A method and system for establishing a parametric tunnel model based on cloud computing

Abstract

The present invention relates to the field of model construction technology, and in particular to a method and system for establishing a parametric tunnel model based on cloud computing. The method comprises the following steps: obtaining the tunnel section type and extracting structural semantics to obtain a section feature description set; constructing a parameter dictionary of the tunnel section based on the section feature description set to obtain a dynamic parameter dictionary data set; and modeling the tunnel section geometry based on the dynamic parameter dictionary data set to obtain an initial three-dimensional tunnel geometry model. The present invention achieves efficient automated modeling and precise simulation optimization of the three-dimensional geometry model of the tunnel section through dynamic parameter management, a self-correcting mechanism, and Latin hypercube sampling technology. At the same time, through unified encapsulation and authority control, it achieves safe and efficient management and collaborative sharing of cloud-based model resources, comprehensively improving the intelligence level and engineering efficiency of tunnel model construction and application.

Description

Technical Field

[0001] The present invention relates to the technical field of model construction, and in particular to a method and system for establishing a parameterized tunnel model based on cloud computing. Background Art

[0002] Tunnel engineering is widely used in urban rail transit, high-speed railways, and highway construction, continuously advancing transportation infrastructure development. Traditional tunnel design methods often rely on two-dimensional drawings and manual modeling, which is not only inefficient but also difficult to achieve detailed simulation and evaluation of structural performance under complex geological conditions and multiple working conditions. The combination of BIM (Building Information Modeling), CAD parameter-driven modeling, multi-physics simulation, and optimization algorithms has promoted the rapid development of three-dimensional tunnel modeling technology. In particular, parameter dictionary-driven modeling methods enable flexible expression and rapid generation of different tunnel sections, structural dimensions, and structural forms. However, traditional modeling and simulation are mostly limited to local computing environments, and suffer from problems such as limited computing resources, complex model management, and low collaborative efficiency, making it difficult to meet the needs of large-scale design tasks and multi-party collaboration.

[0003] At the same time, existing cloud-based modeling methods still have many limitations:

[0004] (1) The modeling process lacks a self-correction mechanism, and geometric inconsistencies often require manual repair;

[0005] (2) Modeling parameters and simulation parameters are not highly integrated, resulting in data fragmentation and affecting modeling efficiency and accuracy;

[0006] (3) The lack of multi-field coupling and intelligent optimization mechanisms makes it difficult to meet the demand for high-precision prediction of tunnel structure responses under complex working conditions;

[0007] (4) Model service resources are not standardized and packaged, and cloud deployment and permission control are imperfect, which limits the flexibility of model sharing and calling. Summary of the Invention

[0008] Based on this, it is necessary for the present invention to provide a method and system for establishing a parameterized tunnel model based on cloud computing to solve at least one of the above technical problems.

[0009] To achieve the above objectives, a method for establishing a parameterized tunnel model based on cloud computing includes the following steps:

[0010] Step S1: Obtain the tunnel section type and extract the structural semantics to obtain a section feature description set; construct a tunnel section parameter dictionary based on the section feature description set to obtain a dynamic parameter dictionary dataset; model the tunnel section geometry based on the dynamic parameter dictionary dataset to obtain an initial 3D tunnel geometry model;

[0011] Step S2: performing a geometric validity check on the initial three-dimensional tunnel geometric model to obtain geometric anomaly identification data; adjusting parameters of the dynamic parameter dictionary data set based on the geometric anomaly identification data to generate a self-correcting three-dimensional tunnel geometric model;

[0012] Step S3: extracting modeling parameters and configuring boundary conditions for the self-correcting three-dimensional tunnel geometric model to obtain a simulation modeling input data set; designing a simulation parameter range based on the simulation modeling input data set, and performing Latin hypercube sampling of the parameter space to obtain an initial sample parameter set;

[0013] Step S4: performing multi-field coupling simulation on the simulation modeling input data set based on the initial sample parameter set to obtain a multi-field coupling simulation result; performing Bayesian optimization modeling of the input-output relationship based on the multi-field coupling simulation result to obtain a surrogate function model;

[0014] Step S5: performing an expected improvement criterion search in the parameter space based on the surrogate function model, and performing iterative simulation and model update based on the search results to obtain a simulation response result with optimal parameter convergence;

[0015] Step S6: The optimal simulation response results of parameter convergence and the self-correcting three-dimensional tunnel geometry model are uniformly packaged, and cloud packaging and permission configuration are performed, and uploaded to the cloud computing platform to obtain cloud tunnel model service resources.

[0016] The present invention can achieve structured semantic expression of tunnel section types, structural forms, and geometric features, improving data organization efficiency in the initial stage of modeling. The construction of a dynamic parameter dictionary data set makes parameter management scalable and flexible, significantly improving the degree of automation of section geometry generation. The self-correction mechanism actively identifies and repairs inconsistencies during the geometric structure construction process, avoiding the traditional inefficient processing method that relies on manual intervention. The integrated extraction method of modeling parameters and boundary conditions ensures the integrity and accuracy of simulation modeling input data, avoiding modeling errors caused by data fragmentation. The input parameter set construction method based on Latin hypercube sampling ensures the uniform distribution of initial samples in the high-dimensional parameter space, which contributes to the representativeness and comprehensiveness of multi-field coupling simulation. The Bayesian optimization modeling process can gradually approximate the input-output response relationship, combining the rapid prediction ability of the proxy function with the parameter guidance mechanism of the expected improvement criterion to achieve efficient target response optimization; through the joint iteration of simulation and proxy function, the parameter tuning and response accuracy are continuously improved, and the final output simulation response results have high reliability; during the encapsulation process, the geometric structure data, response data and permission configuration data are integrated in a unified standard structure, which improves the structured management capability of model resources and the calling efficiency in the cloud; the permission control mechanism ensures the access security and sharing flexibility of model resources, supports the collaborative design and remote calling needs of multiple parties, and comprehensively breaks through the bottlenecks of traditional local modeling processes in computing efficiency, data collaboration and resource deployment.

[0017] Preferably, the present invention further provides a cloud computing-based parameterized tunnel model establishment system for executing the above-mentioned cloud computing-based parameterized tunnel model establishment method, wherein the cloud computing-based parameterized tunnel model establishment system comprises:

[0018] The parameter modeling module is used to obtain the tunnel section type and extract the structural semantics to obtain a section feature description set; construct a tunnel section parameter dictionary based on the section feature description set to obtain a dynamic parameter dictionary data set; and model the tunnel section geometry based on the dynamic parameter dictionary data set to obtain an initial 3D tunnel geometry model.

[0019] The model correction module is used to perform geometric validity detection on the initial 3D tunnel geometry model to obtain geometric anomaly identification data; based on the geometric anomaly identification data, the dynamic parameter dictionary data set is adjusted to generate a self-correcting 3D tunnel geometry model;

[0020] The simulation preparation module is used to extract modeling parameters and configure boundary conditions for the self-correcting 3D tunnel geometry model to obtain a simulation modeling input data set; based on the simulation modeling input data set, the simulation parameter range is designed and Latin hypercube sampling of the parameter space is performed to obtain an initial sample parameter set;

[0021] The response modeling module is used to perform multi-field coupling simulation on the simulation modeling input data set based on the initial sample parameter set to obtain the multi-field coupling simulation results; based on the multi-field coupling simulation results, Bayesian optimization modeling of the input-output relationship is performed to obtain the proxy function model;

[0022] The parameter optimization module is used to perform the expected improvement criterion search in the parameter space based on the surrogate function model, and perform iterative simulation and model update based on the search results to obtain the optimal simulation response result of parameter convergence;

[0023] The cloud deployment module is used to uniformly package the optimal simulation response results of parameter convergence and the self-correcting three-dimensional tunnel geometry model, perform cloud packaging and permission configuration, upload them to the cloud computing platform, and obtain cloud tunnel model service resources.

[0024] The present invention realizes efficient organization and automated processing of the entire process of tunnel geometry modeling through modular design, and has significant advantages in system integration and calculation process consistency. The parameter modeling module enables in-depth analysis of structural semantic information and dynamic management of parameter logical relationships, enhancing the standardization and diversification support capabilities of cross-section parameter definitions. The model correction module ensures the geometric integrity and parameter consistency of modeling data through a closed-loop verification mechanism, effectively avoiding topological conflicts and boundary disconnections, and significantly enhancing the stability of initial modeling results. The simulation preparation module achieves structural reconstruction and unified encoding of high-dimensional input data through systematic extraction of multidimensional parameter fields and classification of boundary attributes, improving the accuracy and scalability of simulation input. The response modeling module adopts efficient distributed sampling and input-output mapping construction methods to significantly improve prediction coverage and response estimation accuracy under limited simulation sample conditions. The parameter optimization module, with its self-driven search and dynamic feedback capabilities, achieves efficient convergence under the constraints of limited computing resources, effectively improving the target response quality and the optimal match between the design parameters. The cloud deployment module, through resource packaging, access control, and platform integration, fully implements structured publishing, unified access, and remote collaborative calling of tunnel simulation resources, improving cross-system resource reuse efficiency and engineering integration capabilities, and addressing key issues in the traditional tunnel modeling process, such as reliance on manual judgment for parameter definition, poor consistency in modeling data, low efficiency in obtaining simulation responses, and insufficient resource sharing. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0026] Figure 1 A schematic flow chart of the steps of a method for establishing a parameterized tunnel model based on cloud computing according to the present invention;

[0027] Figure 2 for Figure 1 Detailed step flow diagram of step S1;

[0028] Figure 3 for Figure 1 Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0029] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0030] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0031] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0032] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for establishing a parameterized tunnel model based on cloud computing, the method comprising the following steps:

[0033] Step S1: Obtain the tunnel section type and extract the structural semantics to obtain a section feature description set; construct a tunnel section parameter dictionary based on the section feature description set to obtain a dynamic parameter dictionary dataset; model the tunnel section geometry based on the dynamic parameter dictionary dataset to obtain an initial 3D tunnel geometry model;

[0034] Step S2: performing a geometric validity check on the initial three-dimensional tunnel geometric model to obtain geometric anomaly identification data; adjusting parameters of the dynamic parameter dictionary data set based on the geometric anomaly identification data to generate a self-correcting three-dimensional tunnel geometric model;

[0035] Step S3: extracting modeling parameters and configuring boundary conditions for the self-correcting three-dimensional tunnel geometric model to obtain a simulation modeling input data set; designing a simulation parameter range based on the simulation modeling input data set, and performing Latin hypercube sampling of the parameter space to obtain an initial sample parameter set;

[0036] Step S4: performing multi-field coupling simulation on the simulation modeling input data set based on the initial sample parameter set to obtain a multi-field coupling simulation result; performing Bayesian optimization modeling of the input-output relationship based on the multi-field coupling simulation result to obtain a surrogate function model;

[0037] Step S5: performing an expected improvement criterion search in the parameter space based on the surrogate function model, and performing iterative simulation and model update based on the search results to obtain a simulation response result with optimal parameter convergence;

[0038] Step S6: The optimal simulation response results of parameter convergence and the self-correcting three-dimensional tunnel geometry model are uniformly packaged, and cloud packaging and permission configuration are performed, and uploaded to the cloud computing platform to obtain cloud tunnel model service resources.

[0039] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a flow chart of a method for establishing a parameterized tunnel model based on cloud computing according to the present invention. In this example, the method for establishing a parameterized tunnel model based on cloud computing includes the following steps:

[0040] Step S1: Obtain the tunnel section type and extract the structural semantics to obtain a section feature description set; construct a tunnel section parameter dictionary based on the section feature description set to obtain a dynamic parameter dictionary dataset; model the tunnel section geometry based on the dynamic parameter dictionary dataset to obtain an initial 3D tunnel geometry model;

[0041] In an embodiment of the present invention, first, based on an established tunnel engineering database, clearly classified standard tunnel section types such as open-cut rectangular section, single circular section, and three-center circle section are selected, and key structural semantic information in the section geometry is extracted using a preset rule template, including arch height, arch thickness, side wall thickness, base width, bottom plate thickness, invert arch radius, headroom height, and side wall retraction angle, and converted into a standardized section feature description set using a structural semantic parsing tool; the section feature description set is stored in JSON format, and fields and values ​​are defined by a standard dictionary with unified units, and field names should comply with the parameter naming rules defined in the "Highway Tunnel Design Code" (JTG D70-2014); after obtaining the section feature description set, a feature attribute and parameter mapping function is constructed based on the Python language, and a two-dimensional mapping table is constructed using a fixed key value method to form a parameter dictionary containing section parameter names, units, value ranges, and initial set values; the parameter dictionary must meet the following requirements: the units of all length parameters are millimeters, the floating point precision is two decimal places, and the parameter range values ​​must refer to the "Railway Tunnel Design Code" (TB 10003-2016) as the boundary limit, and the number of parameters shall not be less than 12 items; based on the parameter dictionary constructed above, the tunnel section geometry structure is modeled using the OpenCASCADE geometric modeling engine or the FreeCAD modeling module encapsulated therein. The modeling method is parameter-driven modeling, in which the parameter dictionary content is used as the input source for sketch dimensioning and constraints, and the solid structure is generated by contour scanning and Boolean operations; for example, when generating the rectangular plus arch section geometry, first construct a two-dimensional sketch based on the parameters such as "clearance height", "side wall thickness" and "arch height" in the parameter dictionary, and then use the "stretch" and "rotate" commands to generate a three-dimensional solid section structure; after the modeling is completed, the initial three-dimensional tunnel geometry generated needs to be exported as a STEP format file.

[0042] Step S2: performing a geometric validity check on the initial three-dimensional tunnel geometric model to obtain geometric anomaly identification data; adjusting parameters of the dynamic parameter dictionary data set based on the geometric anomaly identification data to generate a self-correcting three-dimensional tunnel geometric model;

[0043] In an embodiment of the present invention, based on the initial three-dimensional tunnel geometric model generated in step S1, a geometric analysis tool built based on the OpenCASCADE core library is called to perform geometric legitimacy detection, and the detection content includes entity closure verification, facet continuity check, boundary topology consistency analysis and geometric body self-intersection judgment; entity closure adopts voxel scanning method to perform spatial grid division on each facet of the model to verify whether the facet combination constitutes a closed body; continuity check is performed by calculating the normal angle between adjacent faces and setting a threshold of 5 degrees for error judgment; boundary consistency analysis uses the topological boundary reconstruction method to compare the number of shared edges of faces. If there are unpaired boundaries, it is considered to be topologically inconsistent; self-intersection judgment is based on Boolean intersection operation to detect whether there is a self-overlapping area inside the entity, and a volume threshold of not less than 0.001 cubic millimeter is set as the error judgment standard in the detection; The above detection output results uniformly generate geometric anomaly identification data, including the anomaly type (such as unclosed, intersecting, non-continuous), the corresponding parameter source and the affected geometric area index; based on the geometric anomaly identification data, combined with the dynamic parameter dictionary data set in step S1, the parameter constraint rule function is used to identify the parameter group that causes the geometric anomaly, and its initial value and associated boundary conditions are located; the local parameter search method is used to iteratively adjust the relevant parameters within the range that does not violate the design specifications, wherein the parameter adjustment step size is set to 1 mm or 1 degree, the maximum number of iterations does not exceed 10 times, and the parameter adjustment gives priority to maintaining the cross-section clearance size unchanged; after the adjustment process is completed, the OpenCASCADE modeling interface is re-called to update the tunnel cross-section geometry structure and generate a self-correcting three-dimensional tunnel geometry model; the self-correcting three-dimensional tunnel geometry model is saved as a standard STEP file.

[0044] Step S3: extracting modeling parameters and configuring boundary conditions for the self-correcting three-dimensional tunnel geometric model to obtain a simulation modeling input data set; designing a simulation parameter range based on the simulation modeling input data set, and performing Latin hypercube sampling of the parameter space to obtain an initial sample parameter set;

[0045] In an embodiment of the present invention, based on the self-correcting three-dimensional tunnel geometric model generated in step S2, a parsing interface based on a CAD kernel is first called to perform facet recognition and boundary traversal operations on the entity geometry of the three-dimensional structure, and 13 types of component-level parameters including lining thickness, arch radius, invert arch point, wall foot position, roadbed width, ventilation structure size, and drainage trough cross-section parameters are extracted. A parameter mapping table is constructed by combining the parameter source and entity index relationship recorded in the aforementioned dynamic parameter dictionary data set. On this basis, according to the simulation accuracy requirements and the structural stability analysis objectives, a parameter annotation tool is called to perform dimension normalization and category labeling operations on the extracted geometric parameters to ensure the consistency of the simulation input dimension. Subsequently, according to the geological section division information of the tunnel, the cover thickness, the structure purpose and the construction method, a predefined boundary condition set is called from the engineering specifications, including the surrounding rock elastic modulus ranging from 2 GPa to 15 GPa, the Poisson's ratio ranging from 0.2 to 0.35, and the formation density ranging from 17 kN / m³ to 22 kN / m The initial ground stress vector direction and the principal stress difference are limited to a range of less than 5 MPa, and the displacement and load boundaries applied to the lining inner wall, bottom plate, and side wall are automatically matched according to the cross-section topology. After the boundary conditions are configured, the extracted parameters and boundary information are integrated to form a standardized simulation modeling input data set. The data set is structured and saved in JSON format, and the field integrity and parameter legitimacy are verified by the input checker. A simulation parameter space is constructed based on the simulation modeling input data set, in which key variables such as the surrounding rock modulus, lining elastic modulus, arch foot reaction coefficient, contact friction coefficient, and external load level are clearly set as independent variables, and their value ranges are determined by geological survey results and standard boundaries. Subsequently, the Latin hypercube sampling function library built on Python is called, the sampling dimension is set to 8 dimensions, the number of samples is set to 100, and the spatially uniformly distributed sampling process is executed. The output result is the initial sample parameter set, which is structured in a table format with fields including parameter name, physical unit, upper and lower limits, and sampling value.

[0046] Step S4: performing multi-field coupling simulation on the simulation modeling input data set based on the initial sample parameter set to obtain a multi-field coupling simulation result; performing Bayesian optimization modeling of the input-output relationship based on the multi-field coupling simulation result to obtain a surrogate function model;

[0047] In an embodiment of the present invention, for the initial sample parameter set and simulation modeling input data set generated in step S3, relying on a multi-field coupling solution platform including structural mechanics, geotechnical mechanics and thermal field linkage mechanism, the finite element method is used to perform numerical simulation on the response behavior of the self-correcting three-dimensional tunnel geometric model under each set of sample parameters, wherein the applied mechanical boundary conditions include applying an internal water pressure load in the range of 0.3MPa to the inner wall of the tunnel, applying an equivalent static pressure to the external surrounding rock to simulate the ground stress, and coupling a temperature gradient in the range of 20°C to 40°C to simulate the tunnel thermal-structural coupling effect; during the simulation solution process, for each set of input parameters, corresponding structural response data are generated, including the maximum principal stress of the lining, the maximum displacement of the section, the concentration degree of joint shear stress, the uniformity of the arch foot reaction force distribution, and the bottom plate settlement. and sidewall displacement, a total of 6 structural response indicators; the simulation result data is output in a CSV structure, and the fields are unified and the dimensions are normalized; to construct the numerical relationship between the input parameters and the output response, a sample regression method based on the shortest sample distance and minimum error expectation criteria is adopted. The distribution density of each input parameter group in multidimensional space and the change trend of the output result are used for similarity judgment. The Gaussian function kernel is used to construct the correlation weight matrix between the input points, and then the objective function with a deviation penalty term is constructed in combination with the structural response data. The corresponding weights of the parameters are iteratively updated to finally realize the mapping relationship function from the input parameters to the target response. The function is continuous and differentiable in the parameter space and has the ability to estimate the variance of the predicted output; the final generated surrogate function model contains the input variable vector, the output prediction value and the confidence interval data.

[0048] Step S5: performing an expected improvement criterion search in the parameter space based on the surrogate function model, and performing iterative simulation and model update based on the search results to obtain a simulation response result with optimal parameter convergence;

[0049] In the embodiment of the present invention, based on the surrogate function model constructed in step S4, a parameter point selection mechanism based on the expected improvement criterion is used to perform a global search of the parameter space for the engineering performance constraints of minimizing the maximum principal stress and reducing the maximum displacement of the cross section to less than 5 mm in the target response index. In each round of search, an improved value expectation function is constructed based on the standard deviation of the output prediction value and the confidence interval in the surrogate function model. The function uses the response improvement potential brought by the parameter point as the core evaluation criterion, selects the parameter combination with the highest information gain in the known sample distribution, and uses the parameter combination as input to re-drive the finite element solver to perform a round of structure-thermal field coupling simulation to obtain actual response data; the response data is evaluated based on the expected value. After unit standardization and error evaluation, the data are appended to the original sample set. At the same time, the input-output mapping relationship is recalculated based on the updated sample set, and the weight matrix of the proxy function is adjusted and updated using the minimization of the mean square residual as the judgment criterion; the expected improvement criterion search is performed again with the updated proxy function to generate the next round of parameter input, and parameter selection, simulation execution, data appending and function update operations are continuously performed until the target response change is less than 0.1% and the prediction variance converges to less than 1e-3 during three consecutive rounds of iterations; the final output parameter combination is the optimal parameter value that converges in the target parameter space, and the corresponding finite element simulation output result is recorded as the optimal simulation response result for parameter convergence.

[0050] Step S6: The optimal simulation response results of parameter convergence and the self-correcting three-dimensional tunnel geometry model are uniformly packaged, and cloud packaging and permission configuration are performed, and uploaded to the cloud computing platform to obtain cloud tunnel model service resources.

[0051] In the embodiment of the present invention, the optimal simulation response result of parameter convergence obtained in step S5 and the self-correcting three-dimensional tunnel geometric model generated in step S2 are uniformly packaged and processed. First, the structural response data is sorted and compressed by calling the structure data set processing library (such as pandas and json) through the Python language, and the six simulation output results - maximum principal stress, maximum displacement of the section, joint shear stress, arch foot reaction, bottom plate settlement and side wall inward displacement are uniformly packaged into a JSON structure object, and bound to the corresponding input parameter combination vector; at the same time, the self-correcting three-dimensional tunnel geometric model is output in standard geometric expression in STEP format, and the surface mesh data of the geometric body is converted into VTK format through the meshio tool to form a data unit that can be used for remote visualization rendering; then, a tunnel model packaging container is constructed based on Docker container technology, and the container structure contains ge ometry directory (storing STEP geometry files and VTK mesh files), response directory (storing JSON response data), meta directory (containing YAML description files for simulation boundary condition configuration and parameter space definition); use the SHA-256 digest function to generate a unique data signature for the above container contents and record it in the service index list; after packaging, upload the container file to the cloud computing platform object storage system (such as MinIO) deployed in the Kubernetes environment, and use the platform's built-in access control policy (RBAC) to configure resource access rights, limiting access rights to development members of the registered project team; finally, register service resource information through the platform API, including service path, data signature, resource type and update timestamp, to form a cloud-based tunnel model service resource with access control and data consistency verification mechanism.

[0052] The present invention can achieve structured semantic expression of tunnel section types, structural forms, and geometric features, improving data organization efficiency in the initial stage of modeling. The construction of a dynamic parameter dictionary data set makes parameter management scalable and flexible, significantly improving the degree of automation of section geometry generation. The self-correction mechanism actively identifies and repairs inconsistencies during the geometric structure construction process, avoiding the traditional inefficient processing method that relies on manual intervention. The integrated extraction method of modeling parameters and boundary conditions ensures the integrity and accuracy of simulation modeling input data, avoiding modeling errors caused by data fragmentation. The input parameter set construction method based on Latin hypercube sampling ensures the uniform distribution of initial samples in the high-dimensional parameter space, which contributes to the representativeness and comprehensiveness of multi-field coupling simulation. The Bayesian optimization modeling process can gradually approximate the input-output response relationship, combining the rapid prediction ability of the proxy function with the parameter guidance mechanism of the expected improvement criterion to achieve efficient target response optimization; through the joint iteration of simulation and proxy function, the parameter tuning and response accuracy are continuously improved, and the final output simulation response results have high reliability; during the encapsulation process, the geometric structure data, response data and permission configuration data are integrated in a unified standard structure, which improves the structured management capability of model resources and the calling efficiency in the cloud; the permission control mechanism ensures the access security and sharing flexibility of model resources, supports the collaborative design and remote calling needs of multiple parties, and comprehensively breaks through the bottlenecks of traditional local modeling processes in computing efficiency, data collaboration and resource deployment.

[0053] Preferably, step S1 includes the following steps:

[0054] Step S11: Acquire tunnel section type data;

[0055] Step S12: extracting semantics of structural features based on the tunnel section type data to obtain semantic description data of the section structure;

[0056] Step S13: formatting and parsing the cross-sectional structure semantic description data, extracting parameter information, and obtaining a cross-sectional feature description set;

[0057] Step S14: defining and classifying the tunnel section parameters based on the section feature description set, constructing a mapping structure, and obtaining a dynamic parameter dictionary data set;

[0058] Step S15: Perform parameter modeling on the contour lines, structural areas, and topological relationships of the tunnel section based on the dynamic parameter dictionary data set to obtain an initial three-dimensional tunnel geometric model.

[0059] In the embodiment of the present invention, in step S11, a structured data interface is used to read tunnel section type data from a subway standard atlas database. The type data records the functional area, structural system, and structural form information of the section in the form of a two-dimensional graphic number and a text description. In step S12, a rule-based semantic extraction method is used to identify the combination relationship between keywords such as "vault thickness", "side wall width", and "lining level" and numerical values ​​from the section type data based on part-of-speech tagging and dependency analysis, and outputs a semantic triple data set of structural features. In step S13, the semantic triple is formatted and parsed using a YACC syntax analyzer to extract all parameter description items with physical quantity dimensions and map them into quantitative data items under a unified unit system. For example, "vault thickness 600mm" is parsed into {"part":"vault","attribute":"thickness","value":600,"unit":"mm"}, and the aggregated form is obtained. A cross-sectional feature description set is formed; in step S14, each parameter description item is encoded in the "part-attribute-value" format through a hash mapping structure, and classified and organized according to the structural system type, such as "single-layer lining", "double-layer composite lining", and "open-cut U-section". At the same time, the adjustable parameter boundaries and discrete spacing are set, and the lower limit, upper limit, step size, and default value are set for each type of parameter. A dynamic parameter dictionary dataset is uniformly constructed in JSON format; in step S15, based on the dynamic parameter dictionary dataset, a Python script is used to drive the AutoCAD command line interface. Drawing commands are called in sequence according to the "contour parameter → construction area → construction layout → boundary connection" sequence, such as PLINE to construct the cross-sectional contour, HATCH to generate the structural filling, and REGION to determine the topological boundary. The association definition between contour lines is realized through explicit coordinate calculation and Boolean geometric operations, and the final output is an initial three-dimensional tunnel geometry model in STEP format.

[0060] By systematically acquiring and deeply analyzing tunnel section type data, the present invention achieves accurate semantic expression of section structural features and efficient parameter extraction, significantly improving the structuring and standardization of tunnel section information. The classification organization and mapping based on the dynamic parameter dictionary not only enhances the flexibility and scalability of tunnel section parameter definitions, but also ensures the accuracy and consistency of parametric modeling, thereby effectively supporting the precise reconstruction of complex geometric forms. Overall, this process optimizes the construction efficiency and quality of three-dimensional tunnel geometric models, promotes the intelligence and automation of subsequent design, simulation, and optimization analysis work, and significantly improves the reliability and practical value of engineering applications.

[0061] Preferably, step S2 includes the following steps:

[0062] Step S21: performing a geometric topology consistency check and entity boundary legitimacy verification on the initial three-dimensional tunnel geometric model to obtain geometric anomaly identification data;

[0063] Step S22: Filtering abnormal constraint conditions existing in the dynamic parameter dictionary data set based on the geometric abnormality recognition data to obtain a parameter set to be optimized;

[0064] Step S23: limiting the value range and resetting the parameters of the dynamic parameter dictionary data set based on the parameter set to be optimized to obtain an updated dynamic parameter dictionary data set;

[0065] Step S24: reconstructing the geometric structure of the tunnel section based on the updated dynamic parameter dictionary data set to obtain a self-correcting three-dimensional tunnel geometric model.

[0066] In an embodiment of the present invention, in step S21, a geometric analysis module based on the BREP boundary representation rule is used to perform a geometric topology consistency check and entity boundary legitimacy verification on the initial three-dimensional tunnel geometric model. First, a graph traversal is performed on the connection relationships of vertices, edges, rings and faces in the entity topology structure to check whether there are non-closed contours, overlapping boundaries or non-manifold patches. Then, based on the surface normal vector consistency and entity closure principle, a closure check is performed on the indirect shell of the structural area. Structured geometric anomaly identification data is generated for structural areas with voids, intersecting or repeated boundaries. The data includes the anomaly type, component number, coordinate interval and affected parameter name; in step S22, the parameter name field involved in the geometric anomaly identification data is parsed, and by matching the field index relationship, the parameter entry containing illegal values ​​or invalid constraints is located in the dynamic parameter dictionary data set, and the parameter priority is marked based on the anomaly type, and all key control parameters with geometric interlacing, boundary damage and size conflict are screened to form a list to be optimized. Parameter set; in step S23, the value range of the parameters in the parameter set to be optimized is limited and the parameters are reset according to the type level of the parameters. The value limitation sets hard constraint boundaries based on the "Railway Tunnel Design Code" and engineering experience. For example, the "initial support arch thickness" is controlled within the range of [250,500] mm and the step size is set to 50 mm. At the same time, the back-off step size adjustment method is performed on the value-conflicting parameters. The parameters are reset to a position that no longer triggers an exception through continuous stepping attempts, and the entire reset process is recorded, and finally an updated dynamic parameter dictionary data set is output; in step S24, the updated dynamic parameter dictionary data set is called, and the AutoCAD drawing command sequence is re-executed. The contour line, structural area, structural layout and boundary connection of the tunnel section are reconstructed in a parameter-driven manner. The reconstruction process uses relative coordinate positioning to ensure the topological consistency and dimensional compatibility between the structures. The reconstructed entity is exported in STEP format file and a geometric legitimacy rapid test is performed to confirm the structural consistency, thereby obtaining a self-correcting three-dimensional tunnel geometric model.

[0067] This method ensures the integrity and accuracy of the tunnel geometry model through rigorous geometric consistency and boundary validity verification, effectively identifying and locating potential geometric anomalies. Combined with the optimization and adjustment of the dynamic parameter dictionary, it improves the rationality and stability of parameters and facilitates the fine-tuning and dynamic updating of model parameters. This process achieves high-precision reconstruction and self-correction of the tunnel cross-sectional geometry, significantly enhancing the model's reliability and engineering applicability. It provides a solid foundation for subsequent design optimization and simulation analysis, and significantly improves the automation level and overall quality of tunnel modeling.

[0068] Preferably, S24 includes the following steps:

[0069] Step S241: parsing the tunnel section parameters item by item based on the updated dynamic parameter dictionary data set to obtain a parameter adjustment mapping table;

[0070] Step S242: replacing and updating the parameters of the initial three-dimensional tunnel geometric model to obtain a geometric parameter adjustment model;

[0071] Step S243: executing a geometric construction algorithm based on the geometric parameter adjustment model and reconstructing the tunnel cross-section geometry to obtain an initial self-corrected three-dimensional tunnel geometry model;

[0072] Step S244: performing a topological integrity check on the initial self-correcting three-dimensional tunnel geometric model to obtain topological check result data;

[0073] Step S245: adjusting and optimizing the initial self-correcting three-dimensional tunnel geometric model based on the topology verification result data to obtain a self-correcting three-dimensional tunnel geometric model.

[0074] In the embodiment of the present invention, in step S241, the dynamic parameter dictionary data set is called to update, and the tunnel section parameters are parsed item by item through the field parsing module. According to the preset parameter naming specification and structural semantic classification rules, each parameter entry is associated with its corresponding structural part, influencing geometric item, and control condition, and a two-dimensional mapping table is established. The generated parameter adjustment mapping table records all the correction parameter information in the structure of "parameter key name-control object-target value-correction mark"; in step S242, the parameter adjustment mapping table is called, and the parameter key name is used as the index to perform a value replacement operation on the parameter variables in the initial three-dimensional tunnel geometric model. All parameters marked as "correction" will be updated with the target value specified in the mapping table. The update process uses a dynamic link rebinding mechanism to modify the variable references in the construction instructions one by one and keep the internal constraints unchanged to generate a complete geometric parameter adjustment model; in step S243, based on the geometric parameter adjustment model, the geometric construction algorithm is executed in sequence, including contour line redrawing based on point column interpolation, construction area generation based on region combination, and boundary closure operation based on connection relationship. Each construction command is in the CAD drawing module. The execution is driven by parameters to ensure that the generated entity meets the requirements of the previous adjustment. After completion, the initial self-corrected three-dimensional tunnel geometric model is exported and saved in the standard STEP format. In step S244, the topology consistency detection module is called to perform a topology connection integrity check on each structural area in the initial self-corrected three-dimensional tunnel geometric model. The verification content includes closure verification, normal consistency check between adjacent faces, vertex overlap determination and edge loop connectivity analysis. All structural units that do not meet the connection integrity will be recorded and topology verification result data will be generated. The data format includes error number, structure number, spatial position and correction suggestion. In step S245, the initial self-corrected three-dimensional tunnel geometric model is corrected. The processing logic is based on the structure number given in the topology verification result data. The automatic section adjustment tool is used to compensate for the coordinate offset of the control point positions of overlapping or unclosed components. The compensation range is limited to a numerical range that does not exceed 10% of the maximum width and height change rate of the section, while maintaining structural symmetry. After all adjustments are completed, the section boundary closure path is recalculated and the edge loop connection relationship is updated to output a self-corrected three-dimensional tunnel geometric model that meets the topology consistency requirements.

[0075] Through meticulous parameter analysis and mapping, this invention achieves precise replacement and dynamic updating of geometric model parameters, enhancing the flexibility and accuracy of model adjustments. Combining efficient geometric construction algorithms with cross-sectional structural reconstruction ensures the geometric consistency and structural integrity of the model during the self-correction process. Topological integrity verification and optimization adjustments based on verification results further enhance the stability and reliability of the model. The overall process effectively improves the adaptive correction capability and quality assurance level of the three-dimensional tunnel geometric model, providing solid support for the precise modeling of complex tunnel structures and subsequent engineering applications.

[0076] Preferably, the step S3 of extracting modeling parameters and configuring boundary conditions for the self-correcting three-dimensional tunnel geometric model includes:

[0077] Perform geometric structure hierarchical division on the self-correcting three-dimensional tunnel geometric model to obtain structural component unit division data;

[0078] Based on the structural component unit division data, the material properties, boundary topology and connection mode of each component are extracted to obtain the component modeling parameter data set;

[0079] Classify the parameter fields and unify the units of the component modeling parameter data set to obtain standardized modeling parameter data;

[0080] Based on the self-correcting 3D tunnel geometry model, the loading area and the constraint area are identified to obtain the structural boundary area identification data;

[0081] Based on the structural boundary area identification data, the boundary type, loading method and boundary node position are modeled to obtain the boundary condition configuration data;

[0082] The standardized modeling parameter data and boundary condition configuration data are integrated to generate the simulation modeling input data set.

[0083] In an embodiment of the present invention, a structural component hierarchical parsing operation is first performed on the self-correcting three-dimensional tunnel geometric model. The tunnel entity is regionally decomposed into physical components such as arches, side walls, inverts, secondary linings, and primary supports based on the cross-sectional structural semantics through a geometric structure decomposition module to form structural component unit division data with unique identification numbers, and a spatial bounding box positioning method is used to clarify the geometric boundary range of each component. Subsequently, an attribute extraction operation is performed based on the division data. The material category, elastic modulus, Poisson's ratio, density, thickness, cross-sectional shape, connection node index, and boundary topology form are extracted for each component unit using component identification rules and a standard cross-sectional database to construct a component modeling parameter data set and record it in the form of a two-dimensional structured table. Field classification and unit conversion are then performed on the component modeling parameter data set. All material parameters are classified into a "material physical property class", all geometric parameters are classified into a "component size property class", and connection information is classified into a "structural connection property class". Units are uniformly converted to the international system of units (such as The length is unified into meters, the force is unified into Newtons, and the stress is unified into Pascals) to generate standardized modeling parameter data; in terms of boundary identification, the end boundary, base boundary and load action area in the self-correcting three-dimensional tunnel geometric model are spatially labeled through the spatial position relationship analysis module, and the force area and constraint area are identified by combining the cross-sectional direction and the external force direction, and the structural boundary area identification data is output; based on the boundary area identification data, the boundary nodes are numbered and located, and the boundary type (fixed, sliding, free), loading method (vertical load, lateral pressure, concentrated force or distributed force) and three-dimensional coordinate position of the boundary node are defined respectively to generate complete boundary condition configuration data; finally, the standardized modeling parameter data and the boundary condition configuration data are integrated in a key-value corresponding structure, and the component number and boundary number are associated and bounded through the field matching mechanism to generate a simulation modeling input data set that meets the simulation input format requirements, and the data is encapsulated in a JSON structure or XML structure.

[0084] Through systematic geometric structure hierarchical division and component attribute extraction, this invention achieves refined parameter extraction and standardized processing of tunnel geometry models, effectively improving the consistency and standardization of model parameters. Combined with the precise identification of structural boundary areas and the scientific configuration of boundary conditions, this ensures the true reflection and reasonable expression of the model in terms of loading and constraints. By integrating standardized parameters and boundary condition data, a high-quality simulation input dataset is constructed, greatly improving the accuracy and reliability of simulation modeling, promoting the efficient implementation of subsequent engineering analysis and optimization design, and comprehensively enhancing the engineering practical value and application effect of three-dimensional tunnel geometry models.

[0085] Preferably, in step S3, designing a simulation parameter range based on the simulation modeling input data set and performing Latin hypercube sampling of the parameter space includes:

[0086] Based on the statistical variable parameter dimensions, upper and lower limits, and sampling accuracy requirements of the simulation modeling input data set, the simulation parameter range is obtained;

[0087] Preprocess the input parameter space based on the simulation parameter range and transform the boundary mapping of the high-dimensional space to obtain standardized parameter space data;

[0088] Perform Latin hypercube sampling on the standardized parameter space data to obtain the initial sample parameter index data set;

[0089] The simulation modeling input data set is parameter-bound based on the initial sample parameter index data set to generate an initial sample parameter set.

[0090] In an embodiment of the present invention, firstly, the material physical parameters, component dimensional parameters and loading parameters recorded in the simulation modeling input data set are analyzed field by field, and the dimensional information of all continuous variables and discrete variables including elastic modulus, Poisson's ratio, thickness, load intensity, support stiffness, etc. is statistically calculated, and the upper and lower limit intervals of the value of each variable are determined based on engineering design specifications, field data and historical samples. The accuracy requirements are set by limiting the number of decimal places and the step range, and are uniformly summarized into a simulation parameter range data table; then, the preprocessing operation of the input parameter space is performed based on the simulation parameter range, and the parameter intervals of each dimension are standardized to the unit closed interval of [0,1] through the linear transformation method, and the boundary monotonicity of each parameter distribution is maintained in the space mapping process, so as to obtain a standardized parameter space with uniform packaging characteristics. Data; when performing Latin hypercube sampling operations in the standardized space, the statistical uniform distribution is used as the sampling basis, and each parameter dimension interval is divided into equally spaced sampling sub-segments, and it is ensured that each sub-segment is sampled only once in all dimensions. The initial sample parameter index data set covering the entire parameter space is generated through a multi-dimensional space hierarchical combination construction method. The data set is stored in the form of a two-dimensional matrix, with each row representing a sample parameter combination and each column corresponding to a parameter dimension; the standardized sample index value is then inversely transformed to the original simulation parameter scale, and the parameter binding operation is performed. Each group of sample parameter values ​​is matched and bound to the corresponding field in the simulation modeling input data set according to the variable number, ensuring that the component properties, boundary conditions and loading forms are completely corresponding, and finally an initial sample parameter set with independent simulation input significance is generated.

[0091] This method achieves efficient normalization and boundary mapping of the parameter space through scientific statistics and preprocessing of simulation parameter ranges, ensuring the comprehensiveness and rationality of parameter design. The Latin hypercube sampling method effectively covers the high-dimensional parameter space, improves the uniformity and representativeness of the samples, and avoids sampling bias and redundancy. This process significantly enhances the diversity and distribution balance of parameter samples, improves the exploration efficiency and accuracy of the simulation model, provides a solid data foundation for subsequent simulation analysis and optimization, and significantly promotes the scientific nature and reliability of tunnel simulation modeling.

[0092] Preferably, step S4 includes the following steps:

[0093] Step S41: performing parameter mapping between each set of sample parameters in the initial sample parameter set and the simulation modeling input data set to obtain a simulation task configuration data set;

[0094] Step S42: deploying a multi-physics coupling solver based on batch input of simulation task configuration data sets, executing automated calculation scheduling, and obtaining multi-field coupling simulation results;

[0095] Step S43: performing response index analysis and feature extraction on the multi-field coupling simulation results to obtain a set of simulation response feature parameters;

[0096] Step S44: performing joint normalization processing on the simulation response characteristic parameter set and the initial sample parameter set to obtain an input and output sample alignment data set;

[0097] Step S45: constructing a Bayesian optimization modeling sample library based on the input and output sample alignment data set to obtain a Bayesian optimization training sample data set;

[0098] Step S46: performing Gaussian process fitting modeling on the Bayesian optimization training sample data set, extracting the predicted mean and confidence interval statistics, and obtaining Bayesian response prediction function data;

[0099] Step S47: constructing a mapping relationship from the input space to the response space based on the Bayesian response prediction function data, and performing training to obtain a proxy function model.

[0100] In an embodiment of the present invention, each set of sample parameters in the initial sample parameter set is first mapped item by item according to the field name and type, and the sample parameter values ​​are embedded into the corresponding material parameters, boundary parameters and loading parameter fields in the simulation modeling input data set, and accurately bound through the component index, node number and parameter classification label to form a complete simulation task configuration data set; then the simulation task configuration data set is batch imported into the multi-physics field coupling solver, and the structural mechanics, electromagnetic field, heat conduction and seepage analysis modules are called. The finite element solution method based on the constitutive equation and the boundary control equation is used to perform automated simulation scheduling calculation. The scheduling process performs task slicing, computing node allocation and load balancing through the task distribution engine in the cloud computing platform, and outputs the multi-field simulation result data set such as the simulation displacement response, stress distribution, temperature rise evolution and seepage pressure change corresponding to each sample; then, a response index set is defined based on the simulation result field information and engineering requirements, and the simulation field variable data is statistically analyzed to extract response characteristics such as the maximum principal stress, key node displacement, thermal gradient range, seepage velocity peak, etc., to form a simulation response feature parameter set in a standard format; the parameter set is compared with the initial sample parameter set. The dataset is jointly normalized according to the field order, and the interval mapping method is used to standardize each parameter to the interval [0,1], ensuring that the parameter precision is retained to three decimal places. Finally, an input and output sample alignment dataset with consistent dimensions is generated. This dataset is used as the sample source, and a Bayesian optimization modeling sample library is constructed through sample-by-sample traversal. Each set of input and output pairs is archived and attached with sample number and timestamp information for source traceability, thus obtaining a Bayesian optimization training sample dataset. On this basis, a Gaussian process fitting operation is performed on the sample dataset through maximum likelihood function solution and covariance function kernel structure construction. The predicted response mean and upper and lower confidence interval values ​​corresponding to each input group are calculated respectively, forming Bayesian response prediction function data with interval estimation capability. Finally, a function mapping relationship construction module is used to establish a one-to-one correspondence between the standardized input parameters and the predicted response indicators. The weighted least squares approximation and continuous mapping interpolation methods are used for function training and regression refinement, ultimately obtaining a surrogate function model with the ability to map from input space to response space. An error correction mechanism is used to ensure that its prediction deviation does not exceed 3% within the input parameter perturbation range.

[0101] The present invention realizes the efficient configuration and batch execution of simulation tasks through systematic parameter mapping and automated multi-physics field coupling simulation calculations, significantly improving the automation and simulation efficiency of computational scheduling. In-depth analysis and feature extraction of response indicators, combined with joint normalization processing, ensure high-quality alignment of input and output data, providing a solid data foundation for subsequent Bayesian optimization modeling. Through Gaussian process fitting and confidence interval statistics extraction, a precise Bayesian response prediction function with uncertainty quantification capability is constructed, realizing effective mapping of input space and response space. The overall process greatly improves the prediction accuracy and generalization ability of the proxy model, provides reliable intelligent decision-making support for tunnel simulation optimization, and promotes the scientificity and intelligence level of simulation analysis.

[0102] Preferably, step S5 includes the following steps:

[0103] Step S51: Predicting sampling points in the parameter space based on the surrogate function model to obtain expected lift function value distribution data;

[0104] Step S52: performing a maximum search on the expected lift function value distribution data to obtain the optimal sampling parameter point;

[0105] Step S53: updating the parameters of the simulation modeling input data set based on the optimal sampling parameter points to generate iterative simulation input data;

[0106] Step S54: performing multi-field coupling simulation on the iterative simulation input data to obtain an iterative simulation response result;

[0107] Step S55: incrementally training and updating the agent function model based on the iterative simulation response results to obtain an updated agent function model;

[0108] Step S56: determining the convergence state of the parameters of the updated proxy function model and obtaining a convergence determination result;

[0109] Step S57: Perform iterative control based on the convergence judgment result; if not converged, return to step S51 to continue searching; if converged, output the optimal simulation response result of parameter convergence.

[0110] In an embodiment of the present invention, first, based on the established proxy function model, a uniform grid is divided in the standardized parameter space with a step size of 0.01 to construct a high-dimensional input sample index set, and the proxy function is called for each index point to predict the simulation response value and the response uncertainty value, and then the maximum improvement potential is used as the objective function, and the expected improvement function value is calculated for each sampling point, and the numerical estimation is completed by the integral expression of the Gaussian distribution probability density function to obtain the expected improvement function value distribution data; then, based on the expected improvement function value distribution data, the maximum value is located, and a local search window is constructed in each dimension by the sliding window method, and the local extreme value point is locked in combination with the recursive search strategy and the improvement function value corresponding to each local extreme value is compared in turn, and finally the parameter point with the largest improvement value is extracted as the optimal sampling parameter point; the optimal sampling parameter point is mapped back to the original simulation parameter scale, and the parameter value is written into the corresponding position in the simulation modeling input data set by field matching, replacing the original parameter value, and generating new iterative simulation input data; then, the iterative simulation input data is imported into the cloud simulation task scheduling system, and the task is divided into two parts. The generator divides the task into structural mechanics, electromagnetics, seepage, and thermal coupling subtasks and dispatches them for execution across the computing cluster. It obtains the response distribution and key performance indicators of each physical field based on the finite element numerical solution method, completes the multi-field coupling simulation process, and outputs the iterative simulation response results. The response results are standardized in field format and appended with the parameter source index number. Sample pairs are mapped one-to-one with the optimal sampling parameter points of that round and appended to the original training sample library. Incremental fitting is performed on these sample pairs. Based on the original function structure framework, covariance updates and mean corrections are performed on the current sample set to complete the incremental training of the surrogate function model and generate an updated surrogate function model. The target response values ​​corresponding to three consecutive rounds of surrogate function output results are then calculated with a convergence threshold of 0.001. If the target response value improvement is less than the threshold for three consecutive times, the surrogate function output is considered to have reached a converged state, and the optimal simulation response result with parameter convergence is output. Otherwise, the process returns to the initial step to recalculate the expected improvement function value and execute a new round of sampling search and simulation update until the convergence judgment criteria are met.

[0111] The present invention realizes intelligent sampling and optimization of parameter space through the proxy function model, effectively guiding the sampling points to focus on the optimal area, and significantly improving the efficiency and accuracy of parameter search. The iteratively updated simulation input is closely integrated with the multi-field coupling simulation response, realizing real-time feedback of simulation results and dynamic adaptive training of the model, ensuring continuous optimization and prediction accuracy of the proxy model. The convergence judgment mechanism ensures the stability and reliability of the optimization process, avoiding excessive calculation and waste of resources. The overall process realizes efficient and accurate parameter optimization iteration, strongly supports the optimal design and performance improvement of the tunnel simulation model, and promotes the intelligentization and automation of engineering simulation.

[0112] Preferably, step S6 includes the following steps:

[0113] Step S61: performing data format unification and structural integration on the optimal simulation response results of parameter convergence and the self-correcting three-dimensional tunnel geometric model to obtain a unified encapsulated data package;

[0114] Step S62: Encapsulate the cloud model resources based on the unified encapsulation data packet to generate a cloud tunnel model service package;

[0115] Step S63: performing permission configuration and access control settings on the cloud tunnel model service package to obtain a permission configuration data set;

[0116] Step S64: Execute an upload operation based on the cloud tunnel model service package and the permission configuration data set to complete resource deployment and obtain cloud tunnel model service resources.

[0117] In the embodiment of the present invention, first, the output parameter convergence optimal simulation response result and the self-correcting three-dimensional tunnel geometric model are respectively subjected to data format conversion processing according to the predefined structural field requirements, and the simulation response result is reorganized according to the field standard "component number-response type-numeric unit-timestamp" and uniformly encoded in CSV format; the self-correcting three-dimensional tunnel geometric model is reconstructed according to the three types of data "component topological structure, boundary node coordinates, and attribute label set", and the point, line, and surface data are represented by the B-rep structure, and the geometric information structure is defined in XML format; then, the above two types of data are structurally integrated with the task number and component index as the primary key fields, and the Pandas and lxml libraries in Python are used for data docking and field nesting processing, and finally a unified encapsulated data packet containing the complete topological geometric structure, parameter field information and response data is generated; then, the encapsulated The data package is input into the cloud-based model resource packaging tool, ChainPack. During the packaging process, the data package content is reorganized into four sub-modules: resource header file, geometry file, response file, and index directory through field mapping relationships. The package is then packaged in Zip compression format to generate a cloud-based tunnel model service package. Permission configuration is performed on the service package. Four fields, namely user number, access token, access level (divided into read, modify, and manage permissions), and access validity period, are written through the identity identification interface. A JSON-based permission description template is used to generate a permission control configuration file, which is finally bound to the cloud-based tunnel model service package to generate a permission configuration dataset. The resource upload tool, CloudUploader, is then launched to upload the generated tunnel model service package and the corresponding permission configuration dataset to the designated cloud computing platform resource pool directory. The upload process ensures data security through the encrypted transmission protocol TLS 1.3, and records upload logs and verification information. After the upload is completed, the cloud platform records the resource number, upload time, owner information, and access permission mapping table through the resource registration interface, ultimately forming a cloud-based tunnel model service resource that can be called by the access control engine.

[0118] The present invention achieves efficient encapsulation of tunnel geometry models and simulation response results through unified data format and structural integration, greatly improving the standardization and consistency of data management. The encapsulation and permission configuration of cloud-based model resources ensure the security and controllability of model services, and support flexible access control and multi-user collaboration. The efficient uploading and deployment of resources realizes the cloud-based management of tunnel model services, facilitates remote calling and sharing, and significantly enhances the scalability and application convenience of the model. The overall process promotes the standardized, intelligent and secure management of tunnel model digital resources, providing a solid guarantee for subsequent engineering applications and cloud-based services.

[0119] Preferably, the present invention further provides a cloud computing-based parameterized tunnel model establishment system for executing the above-mentioned cloud computing-based parameterized tunnel model establishment method, wherein the cloud computing-based parameterized tunnel model establishment system comprises:

[0120] The parameter modeling module is used to obtain the tunnel section type and extract the structural semantics to obtain a section feature description set; construct a tunnel section parameter dictionary based on the section feature description set to obtain a dynamic parameter dictionary data set; and model the tunnel section geometry based on the dynamic parameter dictionary data set to obtain an initial 3D tunnel geometry model.

[0121] The model correction module is used to perform geometric validity detection on the initial 3D tunnel geometry model to obtain geometric anomaly identification data; based on the geometric anomaly identification data, the dynamic parameter dictionary data set is adjusted to generate a self-correcting 3D tunnel geometry model;

[0122] The simulation preparation module is used to extract modeling parameters and configure boundary conditions for the self-correcting 3D tunnel geometry model to obtain a simulation modeling input data set; based on the simulation modeling input data set, the simulation parameter range is designed and Latin hypercube sampling of the parameter space is performed to obtain an initial sample parameter set;

[0123] The response modeling module is used to perform multi-field coupling simulation on the simulation modeling input data set based on the initial sample parameter set to obtain the multi-field coupling simulation results; based on the multi-field coupling simulation results, Bayesian optimization modeling of the input-output relationship is performed to obtain the proxy function model;

[0124] The parameter optimization module is used to perform the expected improvement criterion search in the parameter space based on the surrogate function model, and perform iterative simulation and model update based on the search results to obtain the optimal simulation response result of parameter convergence;

[0125] The cloud deployment module is used to uniformly package the optimal simulation response results of parameter convergence and the self-correcting three-dimensional tunnel geometry model, perform cloud packaging and permission configuration, upload them to the cloud computing platform, and obtain cloud tunnel model service resources.

[0126] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is not limited by the above description. Therefore, it is intended that all changes that fall within the meaning and scope of the equivalent elements of the application documents are included in the present invention.

[0127] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for establishing a parametric tunnel model based on cloud computing, characterized in that: The following steps are involved: Step S1: Obtain the tunnel section type and extract the structural semantics to obtain a section feature description set; A parameter dictionary of the tunnel section is constructed based on the section feature description set to obtain a dynamic parameter dictionary data set; the geometric structure of the tunnel section is modeled based on the dynamic parameter dictionary data set to obtain an initial three-dimensional tunnel geometric model; Step S2: Performing a geometric validity check on the initial three-dimensional tunnel geometric model to obtain geometric anomaly identification data; Based on the geometric anomaly identification data, the dynamic parameter dictionary data set is adjusted to generate a self-correcting 3D tunnel geometry model; Step S3: extracting modeling parameters and configuring boundary conditions for the self-correcting three-dimensional tunnel geometric model to obtain a simulation modeling input data set; designing a simulation parameter range based on the simulation modeling input data set, and performing Latin hypercube sampling of the parameter space to obtain an initial sample parameter set; Step S4: performing a multi-field coupling simulation on the simulation modeling input data set based on the initial sample parameter set to obtain a multi-field coupling simulation result; Based on the multi-field coupling simulation results, Bayesian optimization modeling of the input-output relationship is performed to obtain the surrogate function model; Step S5: performing an expected improvement criterion search in the parameter space based on the surrogate function model, and performing iterative simulation and model update based on the search results to obtain a simulation response result with optimal parameter convergence; Step S6: The optimal simulation response results of parameter convergence and the self-correcting three-dimensional tunnel geometry model are uniformly packaged, and cloud packaging and permission configuration are performed, and uploaded to the cloud computing platform to obtain cloud tunnel model service resources.

2. The method for establishing a parameterized tunnel model based on cloud computing according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire tunnel section type data; Step S12: extracting semantics of structural features based on the tunnel section type data to obtain semantic description data of the section structure; Step S13: formatting and parsing the cross-sectional structure semantic description data, extracting parameter information, and obtaining a cross-sectional feature description set; Step S14: defining and classifying the tunnel section parameters based on the section feature description set, constructing a mapping structure, and obtaining a dynamic parameter dictionary data set; Step S15: Perform parameter modeling on the contour lines, structural areas, and topological relationships of the tunnel section based on the dynamic parameter dictionary data set to obtain an initial three-dimensional tunnel geometric model.

3. The method for establishing a parametric tunnel model based on cloud computing according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing a geometric topology consistency check and entity boundary legitimacy verification on the initial three-dimensional tunnel geometric model to obtain geometric anomaly identification data; Step S22: Filtering abnormal constraint conditions existing in the dynamic parameter dictionary data set based on the geometric abnormality recognition data to obtain a parameter set to be optimized; Step S23: limiting the value range and resetting the parameters of the dynamic parameter dictionary data set based on the parameter set to be optimized to obtain an updated dynamic parameter dictionary data set; Step S24: reconstructing the geometric structure of the tunnel section based on the updated dynamic parameter dictionary data set to obtain a self-correcting three-dimensional tunnel geometric model.

4. The method for establishing a parameterized tunnel model based on cloud computing according to claim 3, characterized in that: Step S24 includes the following steps: Step S241: parsing the tunnel section parameters item by item based on the updated dynamic parameter dictionary data set to obtain a parameter adjustment mapping table; Step S242: replacing and updating the parameters of the initial three-dimensional tunnel geometric model to obtain a geometric parameter adjustment model; Step S243: executing a geometric construction algorithm based on the geometric parameter adjustment model and reconstructing the tunnel cross-section geometry to obtain an initial self-corrected three-dimensional tunnel geometry model; Step S244: performing a topological integrity check on the initial self-correcting three-dimensional tunnel geometric model to obtain topological check result data; Step S245: adjusting and optimizing the initial self-correcting three-dimensional tunnel geometric model based on the topology verification result data to obtain a self-correcting three-dimensional tunnel geometric model.

5. The method for establishing a parameterized tunnel model based on cloud computing according to claim 1, characterized in that: The step S3 of extracting modeling parameters and configuring boundary conditions for the self-correcting three-dimensional tunnel geometric model includes: Perform geometric structure hierarchical division on the self-correcting three-dimensional tunnel geometric model to obtain structural component unit division data; Based on the structural component unit division data, the material properties, boundary topology and connection mode of each component are extracted to obtain the component modeling parameter data set; Classify the parameter fields and unify the units of the component modeling parameter data set to obtain standardized modeling parameter data; Based on the self-correcting 3D tunnel geometry model, the loading area and the constraint area are identified to obtain the structural boundary area identification data; Based on the structural boundary area identification data, the boundary type, loading method and boundary node position are modeled to obtain the boundary condition configuration data; The standardized modeling parameter data and boundary condition configuration data are integrated to generate the simulation modeling input data set.

6. The method for establishing a parameterized tunnel model based on cloud computing according to claim 1, characterized in that: In step S3, the simulation parameter range is designed based on the simulation modeling input data set, and Latin hypercube sampling of the parameter space is performed, including: Based on the statistical variable parameter dimensions, upper and lower limits, and sampling accuracy requirements of the simulation modeling input data set, the simulation parameter range is obtained; Preprocess the input parameter space based on the simulation parameter range and transform the boundary mapping of the high-dimensional space to obtain standardized parameter space data; Perform Latin hypercube sampling on the standardized parameter space data to obtain the initial sample parameter index data set; The simulation modeling input data set is parameter-bound based on the initial sample parameter index data set to generate an initial sample parameter set.

7. The method for establishing a parameterized tunnel model based on cloud computing according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing parameter mapping between each set of sample parameters in the initial sample parameter set and the simulation modeling input data set to obtain a simulation task configuration data set; Step S42: deploying a multi-physics coupling solver based on batch input of simulation task configuration data sets, executing automated calculation scheduling, and obtaining multi-field coupling simulation results; Step S43: performing response index analysis and feature extraction on the multi-field coupling simulation results to obtain a set of simulation response feature parameters; Step S44: performing joint normalization processing on the simulation response characteristic parameter set and the initial sample parameter set to obtain an input and output sample alignment data set; Step S45: constructing a Bayesian optimization modeling sample library based on the input and output sample alignment data set to obtain a Bayesian optimization training sample data set; Step S46: performing Gaussian process fitting modeling on the Bayesian optimization training sample data set, extracting the predicted mean and confidence interval statistics, and obtaining Bayesian response prediction function data; Step S47: constructing a mapping relationship from the input space to the response space based on the Bayesian response prediction function data, and performing training to obtain a proxy function model.

8. The method for establishing a parameterized tunnel model based on cloud computing according to claim 1, wherein: Step S5 includes the following steps: Step S51: Predicting sampling points in the parameter space based on the surrogate function model to obtain expected lift function value distribution data; Step S52: performing a maximum search on the expected lift function value distribution data to obtain the optimal sampling parameter point; Step S53: updating the parameters of the simulation modeling input data set based on the optimal sampling parameter points to generate iterative simulation input data; Step S54: performing multi-field coupling simulation on the iterative simulation input data to obtain an iterative simulation response result; Step S55: incrementally training and updating the agent function model based on the iterative simulation response results to obtain an updated agent function model; Step S56: determining the convergence state of the parameters of the updated proxy function model and obtaining a convergence determination result; Step S57: Perform iterative control based on the convergence judgment result; if not converged, return to step S51 to continue searching; if converged, output the optimal simulation response result of parameter convergence.

9. The method for establishing a parameterized tunnel model based on cloud computing according to claim 1, characterized in that: Step S6 includes the following steps: Step S61: performing data format unification and structural integration on the optimal simulation response results of parameter convergence and the self-correcting three-dimensional tunnel geometric model to obtain a unified encapsulated data package; Step S62: Encapsulate the cloud model resources based on the unified encapsulation data packet to generate a cloud tunnel model service package; Step S63: performing permission configuration and access control settings on the cloud tunnel model service package to obtain a permission configuration data set; Step S64: Execute an upload operation based on the cloud tunnel model service package and the permission configuration data set to complete resource deployment and obtain cloud tunnel model service resources.

10. A cloud computing-based parametric tunnel model establishment system, characterized in that: The method for establishing a parameterized tunnel model based on cloud computing according to claim 1 is used to execute the system for establishing a parameterized tunnel model based on cloud computing, comprising: The parameter modeling module is used to obtain the tunnel section type and extract the structural semantics to obtain a section feature description set; construct a tunnel section parameter dictionary based on the section feature description set to obtain a dynamic parameter dictionary data set; and model the tunnel section geometry based on the dynamic parameter dictionary data set to obtain an initial 3D tunnel geometry model. The model correction module is used to perform geometric validity detection on the initial 3D tunnel geometry model to obtain geometric anomaly identification data; based on the geometric anomaly identification data, the dynamic parameter dictionary data set is adjusted to generate a self-correcting 3D tunnel geometry model; The simulation preparation module is used to extract modeling parameters and configure boundary conditions for the self-correcting 3D tunnel geometry model to obtain a simulation modeling input data set; based on the simulation modeling input data set, the simulation parameter range is designed and Latin hypercube sampling of the parameter space is performed to obtain an initial sample parameter set; The response modeling module is used to perform multi-field coupling simulation on the simulation modeling input data set based on the initial sample parameter set to obtain the multi-field coupling simulation results; based on the multi-field coupling simulation results, Bayesian optimization modeling of the input-output relationship is performed to obtain the proxy function model; The parameter optimization module is used to perform the expected improvement criterion search in the parameter space based on the surrogate function model, and perform iterative simulation and model update based on the search results to obtain the optimal simulation response result of parameter convergence; The cloud deployment module is used to uniformly package the optimal simulation response results of parameter convergence and the self-correcting three-dimensional tunnel geometry model, perform cloud packaging and permission configuration, upload them to the cloud computing platform, and obtain cloud tunnel model service resources.

Images