Visual analysis method and system for grouting of non-newtonian fluid in rough fissure of goaf

Abstract

The present application relates to a kind of mined-out area rough fissure non-Newtonian fluid grouting visual analysis method and system, the method includes: generating roughness controllable fissure surface digital model and preparing transparent fissure physical model;Non-Newtonian fluid and cement-based slurry are used respectively to carry out visual grouting test and collect image data;Image is identified and processed to extract slurry diffusion boundary profile and calculate diffusion area;Traditional filling rate and actual filling rate are used as grouting effect evaluation index, wherein actual filling rate is the modified filling rate considering the filling degree of fissure opening direction;Rough single fissure grouting numerical model is constructed and cross-validated with test results;Based on the quantitative corresponding relationship of each grouting parameter and grouting effect evaluation index is determined and sorted based on the calculation results of multiple working conditions extension.This application realizes the quantitative characterization and parameter optimization of rough fissure grouting diffusion process, solves the problem of single evaluation index and unknown parameter influence mechanism in prior art.

Description

Technical Field

[0001] This invention relates to the field of grouting reinforcement technology for fractured rock masses, and in particular to a visualization analysis method and system for non-Newtonian fluid grouting in rough fractures of goaf areas. It can be applied to technical fields such as grouting and filling treatment of goaf areas in mines, reinforcement of fractured rock masses in underground engineering, and optimization design of grouting parameters. Background Technology

[0002] Large-scale coal mining has created numerous goaf areas. By the end of 2015, the total volume of goaf areas nationwide had reached 1.28 billion cubic meters. The overlying strata of these goaf areas are disturbed by mining, and under complex stress, gradually form a "three-zone structure": a bending zone, a fracture zone, and a collapse zone. After the roof collapses, the fracture zone deepens, and the bending zone and overlying strata experience increased subsidence, drastically increasing the risk of further collapse. In recent years, grouting and filling methods have become a common approach for goaf remediation due to their ability to reuse solid waste.

[0003] However, goaf areas are widely distributed in my country, and the geological structures vary significantly across different regions, resulting in complex and variable geological conditions for grouting treatment. Furthermore, goaf grouting typically takes place deep underground, making it difficult to achieve real-time visual monitoring of the grouting process and comprehensively assess its effectiveness using current technologies. Existing relevant standards are primarily based on experience, and the determination of key grouting parameters in engineering practice often relies on empirical judgment, lacking systematic scientific basis and theoretical guidance.

[0004] In the research of grouting simulation tests, the existing technology has the following main shortcomings:

[0005] First, the grouting process cannot be visualized in real time. Underground grouting projects are concealed, and existing technologies make it difficult to achieve real-time observation and recording of the grout diffusion process. Although some studies have used transparent materials to create fracture models for visualization experiments, most are limited to smooth parallel plates or simple artificial rough surfaces, which cannot truly reflect the influence of the morphological characteristics of rough fractures on grout diffusion.

[0006] Second, there is a contradiction between the uncontrollable roughness of the fractures and the insufficient realism of the models. The fracture models used in existing laboratory tests present a dilemma: although real rock fractures obtained by the splitting method have real roughness characteristics, the roughness is uncontrollable, making it impossible to systematically study the influence of roughness on the grouting diffusion law; although artificially processed smooth fractures or simple rough surfaces can achieve a certain degree of control, they are fundamentally different from the roughness characteristics of real rock fractures, and the engineering representativeness of the test conclusions is questionable.

[0007] Third, the evaluation index is too simplistic. Current grouting effectiveness evaluations primarily rely on the "filling rate"—the ratio of grout diffusion area to the total fracture area. This index implicitly assumes that "wherever the grout reaches is effectively filled." However, when grout diffuses in rough fractures, due to the unevenness of the fracture wall surface and the non-uniformity of the fracture opening, although the grout front reaches a certain area, the fracture cavities within that area may not be completely filled by the grout. Traditional filling rates cannot reflect this "incomplete filling" phenomenon.

[0008] Fourth, the effectiveness of alternative fluids lacks systematic verification. Because cement grout can cause irreversible damage to precision grouting equipment, existing studies often use alternative fluids such as xanthan gum solution to conduct high-precision tests. However, the similarity between the diffusion behavior of alternative fluids and real cement grout lacks systematic demonstration, and the sources of error are unclear.

[0009] Fifth, the mechanisms by which key parameters affect the grouting effect are unclear. There is a lack of systematic research on the coupled influence mechanisms of multiple parameters such as grouting pressure, crack opening, grout mix ratio, and crack roughness on the grouting effect. The order of importance of these parameters is unclear, making it impossible to provide a quantitative basis for optimizing engineering parameters.

[0010] Therefore, in-depth research on the grouting diffusion process in the overlying rock fractures of goaf areas, and the establishment of a complete analytical method from fracture model preparation to quantitative evaluation of grouting effect and parameter optimization, has important engineering practical significance. Summary of the Invention

[0011] To address the shortcomings of the existing technologies, this invention provides a visualization analysis method and system for non-Newtonian fluid grouting in rough fractures of goaf areas, aiming to achieve a scientific analysis of the entire process from fracture model preparation to quantitative evaluation of grouting effect and parameter optimization.

[0012] The first aspect of this invention discloses a visualization analysis method for non-Newtonian fluid grouting in rough fractures of goaf areas, characterized by the following steps: S1, generating a digital model of the fracture surface with controllable roughness, and preparing a transparent fracture physical model based on the digital model of the fracture surface, wherein the transparent fracture physical model has roughness characteristics statistically equivalent to those of real rock fractures; S2, conducting visualization grouting tests using a first fluid and a second fluid, wherein the first fluid is an alternative non-Newtonian fluid and the second fluid is a cement-based grout, and collecting image data of the grout diffusion process in the transparent fracture physical model during the grouting process; S3, performing recognition processing on the image data, extracting the boundary contour of the grout diffusion area, and calculating the grout diffusion area. S4. Based on the geometric characteristics of the boundary contour, the traditional filling rate and the actual filling rate are used as evaluation indicators for grouting effect. The traditional filling rate is the ratio of grout diffusion area to the total area of ​​the crack, and the actual filling rate is the corrected filling rate that takes into account the filling degree in the crack opening direction. S5. A rough single-crack grouting numerical model is constructed, and the numerical simulation results are cross-validated with the grout diffusion area over time obtained in step S3. Based on the validated numerical model, multi-condition extended calculations are performed. S6. Based on the calculation results of the multi-condition extended calculations, the quantitative correspondence between each grouting parameter and each grouting effect evaluation indicator is determined, and the importance of multiple grouting parameters is ranked.

[0013] Specifically, in step S1, the geometric parameters of the fracture model are first determined based on the actual survey data of the fracture zone in the overlying rock strata of the goaf. Then, a digital model of a rough fracture surface with a specific Hearst exponent is generated using a successive random accumulation algorithm. This digital model is verified by both the root mean square value of the three-dimensional slope and the Hearst exponent, proving its statistical similarity to the real rock splitting fracture surface. A transparent fracture physical model is then prepared using a 3D printed mold and an epoxy resin molding process. This step resolves the contradiction between uncontrollable fracture roughness and unrealistic models in existing technologies. In step S2, visual grouting tests are conducted using xanthan gum solution (an alternative non-Newtonian fluid used with a high-precision plunger pump) and cement grout (an engineering verification fluid used with an air compressor-grouting tank system). The engineering applicability of the alternative fluid test results is verified through dual-fluid comparison. During the grouting process, a red tracer is added to the grout, and the diffusion process of the grout in the transparent fracture model is recorded throughout using a high-speed camera. In step S3, the acquired image sequence is segmented using the HSV spatial red threshold to extract the boundary contour of the grout diffusion area. The grout diffusion area and its time-varying relationship are calculated using the pixel ratio conversion principle, converting the visual image into quantitative data. In step S4, based on the "incomplete filling phenomenon" observed in the visualization experiment, the traditional filling rate and the actual filling rate are used as evaluation indicators for the grouting effect. The actual filling rate is defined as η. eff=λ·η, where η is the traditional filling rate and λ is the filling correction coefficient, determined based on the fractal dimension of the grout diffusion boundary, reflecting the completeness of grout filling the fracture cavity in the fracture aperture direction. In step S5, a coarse single-fracture grouting numerical model is constructed based on the level set method module of COMSOL software. The numerical simulation results are cross-validated with physical test results to ensure the reliability of the model before multi-condition extended calculations are performed. In step S6, based on the numerical simulation data, a triple method of Pearson correlation coefficient, random forest, and gradient boosting tree is used to perform parameter sensitivity analysis, determine the quantitative correspondence between each grouting parameter and evaluation index, and complete the importance ranking.

[0014] This achieved the following beneficial effects: 1) It enabled precise control of crack roughness and statistical proof of its engineering representativeness; 2) Through dual-fluid, dual-platform comparative experiments, the engineering applicability of the alternative fluid test results was systematically verified and the source of error was clarified; 3) Automatic and precise quantitative determination of slurry diffusion range was achieved through HSV spatial image recognition; 4) The proposed new index of actual filling rate breaks through the limitations of traditional two-dimensional evaluation and realizes a scientific evaluation of the three-dimensional filling effect of slurry; 5) Through cross-validation of numerical simulation and physical experiments, a reliable data foundation was provided for multi-condition parameter analysis; 6) Through the triple sensitivity analysis method, the quantitative ranking of parameter importance was achieved, providing a data-driven basis for engineering parameter optimization.

[0015] According to the method disclosed in the first aspect of the present invention, the step of generating a digital model of a cracked surface with controllable roughness in step S1 includes: generating the cracked surface using a successive random accumulation algorithm, using the Hearst exponent as a roughness control parameter, wherein the value of the Hearst exponent ranges from 0 to 1, and the closer the Hearst exponent is to 0, the rougher the cracked surface is, and the closer it is to 1, the smoother the cracked surface is.

[0016] The Hearst index enables continuous quantitative control of roughness, overcoming the limitation of uncontrollable roughness in traditional fracturing methods. The Hearst index reflects the fractal self-similarity and scale invariance of rough surfaces, representing the essential characteristics of real rock fracture roughness. Compared to single amplitude statistical parameters (such as root mean square height), the Hearst index better characterizes the distribution of roughness at different scales, providing a more direct effect on understanding the distribution of flow resistance in rough fractures and the formation of localized dominant seepage channels.

[0017] According to the method disclosed in the first aspect of the present invention, the Hearst exponent is obtained through R / S analysis, including: calculating the cumulative deviation, range, and standard deviation of the high-order sequence of the fracture surface to obtain the R / S ratio; and performing logarithmic domain linear regression on the R / S ratio at different scales, using the regression slope as the Hearst exponent. This ensures the objectivity and repeatability of roughness characterization, providing a unified quantitative standard for roughness comparison between different fracture surfaces.

[0018] According to the method disclosed in the first aspect of the present invention, in step S1, the three-dimensional root mean square value of slope and the Hurst exponent of the digital model of the fracture surface are calculated respectively, and compared with the three-dimensional root mean square value of slope and the Hurst exponent of the real rock split fracture surface obtained by the Brazilian splitting method. If both of the above two indices of the digital model of the fracture surface fall within the statistical distribution interval determined by multiple sets of real rock split fracture surface measurement values, then the two are judged to have statistical similarity.

[0019] While existing technologies exist for generating controllable rough surfaces (such as the Fourier transform method), none have verified the statistical similarity between the generated rough surfaces and real rock fissures. Controllability does not equate to effectiveness, because the grouting diffusion pattern of an arbitrarily designed rough surface may be completely different from that of a real rock fissure. This invention establishes a quantitative equivalence bridge between controllability and reality through dual verification using the three-dimensional root mean square slope value and the Hearst exponent, ensuring the engineering representativeness of the model test results.

[0020] According to the method disclosed in the first aspect of the present invention, the method for calculating the root mean square value of the three-dimensional slope is as follows: based on the elevation data of each point on the fracture surface, the slope value of each point in two orthogonal directions is calculated, and the root mean square value of all slope values ​​is obtained.

[0021] According to the method disclosed in the first aspect of the present invention, the real rock fracture surface is obtained by the Brazilian fracturing method, and the point cloud data of the fracture surface is acquired by three-dimensional scanning. The Brazilian fracturing method combined with three-dimensional scanning technology can obtain a high-precision digital model of the fracture surface, providing a reliable real rock data benchmark for comparative verification.

[0022] According to the method disclosed in the first aspect of the present invention, the step of preparing the transparent crack physical model in step S1 includes: preparing a crack surface mold based on the digital model of the crack surface using 3D printing; using epoxy resin as raw material to mold and cure in the mold to obtain a transparent crack surface; stacking and fixing the upper and lower transparent crack surfaces in a mechanical limiting manner; controlling the crack opening through gaskets to form a transparent crack physical model with a set opening. The specific implementation process can be as follows: first, exporting the rough crack surface digital model generated by the SRA algorithm as an STL format file; printing the rough crack surface mold using a 3D printer; then mixing epoxy resin and curing agent in proportion, stirring evenly, and pouring into the mold; curing under a specific temperature control program; obtaining a transparent crack surface after demolding; finally, setting limiting bolts at the four corners of the transparent crack surface and placing gaskets at each bolt to precisely control the crack opening.

[0023] This invention employs a two-step method of "3D printing mold + epoxy resin molding", which has the following advantages: 1) Epoxy resin has better optical transparency than photosensitive resin, resulting in better visualization; 2) The mechanical properties of epoxy resin can be controlled through heat treatment processes, making it closer to real rock; 3) 3D printing molds can be reused, reducing experimental costs; 4) The molding process can more accurately replicate the fine rough structure of the mold surface.

[0024] According to the method disclosed in the first aspect of the present invention, the geometric parameters of the digital model of the fracture surface in step S1 are determined based on the actual survey data of the fracture zone in the overlying strata of the goaf. The actual survey data includes the distribution range of fracture aperture and fracture burial depth. Determining the model parameters based on actual engineering survey data ensures that the experimental model corresponds to the geological conditions of the engineering site, enabling the experimental results to directly serve engineering practice.

[0025] According to the method disclosed in the first aspect of the present invention, in step S2, the first fluid is a xanthan gum solution with a mass fraction of 0.3% to 1.2%; the second fluid is ordinary silicate cement slurry with a water-cement ratio of 0.5 to 1.0.

[0026] According to the method disclosed in the first aspect of the present invention, the xanthan gum solution is degassed by a vacuum pump after preparation and then allowed to stand for 24 hours before use; the ordinary silicate cement slurry is stirred for no less than 15 minutes and immediately transferred to a sealed container after mixing to prevent moisture evaporation. Vacuum degassing removes air bubbles from the solution, ensuring the homogeneity and optical transparency of the slurry; standing for 24 hours allows the xanthan gum molecules to fully hydrate, ensuring the stability of rheological properties. Thorough stirring of the cement slurry ensures homogeneity, and storage in a sealed container prevents moisture evaporation from affecting the water-cement ratio.

[0027] According to the method disclosed in the first aspect of the present invention, the step S2 of conducting a visual grouting test using a first fluid includes: injecting the first fluid into the transparent fracture physical model using a high-precision plunger pump in a constant flow mode or a constant pressure mode, wherein the flow rate accuracy of the high-precision plunger pump is greater than or equal to 0.5% and the pressure accuracy is greater than or equal to 0.5%FS; the step of conducting a visual grouting test using a second fluid includes: injecting the second fluid into the transparent fracture physical model at a constant air pressure using an air compressor-grouting tank system.

[0028] The first fluid test uses a high-precision plunger pump, while the second fluid test uses an air compressor as the power source, connected to a sealed grouting tank. A precision pressure-reducing valve controls the output of constant air pressure, which forces cement slurry into the fracture model. The high-precision plunger pump provides accurate constant flow / constant pressure control conditions for the xanthan gum solution test, ensuring the accuracy of the mechanism study. The air compressor-grouting tank system can directly use cement slurry without damaging the equipment. The two complement each other, forming a dual-platform test system that combines high-precision mechanism research with engineering verification.

[0029] According to the method disclosed in the first aspect of the present invention, the step of acquiring image data of the slurry diffusion process in step S2 includes: adding a red tracer to the first fluid and the second fluid, and using a camera device to record the complete diffusion process of the slurry from the start of grouting to the point where the slurry front reaches the boundary of the transparent fracture physical model. The red tracer makes the slurry area form a significant color feature in the image, which facilitates subsequent image recognition. Recording the entire process until the slurry front reaches the boundary ensures the integrity of the diffusion process data.

[0030] According to the method disclosed in the first aspect of the present invention, the visualization grouting test in step S2 is conducted in a water-saturated environment. The transparent fracture physical model is fixed in a water tank, and the height difference between the liquid level in the water tank and the highest point of the transparent fracture physical model is controlled within 3 cm. The water-saturated environment simulates the groundwater level conditions of the fracture zone in the goaf, making the test environment closer to the actual engineering situation. Controlling the height difference within 3 cm ensures the consistency of water pressure conditions and improves the repeatability of the test.

[0031] According to the method disclosed in the first aspect of the present invention, the step S3 of recognizing and processing the image data includes: extracting single-frame images from the acquired image sequence at set time intervals; converting each extracted frame image from the RGB color space to the HSV color space; generating a binary mask based on a preset threshold range for red in the HSV space; extracting the red region in the image using the binary mask; and counting the number of pixels in the red region. The slurry diffusion area is calculated based on the principle of proportional conversion, using the ratio of the number of pixels to the total number of pixels in the image combined with the actual planar geometric area of ​​the transparent crack physical model. The HSV space provides a color description closer to human visual perception, and the accuracy and robustness of red threshold segmentation are superior to direct segmentation in the RGB space. Furthermore, the proportional conversion principle eliminates the need for complex camera calibration, making calculation simple and highly accurate. The entire process achieves automated quantitative extraction of the slurry diffusion area.

[0032] According to the method disclosed in the first aspect of the present invention, a frame is extracted every 2 seconds. Extracting frames every 2 seconds ensures sufficient temporal resolution to capture diffusion dynamics while effectively controlling the number of images to be processed, thus balancing analysis accuracy and computational efficiency.

[0033] According to the method disclosed in the first aspect of the present invention, the actual filling rate in step S4 is set as: η eff = λ·η, where η is the traditional filling rate and λ is the filling correction coefficient, which reflects the degree of complete filling of the crack cavity by the grout in the direction of crack opening; the value of the filling correction coefficient ranges from 0 to 1, and the filling correction coefficient is determined based on the fractal dimension of the grout diffusion boundary.

[0034] Traditional filling ratios only consider the diffusion coverage area of ​​the grout on the fracture plane. Actual filling ratios introduce a filling correction factor on top of this. The filling correction factor is determined based on the fractal dimension of the grout diffusion boundary profile: a larger fractal dimension indicates a more irregular grout diffusion pattern (more branches and outward protrusions), suggesting that the grout flows more along locally dominant seepage channels rather than uniformly, corresponding to a lower degree of filling integrity; a smaller fractal dimension indicates a more regular circular diffusion pattern, suggesting uniform grout propagation, corresponding to a higher degree of filling integrity.

[0035] Traditional filling rate methods implicitly assume that "the area reached by the grout is effectively filled." However, this invention, through visual experiments, observes for the first time that in rough cracks, due to wall undulations and uneven aperture, although the grout front reaches a certain area, the crack cavities within that area may not be completely filled by the grout. The actual filling rate, by introducing a filling correction coefficient based on fractal dimension, incorporates this phenomenon, which is masked by traditional indicators, into the evaluation system, achieving a leap from two-dimensional to three-dimensional evaluation of grouting effect.

[0036] According to the method disclosed in the first aspect of the present invention, the filling correction coefficient λ and the fractal dimension D f Negative correlation, D f A larger λ value indicates a more irregular grout diffusion pattern, and a smaller λ value corresponds to a smaller λ value. This quantitatively establishes a mapping relationship between diffusion pattern and filling capacity; that is, the more irregular the diffusion pattern (the larger the fractal dimension), the more significant the grout flow along the dominant channels, the worse the overall filling uniformity, and the smaller the λ value. This quantitative relationship provides a calculable basis for evaluating grouting effectiveness.

[0037] According to the method disclosed in the first aspect of the present invention, when the filling correction coefficient λ approaches 1, it indicates that the grout has completely filled the fracture cavity in the diffusion region, and when the filling correction coefficient λ approaches 0, it indicates that the grout only spreads along the bottom surface of the fracture in the diffusion region and fails to effectively fill the fracture cavity.

[0038] According to the method disclosed in the first aspect of the present invention, the step S4, which uses the traditional filling rate and the actual filling rate as evaluation indicators of grouting effect, includes: calculating the fractal dimension of the boundary profile, wherein the fractal dimension is used to quantify the degree of geometric irregularity of the grout diffusion morphology, the larger the fractal dimension, the more irregular the diffusion morphology, and the smaller the fractal dimension, the more the diffusion morphology tends to be regular and circular; determining the filling correction coefficient λ based on the fractal dimension, thereby obtaining the actual filling rate.

[0039] Fractal dimension is an objective quantification of the degree of irregularity in complex geometric shapes, independent of subjective judgment. By using fractal dimension, the qualitative observation of the degree of irregularity in diffusion patterns is transformed into a calculable quantitative indicator, which is then mapped to a filling correction coefficient, thus achieving a complete transformation from visual phenomena to quantitative evaluation.

[0040] According to the method disclosed in the first aspect of the present invention, the specific method for constructing the numerical model of rough single-fracture grouting in step S5 is as follows: a numerical calculation model of two-phase flow of non-Newtonian fluid in a rough single fracture is established based on the level set method. The geometric dimensions and boundary conditions of the numerical model are determined according to engineering specifications and actual engineering survey data. The fracture roughness in the numerical model is set according to the roughness characteristics of the fracture surface digital model in step one. The level set method can accurately capture the topological changes of the two-phase flow interface and is suitable for simulating the dynamic interface evolution of grout displacing water during grouting. The experimentally generated rough fracture surface is directly imported into the numerical model, ensuring the consistency of the numerical model and the physical model in terms of geometric characteristics, laying a geometric foundation for subsequent experimental-simulation cross-validation.

[0041] According to the method disclosed in the first aspect of the present invention, the step of cross-validating the numerical simulation results with the grout diffusion area obtained in step S3 over time in step S5 includes: comparing the grout diffusion area over time curves obtained from numerical simulation and visual grouting test under the same working conditions; if the degree of agreement between the two reaches a preset threshold, the numerical model is determined to have passed the validation.

[0042] According to the method disclosed in the first aspect of the present invention, the multiple grouting parameters in step S6 include at least grouting pressure, fissure aperture, grout water-cement ratio, and fissure roughness; the grouting effect evaluation indicators include traditional filling rate and actual filling rate. Grouting pressure, fissure aperture, grout water-cement ratio, and fissure roughness are four key parameters affecting the grouting effect, covering three dimensions: grouting construction conditions (pressure), geological conditions (aperture, roughness), and grout properties (water-cement ratio). Using both traditional filling rate and actual filling rate as evaluation indicators can comprehensively reflect the planar diffusion effect and spatial filling effect of the grouting.

[0043] According to the method disclosed in the first aspect of the present invention, step S6, which involves ranking the importance of multiple grouting parameters, includes: calculating the linear correlation coefficient between each grouting parameter and the grouting effect evaluation index using the Pearson correlation coefficient method, and establishing a correlation coefficient heatmap to initially rank the parameter importance; simultaneously calculating the feature importance value of each grouting parameter to the grouting effect evaluation index using both the random forest model and the gradient boosting tree model, and ranking the parameter importance based on the feature importance value; and cross-validating the results calculated by the Pearson correlation coefficient method, the random forest model, and the gradient boosting tree model to obtain the final parameter importance ranking. By systematically integrating these three methods, the reliability of the parameter importance ranking is far higher than that of a single method when the conclusions tend to be consistent; even if there are discrepancies, the nonlinear characteristics and interactive dependencies of the parameter influence can be revealed from the differences in methods. This quantitative ranking result provides a data-driven scientific basis for the optimization priority allocation of engineering grouting parameters.

[0044] According to the method disclosed in the first aspect of the present invention, the mechanical limiting method is configured such that limiting bolts are set at the four corners of the transparent crack physical model, and shims are arranged at each limiting bolt to precisely control the crack aperture. Crack models with different apertures are obtained by adjusting the number of shims.

[0045] According to the method disclosed in the first aspect of the present invention, the engineering scale parameters of the rough single-fracture grouting numerical model are determined based on the actual survey data of the fracture zone of the overlying strata in the goaf, and the actual survey data includes the distribution range of fracture aperture and fracture burial depth.

[0046] According to the method disclosed in the first aspect of the present invention, the method further includes: providing optimization suggestions for grouting engineering parameters based on the parameter importance ranking results obtained in step S6, wherein the optimization suggestions include at least a suitable range for grouting pressure and a suitable range for grout water-cement ratio.

[0047] The second aspect of this invention also discloses a system for implementing the visualization analysis method for non-Newtonian fluid grouting in rough fractures of goaf areas disclosed in the first aspect of this invention. The system includes: a transparent fracture physical model, which has roughness characteristics based on real rock fractures and a set aperture; a grouting system, including a first grouting unit and a second grouting unit, wherein the first grouting unit is used to inject a first fluid into the transparent fracture physical model in a constant flow or constant pressure mode, the first fluid being an alternative non-Newtonian fluid; the second grouting unit is used to inject a second fluid into the transparent fracture physical model at a constant air pressure, the second fluid being a cement-based grout; and an image acquisition system for acquiring image data of the grout diffusion process in the transparent fracture physical model during the grouting process, until the grout... The liquid front reaches the boundary of the physical model of the transparent fracture; the image processing system is used to identify and process the image data, extract the boundary contour of the grout diffusion area, and calculate the grout diffusion area and its change over time; the numerical simulation system is used to construct a numerical model of rough single-fracture grouting, cross-validate the numerical simulation results with the grout diffusion area change over time obtained by the image processing system, and perform multi-condition extended calculations based on the validated numerical model; the parameter analysis system is used to determine the quantitative correspondence between each grouting parameter and each grouting effect evaluation index based on the calculation results of the multi-condition extended calculation, rank the importance of multiple grouting parameters, and output the characteristic importance weight value or ranking result of each grouting parameter to the grouting effect evaluation index.

[0048] Compared with the prior art, the present invention has the following technical effects:

[0049] This study effectively addresses the diffusion patterns and grouting efficiency characterization of non-Newtonian fluids injected into rough fissures at both theoretical and technical levels. It clarifies the influence of grouting pressure, fissure aperture, grout water-cement ratio, and fissure roughness on the grout diffusion radius and filling rate. It also determines the relationship between the grout diffusion pattern and the aforementioned grouting conditions and fissure morphology. The grout diffusion test under complex conditions avoids the blindness and randomness of purely theoretical analysis. The combined analysis method of refined experimental analysis, theoretical modeling, and numerical simulation has outstanding advantages such as clear description of grout diffusion effects and high reliability in characterizing the filling rate.

[0050] The method of this invention is a reliable method that directly reflects the diffusion law and filling degree during the grouting process in goaf areas. It describes the diffusion and filling efficiency of grout in goaf areas, provides a scientific basis for the grouting reinforcement of deep rock masses and the grouting filling efficiency of goaf areas, and improves reinforcement efficiency and engineering safety performance.

[0051] The present invention discloses in detail the visualization analysis method and system for non-Newtonian fluid grouting in rough fractures of goaf areas, with reference to the embodiments shown in the accompanying drawings and the reference numerals. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the steps of the visualization analysis method for non-Newtonian fluid grouting in rough fractures of goaf areas in this invention.

[0053] Figure 2 The drilling television image is taken at a depth of 40.5-41.6 meters in borehole Z108.

[0054] Figures 3a-3c These are schematic diagrams showing the three roughness levels of the crack surfaces generated when H values ​​are 0.3, 0.5, and 0.7, respectively.

[0055] Figures 4a-4c A schematic diagram for modeling randomly generated rough cracks; where, Figure 4a For H=0.3, Figure 4b For H=0.5, Figure 4c H=0.7;

[0056] Figure 5 A schematic diagram of scanning and modeling rock fracture surfaces;

[0057] Figure 6 A visual testing platform for cement grouting;

[0058] Figure 7 Image showing the effect of grouting area identification;

[0059] Figure 8 Comparison of cement grout diffusion effects under different grouting pressures;

[0060] Figures 9a-9bComparison of cement slurry diffusion effects under different crack apertures and roughnesses; among them... Figure 9a For different crack apertures, Figure 9b For different crack roughness;

[0061] Figure 10 Comparison of diffusion morphology between xanthan gum solution and cement grout during grouting

[0062] Figure 11 This represents the radial diffusion path of the slurry.

[0063] Figure 12a The diffusion cloud map of grout under different grouting pressures; Figure 12b Curves showing grouting area versus time under different grouting pressures;

[0064] Figure 13 The diffusion contour maps of grout under different crack apertures;

[0065] Figure 14 Diagrams showing the diffusion effect of slurry under different water-cement ratios;

[0066] Figure 15 Crack grouting diffusion cloud maps under different roughness conditions;

[0067] Figure 16a This is a distribution chart of the sensitivity analysis data. Figure 16b Heatmap of Pearson correlation coefficient for fill rate;

[0068] Figure 17 Heatmap of Pearson correlation coefficient for actual fill rate;

[0069] Figure 18 This is a schematic diagram illustrating the feature importance of each key parameter of the evaluation index in the RF model.

[0070] Figure 19 This is a schematic diagram illustrating the feature importance of each key parameter of the evaluation index in the GBDT model. Detailed Implementation

[0071] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. 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.

[0072] It should be noted that all directional indicators (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of the components in a specific posture (as shown in the figure). If the specific posture changes, the directional indicator will also change accordingly.

[0073] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. If the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0074] according to Figure 1 As shown, the first aspect of this invention discloses a visualization analysis method for grouting of rough fractures in goaf areas using non-Newtonian fluids. The method comprises the following steps: S1, generating a digital model of the fracture surface with controllable roughness, and preparing a transparent fracture physical model based on the digital model, wherein the transparent fracture physical model has roughness characteristics statistically comparable to real rock fractures; S2, conducting visualization grouting tests using a first fluid and a second fluid, wherein the first fluid is an alternative non-Newtonian fluid and the second fluid is a cement-based grout, and collecting image data of the grout diffusion process within the transparent fracture physical model during the grouting process; S3, performing recognition processing on the image data, extracting the boundary contour of the grout diffusion area, and calculating... S4. Based on the geometric characteristics of the boundary contour, the traditional filling rate and the actual filling rate are used as evaluation indicators for grouting effect. The traditional filling rate is the ratio of grout diffusion area to the total area of ​​the crack, and the actual filling rate is the corrected filling rate that takes into account the filling degree in the crack opening direction. S5. A rough single-crack grouting numerical model is constructed, and the numerical simulation results are cross-validated with the grout diffusion area over time obtained in step S3. Based on the validated numerical model, multi-condition extended calculations are performed. S6. Based on the calculation results of the multi-condition extended calculations, the quantitative correspondence between each grouting parameter and each grouting effect evaluation indicator is determined, and the importance of multiple grouting parameters is ranked.

[0075] Specifically, in step S1, based on the exploration results of an underlying goaf area, the characteristics of the fracture model are determined. The SRA algorithm is used to randomly generate fracture surfaces with controllable roughness. Based on engineering data, epoxy resin-coated transparent rock rough fractures are fabricated using a 3D printing mold replication method to verify their authenticity. In step S2, a visual grouting test of rock mass rough fractures is designed and conducted. Xanthan gum solution is selected as the main test fluid, and cement grout is used as the verification test fluid. A visual grouting test platform is built to conduct grouting tests under different grout concentrations, fracture roughness, grouting pressures, and rates, revealing the diffusion law of the grout in goaf fracture grouting. In step S3, the results of the visual grouting test are analyzed to obtain the grout diffusion range and diffusion rate, clarifying the grout diffusion law in rock mass fracture grouting and revealing the comprehensive mechanism of the combined effect of grout concentration, fracture roughness, and grouting input conditions on the grouting effect. In step S4, the grout diffusion pattern is characterized using phase interface theory, revealing incomplete filling of fractures during the grout diffusion process. Based on this, the "actual filling rate" index is proposed to quantitatively reflect the grout's filling capacity. In step S5, a grouting model for a single fracture in rough rock mass is established. The model dimensions and boundary conditions are determined based on relevant specifications and engineering data, and the model is validated by comparing it with experimental results. In step S6, combined with actual engineering projects in goaf areas, the mechanism of action and importance of key parameters in fracture grouting are explored. The influence mechanism of each grouting parameter on grouting evaluation indicators is precisely elucidated. Sensitivity analysis is conducted to rank the importance of parameters, and engineering grouting recommendations are provided.

[0076] Furthermore, in step S1, based on the exploration results of the mining goaf engineering data, important conditions such as fracture morphology and grouting environment were determined. The SRA algorithm was used to randomly generate fracture surfaces with controllable roughness, and the root mean square value of the three-dimensional slope and the Hurst exponent were used as quantitative indicators to compare and verify the roughness similarity between the fracture surface and the fractured rock. In step S2, using epoxy resin as the main raw material and combined with 3D printing molds, transparent rock rough fracture specimens were prepared by molding. Two visual grouting test platforms were built by combining the transparent rock rough fractures with different grout output devices, corresponding test fluids, and camera devices. In step S3, the test conditions were set and grouting tests were carried out using fracture roughness, aperture, grouting pressure, rate, and grout concentration as control parameters. The digital image results were processed using an image recognition program, and the test results provide an analytical basis for subsequent research on the diffusion law of grout in rock fractures.

[0077] In step S2, the test scheme was carried out using two different fluids and supporting equipment. The first part of the test used xanthan gum solution as the test fluid. The specific test procedure is as follows: 1) Fix the transparent fracture model in a cuboid water tank, and control the height difference between the water level in the tank and the highest point of the fracture within 3 cm; 2) Prepare xanthan gum solution with xanthan gum and deionized water according to the preset concentration (0.3%, 0.5%, 1.2%), use a vacuum pump to remove air bubbles in the solution, let it stand for 24 hours, and then draw it into the pump chamber of the ISCO pump through the inlet pipe to ensure fluid stability; 3) Connect the grouting pipe to the fracture model, and control the constant flow or constant pressure grouting conditions through the servo control system to simulate different engineering conditions; 4) Use a high-speed camera to record the entire grout diffusion process until the grout front reaches the fracture boundary; 5) After each test, disassemble the fracture model, use high-pressure water to thoroughly clean the residual grout in the fracture, and reinstall it to ensure that the model can be reused.

[0078] The experiment included three different crack models with different roughness, three different grout concentrations, four different grouting rates, and three different grouting pressures.

[0079] Step S2, Part Two, uses cement slurry as the test fluid. This part of the experiment uses P·O42.5 ordinary Portland cement to prepare cement-based slurry, and conducts experimental research using an air compressor-grouting tank system. The specific experimental procedure is as follows: 1) Accurately weigh cement and water according to a water-cement ratio (W / C) of 0.5, and mix thoroughly for 15 minutes to ensure the uniformity of the slurry; 2) Immediately transfer the mixed slurry to a sealed grouting tank to avoid water evaporation affecting rheological properties; 3) Use an air compressor as the power source, connect the grouting tank to the grouting pipe of the transparent rough fracture model, and ensure good airtightness of the equipment. 4) Control the air compressor to output a constant air pressure by using a precision pressure reducing valve; 5) Fix the transparent crack model horizontally on the test bench, connect the pipeline, and apply pre-pressure to remove air from the system; 6) Use a high-speed camera to record the grout diffusion process throughout until the grout front touches the edge of the crack and stops; 7) After each test, disassemble the rough crack model and connecting pipeline, carefully clean the crack and the residual cement grout in the grouting pipe, and collect the waste liquid into the pH neutralization sedimentation tank.

[0080] The experiment included three different crack models with different roughness, three different crack openings, and three different grouting pressures, with a grouting pipe diameter of 5 mm.

[0081] In step S4, after capturing a video of the grouting diffusion, frames were extracted every two seconds, and a MATLAB program was used to identify the grout diffusion range. Since a red tracer was added to the grout, identifying the red areas in the image yielded the grout diffusion morphology and outline. The following steps were used for grouting area image recognition: 1) Read the image to be processed and load it into the MATLAB working environment: Use the `imread` function to read the image file as a matrix. Each element of this matrix corresponds to the RGB value of a pixel in the image. 2) While the RGB space is usually sufficient for color detection, sometimes converting to other color spaces (such as HSV) can make color separation more obvious. In the HSV (Hue, Saturation, Value) space, hue (H) usually reflects color more intuitively, while saturation (S) and value (V) can help further define the vividness and brightness of the color. 3) Set a threshold range for red. Based on the characteristics of the image and the lighting conditions, determine the range of red for subsequent filtering. Typically, red pixels in the RGB space have the following characteristics: relatively high R (red channel) values, and relatively low G (green channel) and B (blue channel) values. 4) Create a binary mask to mark pixels that meet the red condition. The white parts of the mask represent red areas, and the black parts represent other areas. Use logical operators to create the mask, comparing the red channel values ​​with a threshold. 5) Extract the red areas from the original image using the mask. Create a new image matrix and apply the mask to the original image. Pixels in the original image are only preserved when the mask is true (i.e., a red area). Iterate through the three RGB channels, preserving the values ​​of the red areas and setting other areas to zero.

[0082] The red area recognition program can accurately identify the grout diffusion range and calculate the number of pixels in the red area. The ratio of the number of pixels in the red area to the total number of pixels in the image is equal to the ratio of the area of ​​the red area to the geometric area of ​​the model plane. The area of ​​the red area, i.e., the grout diffusion range, can be calculated from this ratio.

[0083] Based on the above methodology, to verify the reliability of the alternative fluid test results and elucidate the influence of different parameters on the grout diffusion law, this invention, by linking the diffusion form with the grout filling capacity, proposes a new grout filling capacity evaluation index to explore the underlying grout diffusion law. First, supplementary tests using cement grout as the test fluid further enhance the practical guidance of xanthan gum solution test results for engineering applications. High-precision grouting tests using xanthan gum solution as the test fluid clarify that the grout diffusion law is comprehensively influenced by factors such as fracture morphology, grout rheological properties, and grouting input conditions. The phase interface theory of multiphase flow is introduced to classify and characterize different grout diffusion forms, and fractal dimension is introduced for quantitative analysis of grout diffusion forms, revealing that different diffusion forms represent different grout filling capacities. Furthermore, a new index, "actual filling rate," is proposed to quantify the grout filling capacity, providing a scientific basis for the synergistic optimization of grout hole spacing, grouting pressure, and grout mix ratio.

[0084] In steps S5 and S6, numerical simulation combined with sensitivity analysis is used, with an engineering application focus, to calculate and analyze the impact of key parameters on grouting effectiveness in grouting projects. First, numerical calculations are performed using COMSOL software to clarify the theoretical basis of its level set method module for simulating grout diffusion and flow in fractures. An engineering-scale numerical model for grouting single fractures in rough rock mass is constructed, and its agreement with experimental results is verified. Based on the theoretical foundation and experimental results, and using evaluation indicators such as filling rate and the actual filling rate proposed in this paper, the relationship between four parameters—grouting pressure, fracture aperture, grout water-cement ratio, and fracture roughness—and grouting efficiency and the ability of grout to fill fractures is analyzed. Simultaneously, statistical methods and machine learning methods are introduced to further quantify the correlation between key parameters and evaluation indicators.

[0085] This invention analyzes the morphological characteristics and grouting migration process of rough fractures in goaf areas. It uses drilling to obtain fracture structural features, providing a basis for characterizing the morphological features of goaf fractures. A visual grouting experiment is then conducted on the rough fractures, emphasizing the influence of fracture structural features and grouting parameters on the grout diffusion radius and filling rate. Based on the obtained grouting area versus time curves under different grouting pressures, grout proportion contour distributions under different grouting pressures, and filling rate and actual filling rate versus time curves under different grouting pressures, expressions for the filling rate and actual filling rate are established. The diffusion law and filling efficiency of grout injected into rough fractures by non-Newtonian fluids are explored, and optimization suggestions for goaf grouting parameters are proposed.

[0086] refer to Figure 2 The image shows the joints and fractures in the borehole television image of hole Z108 in the exploration results. The image was read using CAD software and the fracture aperture was measured according to the scale. The aperture distribution was found to be approximately 1-3 cm. Figure 2 The image shows a borehole television image at a depth of 40.5-41.6 meters in borehole Z108. The light blue lines within the red box mark the rock mass fissures. As can be seen from the image, the rock mass fissures are distributed in a network-like and relatively dense manner, and the fissures are well connected.

[0087] The fracture morphology from only a few boreholes is insufficient to accurately model the overall fracture structure. Therefore, a simplified model of single-fracture grouting was proposed, with additional fractures generated. Using the same analytical method as in the previous example, analysis of all borehole television images revealed the presence of a fracture zone 38-42m below the surface in the goaf. The fractures were clearly developed, with fracture apertures ranging from 1-3cm.

[0088] Most studies on rock mass fissure grouting have used fissure aperture values ​​ranging from 1 to 10 mm. Therefore, this invention is applicable to grouting of large-aperture fissures, such as those in goaf areas. Furthermore, the groundwater level is approximately 38 m below the surface, and the rock composition here consists of alternating mudstone and sandstone. The core samples obtained from the boreholes are relatively soft and require grouting reinforcement to prevent sudden disasters such as ground subsidence.

[0089] Therefore, based on the engineering survey results and taking into account the limitations of the indoor test model, the planar geometric dimensions of the grouting simulation test model were set as a square of 300mm*300mm, the model crack opening was 1.2mm, and the experimental device was kept in a water-saturated environment.

[0090] In practical engineering, the surface roughness of fracture walls in goaf areas is difficult to observe. Furthermore, the roughness of rock fracture surfaces obtained by the Brazilian fracturing method is uncontrollable. To highlight the influence of fracture surface undulation on grouting effectiveness, this invention does not consider the anisotropy and size effect of the fracture surface and uses randomly generated, controllable roughness fracture surfaces for testing. A modified Successive Random Addition (SRA) algorithm is used to generate rough fracture surfaces. Different roughness levels are generated by controlling the Hurst exponent (H, hereinafter referred to as H value). The H value ranges from 0 to 1; values ​​closer to 1 indicate a smoother fracture surface, while values ​​closer to 0 indicate a rougher surface. Based on these settings, MATLAB software is used to generate controllable roughness fracture surfaces, as shown in Figure 3. H values ​​of 0.3, 0.5, and 0.7 are used, resulting in three different roughness levels of fracture surfaces.

[0091] Referring to Figure 4, as the H value decreases, the peak height difference of the generated fracture surface increases from 0.6 to 1.1, and its undulation distribution tends to become more discrete and inconsistent, reflecting a significant increase in the roughness of the entire fracture surface. The randomly generated fracture surface is copied and "lifted" by 1.2 mm to obtain the upper surface of the fracture, which is then sandwiched with the lower surface to obtain a rough fracture model with equal opening. Due to the limited precision of 3D printing, a combination of linear interpolation, copying, and "lifting" is performed on the point cloud data of the randomly generated fracture surface to obtain a smoother rough fracture model with a fixed opening, as shown in Figure 4, for subsequent 3D printing. However, this operation is similar to reducing the sampling interval, resulting in some loss of roughness. Figure 4 shows that after interpolation smoothing, the roughness undulation of the fracture at H=0.3 is still significantly greater than that at H=0.5 and H=0.7, while the roughness of the fracture surface gradually decreases as the H value increases. When H=0.7, the roughness undulation of the fracture surface is almost similar to that of a smooth parallel plate. Therefore, three different roughness levels of fracture surfaces can simulate rock fracture walls with different roughness levels, providing more references for engineering projects under different geological conditions.

[0092] To further demonstrate the reliability of randomly generated fracture surfaces, this invention introduces the three-dimensional root mean square (RMS) slope value as another quantitative evaluation index for fracture roughness. That is, this invention comprehensively uses both the Hearst index and the RMS slope value to evaluate the roughness of rock fracture surfaces. Specifically, multiple 50mm granite and shale standard cube specimens were processed using the Brazilian splitting method to obtain rough fracture surfaces, with both granite and shale specimens taken from the same batch of stone. A Matlab program was used to calculate different evaluation indices for fractures, and the generated fracture surfaces were compared with the fracture surfaces of split rocks to verify the similarity between randomly generated and real fracture surfaces, thus simulating underground rock fractures. A universal testing machine and a customized high-strength blade clamp were used to perform Brazilian splitting on the granite and shale specimens, and the point cloud data of the fracture surfaces was obtained using a Reeyee 3D object scanner. The 3D scanning process of the rough fracture surface is as follows: Figure 5 As shown.

[0093] Three-dimensional linear interpolation is performed on the 3D scanned point cloud data to obtain a visualized rock fracture surface. Figure 5 For example, the scan results of the fracture surface after a shale rock was split are shown.

[0094] After obtaining the point cloud data of rock fractures, the Hearst exponent and root mean square (RMS) value of the fracture surface are calculated. The RMS value of the slope of the three-dimensional surface is also calculated. The calculation formula is as follows:

[0095] (2.1);

[0096] For the root mean square value of slope of 3D point cloud Equation (2.1) can be expressed as:

[0097] (2.2);

[0098] In the formula, , , and This represents the number of points along the x-axis and y-axis. and y is the distance between two points along the x-axis and y-axis, and z is the point cloud coordinate elevation.

[0099] Finally, point cloud data of the upper and lower fracture surfaces of 60 granite 50mm cube specimens and 15 shale 50mm cube rock specimens were obtained after splitting. Matlab code was used to first calculate the root mean square values ​​of the three-dimensional slope of the fracture surface randomly generated using the SRA algorithm, as shown in Table 2.1.

[0100] Table 2.1 Root mean square values ​​of slope of randomly generated fracture surfaces using the SRA algorithm

[0101] <![CDATA[Root mean square value Z of fracture surface slope 2s > 0.3070 0.2513 0.0810

[0102] From the perspective of the three-dimensional slope root mean square value evaluation index, the roughness of randomly generated cracks is relatively low. This may be because the random crack surface is generated based on the Hearst exponent. After linear interpolation, the average elevation difference on the geometric level is further reduced, resulting in a lower three-dimensional slope root mean square value.

[0103] The H-value is an important statistic used to describe the self-similarity and long-term memory properties of time or spatial series. The Hearst exponent is usually calculated based on the Range over Standard Deviation (R / S) analysis method.

[0104] The steps of R / S analysis are as follows: First, remove the mean of the sequence to obtain a zero-mean sequence. This is to eliminate the influence of the mean on volatility. Then, calculate the cumulative deviation of the sequence. :

[0105] (2.3);

[0106] In the formula, It is the first in the sequence One value, It is the mean.

[0107] R (range) is the difference between the maximum and minimum cumulative deviations in a sequence of length n.

[0108] (2.4);

[0109] S (standard deviation) is used to calculate the standard deviation of the sequence: (2.5);

[0110] Calculate the R / S ratio: (2.6).

[0111] The core of the Hearst exponent lies in the relationship between the R / S ratio and the sample size. According to Hearst theory, the R / S ratio typically follows a power law relationship: (2.7); where, It is the Hearst index.

[0112] Through the and Performing linear regression yields a linear model:

[0113] (2.8).

[0114] in It is a constant term. The slope obtained through linear regression. This refers to the Hearst index. Based on the above calculation process, the roughness evaluation index of the cleaved surface of granite and shale was calculated, and some of the results are shown in Tables 2.2 and 2.3 below.

[0115] Table 2.2 Hearst index (H) and root mean square value of three-dimensional slope of some granite samples ( )

[0116] H 0.2153 0.4050 0.4780 0.5353 0.7544 0.7395 <![CDATA[Z 2s ]]> 0.3052 0.3692 0.3630 0.3138 0.3390 0.3458

[0117] Table 2.3 Hearst index (H) and root mean square value of three-dimensional slope of some shale rock samples ( )

[0118] H 0.5057 0.5197 0.3971 0.7189 0.6698 0.2990 <![CDATA[Z 2s ]]> 0.3111 0.2905 0.3713 0.3050 0.2698 0.3221

[0119] Analysis of the results showed that the H-values ​​of the granite fracture surfaces ranged from 0.2 to 0.7, which was relatively wide, mainly concentrated around 0.5. The values ​​are generally high, mainly ranging from 0.33 to 0.34. The Hurst values ​​of shale fracture surfaces range from 0.3 to 0.7, mainly concentrated around 0.5. The values ​​are slightly lower than those of granite, mainly ranging from 0.31 to 0.32. The randomly generated fracture surfaces control H values ​​of 0.3, 0.5, and 0.7, which basically conform to the H value distribution range of fracture surfaces in split rocks. The value is relatively lower than that of real rock. Randomly generated fracture surfaces focus more on properties such as the self-similarity of the fracture surface, therefore its... The values ​​will be slightly lower than those of real rock, but still have high reference value. Therefore, based on the comparison results of the two evaluation indicators, Hearst index and root mean square slope, it can be preliminarily judged that the randomly generated rock fracture surface has a high similarity to the real rock fracture, and is suitable for subsequent visualization grouting tests.

[0120] The reliability of grouting simulation test results is highly dependent on the precision of the grout output equipment. The more precise the control of grout output by the grouting equipment, the higher the credibility of the grout diffusion law revealed by the test results.

[0121] The Teledyne Isco 260D grouting pump (hereinafter referred to as the ISCO pump) has a flow rate range of 0.001~107 ml / min with an accuracy of 0.5% and a pressure range of 0.7~517.1 bar with an accuracy of 0.5%FS. The ISCO pump can provide stable and highly accurate constant flow rate and high grouting pressure conditions, making it suitable for studying the diffusion law of grout under constant flow rate and high grouting pressure conditions for small-aperture cracks.

[0122] To increase the reliability of the grouting pattern, an additional grout output device was installed, which can directly use cement grout. Cement grouting tests were conducted to verify the feasibility of xanthan gum solution as an alternative fluid. The additional grout output device uses a combination of an air compressor and a pressure reducing valve to obtain a more stable and constant air pressure output source. The output air pressure pumps the cement grout into the cracks in an airtight environment. Therefore, a 3L sealed pressure tank was ordered from a machining manufacturer. The pressure tank mainly consists of two parts: the tank body and the pressure cover. The pressure cover has four ports: an outlet, a liquid (air) inlet equipped with a pressure gauge, a safety valve, and a ball valve equipped with a silencer. The safety valve ensures that the pressure inside the tank does not exceed 0.8MPa; it automatically releases pressure when the pressure inside the tank is too high. The ball valve can be manually opened to control the pressure inside the tank. A pressure gauge is attached to the liquid inlet to check the input air pressure. A flexible hose extends into the bottom of the tank at the outlet; the pressurized grout inside the tank is pumped out through the hose at the outlet. Grout output equipment consisting of an air compressor and a grouting tank, such as Figure 6 As shown, cement grout can be used directly for grouting, but the accuracy of grouting pressure control is low, and the test results are only used as a feasibility verification.

[0123] Building such Figure 6The visualized grouting test platform for goaf areas shown is used to conduct visualized grouting experiments. This platform mainly includes a plunger grouting pump, an air compressor 1, a grouting tank 2, a pressure reducing valve 3, a pressure gauge 4, a grouting pipe, a transparent rough rock fissure grouting device 6, and a camera device 5. It can be used to study the diffusion law of grout in goaf fissures, primarily using xanthan gum solution as a substitute fluid for cement grout. Simultaneously, a combination of an air compressor, pressure reducing valve, and grouting tank is added as a grout output device. Since the air compressor output pressure is unstable, the pressure reducing valve is used to control the pressure. The grouting tank contains an appropriate proportion of cement grout, which is injected into the transparent rock fissures using air pressure. The constructed visualized cement grouting test platform is shown below. Figure 6 As shown, the results are mainly used to verify the feasibility of xanthan gum solution as an alternative fluid.

[0124] Random rock fissure surfaces generated using the SRA algorithm allow for roughness control. Ordinary resin is used as the material, and the fissure surfaces obtained through 3D printing serve as the mold and bottom surface. A transparent top surface is then created by replicating the 3D-printed surface using transparent epoxy resin. The advantages of this method are relatively low cost and the layer formed by the Vaseline release agent during resin molding, resulting in a rougher surface compared to direct 3D printing and silicone replication. This allows for better replication of the rough surface of rock fissures, making it more similar to real rock fissures.

[0125] The following describes a method for creating transparent, rough rocks using epoxy resin.

[0126] First, the rough, cracked test piece, printed using ordinary resin, was 300mm x 300mm in size. A randomly generated rough cracked surface was designated as the top surface, and a smooth bottom surface with a thickness of approximately 1cm. After modeling, a 3D model of the mold was obtained. This model was then exported as an STL file and printed using a 3D printer to produce the mold. Figure 6 As shown.

[0127] The epoxy resin and curing agent used in this invention are mixed in a 3:1 ratio. After preparing the epoxy resin solution according to the ratio, the solution needs to be cured at 25°C for 24 hours, then cured at 75°C for 36 hours after demolding, and subsequently cured in a -30°C ultra-low temperature freezer for 48 hours. Therefore, constant high temperature and constant low temperature environments are required. The constant high temperature environment is provided by an electrically heated constant temperature drying oven. The constant low temperature environment is provided by an ultra-low temperature freezer. The prepared epoxy resin solution is placed in a large volumetric container according to the ratio, stirred in one direction for about 20 minutes, and then allowed to stand for 15 minutes. A layer of petroleum jelly is evenly applied to the inside of the mold beforehand to facilitate demolding. The epoxy resin is guided into the mold using a stirring rod (using a vacuum chamber or other means to reduce air bubbles). The temperature of the drying oven is set to 25°C. The mold filled with resin is placed stably in the drying oven for 24 hours of curing, after which the mold is demolded. Figure 6As shown. Take out the cured sample and use an angle grinder to grind down any uneven or sharp areas on the crack surface.

[0128] The drying oven temperature was then set to 75℃, and the sample was placed in the drying oven for 36 hours of curing. After curing, the sample was placed in a -30℃ low-temperature freezer for 48 hours. It could then be used before the test. In addition, grouting holes needed to be pre-reserved on the upper fracture surface. The grouting holes were formed using a glass rod pre-embedding method. A glass rod with the same diameter as the pre-reserved grouting hole was prepared. Vaseline was applied to the lower part of the glass rod, and it was placed vertically in the center of the fracture surface. After the resin cured, the glass rod was pulled out, and the grouting hole area was smoothed using an electric drill and angle grinder. A transparent, rough fracture surface of the rock was obtained, as shown in the image. Figure 6 Touching the fracture surface with your hand, you can clearly feel that the fracture surface is relatively rough, similar to the fracture surface of a rock. Tapping the fracture surface with your finger does not cause obvious deformation, and the initial judgment is that the hardness is good and the brittleness is basically in line with the physical properties of rock.

[0129] To simplify the experimental model, the aperture of the entire fracture was kept constant throughout the experiment, resulting in an equal-aperture fracture model. The equal-aperture fracture was obtained using the mechanical constraint method, such as... Figure 6 As shown, the upper and lower crack surfaces are stacked and fixed. Holes are drilled at the four corners of the crack model using an electric drill. The contact surfaces between the upper and lower cracks at each hole are smoothed. A 1.2mm flat washer is inserted, and then it is fitted onto the M6 ​​limit bolt to fix the upper and lower crack surfaces and control the crack opening to remain constant at 1.2mm. By controlling the number of 1.2mm flat washers, the crack opening can be controlled to be an integer multiple of 1.2mm.

[0130] During the experiment, after capturing a video of the grouting diffusion, frames were extracted every two seconds, and a MATLAB program was used to identify the grout diffusion range. Since a red tracer was added to the grout, identifying the red areas in the image revealed the grout diffusion morphology and outline. The following steps were taken for grouting area image recognition: 1) Read the image to be processed and load it into the MATLAB working environment: Use the `imread` function to read the image file as a matrix. Each element of this matrix corresponds to the RGB value of a pixel in the image. 2) Although the RGB space is usually sufficient for color detection, sometimes converting to other color spaces (such as HSV) can make color separation more obvious. In the HSV (Hue, Saturation, Value) space, hue (H) usually reflects color more intuitively, while saturation (S) and value (V) can help further define the vividness and brightness of the color. 3) Set a threshold range for red. Based on the characteristics of the image and the lighting conditions, determine the range of red for subsequent filtering steps. Typically, red pixels in the RGB space have the following characteristics: relatively high R (red channel) values, and relatively low G (green channel) and B (blue channel) values. 4) Create a binary mask to mark pixels that meet the red condition. The white parts of the mask represent red areas, and the black parts represent other areas. Use logical operators to create the mask, comparing the red channel values ​​with a threshold. 5) Extract the red areas from the original image using the mask. Create a new image matrix and apply the mask to the original image. Pixels in the original image are only preserved when the mask is true (i.e., a red area). Iterate through the three RGB channels, preserving the values ​​of the red areas and setting other areas to zero.

[0131] like Figure 7 As shown, the red area recognition program can accurately identify the grout diffusion range and calculate the number of pixels in the red area. The ratio of the number of pixels in the red area to the total number of pixels in the image is equal to the ratio of the area of ​​the red area to the geometric area of ​​the model plane. Therefore, the area of ​​the red area, i.e., the grout diffusion range, can be calculated.

[0132] See Figure 8The figure shows the effect of cement grout flow in cracks under different grouting pressures. From left to right, the three columns represent grouting pressures of 0.1, 0.15, and 0.2 MPa, and from bottom to top, they represent grout diffusion effects at 0, 0.5, 1, 1.5, 2, 2.5, and 3 seconds. The figure clearly shows that grouting pressure has a significant impact on grout diffusion. As the grouting pressure increases from 0.1 MPa to 0.2 MPa, the grout diffusion area changes significantly at the same time point. Most research indicates that grouting pressure is a key factor controlling the grout diffusion rate; higher grouting pressure can effectively shorten the grouting time to reach the target grout volume. In the cement grout test, a larger ratio of grout hole to crack opening resulted in a higher grout volume per unit time through the grout hole at a grouting pressure of 0.1 MPa, leading to a faster grout diffusion rate.

[0133] Referring to Figure 9, it shows the effect of cement grout flowing in the cracks under different crack openings. From left to right, the crack openings are 1.2, 2.4, and 3.6 mm, and from bottom to top, the grout diffusion effect is shown at 0, 2, 4, 6, 8, and 10 s. Figure 9(b) shows the grout diffusion effect under different crack roughnesses. From left to right, the crack roughnesses are H=0.7, 0.5, and 0.3, and from bottom to top, the grout diffusion effect is shown at 0, 2, 4, 6, 8, and 10 s.

[0134] It is evident that the crack aperture has a limited impact on the grout diffusion effect; changes in crack aperture do not significantly alter the diffusion area and radius. That is, under the same grout and grouting pressure, different crack apertures require the same grouting time to achieve the same grout diffusion effect. However, when the crack aperture is 3.6 mm, a significant interfacial stretching phenomenon appears at the grout front. This indicates that although the rate of change in the grout diffusion area does not change significantly with the crack aperture, a larger aperture results in a larger crack volume to be filled. The grout cannot completely fill the cracks within the already diffused area, leaving the upper part of the crack unfilled. When the crack aperture increases, the grout's filling capacity is insufficient to completely fill cracks with larger apertures.

[0135] From the perspective of grout diffusion range, the roughness of the fissure has a relatively low impact on the grout diffusion rate. This is reflected in the fact that, at the same time point, the grout diffusion area for different roughnesses is not significantly different. When the fissure roughness H=0.7, the roughness of the fissure is low, and the resistance of the fissure to grout flow is small. Therefore, the diffusion radius of the grout is slightly larger at H=0.7. However, from the perspective of grout diffusion morphology, there are significant differences between different roughnesses. When the fissure roughness is low, the resistance of the fissure to grout flow is small, and the grout forms multiple locally dominant seepage channels within the fissure. This is reflected in the fact that, at H=0.7, the irregularity of the grout diffusion morphology at 10 seconds is significantly higher than that at H=0.3. In two-dimensional rough fissures, the grout flow rate is not affected by the fissure roughness. In the experiments of this invention, the diffusion rate of cement grout at different roughnesses is not significantly different, which is consistent with the aforementioned understanding.

[0136] Considering the influencing factors such as different grouting pressures, crack roughness, and crack opening, the grout diffusion law of the cement grout test results of this invention is basically consistent with existing research results, proving that the cement grout test has reference value and can be further compared with the xanthan gum solution test results.

[0137] The diffusion pattern of cement grout is basically consistent with existing research results. To verify that the test results of xanthan gum solution can provide a reference for engineering grouting, the diffusion patterns of cement grout and xanthan gum solution at different roughness levels when the grout front diffuses to the edge of the model are compared. Figure 10 As shown, when the fissure roughness H=0.7, the diffusion morphology and diffusion range of the two are highly similar. When the fissure roughness H=0.5, there is a certain difference in the diffusion range, but the irregularity of the cement slurry diffusion morphology is significantly lower than that of the xanthan gum solution, showing a considerable difference. When the fissure roughness H=0.3, the diffusion range of the two is basically the same, but the irregularity of the cement slurry diffusion morphology is extremely low, almost circular. Overall, the xanthan gum solution basically reflects the diffusion law of the cement slurry.

[0138] A single grouting test condition was selected, and the grout diffusion path was analyzed, showing that the diffusion forms are different. Taking the grouting condition with an injection rate of 40 ml / min, a grout concentration of 1.2%, and a crack roughness H=0.5 as an example, the diffusion path of the grout in the rough crack was analyzed. Figure 11 This diagram illustrates the diffusion path of the slurry in rough fractures, the diffusion range of the slurry at different time points, and the flow path of the slurry over time. The red area represents the diffusion region of the xanthan gum solution with added red pigment. The dashed lines of different colors represent the boundaries of the slurry diffusion range at different times, through which the diffusion rate and diffusion range of the slurry at different times can be clearly observed.

[0139] The boundary profile of the grout diffusion range, centered on the grouting hole, gradually increases over time, forming a roughly circular shape. The grout diffusion boundary extends outward through the dominant seepage channels in the fracture. In the initial stage of grouting, the grout diffusion rate is fast, the pressure at the grout diffusion boundary is high, and the boundary profile is relatively irregular. As the grouting process progresses, the grout diffusion is affected by the coupling effect of fracture morphology, increased flow resistance, and decreased pressure gradient, causing the diffusion range to tend towards a circular shape.

[0140] Figure 11 The middle arrow indicates the dominant path of slurry flow, indicating that the slurry does not diffuse uniformly in the fracture network, but is influenced by fracture roughness, pressure gradient and connectivity, and preferentially flows along the path with less resistance, i.e. the dominant seepage channel. Figure 11 The variation in the intervals of the dashed lines reflects the diffusion rate at different time points. The diffusion rate gradually slows down from the area near the grouting hole to the outer edges, likely due to the gradual decay of the pressure gradient. Diffusion is faster and more uneven in the central region, while slower and more circular diffusion occurs at the edges, indicating that the influence of the fracture morphology on the grout diffusion range gradually weakens as the grouting pressure gradient decreases.

[0141] The above example is a visualization analysis of grout diffusion in a single working condition of a grouting test, intuitively showing the path and final distribution of grout flow and diffusion in the fracture. It is evident that grout diffusion in fractures is not a simple displacement process; not only are there locally dominant seepage paths, but it also exhibits different grout diffusion patterns depending on factors such as fracture morphology and grouting pressure.

[0142] To highlight the evolution of the grout diffusion morphology, a circular area with a radius of 10m centered on the grouting hole was selected as the representation of the grout diffusion range cloud map. With other factors such as crack aperture and grout water-cement ratio kept constant, numerical simulations were performed at grouting pressures P of 0.2MPa, 0.3MPa, 0.4MPa, and 0.5MPa. The grout diffusion effect cloud maps under different grouting pressures are shown below. Figure 12a As shown in the figure, the grout diffusion area varies significantly under different grouting pressures, indicating that increasing the grouting pressure greatly increases the grout diffusion rate. This result demonstrates that grouting pressure has a significant impact on grouting efficiency. According to the Navier-Stokes equations for fluid dynamics, increasing the grouting pressure directly leads to an increase in the pressure gradient term, thus driving the grout to flow rapidly within the fracture. However, although increasing the grouting pressure significantly alters the grout diffusion area, the grout still preferentially diffuses within the fracture along local seepage channels. This indicates that increasing the grouting pressure does not cause the pressure gradient to become absolutely dominant relative to the flow resistance; the grout diffusion range profile remains roughly similar under different grouting pressures.

[0143] Based on the regularity analysis results in the cloud map, the numerical simulation results were further analyzed. The grouting area values ​​as a function of time under different grouting pressures were extracted, and the grouting area-time curve was plotted as follows: Figure 12b As shown in the figure, the grout diffusion area and its rate of change vary significantly under different grouting pressures. At 50 seconds, when the grouting pressure P is 0.2 MPa, 0.3 MPa, 0.4 MPa, and 0.5 MPa, the grouting areas are 174.03 m², 210.35 m², 238.54 m², and 265.89 m², respectively. Compared to 0.2 MPa, the grouting areas under 0.3 MPa, 0.4 MPa, and 0.5 MPa pressures increase by 20.87%, 37.07%, and 52.78%, respectively. These data indicate that the grouting pressure has a significant impact on grouting efficiency, which is consistent with existing research results and experiments. Furthermore, the grouting area and grouting pressure have a non-linear positive relationship.

[0144] Numerical simulations were performed with other factors such as the water-cement ratio and grouting pressure kept constant, and the crack aperture d was set to 1cm, 1.5cm, 2cm, 2.5cm, and 3cm. The cloud diagrams showing the grout diffusion effect under different crack apertures are as follows: Figure 13 As shown in the figure, the most significant difference under different fracture openings is the dramatic change in the shape of the grout diffusion area. Observing the grout diffusion shape under different openings reveals that as the fracture opening increases, the grout no longer flows along the locally optimal seepage channel, but directly selects the globally optimal seepage channel as the diffusion path. Specifically, the grout front changes from a few local protrusions to multiple outwardly extending protrusions. This phenomenon indicates that as the fracture opening increases, the restrictive effect of the fracture opening on the grout gradually decreases. When the grout front flows forward, it no longer simultaneously contacts the upper and lower surfaces of the fracture. Its flow pattern changes from flow within a narrow channel to unrestricted natural flow. The resistance and surface tension effects brought about by flow along the upper and lower fracture surfaces decrease, while the influence of gravity on the fluid increases accordingly. Therefore, the grout seepage channel extends along the lower elevation of the fracture morphology, causing the dramatic change in the grout diffusion shape shown in the figure. Due to the dramatic change in the grout diffusion shape, it is impossible to directly determine the trend of the grout diffusion area. It is necessary to identify and calculate the grout diffusion area to obtain the law of diffusion area change.

[0145] Based on the results in the cloud map, the numerical simulation results were further analyzed. The grouting area values ​​over time were extracted for different crack openings, and grouting area-time curves were plotted. Even with significant changes in the shape of the grout diffusion range under different crack openings, the grout diffusion area and its rate of change did not change considerably. At 50 seconds, when the crack opening d was 1cm, 1.5cm, 2cm, 2.5cm, and 3cm, the grouting areas were 174.03m², 175.49m², 176.14m², 171.96m², and 164.60m², respectively. These data indicate that the impact of crack opening on the efficiency of increasing the grouting coverage area is relatively limited, consistent with the cement grout test results under different crack openings. As the crack opening increases, the grouting area shows a very small decreasing trend. Because the fracture aperture changes, the volume of the fracture cavity per unit area also changes. Considering only the grouting area is insufficient to accurately determine the grouting rate. Therefore, the grouting volume over time under different fracture apertures was statistically analyzed to obtain a curve. Specifically, at 50 seconds, when the fracture aperture d was 1cm, 1.5cm, 2cm, 2.5cm, and 3cm, the grouting volumes were 1.483m³, 2.126m³, 2.759m³, 3.305m³, and 3.754m³, respectively. Compared to the aperture d=1cm, the grouting volume increased by 43.36%, 86.04%, 122.86%, and 153.14% when the aperture d was 1cm, from 1.5cm to 3cm.

[0146] Keeping other factors such as grouting pressure and crack morphology constant, numerical simulations were performed with cement paste water-cement ratios (W / C) of 0.5, 0.8, and 1.0. The cloud maps showing the grout diffusion effect under different water-cement ratios are shown below. Figure 14 As shown. Figure 14 As shown, when the water-cement ratio of the cement-based grout is high, the apparent viscosity of the liquid is low, the viscous flow resistance is small, and the grouting pressure is absolutely dominant, driving the grout to flow rapidly, resulting in an irregular flow shape in the diffusion range. Conversely, when the water-cement ratio of the cement grout is low, the apparent viscosity of the liquid is high, the flow resistance is large, the resistance from the crack morphology is relatively small, and the diffusion range exhibits a geometrically regular flow shape. Since the shape of the grout diffusion range changes, the change in grouting area cannot be directly judged by comparison. Therefore, it is considered to calculate the grouting area value for quantitative and precise analysis.

[0147] The grouting area values ​​over time under different water-cement ratios were extracted. It is evident that the water-cement ratio not only affects the shape of the grout diffusion range but also significantly influences the diffusion area and its rate of change. At 50 seconds, when the water-cement ratio (W / C) of the cement paste was 0.5, 0.8, and 1, the grouting areas were 98.65 m², 142.05 m², and 177.70 m², respectively. Specifically, the grouting area at a water-cement ratio of 0.5 decreased by 30.55% compared to 0.8, while the grouting area at a water-cement ratio of 1 increased by 25.10% compared to 0.8. These data indicate that an increase in water-cement ratio leads to an increase in grouting area and grouting efficiency, due to a decrease in grouting resistance. Changes in the water-cement ratio directly affect the rheological properties of the grout, and controlling the water-cement ratio essentially alters the rheological capacity of the grout. Existing research shows that the properties of slurry-based fillers can be modified by controlling the water-cement ratio, adding fly ash and water-reducing agents, etc., to meet different crack filling needs. For small cracks with an aperture of 1 cm or less, appropriately increasing the water-cement ratio and adding water-reducing agents can improve the fluidity of the slurry and help it to flow and diffuse fully within the crack.

[0148] With other factors such as crack opening and grouting pressure kept constant, numerical simulations were performed with crack roughness H set to 0.3, 0.5, and 0.7. The cloud diagrams showing the grout diffusion effect under different roughnesses are as follows: Figure 15 As shown, different fracture roughnesses result in different grout diffusion patterns, indicating that the grout diffusion path is influenced by the fracture morphology distribution. In different fractures, the grout tends to diffuse along locally dominant seepage channels, forming convex grout diffusion regions, thus transforming the overall grout diffusion profile from a regular to an irregular shape. As the fracture roughness increases, the fracture morphology becomes more complex and variable, with fracture undulations forming more locally dominant seepage channels. The grout front continuously advances within the fracture, interacting with these channels and forming more branches, resulting in a more irregular geometric profile of the grout diffusion. Therefore, as the H value decreases and approaches 0, the fracture roughness increases, and the geometric irregularity of the grout diffusion profile increases. The cloud map reflects a correlation between the geometric irregularity of the grout diffusion profile and the fracture roughness, revealing the underlying reason: as fracture roughness increases, the fracture morphology becomes more undulating, the distribution of localized seepage channels within the fracture becomes denser, and the grout diffusion area exhibits a more convex and irregular shape. However, the cloud map shows no significant difference in grout diffusion area and velocity under different fracture roughnesses. Therefore, it can be preliminarily concluded that fracture roughness has a relatively low impact on grouting efficiency, but further quantitative analysis based on specific data is still needed.

[0149] Based on the regularity analysis results in the cloud map, the numerical simulation results were further analyzed. As shown in the figure, the rate of change of grout diffusion area is similar under the three crack roughness conditions, indicating that the grouting efficiency is almost the same under different crack roughness conditions. Only when the roughness H is 0.7 does the grouting area and growth rate show a slight advantage over the other cases. The reason for this phenomenon is that as the H value approaches 1, the crack becomes smoother, the undulation is reduced, and the flow resistance term of the grout diffusion behavior decreases, thus making the grout diffusion range slightly wider. However, the difference in grouting area under different roughness conditions is extremely small: at 50s, when the H values ​​are 0.3, 0.5, and 0.7, the grouting areas are 177.17 m², 174.03 m², and 180.04 m², respectively. This shows that the flow resistance term caused by crack roughness accounts for a small proportion of the overall resistance to grout diffusion, and its influence is relatively small compared to the pressure gradient driving grout diffusion.

[0150] To calculate the Pearson correlation coefficient between two variables, let the variables be respectively... and Pearson correlation coefficient Defined as the ratio of covariance to standard deviation, as shown in equation (4.13):

[0151] (4.13);

[0152] In the formula, and They are respectively and The mean, For covariance, and The standard deviation is denoted as .

[0153] The calculated Pearson correlation coefficient The value range is from -1 to 1, when When =1, it indicates and Perfectly positive linear correlation; when When =-1, it indicates and Perfectly negative linear correlation; when When =0, it indicates and Wireless correlation.

[0154] Based on existing data, Pearson correlation coefficients were calculated between each of the four key parameters and two evaluation indicators. The results are presented as heatmaps in Figures 16 and 17. As can be seen from the figures, there is indeed no correlation between the four key parameters. The influence on the grouting filling rate, from largest to smallest, is: grouting pressure > water-cement ratio > crack aperture > crack roughness. The influence on the actual grouting filling rate, from largest to smallest, is: water-cement ratio > grouting pressure > crack aperture > crack roughness. This result is highly consistent with the experimental and numerical simulation regularity analysis results. The Pearson correlation coefficient between grouting pressure and filling rate is 0.82, indicating a high linear positive correlation between grouting pressure and filling rate. The Pearson correlation coefficient between grout water-cement ratio and actual filling rate is -0.88, indicating a high linear negative correlation between grouting pressure and actual filling rate. While the Pearson correlation coefficient quantitatively characterizes the correlation between key parameters and evaluation indicators, it also has limitations; it only captures the linear relationship between variables in the calculation, ignoring nonlinear effects. Therefore, we should consider using other methods for cross-validation.

[0155] Decision trees are tree-like classification models that achieve classification or regression by recursively partitioning the data space. Building upon the Pearson correlation coefficient, which reveals linear relationships between variables, decision trees are further introduced to capture nonlinear interactions and conditional dependencies between indicators. As a typical white-box model, decision trees construct rule sets through recursive feature partitioning, and their node splitting process directly reflects the ranking of feature importance: driven by specific branching criteria, indicators with stronger explanatory power for the target variable will participate in data partitioning first. This feature selection mechanism based on information purity can effectively identify sensitive indicators with discriminative power in the local data space, and is particularly suitable for research scenarios with threshold effects or interactions. Compared to the global linear assumption of the Pearson coefficient, decision trees achieve hierarchical deconstruction of complex association patterns through multi-level conditional judgments, providing an interpretable path analysis framework for understanding the coupling effects of multiple indicators. However, the feature importance assessment of a single decision tree is susceptible to random perturbations in the training data. Therefore, this study will employ two ensemble learning methods to enhance the robustness of the assessment: in Random Forest, the global impact of the importance metric features on prediction accuracy is assessed through Out-of-Bag (OOB) permutation, i.e., the performance degradation of the model is calculated by randomly permuting features from out-of-bag samples; in Gradient Boosting Tree (GBDT), the splitting frequency of each feature across all base learner nodes is statistically analyzed to quantify the cumulative contribution of features to error optimization in an incremental manner. Cross-validation of these two methods will enhance the reliability of the sensitivity analysis conclusions.

[0156] Random Forest (RF) is an ensemble learning algorithm based on decision trees. By constructing multiple decision trees and aggregating predictions, Random Forest significantly improves the robustness and generalization ability of the model. In sensitivity analysis, the feature importance score of Random Forest can quantify the nonlinear and interactive effects of input parameters on output metrics.

[0157] One method for evaluating feature importance in RF models is permutation importance. This involves randomly shuffling the values ​​of a feature and calculating the decrease in classification accuracy for out-of-bag (OOB) data. A greater decrease indicates a more important feature. The formula for calculating OOB permutation importance is shown in equation (4.18).

[0158] (4.18);

[0159] In the formula, The total number of trees, and The first The original OOB error and the OOB error after feature permutation for each tree.

[0160] Based on the above method, the RF model in Matlab software was trained using a multi-condition numerical simulation dataset of rock mass grouting, with the number of trees set to 100, and feature importance was extracted as follows: Figure 18 As shown.

[0161] The above results show that grouting pressure has the most significant impact on filling rate (0.7964), similar to the result of Pearson coefficient (0.82). The water-cement ratio has the same impact on grout (0.146), which is somewhat different from the result of Pearson coefficient (0.48). The impact of crack aperture is relatively small (0.0553). Roughness is almost unrelated to filling rate. On the other hand, the water-cement ratio has the most prominent impact on actual filling rate (0.8085), similar to the result of Pearson coefficient (-0.88). The impact of crack aperture is the second largest (0.0977). Roughness has very little correlation with actual filling rate.

[0162] Overall, the relative importance of the key parameters calculated by the RF model to the filling rate is ranked in the same order as that calculated by the Pearson coefficient. However, relative to the actual filling rate, both the RF model and the Pearson coefficient results indicate that the water-cement ratio has the greatest impact, while the RF model shows that the impact of crack aperture is stronger than that of grouting pressure, whereas the Pearson coefficient calculation results show the opposite conclusion.

[0163] Gradient Boosting Decision Tree (GBDT) is a supervised learning algorithm based on ensemble learning. It aims to progressively optimize an objective function by constructing a series of decision tree models. It models complex relationships step-by-step through residual optimization and cross-validates with Random Forest. The core idea of ​​this method is to minimize the loss function through gradient descent. Similar to RF models, GBDT uses decision trees as base learners to capture complex nonlinear relationships in the data. It performs well in both regression and classification problems, especially when handling complex data features. Furthermore, GBDT can assess the importance of features by analyzing the frequency of feature usage in each decision tree. By combining gradient descent and decision trees, GBDT can effectively handle nonlinear relationships in complex data.

[0164] The basic idea of ​​GBDT is to build a set of decision tree models iteratively, attempting to correct the residuals (i.e., prediction errors) of the previous model at each step. Specifically, GBDT optimizes the model step by step through the following steps: ① Initialize a simple model (usually a constant or a single decision tree). ② In each iteration, calculate the residuals of the current model (i.e., the difference between the predicted and actual values). ③ Use gradient descent to build a new decision tree to fit the residuals. ④ Add the newly built decision tree to the model, forming a weighted sum model. ⑤ Repeat the above steps until the preset number of iterations is reached or the model converges.

[0165] The mathematical model of GBDT is established under the guidance of the above ideas. Let the training set be... ,,in The input features (i.e., the key parameters in this study) are used as input features. The target value is denoted as (i.e., the evaluation index in this study). The objective of GBDT is to minimize the loss function as shown in equation (4.19):

[0166] (4.19);

[0167] In the formula, It is the final model, by Decision Tree composition, It is a loss function. It is a regularization term used to prevent overfitting.

[0168] During the iteration process, the loss function is optimized using gradient descent, and the negative gradient of the current model is calculated as shown in equation (4.20):

[0169] (4.20);

[0170] Then, construct a new decision tree. gradient The model is then fitted. Finally, the model is updated as shown in equation (4.21):

[0171] (4.21);

[0172] In the formula, It is the learning rate, used to control the update magnitude at each step.

[0173] Based on the aforementioned GBDT mathematical model, the GBDT function `fitrensemble` from the Statistics and Machine Learning Toolbox in Matlab was called. The hyperparameters of GBDT were set as follows: number of trees = 100, learning rate = 0.2, maximum depth of a single tree = 3, minimum number of samples per leaf node = 10. The GBDT model was trained using a numerical simulation model dataset of grouting in a single fracture of rough rock mass, and feature importance was calculated. The results are as follows. Figure 19 As shown.

[0174] This invention also discloses a system for implementing the visualization analysis method of non-Newtonian fluid grouting in rough fractures of goaf areas disclosed in the first aspect of this invention. The system includes: a transparent fracture physical model, which has roughness characteristics based on real rock fractures and a set aperture; a grouting system, including a first grouting unit and a second grouting unit, wherein the first grouting unit is used to inject a first fluid into the transparent fracture physical model in a constant flow or constant pressure mode, the first fluid being an alternative non-Newtonian fluid; the second grouting unit is used to inject a second fluid into the transparent fracture physical model at a constant air pressure, the second fluid being a cement-based grout; and an image acquisition system, used to acquire image data of the grout diffusion process in the transparent fracture physical model during the grouting process, up to the point where the grout... The system is designed to: 1) Reach the boundary of the transparent fracture physical model; 2) Image processing system to identify and process image data, extract the boundary contour of the grout diffusion area, and calculate the grout diffusion area and its time-varying relationship; 3) Numerical simulation system to construct a rough single-fracture grouting numerical model, cross-validate the numerical simulation results with the grout diffusion area time-varying relationship obtained from the image processing system, and perform multi-condition extended calculations based on the validated numerical model; 4) Parameter analysis system to determine the quantitative correspondence between each grouting parameter and each grouting effect evaluation index based on the calculation results of the multi-condition extended calculations, rank the importance of multiple grouting parameters, and output the characteristic importance weight value or ranking result of each grouting parameter to the grouting effect evaluation index.

[0175] In summary, the present invention has the following overall technical advantages over the prior art:

[0176] First, it achieves a balance between precise controllability of crack roughness and engineering representativeness. Through a closed-loop method combining SRA algorithm generation, 3D printing replication, and Z2+H dual-index verification, it achieves for the first time fractal-level precise control of crack wall roughness and statistical proof of engineering representativeness, breaking the technical prejudice in this field that "controllability and realism cannot be achieved simultaneously."

[0177] Second, a three-dimensional spatial evaluation system for grouting effect was established. The proposed new index of "actual filling rate" elevates the traditional two-dimensional evaluation, which only considers the planar diffusion coverage area, to a three-dimensional spatial evaluation that comprehensively considers the degree of filling in the opening direction. This fundamentally solves the inherent defect of the traditional filling rate in being unable to reflect the true degree of filling of the crack cavity by the grout.

[0178] Third, a systematic verification method for alternative fluids was pioneered. Through a dual-fluid, dual-platform comparative test of xanthan gum solution (high-precision pump) and cement slurry (air compressor-grouting tank), the engineering applicability of the alternative fluid test results was verified for the first time, and the sources of error were quantitatively analyzed, providing a methodological reference standard for subsequent research.

[0179] Fourth, it achieves a quantitative ranking of parameter importance. Through cross-validation using three different mathematical principles—Pearson correlation coefficient, random forest, and gradient boosting tree—machine learning is introduced into the sensitivity analysis of fracture grouting parameters for the first time, achieving a leap from qualitative description of parameter influence to quantitative ranking.

[0180] Fifth, a complete research chain of "experimentation-theory-numerical simulation-machine learning" has been constructed. From the preparation of the fracture model to the quantitative evaluation of grouting effect and parameter optimization, each link in this invention has a strict logical progression relationship, forming a synergistic and efficient systematic methodology, and providing a complete scientific tool for the parameter optimization design of fracture grouting projects in goaf areas.

[0181] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A visual analysis method for non-Newtonian fluid grouting in rough fractures of goaf areas, characterized in that, Includes the following steps: S1, Generate a digital model of the crack surface with controllable roughness, and prepare a physical model of the transparent crack based on the digital model of the crack surface; S2, Visual grouting tests were conducted using a first fluid and a second fluid. The first fluid was an alternative non-Newtonian fluid, and the second fluid was a cement-based grout. Image data of the diffusion process of the grout in the transparent fracture physical model were collected during the grouting process. S3, perform recognition processing on the image data, extract the boundary contour of the slurry diffusion area, and calculate the slurry diffusion area and its change over time; S4. Based on the geometric morphological characteristics of the boundary contour, the traditional filling rate and the actual filling rate are used as evaluation indicators for grouting effect. The traditional filling rate is the ratio of grout diffusion area to the total area of ​​the crack, and the actual filling rate is the corrected filling rate that takes into account the degree of filling in the crack opening direction. S5. Construct a rough single-crack grouting numerical model, cross-validate the numerical simulation results with the relationship between the grout diffusion area and time obtained in step S3, and perform multi-condition extended calculations based on the validated numerical model. S6. Based on the calculation results of the multi-condition extended calculation, determine the quantitative correspondence between each grouting parameter and each grouting effect evaluation index, and rank the importance of the multiple grouting parameters.

2. The method according to claim 1, characterized in that, The actual filling rate in step S4 is set as: η eff =λ·η, where η is the traditional filling rate and λ is the filling correction coefficient, which reflects the degree of complete filling of the crack cavity by the grout in the direction of crack opening; The filling correction coefficient ranges from 0 to 1, and is based on the fractal dimension D of the slurry diffusion boundary. f It is determined that the filling correction coefficient λ is negatively correlated with the fractal dimension, D f A larger value indicates a more irregular slurry diffusion pattern, and a smaller corresponding λ value.

3. The method according to claim 2, characterized in that, Step S4, which uses the traditional filling rate and the actual filling rate as evaluation indicators for grouting effect, includes: The fractal dimension of the boundary profile is calculated. The fractal dimension is used to quantify the degree of geometric irregularity of the slurry diffusion pattern. The larger the fractal dimension, the more irregular the diffusion pattern. The smaller the fractal dimension, the more regular and circular the diffusion pattern. The filling correction coefficient λ is determined based on the fractal dimension, thereby obtaining the actual filling rate.

4. The method according to claim 1, characterized in that, In step S1, the root mean square value of the three-dimensional slope and the Hurst exponent of the digital model of the fracture surface are calculated respectively, and compared with the root mean square value of the three-dimensional slope and the Hurst exponent of the real rock split fracture surface obtained by the Brazilian splitting method. If both of the above two indicators of the digital model of the fracture surface fall within the statistical distribution interval determined by multiple sets of real rock split fracture surface measurement values, then the two are judged to have statistical similarity.

5. The method according to claim 1, characterized in that, The step of preparing the physical model of the transparent crack in step S1 includes: A crack surface mold was prepared based on the digital model of the crack surface using 3D printing. Epoxy resin was used as the raw material to be molded and cured in the mold to obtain a transparent crack surface. The upper and lower transparent crack surfaces are stacked and fixed in a mechanical limiting manner, and the crack opening is controlled by a shim to form a physical model of a transparent crack with a set opening.

6. The method according to claim 1, characterized in that, In step S2, the first fluid is a xanthan gum solution with a mass fraction of 0.3% to 1.2%; the second fluid is ordinary silicate cement slurry with a water-cement ratio of 0.5 to 1.

0. The visualization grouting test described in step S2 is conducted in a water-saturated environment. The transparent crack physical model is fixed in a water tank, and the height difference between the liquid level in the water tank and the highest point of the transparent crack physical model is controlled within 3 cm.

7. The method according to claim 1, characterized in that, The steps for recognizing and processing image data described in step S3 include: The system extracts single-frame images from the acquired image sequence at set time intervals and converts each extracted frame image from the RGB color space to the HSV color space. A binary mask is generated based on a preset threshold range for red in the HSV space. The red region in the image is extracted using the binary mask, and the number of pixels in the red region is counted. The slurry diffusion area is calculated based on the principle of proportional conversion, using the ratio of the number of pixels to the total number of pixels in the image, combined with the actual planar geometric area of ​​the transparent crack physical model.

8. The method according to claim 1, characterized in that, The step of cross-validating the numerical simulation results with the grout diffusion area obtained in step S3 over time in step S5 includes: comparing the grout diffusion area over time curves obtained from numerical simulation and visual grouting test under the same working conditions; if the degree of agreement between the two reaches a preset threshold, the numerical model is deemed to have passed the validation.

9. The method according to claim 1, characterized in that, Step S6, which involves ranking the importance of multiple grouting parameters, includes: The Pearson correlation coefficient method was used to calculate the linear correlation coefficient between each grouting parameter and the grouting effect evaluation index, and a correlation coefficient heat map was established to preliminarily rank the importance of the parameters. Simultaneously, random forest model and gradient boosting tree model are used to calculate the feature importance value of each grouting parameter to the grouting effect evaluation index, and the importance of the parameters is sorted in a second time based on the feature importance value; The results calculated by the Pearson correlation coefficient method, the random forest model, and the gradient boosting tree model were cross-validated to obtain the final parameter importance ranking.

10. A system for implementing a visualization analysis method for non-Newtonian fluid grouting in rough fractures of goaf areas according to any one of claims 1-9, characterized in that, include: A transparent fracture physical model, wherein the transparent fracture physical model has roughness characteristics based on real rock fractures and has a set opening degree; The grouting system includes a first grouting unit and a second grouting unit. The first grouting unit is used to inject a first fluid into the transparent fracture physical model in a constant flow mode or a constant pressure mode. The first fluid is an alternative non-Newtonian fluid. The second grouting unit is used to inject a second fluid, which is a cement-based grout, into the transparent fracture physical model at a constant air pressure. An image acquisition system is used to acquire image data of the diffusion process of grout in the transparent fracture physical model during the grouting process, until the grout front reaches the boundary of the transparent fracture physical model; The image processing system is used to recognize and process the image data, extract the boundary contour of the slurry diffusion area, and calculate the slurry diffusion area and its change over time. The numerical simulation system is used to construct a numerical model of rough single-crack grouting. The numerical simulation results are cross-validated with the relationship between the grout diffusion area and time obtained by the image processing system. Based on the validated numerical model, multi-condition extended calculations are performed. The parameter analysis system is used to determine the quantitative correspondence between each grouting parameter and each grouting effect evaluation index based on the calculation results of the multi-condition extended calculation, and to rank the importance of the multiple grouting parameters, and output the characteristic importance weight value or ranking result of each grouting parameter to the grouting effect evaluation index.

Images