BIM-based Intelligent Analysis System for Cavern Excavation Stability

Abstract

This invention provides a BIM-based intelligent analysis system for cavern excavation stability. Through deep fusion of BIM models and multi-source monitoring data, it achieves dynamic control of cavern excavation stability throughout the entire process. The system employs lightweight model processing technology to significantly improve computational efficiency, while the intelligent inversion algorithm effectively solves the subjectivity problem of traditional parameter determination. The interactive verification mechanism between 3D numerical simulation and real-time monitoring data greatly improves the reliability of the analysis results, and the dynamic risk assessment system achieves precise risk warning through multi-index fusion analysis. The final support scheme optimization module combines engineering experience with intelligent algorithms, ensuring both safety and economy, forming an intelligent solution that can adapt to changes in engineering conditions. The entire system achieves seamless integration of each link through modular design, providing full-chain technical support from data to decision-making for safe underground engineering construction.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for geotechnical engineering, and in particular to a BIM-based intelligent analysis system for the stability of tunnel excavation. Background Technology

[0002] Geotechnical stability analysis technology has undergone three main development stages: the early stage of qualitative assessment based on empirical formulas, the middle stage of numerical simulation based on finite element analysis, and the current stage of intelligent analysis integrating multi-source monitoring data. Existing technologies have achieved technological leaps from single-point monitoring to distributed monitoring networks, and from static analysis to dynamic simulation. The introduction of BIM technology has made 3D visualization modeling an industry standard. In parameter inversion, traditional methods mainly rely on mathematical optimization techniques such as the least squares method, while modern technologies have begun to apply machine learning algorithms to improve inversion accuracy. Support scheme optimization has also evolved from manual experience-based judgment to intelligent recommendation systems based on case libraries.

[0003] The fundamental flaw in the current system lies in the data fragmentation and process disruption between its various technical modules: the lack of a dynamic coupling mechanism between the BIM model and monitoring data causes model updates to lag behind the actual engineering conditions; the separate computational framework used for parameter inversion and numerical analysis results in wasted computational resources and loss of timeliness. This fragmented technical architecture makes it difficult for the existing system to meet the requirements of modern geotechnical engineering for real-time performance, accuracy, and adaptability, especially when dealing with dynamic excavation processes under complex geological conditions, which can easily lead to major problems such as delayed early warnings and mismatches between support schemes and actual conditions. Summary of the Invention

[0004] In view of this, the present invention provides a BIM-based intelligent analysis system for the stability of tunnel excavation, in order to solve the technical defects existing in the prior art.

[0005] Specifically, the present invention provides a BIM-based intelligent analysis system for the stability of tunnel excavation, comprising a BIM model preprocessing module, a multi-source monitoring data acquisition module, a mechanical parameter intelligent inversion module, a three-dimensional numerical analysis and calculation module, a stability dynamic evaluation module, and a support scheme optimization module connected sequentially via data flow.

[0006] The BIM model preprocessing module extracts engineering entities from the original BIM model and outputs a lightweight model.

[0007] The multi-source monitoring data acquisition module collects surrounding rock condition data in real time, processes it, and then transmits it to the intelligent mechanical parameter inversion module.

[0008] The intelligent mechanical parameter inversion module inverts the set of mechanical parameters based on the monitoring data and inputs it into the three-dimensional numerical analysis and calculation module.

[0009] The three-dimensional numerical analysis and calculation module outputs engineering status indicators, which are then transmitted to the dynamic stability assessment module.

[0010] Risk assessment results generated by the stability dynamic assessment module;

[0011] The support scheme optimization module generates and outputs optimized support schemes based on the risk assessment results.

[0012] In some implementations, the BIM model preprocessing module includes:

[0013] The engineering entity extraction unit is used to extract geological entities and support structure entities from the original BIM model through the building information modeling application programming interface.

[0014] The feature filtering unit uses a feature preservation algorithm based on geometric topology to automatically identify and remove decorative component entities. The feature preservation algorithm includes a component functional attribute identification submodule and a structural importance assessment submodule.

[0015] The surface processing unit reconstructs complex surface entities by piecewise linearization using an adaptive algorithm based on curvature features. The piecewise linearization process preserves the continuity of key feature lines and curvature distribution characteristics of the original surface.

[0016] The model output unit is used to generate a lightweight building information model that conforms to the industry standard format, and simultaneously outputs a material property mapping table associated with the model components. The material property mapping table contains the bidirectional relationship between the material mechanical parameters and the model components.

[0017] In some implementations, the engineering entity extraction unit supports entity extraction operations at multiple levels of precision, including two modes: complete geometric entity extraction and simplified wireframe extraction.

[0018] In some implementations, the multi-source monitoring data acquisition module includes a hybrid monitoring network and an industrial IoT gateway, deploys various sensors to monitor surrounding rock condition parameters, achieves unified data acquisition through multi-protocol access, uses joint algorithms for data denoising, and establishes a monitoring database with unified spatiotemporal reference.

[0019] In some implementations, the intelligent mechanical parameter inversion module constructs an inversion vector containing multiple key parameters, uses an improved optimization algorithm to establish an error function between monitoring data and numerical simulation results, and accelerates the inversion process through parallel computing.

[0020] In some implementations, the fitness function of the improved optimization algorithm is:

[0021]

[0022] Where F is the fitness and N is the total number of monitoring points. The weight coefficient for the i-th monitoring point is derived from the reliability assessment results of the monitoring data. The measured value of the i-th monitoring point comes from the monitoring database. Here, M represents the simulated value for the i-th monitoring point, derived from the numerical simulation output, and M is the number of inversion parameters. Let j be the change in the j-th parameter. Let be the constraint coefficient of the j-th parameter, and α be the penalty exponent. And α are determined based on parameter sensitivity analysis.

[0023] In some implementations, the formula for calculating the weighting coefficient is:

[0024]

[0025] Where K is the number of sensor types. The confidence factor for the k-th type of sensor is derived from sensor calibration data. Let L be the signal strength of the k-th type sensor at point i, derived from real-time monitoring values, and L be the number of environmental influencing factors. The degree of interference from the l-th environmental factor at point i. γ is the correction coefficient for the l-th environmental factor, and γ is the environmental sensitivity index; γ was obtained through statistical analysis of historical environmental monitoring data.

[0026] In some implementations, the three-dimensional numerical analysis and calculation module uses a coupled algorithm to simulate the excavation process, realize automatic load step adjustment, output the distribution of stress field, displacement field and plastic zone, and generate a report containing multiple safety indicators.

[0027] In some implementations, the stability dynamic assessment module establishes a risk level evaluation system based on fuzzy logic, uses neural networks to predict the deformation development trend, and triggers a graded early warning signal when the risk index exceeds a threshold.

[0028] In some implementations, the support scheme optimization module constructs a case knowledge base containing multiple support combinations, recommends an initial support scheme based on case reasoning technology, adjusts the support parameters through optimization algorithms, and outputs the optimal scheme that balances economy and safety.

[0029] At least one embodiment of this invention achieves dynamic control of tunnel excavation stability throughout the entire process through deep fusion of BIM models and multi-source monitoring data. The system employs lightweight model processing technology, significantly improving computational efficiency, while the intelligent inversion algorithm effectively solves the subjectivity problem inherent in traditional parameter determination. The interactive verification mechanism between 3D numerical simulation and real-time monitoring data greatly enhances the reliability of analysis results, and the dynamic risk assessment system achieves precise risk warning through multi-index fusion analysis. The final support scheme optimization module combines engineering experience with intelligent algorithms, ensuring both safety and economy, forming an intelligent solution adaptable to changing engineering conditions. The entire system achieves seamless integration of all stages through modular design, providing full-chain technical support from data to decision-making for safe underground engineering construction. Attached Figure Description

[0030] Figure 1 This is a structural block diagram of a BIM-based intelligent analysis system for cavern excavation stability provided by the present invention. Detailed Implementation

[0031] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0032] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items. The modifications “a” and “a plurality” as used in this disclosure are illustrative and not restrictive, and those skilled in the art will understand that they should be understood as “one or more” unless the context clearly indicates otherwise.

[0033] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0034] In traditional tunnel excavation engineering practice, stability analysis has long faced three major technical bottlenecks: First, the data fragmentation between BIM models and numerical analysis software leads to repeated model reconstruction; a tunnel engineering case showed that model conversion alone took 37 working days. Second, there is a serious lag between monitoring data and mechanical parameter inversion; statistical data shows an average inversion cycle of 5.3 days. Third, existing assessment methods are unable to dynamically respond to changes in excavation conditions; of the 12 tunnel collapse accidents recorded, 9 were directly related to the failure of the static assessment system. To address these problems, this invention considers the deep integration of Building Information Modeling (BIM) technology and intelligent algorithms. By introducing an adaptive reconstruction technology based on curvature features, it successfully achieves lossless conversion from BIM models to analysis models.

[0035] See Figure 1 , Figure 1 This document illustrates a structural block diagram of a BIM-based intelligent analysis system for cavern excavation stability according to some embodiments of this specification. The system includes, sequentially connected via data flow, a BIM model preprocessing module, a multi-source monitoring data acquisition module, a mechanical parameter intelligent inversion module, a three-dimensional numerical analysis and calculation module, a stability dynamic assessment module, and a support scheme optimization module. The BIM model preprocessing module extracts engineering entities from the original BIM model and outputs a lightweight model. The multi-source monitoring data acquisition module collects and processes surrounding rock condition data in real time, then transmits it to the mechanical parameter intelligent inversion module. The mechanical parameter intelligent inversion module inverts the mechanical parameter set based on the monitoring data and inputs it into the three-dimensional numerical analysis and calculation module. The three-dimensional numerical analysis and calculation module outputs engineering condition indicators and transmits them to the stability dynamic assessment module. The stability dynamic assessment module generates risk assessment results. The support scheme optimization module generates and outputs optimized support schemes based on the risk assessment results.

[0036] **Engineering Entities:** This refers to the geometric objects representing the engineering structure in the BIM model. It's used to analyze the topological relationships of components such as beams, columns, and anchors using IFC standards. Feature recognition algorithms filter decorative non-load-bearing elements, accurately extracting the computational model of the support system. **Lightweight Models:** These are optimized 3D computational models that use curvature-adaptive mesh simplification algorithms to compress the number of facets. Key mechanical features are preserved through parameterization, reducing the computational load of subsequent numerical analysis. **Surrounding Rock Condition Data:** This refers to monitoring datasets reflecting stratum changes. It can include physical quantities such as fiber optic displacement values ​​and anchor axial force values. Data points are collected per second using a distributed sensor network, providing the raw input for inversion calculations. **Mechanical Parameter Sets:** This refers to a set of physical quantities characterizing soil and rock properties. It includes key parameters such as cohesion and internal friction angle. The parameter matrix is ​​dynamically updated using intelligent inversion algorithms, eliminating the accumulation of errors from traditional empirical values. **Engineering Condition Indicators:** These are quantitative results from numerical simulation outputs, expressed in visual forms such as stress cloud maps and displacement vector fields. A dynamic mechanical field is generated through a real-time rendering engine, intuitively reflecting the stress redistribution trend in the surrounding rock. Risk assessment results can refer to the decision-making basis for stability determination. They integrate indicators such as safety factor and plastic zone ratio, and use fuzzy comprehensive evaluation to generate five-level risk labels, enabling automatic graded early warning of engineering hazardous states. Optimized support schemes can refer to reinforcement measures generated by algorithms. Using support type, layout density, and construction sequence as optimization variables, a multi-objective genetic algorithm outputs a Pareto optimal solution set, simultaneously satisfying safety reserve requirements and cost control objectives.

[0037] The following is a principle explanation:

[0038] When engineers start the system, the BIM model preprocessing module first establishes a data channel with the original building information model, extracting engineering entities such as geological bodies and supporting structures through the application programming interface. The feature filtering unit uses geometric topology analysis technology to automatically identify and separate decorative components from load-bearing structures, a process that preserves the key connection features of beam-column joints. The surface processing unit then decomposes complex surfaces into several planar elements based on curvature variation characteristics. This processed data flows to the model output unit, ultimately generating a lightweight model that includes material property mapping relationships.

[0039] The multi-source monitoring data acquisition module captures real-time signals of rock deformation and stress changes through a sensor network deployed at key locations in the surrounding rock. Data from different types of sensors is converted using a unified protocol from the industrial IoT gateway before entering a joint denoising process. A wavelet transform algorithm first separates the signal frequency bands, followed by a Kalman filter dynamically correcting the time-domain data. The purified data stream is then synchronously written into a unified spatiotemporal benchmark monitoring database, providing a reliable data source for subsequent analysis.

[0040] The intelligent inversion module for mechanical parameters acquires the latest data stream from the monitoring database and constructs an inversion vector space containing the rock mass mechanical properties. An improved optimization algorithm intelligently searches this high-dimensional space, dynamically adjusting parameter combinations by continuously comparing the differences between the monitoring data and numerical simulation results. When the inversion accuracy reaches a preset threshold, the optimal parameter set is automatically pushed to the three-dimensional numerical analysis module, forming a closed-loop feedback mechanism.

[0041] After receiving the inversion parameters, the 3D numerical analysis module initiates an explicit-implicit coupled calculation process. The simulation of the excavation process automatically adjusts the calculation step size based on the expansion of the plastic zone; this adaptive algorithm ensures more accurate stress distribution results in areas of severe deformation. The calculated engineering state indicators are transmitted to the stability dynamic evaluation module via a standardized interface, forming a complete numerical analysis chain.

[0042] The stability dynamic assessment module integrates two assessment methods: fuzzy logic and neural networks. The fuzzy inference system fuzzifies indicators such as displacement growth rate and outputs a comprehensive evaluation of risk level; at the same time, the neural network predicts future deformation trends based on historical data. The assessment results of the two methods are cross-validated, and when the consistency requirement is met, an early warning signal of the corresponding level is triggered.

[0043] The support scheme optimization module begins with the reception of early warning signals. The system first searches the case library for support schemes under similar engineering conditions, forming an initial solution set. The optimization algorithm then explores the parameter space of these schemes, performing multi-objective optimization that balances safety and economy, ultimately outputting a recommended scheme that includes construction details. The entire process embodies a complete intelligent chain from monitoring and early warning to decision support.

[0044] The present invention will be further described below through a detailed embodiment:

[0045] The BIM-based intelligent analysis system for tunnel excavation stability, based on this invention, was implemented in a typical tunnel excavation project scenario. Upon system startup, the BIM model preprocessing module extracts the main tunnel structure and surrounding rock geological features through the Building Information Modeling (BIM) application interface. A feature retention algorithm based on geometric topology automatically filters decorative components while retaining key structural components such as support anchors and initial lining concrete. For example, when processing the arch surface, the curvature adaptive algorithm decomposes it into 12 polylines, controlling the maximum curvature deviation within 0.15%, ultimately resulting in a lightweight model with an 82% volume reduction.

[0046] The multi-source monitoring data acquisition module is equipped with 8 sets of vibrating wire stress gauges and 16 fiber optic displacement sensors, integrating monitoring data from different sampling frequencies through an industrial IoT gateway. When the tunnel reached kilometer DK15+360, the system detected an abnormal settlement rate at the crown. It then initiated a wavelet transform-Kalman filter joint algorithm to purify the data, eliminating noise interference caused by blasting vibrations and improving the signal-to-noise ratio of the displacement monitoring data to over 35dB.

[0047] The intelligent inversion module for mechanical parameters constructs an inversion vector containing seven parameters, including elastic modulus and cohesion, and optimizes the parameters using an improved multi-objective particle swarm optimization algorithm. During the inversion process in the third excavation stage, the system's parallel computing cluster completed 36 iterations in 14 minutes and 38 seconds, obtaining an elastic modulus value of 2.15 GPa for the surrounding rock, with a deviation of only 3.7% from the results of on-site sampling tests.

[0048] The three-dimensional numerical analysis and calculation module uses an explicit-implicit coupling algorithm to simulate the excavation process. When the expansion rate of the plastic zone at section DK15+375 exceeds 0.12 mm per second, the load step is automatically reduced from 30 seconds to 5 seconds to accurately capture the stress redistribution process of the support structure. The calculated displacement field cloud map shows that the maximum displacement is located at the arch waist, reaching 8.7 mm.

[0049] The LSTM neural network in the stability dynamic assessment module predicts the deformation trend for the next 3 days based on data from the past 72 hours. When the predicted displacement growth rate exceeds the threshold of 0.25 mm / hour, a level-two yellow warning is triggered. Simultaneously, the fuzzy logic assessment system, considering indicators such as the plastic zone ratio of 18.7% and principal stress deflection of 9.3 degrees, gives an overall risk coefficient of 0.68.

[0050] The support scheme optimization module matches three similar working conditions from the case library. After optimization by the genetic algorithm, it recommends a reinforcement scheme that adjusts the system anchor spacing from 1.5 meters to 1.2 meters and increases the thickness of the sprayed concrete by 50 millimeters.

[0051] The beneficial effects of one of the embodiments in this specification include at least the following: Through deep fusion of BIM models and multi-source monitoring data, dynamic control of the stability of tunnel excavation is achieved throughout the entire process. The system employs lightweight model processing technology, significantly improving computational efficiency, while the intelligent inversion algorithm effectively solves the subjectivity problem of traditional parameter determination. The interactive verification mechanism between three-dimensional numerical simulation and real-time monitoring data greatly improves the reliability of analysis results, and the dynamic risk assessment system achieves precise risk warning through multi-index fusion analysis. The final support scheme optimization module combines engineering experience with intelligent algorithms, ensuring safety while considering economy, forming an intelligent solution that can adapt to changes in engineering conditions. The entire system achieves seamless connection between each link through modular design, providing full-chain technical support from data to decision-making for safe construction of underground engineering projects.

[0052] In some implementations, the BIM model preprocessing module includes: an engineering entity extraction unit, used to extract geological entities and support structure entities from the original BIM model through a building information modeling application programming interface; a feature filtering unit, which automatically identifies and removes decorative component entities using a geometric topology-based feature preservation algorithm, wherein the feature preservation algorithm includes a component functional attribute identification submodule and a structural importance assessment submodule; a surface processing unit, which performs piecewise linearization reconstruction of complex surface entities using an adaptive algorithm based on curvature features, wherein the piecewise linearization process maintains the continuity of key feature lines and curvature distribution characteristics of the original surface; and a model output unit, used to generate a lightweight building information model conforming to industry standard formats and simultaneously output a material property mapping table associated with the model components, wherein the material property mapping table contains a bidirectional association between material mechanical parameters and model components.

[0053] Engineering entity extraction units refer to components that separate key structures from the BIM model. These are used to parse the model database via the IFC standard interface, filtering geological bodies and support structures by entity type tags to ensure that subsequent analysis focuses solely on the load-bearing system. Geological body entities refer to model objects representing engineering geological conditions. Voxelization is used to convert stratigraphic information into a 3D mesh, and different geological units are distinguished by rock stratum attribute coding, accurately reflecting the distribution of the mechanical properties of the surrounding rock. Support structure entities refer to assemblies of artificially reinforced components, used to extract the spatial topological relationships of support components such as anchor bolts and linings. Component IDs are used to automatically associate these components with design parameters, preserving the mechanical structural features from construction drawings.

[0054] The feature filtering unit refers to the model purification module, which uses graph theory algorithms to construct a component relationship network and identifies non-load-bearing decorative components based on centrality analysis, significantly improving model computational efficiency. The component functional attribute identification submodule refers to the intelligent unit that determines the component's purpose. It parses the functional type attributes in the IFC file and matches the component's engineering purpose against a preset rule base, automatically distinguishing between structural and architectural decorative components. The structural importance assessment submodule refers to the component criticality analysis system. It uses finite element pre-calculation to determine the stress contribution of each component and removes low-impact components through threshold filtering, ensuring that the simplified model retains its original mechanical properties.

[0055] Surface processing units can refer to dedicated processors for geometric optimization. They employ moving least squares to fit complex surfaces and dynamically adjust mesh density based on curvature sensitivity, achieving high-precision surface approximation. Piecewise linear reconstruction refers to surface discretization methods that set fixed dividing points at feature lines and generate piecewise planes through constrained optimization algorithms, preserving the mechanical transmission path of the original surface. Key feature line continuity refers to the transferability of geometric features. Curvature extrema are extracted using differential geometry algorithms, and the feature line topology is forcibly preserved during mesh generation, accurately reflecting the distribution of stress concentration regions. Curvature distribution characteristics refer to the quantitative description of surface geometry. By establishing a curvature-mesh size mapping relationship and densifying discrete nodes in curvature abrupt change regions, geometric errors in simplified models can be effectively controlled.

[0056] The model output unit can refer to a standardized interface generator, which uses an open-source IFC library for format conversion and employs a data validation mechanism to ensure model integrity, providing standardized input for downstream analysis. The material property mapping table can refer to a component-parameter association database, used to establish an index relationship between material IDs and parameters such as elastic modulus. It uses a hash table for fast bidirectional lookups and supports rapid parameter replacement across multiple working conditions. The bidirectional association relationship can refer to a data coupling link mechanism, using component GUIDs as keys to achieve reverse parameter tracing. Event listeners maintain synchronized data updates, ensuring real-time consistency between the model and the parameter set.

[0057] This preprocessing module effectively solves core problems such as information redundancy, format barriers, and feature loss in the traditional BIM model conversion process through intelligent and standardized data processing workflow. It provides a high-fidelity and lightweight basic model for numerical analysis of geotechnical engineering, and greatly improves the accuracy and efficiency of digital analysis of underground engineering.

[0058] In some implementations, the engineering entity extraction unit supports entity extraction operations at multiple levels of precision, including two modes: complete geometric entity extraction and simplified wireframe extraction.

[0059] Complete geometric entity extraction refers to an extraction method that preserves all geometric details, exporting NURBS surfaces and solid models in the native BIM format. Parametric transformation maintains design accuracy, meeting the needs of high-precision finite element analysis. Simplified wireframe extraction refers to a lightweight geometric representation method that uses feature line recognition algorithms to extract key contour lines and generates a skeleton model through topological simplification. This method reflects the main stress characteristics of the structure with extremely low computational cost.

[0060] The multi-precision extraction mechanism enables adaptive processing of BIM model data. The complete geometric mode ensures the analysis accuracy of key nodes, while the simplified wireframe mode significantly improves the processing efficiency of large-scale models. The two work together to form a complete solution covering different engineering scenarios, effectively solving the industry problem that traditional single-precision extraction methods cannot balance computational accuracy and efficiency.

[0061] In some implementations, the multi-source monitoring data acquisition module includes a hybrid monitoring network and an industrial IoT gateway, deploys various sensors to monitor surrounding rock condition parameters, achieves unified data acquisition through multi-protocol access, uses joint algorithms for data denoising, and establishes a monitoring database with unified spatiotemporal reference.

[0062] Hybrid monitoring networks refer to integrated systems of heterogeneous sensing devices, used to deploy various types of monitoring equipment such as fiber optic sensors and inclinometers. Optimal deployment locations are determined through topology optimization algorithms, enabling full-section coverage of surrounding rock deformation data. Surrounding rock state parameters refer to the set of physical quantities reflecting stratum stability, including monitoring indicators such as displacement, stress, and seepage pressure. Synchronous sampling at a preset frequency by distributed acquisition nodes allows for the construction of a complete characteristic spectrum of the surrounding rock's dynamic response. Unified data acquisition refers to a standardized data aggregation process, employing timestamp alignment mechanisms to process asynchronous data streams and balancing acquisition rate differences through buffer queues, achieving clock synchronization of multi-source data. Joint algorithms refer to multimodal data fusion methods, combining wavelet transform and Kalman filtering for cascaded denoising. Adaptive adjustment of filtering parameters through residual analysis effectively suppresses environmental interference in monitoring data. Unified spatiotemporal references refer to a data standardization framework used to establish a mapping relationship between the engineering coordinate system and UTC time. Positioning deviations are eliminated through coordinate transformation matrices, ensuring the comparability of monitoring data.

[0063] By constructing a complete surrounding rock monitoring data chain, forming a closed loop from multi-source heterogeneous acquisition to standardized processing, it not only retains the technical advantages of various sensors, but also improves data quality through intelligent fusion, providing a real and reliable data foundation for engineering stability analysis, which is significantly better than the performance of traditional single-source monitoring systems in terms of data integrity and accuracy.

[0064] In some implementations, the intelligent mechanical parameter inversion module constructs an inversion vector containing multiple key parameters, uses an improved optimization algorithm to establish an error function between monitoring data and numerical simulation results, and accelerates the inversion process through parallel computing.

[0065] Inversion vectors can refer to the mathematical expression of the parameters to be solved, used to encode parameters such as the elastic modulus of rock mass into high-dimensional vectors. The vector dimension is determined through parameter sensitivity analysis, enabling a systematic organization of the inversion calculation process. Improved optimization algorithms can refer to enhanced solution methods, combining the global search capability of genetic algorithms with the convergence speed of particle swarm optimization, employing dynamic weight adjustment strategies to effectively avoid traditional algorithms getting trapped in local optima. Error functions can refer to the quantitative index of model fit, used to construct the sum of squared residuals between monitored displacements and simulation results. Adding regularization terms controls the parameter fluctuation range, balancing the accuracy and stability of the inversion results. Numerical simulation results can refer to the output data of finite element calculations. Multi-threading technology is used to execute simulation calculations with different parameter combinations in parallel, and a result caching mechanism reduces redundant calculations, significantly improving inversion efficiency. Parallel computing can refer to a distributed processing architecture, using the MPI protocol to allocate tasks across multiple nodes. Dynamic load balancing algorithms optimize computational resource utilization, reducing the traditional serial inversion time by several orders of magnitude.

[0066] The intelligent inversion method achieves accurate identification of rock mechanics parameters. The improved optimization algorithm ensures the global convergence of parameter search. The parallel computing architecture solves the computational bottleneck of large-scale inversion and provides a reliable parameter basis for engineering stability assessment. Compared with the traditional trial and error method, it has higher efficiency and accuracy.

[0067] In some implementations, the fitness function of the improved optimization algorithm is:

[0068]

[0069] Where F is the fitness and N is the total number of monitoring points. The weight coefficient for the i-th monitoring point is derived from the reliability assessment results of the monitoring data. The measured value of the i-th monitoring point comes from the monitoring database. Here, M represents the simulated value for the i-th monitoring point, derived from the numerical simulation output, and M is the number of inversion parameters. Let j be the change in the j-th parameter. Let α be the constraint coefficient for the j-th parameter, and α be the penalty exponent. And α are determined based on parameter sensitivity analysis.

[0070] Fitness F is a quantitative indicator that directly reflects the overall quality (accuracy and rationality) of a set of inversion parameters. The core task of the optimization algorithm is to find the optimal set of parameters that minimizes fitness F through iteration. Fitness describes the ability of the parameter set to "adapt" to real engineering scenarios; the stronger the ability, the lower the fitness value (i.e., error and penalty). The total number of monitoring points can refer to the scale of spatial sampling points, used to determine the data dimension for error calculation. By balancing coverage density and computational load through point optimization algorithms, it can comprehensively reflect the response characteristics of the engineering structure. Weighting coefficients can refer to a quantitative indicator of data reliability. Using sensor calibration error and historical stability as evaluation criteria, the contribution of each monitoring point is determined by the analytic hierarchy process (AHP), highlighting the decision weight of highly reliable data. The number of inversion parameters can refer to the dimension of the variables to be identified. Sensitive parameter sets are screened based on orthogonal experimental design, and parameter correlation is reduced through principal component analysis, optimizing the well-posedness of the inversion problem. Constraint coefficients can refer to the strength of the restriction on parameter changes. The feasible region is determined based on the physical meaning of the parameters, and expert experience is quantified using fuzzy mathematics methods to prevent the inversion results from exceeding the reasonable range of engineering. The penalty index can refer to the severity of nonlinear constraints. By adaptively adjusting the shape of the index curve based on parameter sensitivity and using a trial-and-error method to determine the optimal penalty strength, abnormal parameter fluctuations can be effectively suppressed. Parameter sensitivity analysis refers to the study of input-output correlation. Using the Morris screening method to identify key parameters and quantifying the contribution rate of each parameter through variance decomposition, it can guide the targeted optimization of the inversion strategy.

[0071] This fitness function integrates both data-driven and physical constraints, ensuring simulation accuracy through weighted residuals while maintaining engineering rationality by introducing parameter penalty terms, thus forming a composite objective function with clear physical meaning. This provides a solution for complex engineering inversion problems that combines mathematical rigor with engineering practicality.

[0072] In some implementations, the formula for calculating the weighting coefficient is:

[0073]

[0074] Where K is the number of sensor types. The confidence factor for the k-th type of sensor is derived from sensor calibration data. Let L be the signal strength of the k-th type sensor at point i, derived from real-time monitoring values, and L be the number of environmental influencing factors. The degree of interference from the l-th environmental factor at point i. γ is the correction coefficient for the l-th environmental factor, and γ is the environmental sensitivity index; γ was obtained through statistical analysis of historical environmental monitoring data.

[0075] The number of sensor types refers to the number of heterogeneous sensing devices, used to determine the dimensions of data fusion. By evaluating the characteristics of various sensors through a device selection matrix, a complementary monitoring network architecture can be constructed. The confidence factor is a quantitative indicator of sensor reliability, based on factory calibration errors and long-term stability test data, dynamically corrected using a Bayesian update algorithm, reflecting the actual operating state of the sensor. The number of environmental influencing factors refers to the total number of interference sources. Based on multi-dimensional data such as temperature and humidity collected by the on-site environmental monitoring system, key factors are screened through correlation analysis, allowing for focused control of major interference sources. The correction coefficient is an adjustment parameter for environmental compensation, used to establish a regression model of environmental parameters and data drift. Using the least squares method to fit historical data, it accurately quantifies the interference patterns of environmental factors. The interference level refers to the intensity measure of environmental noise. By separating the signal and environmental noise components through time-frequency analysis, the reciprocal of the signal-to-noise ratio is calculated as a quantitative indicator, objectively assessing the severity of environmental conditions. The environmental sensitivity index is a parameter of the system's anti-interference capability. Based on the data volatility under different environmental conditions, the index value is determined using fuzzy inference methods, enabling adaptive adjustment of the environmental compensation level. Sensor calibration data refers to the baseline values ​​of metrological characteristics. Multi-point calibration tests are conducted under standard experimental conditions, and input-output relationship curves are established through polynomial fitting, providing an objective basis for confidence assessment. Historical environmental monitoring data refers to long-term accumulated operating condition records, used to construct a time-series database containing multi-dimensional parameters such as temperature and vibration. Statistical analysis identifies environmental disturbance patterns, improving the prediction accuracy of corrective models.

[0076] By integrating both equipment reliability and environmental adaptability, this approach considers both the performance differences of the sensors themselves and the dynamic compensation for complex environments. This ensures that the final weighting coefficients accurately reflect the actual reliability of the monitoring data, providing a scientific basis for data quality evaluation for subsequent inversion calculations.

[0077] In some implementations, the three-dimensional numerical analysis and calculation module uses a coupled algorithm to simulate the excavation process, realize automatic load step adjustment, output the distribution of stress field, displacement field and plastic zone, and generate a report containing multiple safety indicators.

[0078] Coupled algorithms refer to multi-physics collaborative computation methods used to simultaneously solve the governing equations of stress and seepage fields. By establishing a two-way data exchange interface, field variables are transmitted in real time, accurately reflecting the fluid-structure interaction effect during excavation. Excavation process simulation refers to the numerical reproduction of construction procedures. Excavation boundary conditions are loaded in a time-series manner, and the element birth-death technique is used to simulate the rock mass removal process, predicting the ground response at each construction stage. Automatic load step adjustment refers to intelligent control of the calculation step size. The incremental step size is dynamically adjusted based on the convergence state, and the calculation accuracy is judged by an error estimator, balancing computational efficiency and result accuracy. The stress field refers to the distribution of internal forces within the rock mass. The calculation results are processed using Gaussian integral point stress smoothing technology, and the principal stress directions are extracted using tensor invariants, providing a visual representation of stress concentration areas. The displacement field refers to the spatial distribution of rock mass deformation. A continuous field cloud map is generated through nodal displacement interpolation, combined with vector arrows to display displacement directions, identifying potential deformation hazard areas. The distribution of the plastic zone refers to the material yield range parameter, used to calculate the yield element ratio under the Mohr-Coulomb criterion. The yield degree is visualized through a color-scale diagram, assessing the stability of the surrounding rock. Safety indicators can refer to the quantitative parameters of engineering stability, including multi-dimensional evaluation factors such as safety factor and proximity to failure. They adopt a weighted comprehensive scoring method and can provide a quantitative basis for engineering decision-making.

[0079] This three-dimensional numerical analysis module accurately simulates excavation response under complex geological conditions through advanced coupling algorithms. Intelligent load step adjustment ensures the stability of the calculation process, and the comprehensive result output system provides multi-dimensional decision-making basis for engineering safety assessment, significantly improving the accuracy and practicality of numerical analysis of underground engineering.

[0080] In some implementations, the stability dynamic assessment module establishes a risk level evaluation system based on fuzzy logic, uses neural networks to predict the deformation development trend, and triggers a graded early warning signal when the risk index exceeds a threshold.

[0081] Fuzzy logic refers to a mathematical tool for handling uncertain knowledge, used to construct a rule base for inference containing membership functions. Through a three-step process of fuzzification, inference, and defuzzification, it can effectively handle fuzzy information in engineering safety assessments. A risk level evaluation system refers to a classification standard for safety status. Based on expert experience, it establishes multi-level evaluation indicators and uses a weighted comprehensive evaluation method to calculate risk values, enabling a scientific classification of engineering stability. A risk index refers to a quantitative representation of safety status, integrating multi-dimensional indicators such as stress ratio and displacement rate. By normalizing these indicators to eliminate dimensional influences, it can intuitively reflect the degree of engineering hazard. Graded early warning signals refer to a risk response mechanism. Different risk levels are indicated by color codes, and multi-level warnings are achieved through audible and visual alarm devices, guiding targeted on-site measures.

[0082] By integrating fuzzy reasoning and neural network prediction technologies, this approach considers both the uncertainties of the assessment process and the ability to learn and predict deformation trends. Combined with an intelligent hierarchical early warning mechanism, it provides a complete solution for engineering safety, from risk identification to early warning response, significantly improving the timeliness and accuracy of underground engineering safety management.

[0083] In some implementations, the support scheme optimization module constructs a case knowledge base containing multiple support combinations, recommends an initial support scheme based on case reasoning technology, adjusts the support parameters through optimization algorithms, and outputs the optimal scheme that balances economy and safety.

[0084] Support combinations refer to different configuration schemes of support structures, used to establish a database including combinations such as anchor bolts and shotcrete. Typical case studies are selected through orthogonal experimental design, covering support needs under various geological conditions. Case knowledge bases refer to the digital storage of engineering experience, building a structured database based on historical successful cases. Feature encoding technology enables rapid retrieval, providing reliable reference for new projects. Case reasoning technology refers to similarity problem-solving methods, used to calculate the similarity index between the current project and the case database. The nearest neighbor algorithm matches the most similar case, quickly generating an initial scheme that meets the project's characteristics. Optimization algorithms refer to parameter optimization calculation methods, using genetic algorithms to iteratively optimize parameters such as support spacing and length. The fitness function evaluates the scheme's merits, finding the optimal parameter combination. Support parameters refer to the quantified values ​​of design variables, including adjustable factors such as material strength and structural dimensions. A response surface model with a safety factor is established, enabling precise parameter control.

[0085] This support scheme optimization module achieves experience transfer through the construction of a comprehensive case knowledge base, ensures the scientific nature of the design scheme through intelligent optimization algorithms, and takes into account both engineering economic and safety requirements through a dual-objective balancing mechanism. It provides a complete solution for underground engineering support design, from experience reference to parameter optimization, and significantly improves the efficiency and quality of support design.

[0086] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this invention. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A BIM-based intelligent analysis system for stability of a cavern excavation, characterized by, The BIM model preprocessing module, the multi-source monitoring data acquisition module, the mechanical parameter intelligent inversion module, the three-dimensional numerical analysis calculation module, the stability dynamic evaluation module and the support scheme optimization module are sequentially connected by data flow. The BIM model preprocessing module extracts engineering entities from an original BIM model and outputs a lightweight model. The multi-source monitoring data acquisition module acquires surrounding rock state data in real time and transmits the processed data to the mechanical parameter intelligent inversion module. The mechanical parameter intelligent inversion module inverses mechanical parameters based on monitoring data and inputs the three-dimensional numerical analysis calculation module, wherein the multi-source monitoring data acquisition module includes a hybrid monitoring network and an industrial Internet of Things gateway, arranges various sensors to monitor surrounding rock state parameters, realizes unified data acquisition through multi-protocol access, adopts a joint algorithm for data denoising processing, and establishes a monitoring database with unified time and space benchmarks. The mechanical parameter intelligent inversion module constructs an inversion vector containing multiple key parameters, adopts an improved optimization algorithm to establish an error function of monitoring data and numerical simulation results, and accelerates the inversion process through parallel computing, wherein the fitness function of the improved optimization algorithm is as follows: where F is fitness, N is the total number of monitoring points, is the weight coefficient of the ith monitoring point, from the monitoring data reliability evaluation result, is the measured value of the ith monitoring point, from the monitoring database, is the simulation value of the ith monitoring point, from the numerical simulation output, M is the number of inversion parameters, is the change amount of the jth parameter, is the constraint coefficient of the jth parameter, and α is the penalty index, and α are determined according to the parameter sensitivity analysis; The three-dimensional numerical analysis calculation module outputs engineering state indicators to the stability dynamic evaluation module. The risk evaluation results generated by the stability dynamic evaluation module. The support scheme optimization module generates and outputs an optimized support scheme according to the risk evaluation results.

2. The system of claim 1, wherein, The BIM model preprocessing module includes: An engineering entity extraction unit for extracting geological body entities and support structure entities in an original BIM model through a building information model application program interface; A feature filtering unit that automatically identifies and removes decorative component entities using a feature retention algorithm based on geometric topology, wherein the feature retention algorithm includes a component function attribute identification submodule and a structure importance evaluation submodule; A curved surface processing unit that performs segmented linearization reconstruction on complex curved surface entities through an adaptive algorithm based on curvature characteristics, wherein the segmented linearization processing process maintains the continuity of key feature lines and the curvature distribution characteristics of the original curved surface; A model output unit for generating a lightweight building information model in conformity with industry standard formats and synchronously outputting a material attribute mapping table associated with model components, wherein the material attribute mapping table contains a bidirectional association relationship between material mechanical parameters and model components.

3. The system of claim 1, wherein, The engineering entity extraction unit supports entity extraction operations at multiple precision levels, including complete geometric entity extraction and simplified wireframe extraction modes.

4. The system of claim 1, wherein, The calculation formula for calculating the weight coefficient is as follows: Wherein, K is the number of sensor types, is the confidence factor of the kth sensor, from the sensor calibration data, is the signal strength of the kth sensor at the ith point, from the real-time monitoring value, L is the number of environmental factors, is the interference degree of the lth environmental factor at the ith point, is the correction coefficient of the lth environmental factor, and γ is the environmental sensitivity index. and γ are obtained according to statistical analysis of environmental monitoring historical data.

5. The system of claim 1, wherein, The three-dimensional numerical analysis calculation module uses a coupling algorithm for excavation process simulation, realizes automatic load step adjustment, outputs the distribution of stress field, displacement field and plastic zone, and generates a report containing multiple safety indicators.

6. The system of claim 1, wherein, The stability dynamic evaluation module establishes a risk level evaluation system based on fuzzy logic, uses a neural network to predict deformation development trends, and triggers a graded early warning signal when the risk index exceeds a threshold value.

7. The system of claim 1, wherein, The support scheme optimization module constructs a case knowledge base containing various support combinations, recommends an initial support scheme based on case reasoning technology, adjusts support parameters through an optimization algorithm, and outputs an optimal scheme balancing economy and safety.

Images