Full-time monitoring system and method for grouting transformation of area under thick unconsolidated formation

Abstract

The invention relates to a full-time monitoring system and method for area grouting transformation under a thick unconsolidated formation, and relates to the technical field of geological engineering and geotechnical engineering. The system constructs a'grid distribution and key area encryption 'monitoring network composed of high-precision three-component micro sensors; underground medium micro-motion signals are collected and uploaded in real time through optical fiber and 4G / 5G double-link redundancy transmission. Fast Fourier transform and wavelet transform are combined to extract dominant frequency, amplitude, energy and phase abrupt change characteristics, and a slurry diffusion boundary, a diffusion velocity and a grouting body mechanical index are obtained based on finite element inversion. And a multi-dimensional grouting effect judgment method based on a geological multi-parameter random seepage model and a machine learning evaluation model is constructed, and self-adaptive regulation and control of grouting pressure, a slurry mixing ratio and dense hole layout are driven through comprehensive scores. According to the method, high-precision, lossless and real-time dynamic monitoring and quantitative evaluation of the thick unconsolidated formation grouting process are achieved, and the reinforcement reliability and the scientificity of the whole construction process are improved.

Description

Technical Field

[0001] This invention relates to the fields of geological engineering and geotechnical engineering, and in particular to a real-time monitoring system and method for regional grouting modification under thick loose layers. Background Technology

[0002] In underground engineering projects such as coal mining and tunnel construction, loose layers, due to their loose structure, low strength, and high permeability, often cause safety accidents such as roof collapse and water inrush, and are usually reinforced using surface grouting technology. Grouting technology fills the pores and fissures in loose layers to form a continuous reinforced body, thereby improving the stability and seepage prevention performance of the formation. However, the grouting process is highly concealed, and the diffusion path, range, and morphology of the grout under complex geological conditions are difficult to observe directly.

[0003] Traditional monitoring methods have many limitations: core drilling is a local sampling method and cannot reflect the overall diffusion characteristics; ground-penetrating radar has low resolution and is easily affected by strata interference; other methods are often destructive or lack timeliness, making it difficult to achieve accurate, quantitative, and non-destructive dynamic monitoring of the entire grout diffusion process, which seriously restricts the accurate evaluation of grouting effect and the scientific nature of engineering decisions.

[0004] Micro-motion monitoring technology, as a passive source geophysical method, can non-invasively invert the wave velocity structure, elastic parameters, and dynamic change characteristics of underground media by collecting and analyzing weak surface vibration signals. The innovative application of this technology to regional grouting monitoring under thick loose layers effectively overcomes the limitations of traditional monitoring methods, providing a new technical approach to solving the problems of difficult monitoring of grout diffusion patterns and difficulty in evaluating grouting effects.

[0005] To address the aforementioned issues, there is an urgent need for a real-time monitoring system and method for grouting renovation in areas with thick loose layers, based on election principles. Summary of the Invention

[0006] The purpose of this invention is to solve the existing technical problems mentioned in the background section and to provide a real-time monitoring system and method for grouting renovation in areas with thick loose layers.

[0007] The present invention achieves the above objectives through the following solutions: A real-time monitoring system for grouting modification in areas under thick loose layers includes a high-precision sensor network, a multi-link redundant data transmission system, and an intelligent data analysis and processing center. The high-precision sensor network is used to continuously collect underground micro-motion signals throughout the entire space. The multi-link redundant data transmission system is used to transmit the micro-motion signals stably in real time. The intelligent data analysis and processing center is used to perform preprocessing, time-frequency feature extraction, geological parameter inversion, and diffusion effect evaluation on the signals, and outputs the grout diffusion range, velocity, direction, and mechanical performance indicators of the grout body. Specifically, the high-precision sensor network transmits the collected underground micro-motion signals to the intelligent data analysis and processing center through the multi-link redundant data transmission system. The intelligent data analysis and processing center stores the data and allows remote access via terminal devices to provide data support for on-site decision-making.

[0008] As a preferred technical solution of the present invention, the high-precision sensor network adopts a "grid-based deployment + key area densification" deployment method composed of three-component micro-motion sensors, thereby achieving both full-area coverage and high-resolution monitoring of key areas.

[0009] As a preferred technical solution of the present invention, the multi-link redundant data transmission system includes a combination of fiber optic wired transmission and 4G / 5G wireless private network transmission, and has local caching and off-site backup mechanisms to achieve uninterrupted data acquisition and dual-backup secure storage.

[0010] As a preferred technical solution of the present invention, the intelligent data analysis and processing center includes a signal preprocessing unit, a time-frequency feature extraction unit, a geological parameter inversion unit, and a diffusion result visualization unit. It preprocesses the collected data, including signal denoising, outlier removal, and time synchronization; it automatically extracts the dominant frequency, amplitude, and phase difference characteristic parameters of the vibration signal from the preprocessed data; and based on the extracted features, it generates a dynamic map of the slurry diffusion range, a diffusion velocity curve, and a direction vector map using a finite element slurry diffusion inversion model, supporting remote access via a web interface.

[0011] As a preferred technical solution of the present invention, the time-frequency feature extraction unit extracts the main frequency, amplitude, energy and phase change features based on the combined Fast Fourier Transform and Wavelet Transform, which is used to accurately identify the arrival event of the slurry diffusion front.

[0012] This invention also provides a method for real-time monitoring of grouting modification in areas under thick loose layers, comprising: S1 collects background signals before grouting to establish a baseline model; Vibration signals were continuously collected and normalized during the S2 grouting process; S3 extracts the dominant frequency and abrupt change features of the signal based on Fast Fourier Transform and Wavelet Transform; S4 solves for the slurry diffusion boundary and mechanical properties based on a geological parameter inversion model; S5 calculates grouting quality indicators based on the evaluation model and outputs the grouting effect.

[0013] As a preferred technical solution of the present invention, the slurry diffusion rate is calculated by the time difference of the sudden arrival of adjacent sensor signals, and the main diffusion direction is determined based on the phase difference of multiple points.

[0014] As a preferred technical solution of the present invention, the finite element inversion method is used in combination with a three-dimensional geological model to solve the elastic modulus and diffusion boundary of the grout body, and the compressive strength and density index of the grout body are inverted by combining experimental regression relationship.

[0015] As a preferred technical solution of the present invention, a machine learning model is constructed to predictively evaluate the grouting effect. The machine learning model takes the micro-motion signal feature vector as input and the diffusion area, diffusion rate and compressive strength as output.

[0016] As a preferred technical solution of the present invention, when the evaluation result deviates from the preset threshold, an early warning of diffusion anomaly is automatically triggered, and suggestions for adjusting parameters such as grouting pressure, grout mix ratio and densification hole arrangement are generated.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention deeply integrates high-precision micro-motion monitoring technology with grouting engineering construction, constructs a three-dimensional sensing and perception system of "grid layout + key area densification", and adopts a dual-link redundant transmission structure of optical fiber and 4G / 5G to achieve high spatiotemporal resolution and non-destructive real-time monitoring of the entire grouting process under thick loose layers. It effectively overcomes the spatial limitations and lag problems of traditional core sampling and radar detection, and significantly improves the accuracy of grout diffusion state identification and the completeness of diffusion morphology coverage.

[0018] This invention employs a combined algorithm of Fast Fourier Transform and Wavelet Analysis to extract key features of micro-motion signals. Combined with sensitive inversion of formation elastic parameters, it achieves quantitative characterization of grout diffusion boundaries, diffusion velocity, and mechanical properties. Furthermore, it uses a prediction model based on the Darcy flow equation and geological multi-parameter stochastic characteristics to simultaneously provide confidence intervals for diffusion trends. This solves the problem that existing technologies struggle to grasp the real-time evolution of grout diffusion range and intensity, enabling grouting projects to shift from experience-dependent to data-deterministic.

[0019] This invention constructs an intelligent evaluation system for grouting quality using a support vector machine model and a neural network model. Based on the comprehensive scoring trigger parameters, it adaptively adjusts the grouting pressure, grout mix ratio, and densification hole layout to achieve dynamic optimization, ensuring a stable and controllable construction process and forming a complete closed loop of "monitoring-evaluation-control-feedback". This mechanism not only improves reinforcement efficiency and material utilization, but also provides high reliability and promotional value for grouting applications in complex, thick, loose geological conditions. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a system block diagram of a system and method for monitoring the diffusion law of grout in grouting modification under thick loose layers throughout the entire time period, as per the present invention. Figure 2 This is a flowchart of a method for monitoring the diffusion law of grout in a thick loose layer under a continuous period of time. Detailed Implementation

[0022] The following will refer to the appendices in the embodiments of the present invention. Figure 1-2 The technical solutions in the embodiments of the present invention are clearly and completely described herein. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0023] To more clearly demonstrate the engineering implementation process of this invention in grouting construction of thick loose layers, a high-precision micro-motion monitoring system is first deployed in the grouting construction area. The sensor is a high-precision three-component micro-motion sensor with built-in spatial positioning function. This sensor has high sensitivity and can accurately capture the weak vibration changes of the underground medium caused by the diffusion of grout.

[0024] This invention discloses a real-time monitoring system for grouting modification in areas under thick loose layers (as shown in the attached document). Figure 1 The system comprises a high-precision sensor network, a multi-link redundant data transmission system, and an intelligent data analysis and processing center. The high-precision sensor network utilizes high-precision three-component micro-motion sensors to accurately capture the micro-deformation vibration signals of the underground medium caused by grout diffusion. The deployment method employs a "grid-based deployment + key area densification" approach, meaning sensors are evenly distributed at preset intervals throughout the grouting area, with increased density of sensor deployment in key areas such as within 5 meters of the grouting holes, forming a spatial network combining comprehensive monitoring and focused attention. During installation, the sensor horizontality error is strictly controlled to ≤0.1°, and good coupling between the sensors and the strata is ensured to guarantee the quality of the acquired data.

[0025] The multi-link redundant data transmission system uses advanced digitization technology to convert the collected data into digital signals and stores and transmits them back in real time. Sensors located near the control center are transmitted via high-shielded fiber optic cable (transmission rate 100Mbps, bit error rate <10%). The remote sensor uses a 4G / 5G private network or Wi-Fi wireless transmission (uplink rate ≥20Mbps, latency ≤50ms).

[0026] Furthermore, to avoid data loss, the acquisition terminal is equipped with a 128GB local cache module, and an off-site backup server is established in the data analysis center to achieve automatic retransmission in case of connection failure and dual data backup.

[0027] The data analysis and processing center adopts a combination architecture of multi-core processors, large-capacity memory and high-speed solid-state storage devices to meet the real-time processing needs of massive signals; it includes Fast Fourier Transform (FFT) algorithms for time-domain to frequency-domain signal conversion, wavelet analysis algorithms for mining subtle signals, and finite element inversion algorithms for slurry parameter inversion, which can realize the extraction of time-frequency features of micro-motion signals, inversion of key information and quantitative calculation.

[0028] Specifically, in a real-time monitoring system for grouting renovation in areas under thick loose layers, a high-precision sensor network is used, employing a high-precision three-component micro-motion sensor with built-in spatial positioning function. The three-component micro-motion sensor has high sensitivity and can accurately capture the vibration signals of micro-deformation of underground media caused by grout diffusion.

[0029] Based on the geological conditions, area, and engineering requirements of the grouting area, a layout strategy of "grid-based point distribution + densification in key areas" is adopted: Full coverage: The sensors are evenly distributed in a grid at preset intervals to ensure coverage of the entire grouting area; the preset distance is set to 10m, since the theoretical radius of grout diffusion is 30m, 10m is selected in order to capture the diffusion characteristics of grout; Key tracking: Add more encrypted sensors in key areas such as within 5 m around the grouting hole to form a three-dimensional perception network of "full-area monitoring + key focus".

[0030] During sensor installation, strict control of levelness (error ≤ 0.1°) and ground coupling is required to ensure the accuracy of signal acquisition and provide high-quality raw data for subsequent data analysis.

[0031] To further optimize the above technical solution, a multi-link redundant data transmission system is implemented: the data acquisition and transmission equipment adopts digital technology, enabling rapid conversion of vibration analog signals to digital signals. It employs a "wired + wireless" dual-mode transmission architecture, as detailed below: Short-range transmission: For sensors located near the control room, wired transmission is achieved using fiber optic cables or highly shielded cables to ensure stable data transmission and low loss (transmission rate 100 Mbps, bit error rate ≤10⁻). 9 ); Remote transmission: For sensors in remote or difficult-to-wire areas, 4G / 5G private networks or Wi-Fi are used for wireless transmission to achieve remote real-time data upload (uplink rate ≥20 Mbps, latency ≤50 ms).

[0032] It also integrates multiple data security mechanisms: the device has a built-in 128 GB local cache module, which can temporarily store data when the signal is interrupted and automatically retransmit it after the connection is restored; it also builds an off-site disaster recovery server to achieve real-time dual backup of monitoring data, completely solving the problems of data loss and delay in traditional transmission methods.

[0033] To further optimize the above technical solutions, the intelligent data analysis and processing center adopts a hardware architecture combining a multi-core processor, large-capacity memory, and high-speed storage devices to meet the needs of rapid processing and long-term storage of massive amounts of monitoring data.

[0034] The software system innovatively integrates a multi-algorithm fusion system, with the following specific functions: Algorithm Integration: This includes a Fast Fourier Transform (FFT) algorithm for time-domain to frequency-domain signal conversion, a wavelet analysis algorithm for subtle signal mining, and a finite element inversion algorithm for slurry parameter inversion. It enables the extraction of time-frequency features, inversion of key information, and quantitative calculation of micro-motion signals. The specific process is as follows: Step 1: Time-frequency feature extraction; Step 2: Wavelet analysis and subtle signal processing; Step 3: Key information inversion and quantitative calculation; Specifically, each step involves the following steps: ① Time-frequency feature extraction: Extracting time-domain micro-motion signals from the sensor s(t) Denoising was performed (outliers were removed using the 3σ criterion), with a duration of T=10s, a sampling frequency of fs=1000Hz, and a number of sample points N=fs×T=10000.

[0035] The FFT calculation formula is as follows: ; in The sampling period is f Frequency (Hz); Therefore, three types of features are extracted: clock speed f 0: Take The frequency corresponding to the maximum value; Amplitude spectrum peak value A0: Amplitude value corresponding to the dominant frequency ; Bandwidth energy E Calculate the energy integral within ±1Hz of the main frequency. .

[0036] ② Wavelet analysis of subtle signal processing flow: To deeply extract subtle abrupt changes, the preprocessed signal is decomposed into wavelet multi-scale decomposition. The db4 wavelet is used as the mother wavelet to decompose the signal into five levels, obtaining the approximation coefficients A5 and detail coefficients. Retaining mid-to-low frequency information related to medium deformation (typically And filter out high-frequency interference, and reconstruct the signal through coefficients. : ; Subtle Feature Extraction: Calculating the instantaneous frequency f(t) of the reconstructed signal (based on Hilbert transform) and phase difference. It captures signal abrupt changes caused by slurry diffusion (the abrupt change corresponds to the moment when the slurry front reaches the sensor).

[0037] ③ Key information inversion and quantitative calculation based on finite element inversion algorithm: First, a three-dimensional finite element model of the grouting area is established based on geological exploration data. The grouting area and the non-grouting area are divided into grids of different precision (the grid size of the grouting area is 5m×5m×2m, and the grid size of the non-grouting area is 10mx10mx5m). The initial elastic modulus of the underground medium is set to... GPa and Poisson's ratio are 0.25-0.32 respectively. Assuming the elastic modulus of the grout increases with the degree of diffusion, that is... and set It is directly related to the density of the grout filling. The theoretical dominant frequency is obtained through forward modeling. , with "actual measured main frequency" With theoretical main frequency The inversion is performed with the objective of "minimizing the difference" and iterated until this condition is met. ; The inversion then yields the diffuse projected area on the horizontal plane: ; And based on the time difference of sudden changes in signals from adjacent sensors Spacing between sensors Calculate the diffusion rate: ; Using empirical formulas for strength of laboratory specimens The compressive strength of the grout body was obtained by inversion (based on laboratory data fitting).

[0038] Visualization and remote access: It can display key information such as the diffusion range, speed and direction of slurry in real time in the form of dynamic graphs, trend curves and vector graphics, and supports engineers to access it remotely through terminal devices such as mobile phones and computers, providing intuitive and accurate data support for on-site decision-making.

[0039] To further optimize the above technical solution, the specific steps for high-precision sensor network data acquisition are as follows: Step 1: Preliminary preparation and equipment calibration; Before grouting, a precise sensor layout plan is planned based on geological survey data, the area of ​​the grouting zone, and engineering design requirements. The sensor installation locations are determined using professional measuring instruments and then securely fixed. A comprehensive debugging and calibration of the sensors and transmission equipment is performed, including signal acquisition accuracy calibration, GPS positioning function testing, and transmission link stability verification. Simultaneously, micro-motion signals from the underground medium are continuously collected to establish an initial baseline data model for these signals, providing a benchmark for subsequent signal comparison and analysis.

[0040] Step two: Real-time data collection during the construction process; After grouting begins, sensors continuously capture vibration response signals from the underground medium, which are then digitally processed by transmission equipment and uploaded in real time to the intelligent data analysis and processing center.

[0041] To further optimize the above technical solution, the intelligent data analysis and processing center has the following data processing flow: Data collected by a high-precision sensor network is first stored in a dedicated database to ensure data integrity. The time-domain signal is converted to the frequency domain using the FFT algorithm to identify characteristic frequencies related to slurry diffusion. The signal is decomposed into multiple scales using wavelet analysis to uncover subtle changes. Combined with the finite element inversion algorithm and a pre-set geological model, the diffusion range, speed, and direction of the slurry are quantitatively and accurately inferred.

[0042] The specific steps are as follows: 1) Data preprocessing (standardizing data format and precision): Time synchronization: Based on GPS timestamps, all sensor data are synchronized to the same timeline (synchronization accuracy ≤1ms). Standardization: Normalizing time-domain signals ,in The mean of the signal. The standard deviation is denoted as .

[0043] 2) Dimensional Calculation Implementation: ① Frequency domain feature calculation (FFT dimension); Calculate the main frequency of each sensor according to the above "FFT time-domain to frequency-domain conversion process". f 0. Amplitude A 0. Bandwidth energyE Forming a frequency domain feature matrix ( m (Number of sensors).

[0044] ② Calculation of subtle time-domain features (wavelet analysis dimension); The reconstructed signal of each sensor is obtained through the above "wavelet analysis process". Extracting the mutation time and instantaneous phase Forming a time-domain feature matrix .

[0045] ③ Geological parameter inversion calculation (finite element dimension); Input the characteristic matrices F and T into the finite element inversion model and iteratively solve for the diffusion boundary. diffusion rate The elastic modulus E1 of the grout body is calculated as follows: Diffusion boundary coordinates: obtained through inversion vertex coordinates ..., determine the diffusion range at different depths; Average diffusion rate: (k is the number of adjacent sensor pairs, Let be the spacing between the i-th pair of sensors. (Time difference of signal abrupt change); Compaction assessment: Filling degree calculated based on E1 (Emax is the elastic modulus after the slurry is fully filled). It was determined to be dense.

[0046] 3) Data fusion output: The calculation results from various dimensions are integrated to generate a comprehensive data report that includes diffusion range, speed, direction, intensity, and density, and is displayed in the form of dynamic graphs through a visualization system.

[0047] This invention also includes data visualization and early warning mechanisms: The analysis results are presented in real time in a dynamic and visual form. At the same time, an anomaly early warning system is established: when abnormal grout diffusion rate (exceeding the preset stable range) or diffusion range deviating from the design target is detected, the system immediately triggers an audible and visual early warning signal to remind engineers to take timely adjustment measures to ensure the controllability of the grouting process.

[0048] To further optimize the above technical solution, a dynamic evaluation model is constructed based on monitoring data to achieve quantitative evaluation of grouting effect and optimized control of the entire process.

[0049] The specific steps are as follows: Step 1: Construct a multi-dimensional evaluation index system, including: grout diffusion range index, diffusion rate index, diffusion direction index, grout strength index, and density index. Step two involves constructing a dynamic evaluation model. Based on the principles of fluid mechanics and geotechnical mechanics, and combined with Darcy's law and the law of conservation of mass, a partial differential equation model for grout diffusion is established. Random variables of geological parameters (such as porosity and permeability fluctuations) are introduced to describe the uncertainty of geological conditions and improve the model's adaptability to complex, thick, loose layers. Simultaneously, a machine learning model is constructed, using micro-motion signal characteristics (dominant frequency, amplitude, and phase difference) as input and grouting effect indicators (diffusion range and intensity) as output. Through training with historical engineering data, the nonlinear mapping relationship between micro-motion signals and grouting effects is explored to achieve accurate prediction of grouting effects. Step three involves full-cycle dynamic evaluation and control. A monitoring system continuously collects data, which is then rapidly processed using an evaluation model to monitor grouting progress and effects in real time, promptly identifying anomalies such as uneven grout diffusion and insufficient grout strength. For any anomalies detected, the system automatically generates specific grouting parameter adjustment suggestions (such as grouting pressure, flow rate, and grout mix ratio). Engineers confirm and execute these suggestions via remote terminals, achieving dynamic optimization of the grouting process. The entire grouting process is monitored and evaluated. After grouting is completed, a comprehensive assessment is conducted, resulting in a "Dynamic Evaluation Report on the Effect of Grouting Renovation in Areas Under Thick Loose Layers." The results are then fed back to the design and construction departments to provide technical reference for subsequent projects, ultimately achieving a complete closed loop of "monitoring-evaluation-control-feedback."

[0050] In step one, specifically, a multi-dimensional evaluation index system is constructed, including: ① Grout diffusion range index: the diffusion boundary at different times and depths is determined through inversion analysis, the planar diffusion area is calculated, and the coverage effect is evaluated by comparing it with the design requirements; ② Diffusion rate index: based on the diffusion range data at different time points, the average diffusion rate and its changing trend are calculated to determine the stability of the grouting process; ③ Diffusion direction index: the main diffusion direction of the grout is accurately identified by using the time difference and phase difference of multi-sensor signals, providing a basis for optimizing the grouting scheme; ④ Grout strength index: the compressive, tensile, and shear strength of the grout is comprehensively evaluated by combining laboratory sampling tests and non-destructive testing technology to ensure that the engineering bearing requirements are met; ⑤ Density index: data from borehole core observation, ground-penetrating radar detection, and micro-motion monitoring are integrated to comprehensively determine the degree of grout filling of pores and evaluate the impermeability and stability.

[0051] In step two, a dynamic evaluation model is constructed; ① Integration of multidisciplinary mathematical models: Based on the principles of fluid mechanics and geotechnical mechanics, combined with Darcy's law and the law of conservation of mass, a partial differential equation model for slurry diffusion is established; random variables of geological parameters (such as porosity and permeability fluctuations) are introduced to describe the uncertainty of geological conditions and improve the model's adaptability to complex and thick loose layers; Parameter sampling: 1000 random samples were taken from porosity and permeability (based on their distribution characteristics) to obtain 1000 parameter samples. .

[0052] Equation Solution: For each sample group, the partial differential equations were solved using the finite volume method, yielding 1000 sets of diffusion range simulation results. .

[0053] Statistical analysis: Calculate the mean diffusion range at different times. and standard deviation , A diffusion prediction model with confidence intervals (95% confidence level) was established to compare the deviation between the inversion results and the theoretical predictions in real time.

[0054] ②Integration of machine learning algorithms: Construct machine learning models such as support vector machine (SVM) and neural network, take the characteristics of micro-motion signals (main frequency, amplitude, phase difference) as input and the grouting effect indicators (diffusion range, intensity) as output, train through engineering historical data, explore the nonlinear mapping relationship between micro-motion signals and grouting effect, and realize accurate prediction of grouting effect; Dataset Construction: Historical grouting project data from this mining area was collected, totaling 1000 samples, of which 800 were used as the training set and 200 as the test set. The input feature was micro-motion signal characteristics. The output is the grouting effect index. .

[0055] To further optimize the above technical solution, in addition to the support vector machine model, and to improve the fitting ability of nonlinear relationships, this invention further constructs a three-layer feedforward neural network model to simultaneously predict the grouting diffusion area, compressive strength, and comprehensive performance indicators. The network structure is as follows: Input layer: 3 neurons (corresponding to the main frequency of the micro-motion signal). Amplitude and random perturbation eigenvectors (Comprehensive expression); Hidden layer: 16 neurons in the first hidden layer and 8 neurons in the second hidden layer, both using the ReLU activation function to enhance nonlinear expression; Output layer: 3 neurons corresponding to the diffusion range respectively. diffusion rate With compressive strength The network training objective uses the mean squared error loss function: Loss function: using mean squared error ( For the true value, (Predicted value); Optimizer: Adam optimizer selected, learning rate The number of iterations is 500, and the batch size is 32. Early stopping strategy: When the test set loss no longer decreases after 10 consecutive iterations, stop training and save the optimal model.

[0056] Model validation and dynamic optimization: The model is validated using actual engineering monitoring data. The model performance is continuously optimized through cross-validation, parameter adjustment, and algorithm improvement to ensure the accuracy and reliability of the evaluation results.

[0057] Cross-validation: Five-fold cross-validation is used, dividing the training set into five subsets. Each time, four subsets are used for training and one for validation, repeated five times. The average prediction error is calculated. ; in For predicted values, Based on measured values, this invention determines the evaluation and discrimination threshold as: diffusion range (MAE). MAE (Magnetic Strength Index) The optimization method uses the Adam optimizer, with a learning rate of... The training batch size is 32. The dataset consists of measured data from historical grouting projects, with 800 sets used for training and 200 sets used for testing. A 5-fold cross-validation strategy is used to evaluate the model's generalization ability. Training is stopped early whenever the test set loss does not decrease for 10 consecutive iterations to preserve the optimal model.

[0058] The aforementioned machine learning model and finite element inversion model are used together to continuously update the model structure and parameters by comparing numerical simulation and field monitoring data, making the evaluation of grouting effect more reliable and maintaining a small deviation.

[0059] Evaluation Results Output: To achieve intelligent control throughout the entire process, this invention uses a comprehensive grouting effect score to determine the grouting quality. This score is based on the diffusion range compliance rate. Diffusion rate stability score Compressive strength compliance rate The scores are obtained by weighting and then linearly mapping the range to 0-100 based on engineering experience. ; When Score When the score and all key indicators meet the threshold, the grouting quality is judged to be good; when the score is <80, the system provides optimization suggestions for grouting parameters based on the model gradient direction, and adjusts the grouting pressure and grout mix ratio, wherein the pressure adjustment amount and the grout mix ratio adjustment amount are respectively: Pressure adjustment amount Adjustment amount of slurry mix ratio ; in A risk function is constructed for the model. All adjustment suggestions are executed remotely with authorization and fed back to the control system, forming a closed-loop, regular correction mechanism.

[0060] Specifically, the core of the dynamic evaluation model consists of a slurry diffusion partial differential equation model (integration of multidisciplinary mathematics) and a machine learning prediction model (SVM / neural network), the details of which are as follows: 1) Partial differential equation model for slurry diffusion (based on Darcy's law and the law of conservation of mass) Assume the slurry is an incompressible Newtonian fluid, and its flow follows Darcy's law; The thick loose layer is a porous medium, and the porosity φ and permeability k are spatial random variables (considering the fluctuation of geological parameters). The slurry diffusion process is an isothermal, unsteady flow.

[0061] The mathematical expression is: in, Porosity of the loose layer (random variable, satisfying...) , The average porosity, (For spatial fluctuation term) Time (s); Permeability (m², a random variable, following a log-normal distribution) ); : Dynamic viscosity of slurry (Pa·s); : Grout seepage pressure (Pa); : Spatial coordinate vector (m).

[0062] Derivation process: Darcy's Law: the seepage velocity of slurry in porous media satisfy: Law of Conservation of Mass: For a control volume V, the change in slurry mass per unit time is equal to the net inflow mass, that is: ; in, The density of the slurry (kg / m³) 3 When incompressible (where is a constant, it can be canceled out); the surface integral is converted to a volume integral using Gauss's formula: Since the control volume V is arbitrary, the integrands are equal. Substituting this into Darcy's law yields the final partial differential equation.

[0063] 2) Machine learning prediction models (taking SVM as an example); The evaluation model uses an artificial intelligence model trained on historical grouting data, with the input being a vector composed of micro-motion signal features: ; The main frequency of the micro-motion signal (Hz) The amplitude is (μm). For porosity fluctuations, For penetration rate fluctuations; The outputs are the grout diffusion area and compressive strength: S is the diffusion area (m²) 2 ), The compressive strength (MPa) of the grouting body. Construct a Support Vector Machine (SVM) regression model with the following decision function: ,in: It is a Lagrange multiplier (satisfying the KKT conditions); Output for training samples; Here, the kernel function is a radial basis function: γ>0 is a kernel parameter; The optimal parameters are obtained by the following method: ; Constraints The penalty coefficient is used; the machine learning results are used in real time to predict the evolution of grouting effect, improving the evaluation accuracy and response speed of unknown areas.

[0064] In summary, this invention achieves dynamic identification, precise inversion, and visual control of the entire process of grout diffusion range, diffusion rate, diffusion direction, and compressive strength of grout in thick loose layer areas through the deep integration of high-precision all-time micro-motion monitoring, rapid time-frequency domain feature extraction, finite element geological parameter inversion, geological stochastic partial differential equation prediction, and machine learning adaptive evaluation and control. This solves the shortcomings of traditional methods, such as the inability to track the diffusion front in real time, inversion lag, and coarse determination of compressive strength, ensuring that the grouting reinforcement quality meets the design standards and providing reliable engineering and technical support for the treatment of complex geological conditions in thick loose layers.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the rights and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A real-time monitoring system for grouting renovation in areas under thick loose layers, comprising a high-precision sensor network, a multi-link redundant data transmission system, and an intelligent data analysis and processing center, characterized in that: The high-precision sensor network is used to continuously acquire underground micro-motion signals throughout the entire space. The multi-link redundant data transmission system is used to transmit the micro-motion signals stably in real time. The intelligent data analysis and processing center is used to perform preprocessing, time-frequency feature extraction, geological parameter inversion, and diffusion effect evaluation on the signals, and output the grout diffusion range, velocity, direction, and mechanical performance indicators of the grout body. The high-precision sensor network transmits the acquired underground micro-motion signals to the intelligent data analysis and processing center through the multi-link redundant data transmission system. The intelligent data analysis and processing center stores the data and supports remote access by users through terminal devices.

2. The real-time monitoring system for grouting renovation in areas under thick loose layers according to claim 1, characterized in that, The high-precision sensor network consists of three-component micro-motion sensors that are evenly distributed in a grid at preset intervals to cover the entire grouting area; and additional sensors are added within a preset range of the grouting holes.

3. The real-time monitoring system for grouting renovation in areas under thick loose layers according to claim 1, characterized in that, The multi-link redundant data transmission system includes a combination of fiber optic wired transmission and 4G / 5G wireless private network transmission. It adopts a heterogeneous redundant link with "fiber optic as the main link and 4G / 5G wireless private network as the auxiliary link", has built-in link detection and automatic switching functions, and has local caching and off-site backup mechanisms to achieve uninterrupted data collection and dual backup secure storage.

4. The real-time monitoring system for grouting renovation under thick loose layers according to claim 1, characterized in that, The intelligent data analysis and processing center includes a signal preprocessing unit, a time-frequency feature extraction unit, a geological parameter inversion unit, and a diffusion result visualization unit. It preprocesses the collected data, including signal denoising, outlier removal, and time synchronization. It then extracts time-frequency features from the preprocessed data, automatically extracting the dominant frequency, amplitude, and phase difference characteristic parameters of the vibration signal. Based on the extracted feature parameters and a finite element-based slurry diffusion inversion model, it generates a dynamic map of the slurry diffusion range, a diffusion velocity curve, and a direction vector map, supporting remote access via a web interface.

5. The real-time monitoring system for grouting renovation under thick loose layers according to claim 4, characterized in that, The time-frequency feature extraction unit extracts the main frequency, amplitude, energy, and phase change features based on the combined Fast Fourier Transform and Wavelet Transform, which is used to accurately identify the arrival event of the slurry diffusion front.

6. A method for real-time monitoring of grouting modification in areas under thick loose layers, based on the system described in any one of claims 1-5, characterized in that, include: S1 collects background signals before grouting to establish a baseline model; Vibration signals were continuously collected and normalized during the S2 grouting process; S3 extracts the dominant frequency and abrupt change features of the signal based on Fast Fourier Transform and Wavelet Transform; S4 solves for the slurry diffusion boundary and mechanical properties based on a geological parameter inversion model; S5 calculates grouting quality indicators based on the evaluation model and outputs the grouting effect.

7. The method according to claim 6, characterized in that, The slurry diffusion rate is calculated by the time difference of abrupt changes in signals from adjacent sensors, and the main diffusion direction is determined based on the phase difference at multiple points.

8. The method according to claim 6, characterized in that, The elastic modulus and diffusion boundary of the grout were solved by combining the finite element inversion method with a three-dimensional geological model, and the compressive strength and density index of the grout were inverted by combining experimental regression relationships.

9. The method according to claim 6, characterized in that, A machine learning model is constructed to predictively evaluate the grouting effect. The machine learning model takes the feature vector of the micro-motion signal as input and the diffusion area, diffusion rate and compressive strength as output.

10. The method according to claim 6, characterized in that, When the evaluation result deviates from the preset threshold, an abnormal diffusion warning is automatically triggered, and suggestions for adjusting parameters such as grouting pressure, grout mix ratio, and densification hole arrangement are generated.

Images