Modeling method and system applied to roadway model

Abstract

The invention relates to the technical field of data modeling, in particular to a modeling method and system applied to a roadway model. The method comprises the following steps: deploying a multi-source sensing equipment array on a roadway engineering site to collect surrounding rock deformation data, roadway surface deformation point cloud data and supporting structure state data, carrying out data preprocessing, and constructing a roadway multi-source original data set; performing space-time fusion processing according to the roadway multi-source original data set to generate a synchronized roadway data set; constructing a three-dimensional geomechanical feature space by using the synchronized roadway data set, and generating a three-dimensional roadway feature tensor; therefore, through multi-source data fusion and digital-physical collaborative modeling, the problems of data splitting and insufficient model precision in traditional roadway support design are solved, and the prediction accuracy and the construction safety guarantee level of a roadway support scheme are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data modeling, and in particular to a modeling method and system applied to a roadway model. Background Art

[0002] Existing methods applied to roadway modeling generally have significant technical shortcomings in aspects such as multi-source data fusion, model accuracy control, parameter optimization, and engineering adaptability. Most methods still rely on single physical quantity information in the data acquisition link. For example, the model is constructed only based on surrounding rock stress or surface deformation, ignoring the comprehensive utilization of multi-modal data such as bolt strain, stress state of the support structure, and geometric changes of the roadway point cloud, resulting in one-sidedness in reflecting the actual stress and deformation mechanism inside the roadway. At the same time, due to the lack of high-precision time synchronization and spatial alignment mechanisms, there is a problem of inconsistent spatio-temporal scales between multi-source data, severely restricting the integrity and accuracy of subsequent three-dimensional geomechanical feature construction. At the modeling verification level, existing methods mostly use pure numerical simulation to construct the response model of the surrounding rock-support system, lacking a closed-loop checking mechanism with actual physical model tests, unable to achieve dynamic correction and error compensation of model parameters, resulting in insufficient reflection ability of the digital model for actual engineering conditions. In addition, during the support optimization design process, existing methods often rely on experience to set a small number of parameter combinations, unable to construct a high-dimensional support scheme space covering complex working conditions, and lacking an evaluation mechanism integrating various deformation criteria and structural constraints, making it difficult to meet the refined prediction requirements of multiple objectives and multiple constraints in actual roadway design. More prominently, most current models are not lightweight processed after construction, with problems such as excessive parameter quantity and complex structure, restricting their real-time deployment and dynamic update capabilities in on-site embedded systems or edge computing platforms. Summary of the Invention

[0003] Based on this, it is necessary to provide a modeling method and system applied to a roadway model to solve at least one of the above technical problems.

[0004] To achieve the above object, a modeling method applied to a roadway model, the method includes the following steps: Step S1: Deploy a multi-source sensor device array at the roadway engineering site to collect surrounding rock deformation data, roadway surface deformation point cloud data, and support structure state data, and perform data preprocessing to construct a roadway multi-source original data set; Step S2: Perform spatio-temporal fusion processing on the roadway multi-source original data set to generate a synchronized roadway data set; use the synchronized roadway data set to construct a three-dimensional geomechanical feature space, and generate a three-dimensional roadway feature tensor; generate an initial roadway digital model based on the three-dimensional roadway feature tensor; Step S3: Use the initial digital model of the roadway to simulate the repair support plan and conduct physical model tests to obtain physical test data; perform digital-physical error analysis using the physical test data to obtain model compensation parameters; correct the parameters of the initial digital model of the roadway based on the model compensation parameters to obtain an optimized digital model of the roadway. Step S4: Predict the repair effect based on the optimized digital model of the roadway to generate support optimization prediction data; conduct a full-process modeling evaluation according to the support optimization prediction data and generate a report to obtain a full-process model construction report of the roadway model.

[0005] The beneficial effects of the present invention are as follows. By deploying a multi-source sensor device array at the roadway engineering site, the surrounding rock deformation data, the roadway surface deformation point cloud data, and the support structure state data are collected in real time, forming a data basis with multi-dimensional and multi-temporal resolutions, providing detailed and comprehensive original data support for subsequent data fusion and model construction. The spatio-temporal fusion algorithm is used to synchronize the multi-source original data sets, eliminating the spatio-temporal asynchrony and dimensional differences between the data, ensuring that the spatial positions and timestamps of different types of data are highly consistent, thereby constructing an accurate synchronized roadway data set. On this basis, combined with the geomechanics principle, the fused data is mapped to a three-dimensional feature space, and the high-dimensional characterization of the multi-physical quantity characteristics of the complex roadway structure is realized through tensor expression, effectively capturing the coupling relationship between the surrounding rock and the support structure, and then generating an initial digital model to accurately reflect the mechanical state of the roadway. By combining the digital model with physical model tests, the error analysis method is used to compare the digital and physical test data, extract model compensation parameters to correct the systematic deviation in the digital model, and realize the optimization and precision of the digital model. Based on the optimized digital model, the effect of the repair support plan is predicted, and relying on the structured support optimization prediction data, a full-process modeling evaluation is carried out, systematically quantifying the feasibility and safety of the repair plan, and generating a detailed evaluation report, forming a closed-loop data-driven design and verification process. This process not only realizes the dynamic and refined monitoring of the roadway structure state, but also improves the accuracy and reliability of the model through data-driven methods, promoting the scientific and intelligent support design. Therefore, the present invention solves the problems of data fragmentation and insufficient model accuracy in traditional roadway support design through multi-source data fusion and digital-physical collaborative modeling, and improves the prediction accuracy of the roadway support plan and the construction safety guarantee level.

[0006] Preferably, step S1 includes the following steps: Step S11: Deploy a fiber Bragg grating sensor array on the roof and two sides of the roadway, with a sampling frequency of 10 Hz, a measurement range of 0 - 50 MPa, and an accuracy of ±0.1 MPa, and collect the surrounding rock deformation data. Step S12: Arrange a laser scanner on the tunnel surface with a scanning accuracy of 0.1 mm, and collect deformation point cloud data of the tunnel surface at intervals of 30 minutes; Step S13: embedding strain sensors inside the anchor rods and lining structures, with a sampling frequency of 5 Hz, to collect support structure status data; Step S14: using wavelet transform to perform noise reduction processing on surrounding rock deformation data, tunnel surface deformation point cloud data and support structure status data to obtain a multi-source tunnel noise reduction data set; Step S15: performing data cleaning on the multi-source lane denoising dataset to generate a lane multi-source original dataset.

[0007] The present invention realizes continuous and high-frequency monitoring of surrounding rock deformation by deploying high-precision fiber grating sensor arrays on the top plate and both sides of the tunnel. The sampling frequency reaches 10Hz, the measurement range covers 0-50MPa, and it has a high accuracy of ±0.1MPa, which can carefully capture the stress changes of the surrounding rock at different construction stages. At the same time, a laser scanner is used to collect high-precision (0.1mm) point cloud data on the tunnel surface, and regular scanning is performed at intervals of 30 minutes to ensure the spatial resolution and time continuity of the tunnel surface deformation, thereby obtaining rich three-dimensional spatial deformation information. For the strain sensors embedded in the anchor rods and lining structures, a sampling frequency of 5Hz is adopted, and the range is ±5000 , accuracy ±5 , reflecting the stress state and deformation of the support structure in real time, forming a multi-dimensional, multi-physical quantity sensor data set. For different types of sensor data, wavelet transform technology is used for multi-scale noise reduction processing to effectively remove measurement noise and environmental interference, enhance the signal-to-noise ratio of the data, and ensure the accuracy and reliability of the data. Subsequently, systematic data cleaning is performed on the denoised multi-source data set to remove outliers and missing data, fill data gaps, unify data formats and standards, and form a structured tunnel multi-source original data set. This data set not only covers the multi-dimensional information of the surrounding rock, surface and support structure, but also achieves high precision and temporal and spatial consistency of the data, laying a solid data foundation for subsequent fusion analysis and digital model construction. In general, the present invention ensures the integrity and accuracy of the tunnel structure monitoring data from the source of the data through multi-sensor fusion acquisition and advanced data processing technology, greatly improving the data quality and the effectiveness of subsequent analysis.

[0008] Preferably, in step S2, constructing a three-dimensional geomechanical feature space using the synchronized tunnel data set and generating a tunnel feature tensor comprises: Extract stress distribution based on synchronized tunnel data set, mark rock stress maximum value, and generate tunnel rock stress characteristic data; Conduct convergence analysis on the displacement of the roadway surface points based on the synchronized roadway dataset, and measure the change in the surface bending degree to generate roadway point cloud feature data; Conduct axis deformation analysis on the axis of the support structure based on the synchronized roadway dataset, and calculate the internal shear deformation of the structure to generate roadway internal structure feature data; Map the roadway rock stress feature data, roadway point cloud feature data, and roadway internal structure feature data to a three-dimensional space grid according to stress × displacement × strain to construct a three-dimensional geomechanical feature space, and generate a three-dimensional roadway feature tensor.

[0009] The present invention realizes the systematic extraction and comprehensive expression of the multi-dimensional mechanical characteristics of the roadway based on the synchronized roadway dataset. First, extract the stress distribution of the collected synchronized data, accurately calculate the stress intensity and its maximum and minimum values at different positions of the surrounding rock, and effectively capture the local stress concentration phenomenon through the marking of the maximum and minimum values of the rock stress, forming a high-resolution roadway rock stress feature dataset; Secondly, conduct convergence analysis on the displacement of each key point on the roadway surface based on the point cloud data, quantify the displacement change trend and amplitude on the roadway surface, and combine with the measurement of the change in the bending degree of the surface points to generate point cloud feature data reflecting the overall and local deformation forms of the roadway; Furthermore, by dynamically tracking the axial deformation of the axis and defense line of the support structure, carry out quantitative calculation of the internal shear deformation of the structure, obtain the structural strain distribution and its change law, and then generate roadway internal structure feature data. Map and fuse the above three types of feature data according to the three dimensions of stress, displacement, and strain to construct a unified three-dimensional space grid system, realize the three-dimensional geomechanical feature space description of the complex mechanical behavior of the roadway, and finally generate a three-dimensional roadway feature tensor containing multi-physical quantity coupling information. This tensor not only reflects the spatial distribution characteristics of the surrounding rock stress state, surface deformation, and support structure response, but also realizes the effective fusion and high-dimensional expression of multi-source heterogeneous data, providing accurate data basis and theoretical support for subsequent mechanical analysis, numerical simulation, and optimization design, and significantly improving the scientificity and accuracy of roadway structure health monitoring and safety assessment.

[0010] Preferably, generating the initial digital model of the roadway based on the three-dimensional roadway feature tensor in step S2 includes: Learn the spatial pattern through a preset convolutional layer based on the three-dimensional roadway feature tensor to obtain roadway spatial feature data; Perform geometric reconstruction of the surrounding rock mass on the three-dimensional roadway feature tensor to obtain a three-dimensional roadway structure sub-model; Perform geometric strain state attribute reconstruction of the concrete lining on the three-dimensional roadway feature tensor to obtain a three-dimensional support structure sub-model; Use the roadway spatial feature data to simulate the contact surface force transfer mechanism of the three-dimensional roadway structure sub-model and the three-dimensional support structure sub-model to obtain the initial digital model of the roadway.

[0011] The present invention realizes the deep learning of the multi-dimensional space mechanical characteristics in the roadway and the extraction of spatial patterns by presetting the convolutional layer structure of a multi-layer convolutional neural network based on the three-dimensional roadway feature tensor, systematically captures the complex correlation and potential spatial law between the surrounding rock mass and the support structure, generates high-dimensional roadway spatial feature data, and provides an accurate feature expression for subsequent structural modeling. At the data level, using the stress, displacement, and strain information in the three-dimensional roadway feature tensor, the structural sub-model of the three-dimensional surrounding rock mass is reconstructed through a geometric reconstruction algorithm, accurately restoring the spatial form and mechanical state of the roadway surrounding rock mass, and realizing the unified description of geometric deformation and mechanical response. At the same time, based on the strain state attribute in the tensor, the three-dimensional geometric model of the concrete lining is reconstructed, the strain distribution and local mechanical behavior inside the support structure are refined, and the three-dimensional support structure sub-model is obtained. Combining the roadway spatial feature data, the force transfer mechanism of the contact surface between the surrounding rock mass sub-model and the support structure sub-model is simulated, and through numerical calculation and mechanical coupling analysis, the interface mechanics interaction between the surrounding rock and the support structure is accurately characterized, reflecting the stress transfer path and deformation coordination, and finally generating an initial digital model of the roadway with high precision and multi-physical quantity coupling characteristics. This model fully integrates the structural information and mechanical characteristics of multi-source spatial data, improves the authenticity and detail performance of the digital model, provides a solid data foundation and theoretical support for subsequent simulation analysis, structural optimization, and safety assessment, and significantly enhances the scientificity and practicality of the digital design and management of roadway engineering.

[0012] Preferably, step S3 includes the following steps: Step S31: Use the initial digital model of the roadway to simulate the repair support plan to obtain the simulated data of the designed repair support for the roadway; Step S32: Generate physical mold parameters based on the simulated data of the designed repair support for the roadway and conduct physical model tests to obtain physical test data; Step S33: Use the physical test data and the simulated data of the designed repair support for the roadway to conduct digital-physical error analysis to obtain model compensation parameters; based on the model compensation parameters, correct the parameters of the initial digital model of the roadway to obtain an optimized digital model of the roadway.

[0013] The present invention realizes the pre - evaluation and effect prediction of different support schemes in a digital environment by performing numerical simulation of the repair support scheme based on the initial digital model of the roadway, generating detailed numerical simulation data for the repair support of the roadway design. At the data level, these simulation data are used to construct the parameters of the physical mold, guiding the implementation of the physical model test, and collecting multi - dimensional physical test data including stress, strain, and displacement, which truly reflects the mechanical response and deformation behavior of the roadway structure under the repair support conditions. Subsequently, through quantitative digital - physical error analysis of the differences between the physical test data and the digital simulation data, the error distribution characteristics are extracted and the model compensation parameters are calculated, thereby realizing the correction of the deviations existing in the digital model. This process effectively utilizes the error feedback mechanism, enabling the model compensation parameters to reflect the non - linear mechanical behavior and complex interaction effects in the real physical environment, significantly improving the accuracy and credibility of the digital model. Based on these compensation parameters, the initial digital model is parameter - corrected, and the generated optimized digital model of the roadway more accurately reflects the actual mechanical state and structural response of the roadway, enhancing the prediction ability and adaptability of the digital model to the repair support effect, and reducing the deviations caused by model simplification or assumptions. This method realizes the closed - loop integration of digital simulation and physical test at the data level. Through error - driven model optimization, it effectively improves the scientificity and practicality of the digital design of the roadway structure, provides reliable data support and theoretical basis for the optimization of subsequent support schemes, and significantly promotes the in - depth application of digital twin technology in mine roadway engineering.

[0014] Preferably, step S32 includes the following steps: Step S321: Perform scaling model size processing based on the numerical simulation data of the repair support of the roadway design to obtain the roadway scaling model parameters, where the geometric similarity ratio of the scaling model size processing is 1:50 and the stress similarity ratio is 1:75; Step S322: Extract the bolt position points of the initial digital model of the roadway to obtain the model bolt position data; based on the model bolt position data, reserve the bolt position points to obtain the hole density parameter data; Step S323: Use the roadway scaling model parameters and the hole density parameter data to conduct an equivalent boundary load physical model experiment to obtain the physical test data.

[0015] The present invention effectively realizes the proportional matching and mechanical similarity conversion between the digital model and the physical test model by performing scaling model size processing based on the simulated data of roadway design repair support. Specifically, the scaling ratios of geometric similarity ratio 1:50 and stress similarity ratio 1:75 are used to perform scale transformation on the roadway design data, ensuring a high degree of consistency between the physical model and the actual roadway in terms of spatial dimensions and stress distribution, and laying a foundation for subsequent physical experiments. At the data level, by accurately extracting the bolt position points in the initial digital model of the roadway, a dataset of model bolt positions is constructed, which contains the spatial coordinates and distribution density information of the bolts. Then, based on this data, bolt point reservations are made to generate hole density parameter data, which reflects the density and spatial distribution characteristics of the bolt arrangement. Using the above scaling model parameters and hole density parameter data, equivalent boundary load conditions are constructed in the physical experiment stage to accurately simulate the boundary mechanical loading of the physical model, ensuring a high degree of consistency between the experimental environment and the mechanical state of the actual roadway. The test data collected through the physical model experiment covers multi-dimensional mechanical response information, including stress field distribution, deformation mode, and local bolt response, forming a real physical test dataset. This dataset provides a key measured basis for subsequent digital-physical error analysis, effectively supporting the optimization and correction of digital model parameters. Overall, this method realizes a closed-loop mapping from digital simulation to physical experiment at the data level, promotes the accurate maintenance of geometric and mechanical similarity in scale transformation, improves the representativeness and reliability of physical experiment results, and provides a solid data foundation and experimental guarantee for the scientific design and optimization of roadway support structures.

[0016] Preferably, step S33 includes the following steps: Step S331: Calculate the relative displacement error using the physical test data and the simulated data of roadway design repair support to obtain the relative displacement error data of the roadway; calculate the relative strain error using the physical test data and the simulated data of roadway design repair support to obtain the relative strain error data of the roadway; Step S332: When the relative displacement error data of the roadway is greater than 10% and the relative strain error data of the roadway is less than 0.85, use the genetic algorithm to optimize the parameters of the surrounding rock constitutive model to obtain model compensation parameters; Step S333: Based on the model compensation parameters, correct the parameters of the initial digital model of the roadway to obtain the optimized digital model of the roadway.

[0017] The present invention realizes the dynamic correction between the digital model and the actual physical state by conducting a detailed error analysis on the physical test data and the simulated data of roadway design repair support. Specifically, first, the displacement data of the physical test and the digital simulation results are compared and calculated at the data level to obtain the relative error data of roadway displacement, which reflects the deviation degree between the roadway deformation predicted by the digital model and the actual deformation of the physical model. At the same time, the same comparison process is carried out on the strain data to obtain the relative error data of roadway strain, which reflects the accuracy of the simulation of the strain state of the internal materials of the support structure. Based on the above error data, a judgment threshold mechanism is set. When the relative displacement error exceeds 10% and the relative strain error is lower than 0.85, the genetic algorithm optimization process is started. As a global optimization method based on the evolutionary mechanism, the genetic algorithm encodes the constitutive model parameters of the surrounding rock as chromosomes, and through selection, crossover, and mutation operations, iteratively optimizes for the goal of minimizing errors to search for a set of optimal or approximately optimal model compensation parameters. These compensation parameters are aimed at correcting the deviations of material mechanical properties, boundary conditions, or other key parameters in the digital model to make the model more conform to the actual performance of the physical test. After obtaining the model compensation parameters, the initial digital model of the roadway is systematically adjusted based on the data-driven parameter correction method to generate an optimized digital model of the roadway with higher fitting degree and prediction accuracy. This process realizes the closed-loop feedback between the physical and digital models at the data level, enhances the adaptive ability and reliability of the model, significantly improves the accuracy and practical value of the digital model in actual engineering applications, and promotes the scientific and refined development of the design and optimization of engineering support structures.

[0018] Preferably, step S4 includes the following steps: Step S41: Predict the repair effect based on the optimized digital model of the roadway to generate support optimization prediction data; Step S42: Conduct a full-process modeling evaluation based on the support optimization prediction data and update the die design parameters to obtain the full-process roadway model modeling data; Step S43: Construct a full-process model construction report of the roadway model based on the full-process roadway model modeling data.

[0019] The present invention realizes the prospective data analysis and evaluation of the performance of the support structure by predicting the repair effect based on the optimized digital model of the roadway. Specifically, at the data level, first, through the multi-dimensional parameter input of the optimized digital model, the simulation algorithm is used to generate the support optimization prediction data, which covers the quantitative prediction results of design variables (such as bolt spacing, lining thickness, etc.) and constraint conditions (such as displacement limits), reflecting the potential performance of the repair plan in the actual engineering environment. Subsequently, based on the support optimization prediction data, a full-process modeling evaluation is carried out. This process includes the dynamic adjustment and feedback of the parameters at each stage of the model, and the real-time update of the die design parameters in combination with the prediction data, effectively ensuring the data consistency and accuracy between the physical model and the digital model, and ensuring the continuity and closed-loop control of the modeling process. Through this multi-level and multi-angle parameter adjustment and evaluation, a full-process roadway model modeling data covering design, simulation, test and feedback is formed, fully reflecting the organic integration from data acquisition, model simulation to physical test. On this basis, the full-process roadway model modeling data is further systematically summarized and analyzed to generate a full-process model construction report of the roadway model. The report includes the evolution process of the key parameters of the model, the error correction situation and the quantitative indicators of the optimization effect, providing data support and scientific basis for subsequent engineering decisions. This process ensures the accuracy and adaptability of the model through data-driven iterative optimization, while promoting the intelligence and standardization of roadway support design, and significantly improving the systematicness and transparency of the entire modeling process.

[0020] Preferably, step S41 includes: Predict the repair effect of the optimized digital model of the roadway based on a preset judgment criterion to generate support optimization prediction data, where the repair effect prediction includes design variable prediction and constraint condition prediction; When performing design variable prediction, the preset judgment criterion includes that the horizontal bolt spacing is 0.5 - 1.5 m and the lining thickness is 0.1 - 0.3 m. When the support optimization prediction data is other data, mark it as scrap data; When performing constraint condition prediction, the preset judgment criterion includes that the roof displacement is less than 50 mm and the rib displacement is less than 30 mm. When the support optimization prediction data is other data, mark it as scrap data.

[0021] The present invention realizes systematic data verification and screening of the design variables and constraint conditions of the support plan by predicting the repair effect of the roadway optimization digital model based on a preset judgment criterion, ensuring the effectiveness and reliability of the predicted data. At the data level, first, the support optimization prediction data is generated using the optimization digital model. This data contains the specific numerical information of the design variables (such as the horizontal spacing of bolts and the lining thickness) and constraint conditions (such as the roof displacement and the rib displacement), reflecting the performance characteristics of the repair plan under actual working conditions. Subsequently, according to the preset judgment criterion, the predicted data of the design variables is screened. The horizontal spacing of bolts is limited to the range of 0.5 to 1.5 meters, and the lining thickness is limited to the range of 0.1 to 0.3 meters. All predicted data outside this range is marked as scrapped data, thereby eliminating abnormal data that does not conform to the engineering reality and design specifications, and ensuring the accuracy and rationality of the data basis for subsequent analysis and application. At the same time, in the prediction of constraint conditions, threshold limits of less than 50 millimeters and 30 millimeters are respectively set for the roof and rib displacements. The predicted data exceeding this threshold is also marked as scrapped data to exclude potential data of structural instability risks and improve the prediction accuracy of the model for structural safety. This dual judgment mechanism based on a rigorous threshold system effectively realizes the dynamic verification between the design variables and constraint conditions and data quality control, enhancing the usability of the data and the reliability of the model prediction. The overall process, through strict filtering and classification of multi-dimensional data, makes the support optimization prediction data not only have engineering applicability but also statistical effectiveness, providing a solid data support for subsequent support plan optimization and risk assessment, and thus promoting the development of roadway support design towards intelligence and refinement.

[0022] In this specification, a modeling system applied to a roadway model is provided for implementing the above-mentioned modeling method applied to a roadway model. The modeling system applied to a roadway model includes: A multi-source data acquisition module, configured to deploy a multi-source sensor device array at the roadway engineering site to collect surrounding rock deformation data, roadway surface deformation point cloud data, and support structure state data, and perform data preprocessing to construct a roadway multi-source original data set; A spatio-temporal fusion and three-dimensional modeling module, configured to perform spatio-temporal fusion processing on the roadway multi-source original data set to generate a synchronized roadway data set; construct a three-dimensional geomechanical feature space using the synchronized roadway data set, and generate a three-dimensional roadway feature tensor; generate an initial roadway digital model based on the three-dimensional roadway feature tensor; A physical feedback and model correction module, configured to simulate a repair support plan using the initial roadway digital model, and perform a physical model test to obtain physical test data; perform digital-physical error analysis on the physical test data to obtain model compensation parameters; perform parameter correction on the initial roadway digital model based on the model compensation parameters to obtain an optimized roadway digital model; The modeling evaluation and report generation module is used to predict the repair effect based on the digital model of roadway optimization, generate prediction data for support optimization; conduct a full-process modeling evaluation based on the prediction data for support optimization, and generate a report to obtain the full-process model construction report of the roadway model.

[0023] The beneficial effects of the present invention are as follows: By deploying a multi-source sensor device array at the roadway engineering site, real-time acquisition of surrounding rock deformation data, roadway surface deformation point cloud data, and support structure state data is carried out, forming a data basis with multi-dimensional and multi-temporal resolution, providing detailed and comprehensive original data support for subsequent data fusion and model construction. The spatio-temporal fusion algorithm is used to synchronize the multi-source original data sets, eliminate the spatio-temporal asynchrony and dimensional differences between the data, ensure that the spatial positions and timestamps of different types of data are highly consistent, and thus construct an accurate synchronized roadway data set. On this basis, combined with the principles of geomechanics, the fused data is mapped into a three-dimensional feature space, and the high-dimensional characterization of multi-physical quantity features of complex roadway structures is realized through tensor expression, effectively capturing the coupling relationship between the surrounding rock and the support structure, and then generating an initial digital model to accurately reflect the mechanical state of the roadway. Through the combination of the digital model and physical model tests, the error analysis method is used to compare the digital and physical test data, extract model compensation parameters to correct the systematic deviations in the digital model, and achieve the optimization and precision of the digital model. Based on the optimized digital model, the effect prediction of the repair support plan is carried out, relying on the structured prediction data for support optimization, conducting a full-process modeling evaluation, systematically quantifying the feasibility and safety of the repair plan, and generating a detailed evaluation report, forming a closed-loop data-driven design and verification process. This process not only realizes the dynamic and refined monitoring of the roadway structure state, but also improves the accuracy and reliability of the model through data-driven methods, promoting the scientific and intelligent support design. Description of the Drawings

[0024] Figure 1 It is a schematic diagram of the step process of a modeling method applied to a roadway model; Figure 2 It is Figure 1 A detailed implementation step process schematic diagram of step S4 in The realization of the purpose of the present invention, functional features and advantages will be further described with reference to the embodiments and the drawings. Specific Embodiments

[0025] The technical methods of the present invention patent will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those skilled in the art within the scope of the present invention without creative work based on the embodiments of the present invention belong to the protection scope of the present invention.

[0026] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0027] It should be understood that although terms such as "first" and "second" may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0028] To achieve the above object, please refer to Figures 1 to 2 , a modeling method applied to a roadway model, the method comprising the following steps: Step S1: Deploy a multi-source sensing device array at the roadway engineering site to collect surrounding rock deformation data, roadway surface deformation point cloud data, and support structure state data, and perform data preprocessing to construct a roadway multi-source original data set; Step S2: Perform spatio-temporal fusion processing on the roadway multi-source original data set to generate a synchronized roadway data set; construct a three-dimensional geomechanical feature space using the synchronized roadway data set, and generate a three-dimensional roadway feature tensor; generate an initial roadway digital model based on the three-dimensional roadway feature tensor; Step S3: Use the initial roadway digital model to simulate the repair support plan and perform a physical model test to obtain physical test data; perform digital-physical error analysis using the physical test data to obtain model compensation parameters; correct the parameters of the initial roadway digital model based on the model compensation parameters to obtain an optimized roadway digital model; Step S4: Predict the repair effect based on the optimized roadway digital model to generate support optimization prediction data; perform a full-process modeling evaluation based on the support optimization prediction data and generate a report to obtain a full-process model construction report for the roadway model.

[0029] In an embodiment of the present invention, as shown in reference to Figure 1 , a schematic flow chart of the steps of a modeling method applied to a roadway model according to the present invention is shown. In this example, the modeling method applied to the roadway model includes the following steps: Step S1: Deploy a multi-source sensor device array at the roadway engineering site to collect surrounding rock deformation data, roadway surface deformation point cloud data, and support structure state data, and perform data preprocessing to construct a roadway multi-source original dataset; In the embodiment of the present invention, during the process of deploying the multi-source sensor device array at the roadway engineering site, the deformation behaviors of the surrounding rock and the support structure are mainly monitored with high spatio-temporal resolution by arranging multiple types of sensors, and the point cloud data of the roadway surface deformation is collected synchronously. Specifically, the surrounding rock deformation data is mainly continuously obtained relying on fiber Bragg grating strain gauges, displacement gauges or distributed optical fiber sensing systems to capture the detailed evolution during the micro-deformation process inside the surrounding rock; the roadway surface deformation point cloud data is periodically scanned based on a laser scanner (such as a 3D lidar) and a structured light system to form a multi-viewpoint cloud dataset with spatial density characteristics; the support structure state data is usually jointly collected by multiple types of sensors such as stress gauges, crack gauges and inclinometers to reflect the stress and deformation states of bolts, steel arch frames or shotcrete layers. These original data need to be unified in time stamp and coordinate system after collection. The asynchronous data sources are aligned through the time synchronization module, and the data in different coordinate spaces are standardized and transformed through the spatial projection algorithm. In addition, according to the data characteristics generated by different sensor devices, operations such as format conversion, noise filtering, missing value interpolation and outlier identification need to be carried out respectively. Among them, the point cloud data also needs to be further processed by registration, filtering and sparse reconstruction to improve the accuracy of subsequent feature extraction. Finally, all the processed data is fused and organized into a structured roadway multi-source original dataset, which retains the original measurement dimensions and spatio-temporal distribution characteristics of each sensor source and provides a unified and reusable data basis for subsequent modeling.

[0030] Step S2: Perform spatio-temporal fusion processing according to the roadway multi-source original dataset to generate a synchronized roadway dataset; construct a three-dimensional geomechanical feature space using the synchronized roadway dataset, and generate a three-dimensional roadway feature tensor; generate an initial digital model of the roadway based on the three-dimensional roadway feature tensor; In the embodiments of the present invention, after the construction of the roadway multi-source original dataset is completed, it is first necessary to perform spatio-temporal fusion processing on various types of temporal data and spatial data in the dataset. This process mainly includes two core links: time alignment and spatial reconstruction. In terms of time alignment, through the normalization of the timestamps of multi-source sensing data and the combination of linear interpolation or a non-linear alignment algorithm based on Dynamic Time Warping (DTW), the asynchronously collected data can obtain collaborative expression on a unified time scale. In terms of spatial reconstruction, a multi-source point cloud registration algorithm is adopted, such as a joint method based on Iterative Closest Point (ICP) and voxel grid filtering, to convert the spatial measurement results from different sources into the global geological coordinate system, and multi-scale spatial interpolation technology is used to fill the measurement blind spots and sparse areas, so as to generate a continuous three-dimensional space field. Subsequently, based on the fused data, a three-dimensional geomechanical feature space is constructed. This space expresses the coupling relationship between the surrounding rock stress, displacement field, structural response, and the state of the support system in the form of a tensor field. High-dimensional feature extraction techniques (such as tensor decomposition, principal component analysis, or three-dimensional convolutional kernel embedding) are used to encode the spatio-temporal data, mapping it into a three-dimensional feature tensor with a fixed structure. This feature tensor serves as the basic input for constructing the initial digital model of the roadway. Further, through a data-driven three-dimensional modeling process, such as roadway geometry reconstruction and mechanical response simulation based on a graph neural network or a spatial mapping reconstruction algorithm, an initial digital model of the roadway with structural integrity, spatial consistency, and physical logical rationality is generated, providing basic numerical support for subsequent repair plan simulation and multi-field coupling calculation.

[0031] Step S3: Use the initial digital model of the roadway to simulate the repair support plan and conduct a physical model test to obtain physical test data; use the physical test data to perform digital-physical error analysis to obtain model compensation parameters; based on the model compensation parameters, correct the parameters of the initial digital model of the roadway to obtain an optimized digital model of the roadway; In the embodiments of the present invention, after the initial digital model of the roadway is constructed, to evaluate its response ability to actual working conditions and the adaptability of the support strategy, it is necessary to carry out simulation of the repair support plan based on this model. This process relies on numerical simulation methods such as the finite element method (FEM), the discrete element method (DEM), or the coupled field strength method. By applying different types of loading boundary conditions to the model, the deformation, crack propagation, and stress evolution process of the roadway structure under repair support are simulated, and a multi-time segment mechanical response data sequence is generated. At the same time, an entity physical model test is carried out to compare the reaction of the repair support working conditions under real material and structural conditions, and physical observation data such as experimental stress-strain curves, displacement field evolution maps, and local failure mode images are obtained. To achieve the consistency analysis of the two, it is necessary to quantitatively process and feature extract the experimental data, and compare it with the spatio-temporal response data obtained during the simulation process. The deviation sources of the digital model and the key parameter mismatch intervals are identified through technical means such as error mapping analysis, residual distribution evaluation, and multi-dimensional regression fitting. According to the error analysis results, a set of model compensation parameters is extracted, including correction factors for material mechanical parameters, stiffness adjustment coefficients for support structures, and fine-tuning items for contact boundary conditions, etc., and a compensation parameter mapping table is constructed. Finally, based on this parameter table, the relevant modules in the initial digital model are corrected and updated, and local reconstruction and overall consistency checking are carried out using the parameter replacement and sub-module reconstruction methods, so as to form a corrected optimized digital model of the roadway. This optimized model not only maintains the characteristic distribution consistent with the site at the geometric and structural levels, but also has higher approximation accuracy at the response behavior level, providing a more realistically constrained model basis for the next-stage prediction analysis.

[0032] Step S4: Predict the repair effect based on the optimized digital model of the roadway to generate support optimization prediction data; conduct a full-process modeling evaluation based on the support optimization prediction data and generate a report to obtain a full-process model construction report of the roadway model.

[0033] In the embodiment of the present invention, after completing the construction of the tunnel optimization digital model, it is necessary to carry out the prediction and analysis of the repair support effect based on the model. The core is to use the numerical evolution capability of the coupling of structure, geology and support to simulate virtual loading of different repair strategies and form a complete time series response data set. The specific technical path includes grouping and arranging various support materials and structural parameters (such as anchor length, anchor angle, steel arch stiffness, shotcrete thickness, etc.) input into the optimization model to construct a multi-condition simulation sample set; for each group of parameter combinations, the structural response is predicted by the three-dimensional finite element method (FEM) or multi-field coupling simulation method, and the output includes multi-dimensional spatiotemporal response indicators such as surrounding rock stress redistribution, support structure bearing capacity attenuation trend and crack expansion morphology. On this basis, the prediction results are normalized, and the response data indicator set under typical support conditions is extracted to form support optimization prediction data. Subsequently, the predicted data was compared with the initial support condition response results, and the full-process optimization effect was quantitatively evaluated by constructing an evaluation function system (such as the support safety factor improvement rate, residual deformation reduction rate, plastic zone expansion rate, etc.), and error backtracking and sensitivity analysis were performed to determine the influence weight of different parameters on the repair effect. After the full-process modeling evaluation is completed, the tunnel modeling analysis report is generated through the automatic template generation system in combination with the simulation parameters, response curves, evaluation indicators and model evolution record data. The report presents the model building process, key input parameters, optimized simulation path, predicted response data and evaluation indicator maps in a structured manner, and is accompanied by model stability analysis and numerical accuracy evaluation results, forming a systematic tunnel full-process model construction and evaluation document data set.

[0034] Preferably, step S1 comprises the following steps: Step S11: deploying fiber grating sensor arrays on the top and sides of the tunnel, with a sampling frequency of 10 Hz, a measurement range of 0-50 MPa, and an accuracy of ±0.1 MPa, to collect surrounding rock deformation data; Step S12: Arrange a laser scanner on the tunnel surface with a scanning accuracy of 0.1 mm, and collect deformation point cloud data of the tunnel surface at intervals of 30 minutes; Step S13: embedding strain sensors inside the anchor rods and lining structures, with a sampling frequency of 5 Hz, to collect support structure status data; Step S14: using wavelet transform to perform noise reduction processing on surrounding rock deformation data, tunnel surface deformation point cloud data and support structure status data to obtain a multi-source tunnel noise reduction data set; Step S15: performing data cleaning on the multi-source lane denoising dataset to generate a lane multi-source original dataset.

[0035] In the embodiment of the present invention, in the process of acquiring and processing the monitoring data of the tunnel, firstly, an array of fiber grating sensors is arranged on the top plate and both sides of the tunnel, and the stress change data of the surrounding rock is collected in real time through the principle of wavelength modulation. The operating frequency of the sensor is set to 10 Hz, which can cover the measurement range of 0 to 50 MPa, and the measurement accuracy reaches ±0.1 MPa, which can dynamically track the stress state of the surrounding rock; at the same time, a high-precision laser scanner is arranged on the surface of the tunnel, and a planar array scanning method is used to perform periodic sampling at a time interval of 30 minutes. The scanning accuracy is controlled at 0.1 mm, and the output data is a dense spatial point cloud collection, which records the geometric characteristics of the deformation evolution of the tunnel surface; in addition, resistive strain sensors are embedded in support structures such as anchor rods and concrete linings, and the sampling frequency is set to 5 Hz, the range is ±5000 microstrain, and the accuracy is controlled at ±5 microstrain, which is used to obtain subtle stress change information inside the support structure. After the data of these three types of sensors are collected, they need to enter the signal preprocessing stage respectively. Wavelet transform is used as the main noise reduction algorithm, and its local analysis ability in the time and frequency domain is used to perform multi-scale decomposition and reconstruction of high-frequency interference in non-stationary signals, so as to remove high-frequency noise in the surrounding rock deformation time series, outliers and ranging errors in point cloud data, and random disturbances in strain data. After the noise reduction process is completed, the multi-source data set is cleaned, mainly including missing value identification and interpolation, data format standardization, timestamp synchronization and sensor drift correction, so that all data can be unified and standardized in structure. Finally, the surrounding rock stress data, surface point cloud deformation data and structural strain data processed by wavelet denoising and cleaning are integrated into a structured multi-source original data set of the tunnel, providing a clear, stable and high-precision data foundation for subsequent spatiotemporal fusion modeling.

[0036] It is particularly important that the spatiotemporal fusion processing of the multi-source original data set of the lane in step S2 includes: The multi-source original data sets were time-aligned with a time resolution of 0.1s using the cubic spline interpolation method to obtain multi-source time-aligned data. Perform timestamp synchronization according to multi-source time alignment data to obtain multi-source timestamp synchronization data; Spatial alignment is performed based on multi-source time-stamp synchronization data to obtain a synchronized lane dataset.

[0037] In the embodiments of the present invention, for the multi-source original data set collected on-site in the roadway, the sources of which include fiber Bragg grating strain data, laser point cloud deformation data, and the support structure state data collected by embedded sensors, there are problems such as inconsistent sampling frequencies, different starting points of time collection, and differences in data frame lengths. Therefore, unified time and space fusion processing must be carried out. To achieve consistency in the time domain, the cubic spline interpolation method is used to perform equal-time-step interpolation resampling operations on various data channels. Specifically, the time resolution is set to 0.1 s, so that all types of data have the same sampling point structure on the unified time axis, and the missing segments or asynchronous segments are smoothly complemented, thereby obtaining multi-source time-aligned data. Subsequently, for the inherent acquisition time offset problem of different data channels, timestamp synchronization processing is further carried out on the basis of the multi-source time-aligned data, that is, the local acquisition timestamp attached to each type of data is offset-corrected with the global reference timestamp, and the time difference minimization algorithm is used to align the starting times and update frequencies of multiple channels, thereby generating multi-source timestamp-synchronized data under the unified reference frame structure. After completing the time synchronization, to achieve spatial consistency fusion, it is necessary to perform coordinate transformation and registration on the original spatial position markings according to the spatial reference system of each type of data and its mapping relationship in the roadway coordinate system. Through the rigid transformation matrix and spatial rotation correction, the unified spatial matching of various types of data in the three-dimensional coordinates is completed. Finally, the multi-source data that has been time-aligned, timestamp-synchronized, and spatially registered is integrated to construct a synchronized roadway data set with consistent structure and continuous time and space, providing a stable data basis for the subsequent construction of three-dimensional feature tensors and the reconstruction of the roadway digital model.

[0038] Preferably, in step S2, constructing a three-dimensional geomechanical feature space by using the synchronized roadway data set and generating a roadway feature tensor includes: Extracting the stress distribution according to the synchronized roadway data set, marking the maximum and minimum values of the rock stress, and generating roadway rock stress feature data; Performing a convergence analysis on the displacement amount of the roadway surface points according to the synchronized roadway data set, measuring the change in the surface bending degree, and generating roadway point cloud feature data; Performing an axis change analysis on the axis of the support structure according to the synchronized roadway data set, calculating the internal shear deformation of the structure, and generating roadway internal structure feature data; Constructing a three-dimensional geomechanical feature space by mapping the roadway rock stress feature data, roadway point cloud feature data, and roadway internal structure feature data to a three-dimensional space grid according to stress × displacement × strain, and generating a three-dimensional roadway feature tensor.

[0039] In the embodiments of the present invention, on the basis of synchronizing the roadway dataset, it is first necessary to extract the surrounding rock stress distribution with high precision. This process is based on the time-series stress data at each measuring point. By using spatial interpolation algorithms (such as the inverse distance weighting method and the Kriging interpolation method), a continuous stress distribution field is constructed in a three-dimensional coordinate system, and the maximum principal stress direction is combined with the local gradient calculation method to identify the stress concentration areas and weak zones. Furthermore, the maximum and minimum values of the rock stress are marked, and the specific positions, directions, and amplitudes of the maximum stress point and the minimum stress point in the marked area are generated to obtain the roadway rock stress characteristic data. At the same time, for the laser scanning point cloud data of the roadway surface, the point displacement differential analysis method is used to quantify the deformation of each observation point by calculating the coordinate change amount of the same point at different time nodes. Then, the method based on Bessel surface fitting or polynomial curvature analysis is used to fit and measure the change of the roadway surface contour, and the change distribution of the surface bending degree per unit area is obtained, and the roadway point cloud characteristic data is output. At the structural level, combined with the internal strain data of the bolt and lining structures, a structural deformation path model based on the axis is constructed. The axis displacement and rotation angle changes are quantified by the Fourier difference analysis method or the generalized least squares method to complete the axis deviation (preventing axis change) analysis. Further, based on the strain gradient, the shear strain distribution on the structural cross-section is calculated to form the roadway internal structure characteristic data. Finally, the above three types of characteristic data are mapped into a unified three-dimensional space grid system, and the tensor mapping relationship of the stress, displacement, and strain ternary data is established, where each grid node contains a multi-field coupling eigenvector expressed in tensor form. Through tensor splicing and spatial normalization processing, a three-dimensional geomechanical characteristic space is constructed, and a three-dimensional roadway characteristic tensor with a unified structure is output, providing data-driven support for subsequent digital modeling and mechanical response analysis.

[0040] Preferably, generating the initial roadway digital model based on the three-dimensional roadway characteristic tensor in step S2 includes: Learning the spatial pattern through a preset convolutional layer based on the three-dimensional roadway characteristic tensor to obtain the roadway spatial characteristic data; Performing geometric reconstruction of the surrounding rock mass on the three-dimensional roadway characteristic tensor to obtain a three-dimensional roadway structure sub-model; Performing geometric strain state attribute reconstruction of the concrete lining on the three-dimensional roadway characteristic tensor to obtain a three-dimensional support structure sub-model; Using the roadway spatial characteristic data to simulate the contact surface force transfer mechanism of the three-dimensional roadway structure sub-model and the three-dimensional support structure sub-model to obtain the initial roadway digital model.

[0041] In the embodiments of the present invention, after the construction of the three-dimensional roadway feature tensor is completed, it is necessary to perform a structural analysis on it to support the construction of the digital model. First, a multi-layer three-dimensional convolutional neural network (3D-CNN) is set to perform spatial feature learning on the tensor data. The convolution operation takes the spatial local area of the feature tensor as the input, and extracts the spatial local stress, displacement, and strain coupling modes on multiple channels through the kernel function. After operations such as ReLU activation and max pooling, high-dimensional roadway spatial feature data is formed, which is used to characterize the spatial distribution law of the structural response. Subsequently, according to the data channels related to the surrounding rock mass in the three-dimensional feature tensor, a voxel-based geometric reconstruction method is used to perform three-dimensional reconstruction on the surrounding rock mass. Specifically, the isosurface extraction (such as the Marching Cubes algorithm) or the reconstruction technology based on implicit functions is used to extract the boundary of the surrounding rock mass from the dense tensor field, generating a sub-model of the three-dimensional roadway structure body to completely express its geometric shape and material property distribution. For the concrete lining structure, the area containing high-gradient strain data in the tensor is selected, and through the decomposition of the deformation gradient tensor and the inversion calculation of the material constitutive relation, the geometric strain state attributes of the lining structure under different loading conditions are reconstructed, and a sub-model of the three-dimensional support structure is generated in combination with engineering primitives (such as ellipsoid fitting and tetrahedral meshing). After obtaining the structure body model and the support model, using the contact state information and mechanical property parameters contained in the spatial feature data, the modeling of the contact surface force transfer mechanism between the two is carried out. This modeling uses the surface coupling method based on finite element contact analysis, sets the contact stiffness, friction coefficient, and slip boundary conditions, numerically solves the normal force and tangential force of the contact nodes between different structures, and performs dynamic equilibrium verification to construct its complete contact force field distribution. Finally, the structure body model, the support structure model, and the mechanical relationship of the contact surface are integrated to form the initial digital model of the roadway. This model has a complete geometric configuration, material response characteristics, and mechanical interaction information, and is used in subsequent mechanical simulations and repair strategy simulations in the form of volume data or graph structures.

[0042] Preferably, step S3 includes the following steps: Step S31: Use the initial digital model of the roadway to simulate the repair support plan to obtain the simulated data of the roadway design repair support; Step S32: Generate physical mold parameters based on the simulated data of the roadway design repair support, and conduct a physical model test to obtain physical test data; Step S33: Use the physical test data and the simulated data of the roadway design repair support to perform digital-physical error analysis to obtain model compensation parameters; based on the model compensation parameters, correct the parameters of the initial digital model of the roadway to obtain the optimized digital model of the roadway.

[0043] In the embodiments of the present invention, after the initial digital model of the roadway is constructed, it is first necessary to carry out virtual simulation of the repair support plan based on this model. Specifically, by setting multiple sets of support plan parameter sets, covering multiple design variables such as bolt types and layout methods, arch frame stiffness, shotcrete thickness, etc., and using the three-dimensional stress-strain-displacement tensor field in the digital model as the input boundary condition, the multi-physics coupling finite element simulation method is used to model and calculate the mechanical response process under each support plan, and the roadway design repair support simulation data including the redistribution of surrounding rock stress, the change of the stress state of the support structure, and the evolution of the plastic zone is output. Subsequently, according to the key structural parameters and boundary loading conditions extracted from the simulation data, the die parameters of the scaled model for physical experiments are automatically generated by using the CAD / CAE integrated modeling method, including the geometric surface of the support structure, the material layer structure, and the loading path, etc. Controlling the dimensional accuracy and stress equivalence principle, the corresponding scaled structure loading experiment is completed on the physical experiment platform, and the collected stress, displacement, and crack propagation data form the physical experiment data. After obtaining the above two types of data, by establishing the corresponding relationship of spatial grid nodes and the matching function of the same type of physical quantities, digital-physical error analysis is performed. The root mean square error (RMSE), structural similarity (SSIM), and maximum error point difference degree and other indicators are mainly used to quantitatively evaluate the response differences between the two at the key monitoring points and key loading paths. The error analysis results are used to construct the residual field mapping, extract the spatial distribution characteristics of the model deviation, and calculate the compensation parameter set required for the digital model through the backpropagation optimization algorithm, covering the material stiffness correction coefficient, boundary condition adjustment factor, and local grid refinement compensation parameter, etc. Finally, the compensation parameters are fed back to the initial digital model of the roadway for global or local parameter update, and its mechanical response characteristics are recalibrated to obtain a roadway optimized digital model that is more in line with the actual engineering response.

[0044] Particularly importantly, step S31 includes the following steps: Step S311: Load the boundary conditions using the initial digital model of the roadway to obtain an equivalent mechanical environment; Step S312: Use the equivalent mechanical environment for parametric definition to generate a support simulation plan space; Step S313: Based on the support simulation plan space, perform multi-physics coupling simulation and conduct simulation safety criterion evaluation to obtain the roadway design repair support simulation data.

[0045] In the embodiments of the present invention, boundary condition loading processing is performed based on the constructed initial digital model of the roadway. Specifically, key boundary inputs need to be determined according to the roadway geological environment, surrounding rock category, historical stress evolution data, and on-site monitoring data, including the vertical load at the top, lateral ground pressure, seismic motion simulation disturbance, and the contact friction characteristics between the structure and the surrounding rock, etc. By embedding these mechanical boundary conditions into the initial roadway model in the form of tensor boundary constraints, an equivalent mechanical environment with engineering representativeness is constructed. After obtaining this equivalent environment, further through the parameter space definition method, a support parameter vector space is constructed around dimensions such as bolt length, spacing, angle, anchoring strength, lining thickness, and material modulus. The Latin hypercube sampling, Sobol sequence, or uniform design method is used to generate a high-dimensional support simulation parameter set, forming a support simulation scheme space. Subsequently, a multi-physics field simulation platform, such as FLAC3D, ABAQUS, or UDEC, etc., is used to perform coupled simulation calculations for each parameter combination. The simulation process covers physical behaviors such as non-linear deformation of the surrounding rock, contact at the bolt-lining interface, material fracture, and yield criteria. The simulation results generate multi-dimensional output data including vertical displacement of the roof, horizontal convergence of the rib, bolt stress distribution, and strain nephogram of the lining structure. Based on such output results, using set simulation safety criteria, such as the maximum displacement limit of the roof, the strain limit of the lining, or the bolt yield rate, etc., each set of simulation outputs is quantitatively evaluated, the results that do not meet the engineering safety requirements are eliminated, and the compliant samples are retained as "simulation data for the repair support design of the roadway". This data set provides highly reliable inputs with physical consistency for subsequent physical test modeling and error compensation.

[0046] Preferably, step S32 includes the following steps: Step S321: Perform scaling model size processing based on the simulation data for the repair support design of the roadway to obtain the roadway scaling model parameters, where the geometric similarity ratio for the scaling model size processing is 1:50 and the stress similarity ratio is 1:75; Step S322: Extract the bolt position points from the initial digital model of the roadway to obtain the model bolt position data; based on the model bolt position data, reserve the bolt positions to obtain the hole density parameter data; Step S323: Use the roadway scaling model parameters and the hole density parameter data to conduct an equivalent boundary load physical model experiment to obtain physical test data.

[0047] In the embodiments of the present invention, based on the simulated data of roadway design repair support, the digital model is first subjected to scale conversion to meet the scale requirements of physical experiments. In the specific implementation process, a geometric similarity ratio of 1:50 and a stress similarity ratio of 1:75 are adopted. According to the similarity theory, the three-dimensional geometric dimensions and mechanical loads are proportionally adjusted to ensure the equivalence of the scaled model in terms of spatial structure and mechanical behavior. The geometric similarity ratio conversion is achieved by linearly scaling the three-dimensional grid node coordinates of the roadway digital model to adjust the model size to meet the physical limitations of the experimental platform. At the same time, the stress similarity ratio is processed based on the similarity principle of material mechanics. By adjusting the amplitude and distribution pattern of the loading boundary conditions, the similarity of the stress field is maintained, ensuring the effective representativeness of the experimental data. Subsequently, for the bolt configuration, using the spatial arrangement information of the bolts in the initial roadway digital model, a point cloud extraction and spatial indexing algorithm is used to accurately identify the bolt position points, forming a bolt position data set. Based on this data set, hole position reservation calculations are carried out. The spatial spacing between each bolt and its distribution law are calculated using the hole position density analysis method, generating hole position density parameter data to ensure that the bolt layout in the physical model can truly reflect the density and distribution characteristics of the actual support layout. Finally, the scaled geometric model parameters and hole position density parameters are input into the physical experimental device. Based on the experimental scheme of equivalent boundary loads, a multi-point loading system is used to apply representative mechanical boundary conditions to the model, and the stress, deformation, and failure data of the model during the loading process are collected in real time, forming a complete physical experimental data set to support subsequent digital-physical error analysis and model optimization.

[0048] Preferably, step S33 includes the following steps: Step S331: Calculate the relative displacement error using the physical experiment data and the simulated data of roadway design repair support to obtain the relative displacement error data of the roadway; calculate the relative strain error using the physical experiment data and the simulated data of roadway design repair support to obtain the relative strain error data of the roadway; Step S332: When the relative displacement error data of the roadway is greater than 10% and the relative strain error data of the roadway is less than 0.85, use the genetic algorithm to optimize the parameters of the surrounding rock constitutive model to obtain the model compensation parameters; Step S333: Based on the model compensation parameters, correct the parameters of the initial roadway digital model to obtain the optimized roadway digital model.

[0049] In the embodiments of the present invention, in this step, first, based on the physical test data and the simulated data of the roadway design repair support, the corresponding displacement and strain measurement values are extracted respectively, and the relative displacement error and relative strain error are calculated through comparative analysis. Specifically, the relative displacement error is calculated by using the ratio of the absolute value of the point-by-point displacement difference to the physical test displacement value, forming an error distribution data set containing all monitoring points; the relative strain error is calculated by the ratio of the simulated and test data of the corresponding strain sensing points, obtaining the statistical index of the overall strain error, and further reflecting the spatial distribution characteristics in the form of an error matrix. Subsequently, under the determination condition that the relative displacement error exceeds 10% and the relative strain error index is lower than 0.85, a parameter optimization method based on the genetic algorithm is used to correct the parameters of the surrounding rock constitutive model. The genetic algorithm uses the constitutive model parameters as gene encoding, sets the initial population, uses the error function as the fitness evaluation index, and through selection, crossover and mutation operations, iteratively optimizes to reduce the error between the simulated data and the physical data, and finally obtains a set of optimal compensation parameters, including elastic modulus, Poisson's ratio and plastic yield criterion correction coefficient, etc. Finally, the compensation parameters are fed back to the initial digital model of the roadway, the parameter correction and update are carried out on the description of its mechanical behavior, the material constitutive relationship and boundary conditions are adjusted, so as to improve the fitting accuracy of the digital model to the actual engineering behavior, form an optimized digital model of the roadway after error correction and parameter optimization, have a more accurate mechanical response prediction ability, and support subsequent structural safety assessment and design optimization work.

[0050] As an example of the present invention, refer to Figure 2 As shown, in this example, step S4 includes: Step S41: Perform temperature-strain-pressure cross-sensitivity analysis on the physical quantity characteristic matrix, and perform model parameter fusion to obtain model parameter fusion data; Step S42: Perform backpropagation fine-tuning processing on the model parameter fusion data to obtain backpropagation parameter fine-tuning data; judge the decoupling accuracy according to the backpropagation parameter fine-tuning data. When the decoupling accuracy is less than 1.5%, quantize the backpropagation parameter fine-tuning data according to 8-bit fixed-point, and compress the number of parameters to less than 300,000 to obtain model lightweight parameter data; Step S43: Use the model lightweight parameter data to construct a MEMS fiber optic multi-parameter demodulation model.

[0051] In the embodiments of the present invention, for the collected physical quantity feature matrix, which covers various sensing data such as temperature, strain, and pressure, cross-sensitivity analysis is performed through multivariate statistical analysis methods. Specifically, covariance matrix calculation and principal component analysis (PCA) techniques are used to identify the degree of mutual influence and redundant information among various physical quantities. Based on the cross-sensitivity results, model parameter fusion is implemented. The weighted fusion algorithm is used to combine the sensitivity weights of each sensing channel to integrate the multi-source data features and generate unified model parameter fusion data, which has the ability to comprehensively express multi-parameter information. Subsequently, for the model parameter fusion data, fine-tuning processing is performed using the backpropagation algorithm based on gradient descent. By constructing an error feedback mechanism, the model parameters are optimized, and the prediction error of the model on the training set is gradually reduced to obtain the reverse parameter fine-tuning data. The decoupling accuracy of this data is evaluated, and the multi-parameter decoupling error is calculated using the error analysis method. When the error index is lower than 1.5%, it indicates that the independence and accuracy of the model parameters meet the predetermined requirements. At this time, model lightweight processing is implemented. The reverse parameter fine-tuning data is numerically precision-compressed through 8-bit fixed-point quantization technology, and the pruning algorithm is combined to reduce redundant parameters to ensure that the overall parameter scale is compressed to less than 300,000, thereby achieving the optimization of computing resources and the requirements of model deployment. Finally, based on the lightweight parameter data with compressed and satisfactory accuracy, a MEMS fiber multi-parameter demodulation model is constructed, which can achieve efficient multi-physical quantity signal analysis and real-time demodulation, providing a solid data foundation and algorithm support for the accurate decoding and application of subsequent sensing data.

[0052] Preferably, step S41 includes: Predict the repair effect of the roadway optimization digital model based on a preset judgment criterion to generate support optimization prediction data, where the repair effect prediction includes design variable prediction and constraint condition prediction; When performing design variable prediction, the preset judgment criterion includes that the horizontal spacing of bolts is 0.5 - 1.5 m and the lining thickness is 0.1 - 0.3 m. When the support optimization prediction data is other data, it is marked as scrap data; When performing constraint condition prediction, the preset judgment criterion includes that the roof displacement is less than 50 mm and the rib displacement is less than 30 mm. When the support optimization prediction data is other data, it is marked as scrap data.

[0053] In the embodiments of the present invention, based on the optimized digital model of the roadway, the repair effect is systematically predicted through a preset judgment criterion to generate support optimization prediction data. In the design variable prediction stage, by limiting the numerical ranges of key design parameters in the model, such as the horizontal spacing of bolts and the lining thickness, specifically setting the horizontal spacing of bolts between 0.5 m and 1.5 m, and the lining thickness between 0.1 m and 0.3 m, the interval constraint method is used to screen the simulated output data. When the support optimization prediction data obtained by simulation exceeds the above design variable range, the system automatically marks this data as scrap data to exclude abnormal results that do not meet the design specifications. In the constraint condition prediction stage, aiming at the displacement control requirements of the roadway structure, the maximum allowable displacement of the roof is set to 50 mm, and the maximum allowable displacement of the rib is set to 30 mm. Based on the time series data of the structural displacement, threshold comparison and judgment are carried out. When the predicted displacement data exceeds the corresponding threshold, it is determined that the constraint condition is violated and is also marked as scrap data. During the data processing process, a rule judgment algorithm is used to implement the compliance verification of the prediction data, and combined with a logical screening mechanism, the abnormal data and the compliant data are classified and managed to form a structured support optimization prediction data set. This process ensures the accurate filtering and quality control of the output results of the digital model, supports the subsequent full-process modeling evaluation, and provides an effective data basis that meets the engineering standards for report generation.

[0054] In this specification, a modeling system applied to a roadway model is provided for implementing the above-mentioned modeling method applied to a roadway model. The modeling system applied to a roadway model includes: A multi-source data acquisition module, which is used to deploy a multi-source sensor device array at the roadway engineering site to collect surrounding rock deformation data, roadway surface deformation point cloud data, and support structure state data, and perform data preprocessing to construct a roadway multi-source original data set; A spatio-temporal fusion and 3D modeling module, which is used to perform spatio-temporal fusion processing on the roadway multi-source original data set to generate a synchronized roadway data set; use the synchronized roadway data set to construct a 3D geomechanical feature space, and generate a 3D roadway feature tensor; generate an initial digital model of the roadway based on the 3D roadway feature tensor; A physical feedback and model correction module, which is used to simulate the repair support plan using the initial digital model of the roadway and perform a physical model test to obtain physical test data; perform digital-physical error analysis using the physical test data to obtain model compensation parameters; perform parameter correction on the initial digital model of the roadway based on the model compensation parameters to obtain an optimized digital model of the roadway; A modeling evaluation and report generation module, which is used to predict the repair effect based on the optimized digital model of the roadway to generate support optimization prediction data; perform a full-process modeling evaluation based on the support optimization prediction data, and perform report generation to obtain a full-process model construction report of the roadway model.

[0055] The beneficial effects of the present invention are as follows. By deploying a multi-source sensor device array at the roadway engineering site, the surrounding rock deformation data, the roadway surface deformation point cloud data, and the support structure state data are collected in real time, forming a multi-dimensional and multi-temporal resolution data basis, providing detailed and comprehensive original data support for subsequent data fusion and model construction. The spatio-temporal fusion algorithm is used to synchronize the multi-source original data sets, eliminating the spatio-temporal asynchrony and dimensional differences between the data, ensuring that the spatial positions and timestamps of different types of data are highly consistent, thereby constructing an accurate synchronized roadway data set. On this basis, combined with the geomechanics principle, the fused data is mapped into a three-dimensional feature space, and the multi-physical quantity characteristics of the complex roadway structure are characterized in high dimensions through tensor expression, effectively capturing the coupling relationship between the surrounding rock and the support structure, and then generating an initial digital model to accurately reflect the mechanical state of the roadway. By combining the digital model with the physical model test, the error analysis method is used to compare the digital and physical test data, extract the model compensation parameters to correct the systematic deviation in the digital model, and realize the optimization and precision of the digital model. Based on the optimized digital model, the effect prediction of the repair support scheme is carried out. Relying on the structured support optimization prediction data, the whole process modeling evaluation is carried out, the feasibility and safety of the repair scheme are systematically quantified, and a detailed evaluation report is generated, forming a closed-loop data-driven design and verification process. This process not only realizes the dynamic and refined monitoring of the roadway structure state, but also improves the accuracy and reliability of the model through data-driven methods, promoting the scientific and intelligent support design.

[0056] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes within the meaning and scope of the equivalent elements of the application document within the present invention.

[0057] The above are only the specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A modeling method applied to a roadway model, characterized in that, The following steps are involved: Step S1: deploying a multi-source sensor array at the tunnel engineering site to collect surrounding rock deformation data, tunnel surface deformation point cloud data and support structure status data, and performing data preprocessing to construct a multi-source original data set for the tunnel; Step S2: performing spatiotemporal fusion processing on the multi-source original data sets of the tunnel to generate a synchronized tunnel data set; constructing a three-dimensional geomechanical feature space using the synchronized tunnel data set, and generating a three-dimensional tunnel feature tensor; Generate an initial digital model of the tunnel based on the three-dimensional tunnel characteristic tensor; Step S3: using the initial digital model of the tunnel to simulate the repair support scheme, and conducting a physical model test to obtain physical test data; using the physical test data to conduct a digital-physical error analysis to obtain model compensation parameters; based on the model compensation parameters, the initial digital model of the tunnel is parameter-corrected to obtain an optimized digital model of the tunnel; Step S4: Predict the repair effect based on the tunnel optimization digital model and generate support optimization prediction data; perform full-process modeling evaluation based on the support optimization prediction data, and generate a report to obtain a full-process model construction report for the tunnel model.

2. The modeling method applied to the roadway model according to claim 1, characterized in that Step S1 includes the following steps: Step S11: deploying fiber grating sensor arrays on the top and sides of the tunnel, with a sampling frequency of 10 Hz, a measurement range of 0-50 MPa, and an accuracy of ±0.1 MPa, to collect surrounding rock deformation data; Step S12: Arrange a laser scanner on the tunnel surface with a scanning accuracy of 0.1 mm, and collect deformation point cloud data of the tunnel surface at intervals of 30 minutes; Step S13: embedding strain sensors inside the anchor rods and lining structures, with a sampling frequency of 5 Hz, to collect support structure status data; Step S14: using wavelet transform to perform noise reduction processing on surrounding rock deformation data, tunnel surface deformation point cloud data and support structure status data to obtain a multi-source tunnel noise reduction data set; Step S15: performing data cleaning on the multi-source lane denoising dataset to generate a lane multi-source original dataset.

3. The modeling method applied to the roadway model according to claim 1, wherein In step S2, the synchronized tunnel data set is used to construct a three-dimensional geomechanical feature space, and the tunnel feature tensor is generated, including: Extract stress distribution based on synchronized tunnel data set, mark rock stress maximum value, and generate tunnel rock stress characteristic data; Based on the synchronized roadway data set, the roadway surface point displacement convergence analysis is performed, and the surface curvature change measurement is performed to generate roadway point cloud feature data; According to the synchronized tunnel data set, the support structure axis line deformation analysis is carried out, and the internal shear deformation calculation of the structure is carried out to generate the internal structural characteristic data of the tunnel; The tunnel rock stress characteristic data, tunnel point cloud characteristic data and tunnel internal structure characteristic data are mapped to the three-dimensional space grid according to stress × displacement × strain to construct the three-dimensional geomechanical characteristic space and generate the three-dimensional tunnel characteristic tensor.

4. The modeling method applied to the roadway model according to claim 1, characterized in that, Generating the initial digital model of the tunnel based on the three-dimensional tunnel feature tensor in step S2 includes: Based on the three-dimensional lane feature tensor, the spatial pattern is learned through a preset convolution layer to obtain the lane spatial feature data; The three-dimensional tunnel characteristic tensor is used to reconstruct the surrounding rock mass geometry to obtain a three-dimensional tunnel structure volume model; Reconstruct the geometric strain state attributes of the three-dimensional roadway feature tensor for the concrete lining to obtain a sub-model of the three-dimensional support structure. Use the roadway spatial feature data to simulate the contact surface force transfer mechanism of the sub-model of the three-dimensional roadway structure body and the sub-model of the three-dimensional support structure to obtain the initial digital model of the roadway.

5. The modeling method applied to the roadway model according to claim 1, characterized in that Step S3 includes the following steps: Step S31: Use the initial digital model of the roadway to simulate the repair support plan to obtain the simulated data of the designed repair support of the roadway. Step S32: Generate physical mold parameters based on the simulated data of the designed repair support of the roadway and conduct a physical model test to obtain physical test data. Step S33: Use the physical test data and the simulated data of the designed repair support of the roadway to conduct digital-physical error analysis to obtain model compensation parameters; correct the parameters of the initial digital model of the roadway based on the model compensation parameters to obtain the optimized digital model of the roadway.

6. The modeling method applied to the roadway model according to claim 5, characterized in that Step S32 includes the following steps: Step S321: Process the size of the scaled model based on the simulated data of the designed repair support of the roadway to obtain the parameters of the scaled model of the roadway, where the geometric similarity ratio of the scaled model size processing is 1:50 and the stress similarity ratio is 1:

75. Step S322: Extract the bolt position points of the initial digital model of the roadway to obtain the model bolt position data; reserve bolt positions based on the model bolt position data to obtain the hole density parameter data. Step S323: Conduct an equivalent boundary load physical model experiment using the parameters of the scaled model of the roadway and the hole density parameter data to obtain physical test data.

7. The modeling method applied to the roadway model according to claim 5, wherein Step S33 includes the following steps: Step S331: Use the physical test data and the simulated data of the designed repair support of the roadway to calculate the relative displacement error to obtain the relative displacement error data of the roadway; use the physical test data and the simulated data of the designed repair support of the roadway to calculate the relative strain error to obtain the relative strain error data of the roadway. Step S332: When the relative displacement error data of the roadway is greater than 10% and the relative strain error data of the roadway is less than 0.85, use the genetic algorithm to optimize the parameters of the surrounding rock constitutive model to obtain model compensation parameters. Step S333: Correct the parameters of the initial digital model of the roadway based on the model compensation parameters to obtain the optimized digital model of the roadway.

8. The modeling method applied to the roadway model according to claim 1, characterized in that Step S4 includes the following steps: Step S41: Predict the repair effect based on the optimized digital model of the roadway to generate the predicted data of the optimized support. Step S42: Evaluate the whole process modeling based on the predicted data of the optimized support and update the mold design parameters to obtain the modeling data of the whole process roadway model. Step S43: Construct a report on the construction of the whole process model of the roadway model based on the modeling data of the whole process roadway model.

9. The modeling method applied to the roadway model according to claim 8, wherein Step S41 includes: Predict the repair effect of the optimized digital model of the roadway based on the preset judgment criteria to generate the predicted data of the optimized support, where the repair effect prediction includes the prediction of design variables and the prediction of constraint conditions. When predicting the design variables, the preset judgment criteria include that the horizontal spacing of bolts is 0.5 - 1.5 m and the lining thickness is 0.1 - 0.3 m. When the predicted data of the optimized support is other data, mark it as scrap data. When making constraint condition predictions, the preset judgment criteria include that the roof displacement is less than 50 mm and the rib displacement is less than 30 mm. When the support optimization prediction data is other data, it is marked as scrap data.

10. A modeling system applied to a roadway model, characterized in that, For implementing the modeling method applied to the roadway model as described in claim 1, the modeling system applied to the roadway model includes: A multi-source data acquisition module, which is used to deploy a multi-source sensor device array at the roadway engineering site to collect surrounding rock deformation data, roadway surface deformation point cloud data, and support structure state data, and perform data preprocessing to construct a roadway multi-source original data set; A spatio-temporal fusion and 3D modeling module, which is used to perform spatio-temporal fusion processing on the roadway multi-source original data set to generate a synchronized roadway data set; use the synchronized roadway data set to construct a 3D geomechanical feature space and generate a 3D roadway feature tensor; generate an initial digital model of the roadway based on the 3D roadway feature tensor; A physical feedback and model correction module, which is used to simulate the repair support plan using the initial digital model of the roadway and conduct physical model tests to obtain physical test data; perform digital-physical error analysis using the physical test data to obtain model compensation parameters; correct the parameters of the initial digital model of the roadway based on the model compensation parameters to obtain an optimized digital model of the roadway; A modeling evaluation and report generation module, which is used to predict the repair effect based on the optimized digital model of the roadway to generate support optimization prediction data; conduct a full-process modeling evaluation based on the support optimization prediction data and generate a report to obtain a full-process model construction report of the roadway model.

Images