle swarm optimization algorithm is an improved multi-objective particle swarm optimization algorithm; the improved multi-objective particle swarm optimization algorithm includes introducing adaptive inertia weight AIW and chaotic perturbation mechanism into the multi-objective particle swarm optimization algorithm to perform intelligent optimization of grouting parameters.

6. The intelligent grouting method for underground engineering according to claim 5, characterized in that: The calculation formula of the adaptive inertia weight AIW is as follows: AIW=(in max -In min )×(itermax-iter) / itermax+in min ; Among them, w max and w min are the maximum and minimum inertia weights respectively, itermax is the maximum number of iterations, and iter is the current number of iterations; The chaotic perturbation mechanism adopts Logistic mapping, and the formula is as follows: x n+1 =μx n (1-x n ); Among them, x n is the state value of the multi-objective particle swarm optimization algorithm at the nth step; μ is the control parameter, which is usually taken as 4 to obtain a completely chaotic state.

7. The intelligent grouting method for underground engineering according to claim 1, characterized in that: Step 4) includes using the parameters intelligently optimized in step 3) to perform grouting operations, using an anomaly detection algorithm based on a long short-term memory network to analyze the monitoring data during the grouting process in real time, and using a Deep Q-Network reinforcement learning algorithm to dynamically adjust the grouting parameters.

8. The intelligent grouting method for underground engineering according to claim 1, characterized in that: Step 5) includes: scanning the grouting treatment area to evaluate the grouting effect; establishing a long-term monitoring system based on the Internet of Things, and using an early warning model based on a random forest algorithm for real-time evaluation and early warning; the long-term monitoring system includes distributed optical fiber sensors and intelligent microseismic monitoring equipment.

9. An intelligent grouting system for underground engineering for implementing the method according to any one of claims 1 to 8, characterized in that: include: A data acquisition unit, used for collecting rock mass fracture data; Data processing unit for discrete fracture network model construction and parameter optimization; Intelligent control unit for real-time optimization and adjustment based on deep reinforcement learning; Monitoring unit, used to collect grouting pressure and flow in real time; A grouting execution unit, used for executing grouting operations; Long-term monitoring unit, used for long-term monitoring of grouting effect; Among them, the intelligent control unit uses edge computing devices to achieve real-time optimization and adjustment based on deep reinforcement learning.

10. The intelligent grouting system for underground engineering according to claim 9, characterized in that: The data acquisition unit includes ground penetrating radar, 3D laser scanning, drone aerial photography, and ground three-dimensional seismic exploration; the data processing unit is a high-performance computing server; the intelligent control unit is an edge computing device; the monitoring unit includes a high-precision pressure sensor, a flow sensor, and a data acquisition system; the grouting execution unit includes an intelligent grouting pump and an automatic grouting system; the long-term monitoring unit includes a distributed fiber optic sensing system and an intelligent microseismic monitoring device.

Images