Early warning method and system for compartmental analysis and construction of large foundation pits for subway double-sided expansion

Abstract

The present invention proposes an early warning method and system for compartmental analysis of large-scale foundation pits for bilateral extension of subways. Multi-source heterogeneous data of subway station underground complexes are subjected to multi-dimensional fusion processing to achieve effective integration and utilization of data. The Kriging interpolation algorithm is used for spatial interpolation and dynamic characteristic simulation to generate a three-dimensional geological structure that changes dynamically with construction, thereby improving the accuracy and predictive ability of the geological model. Combined with parametric multi-scale modeling technology, seamless conversion from two-dimensional engineering drawings and three-dimensional scanning data to three-dimensional representation data of subway station underground complexes is achieved, enhancing the practicality and integrity of the model. A multi-objective optimization algorithm is used to perform three-dimensional finite element numerical simulation and compartmental analysis to optimize the construction planning of subway station underground complexes and improve construction efficiency. The construction process is digitally simulated and monitored in real time to achieve trend prediction and anomaly detection, ensuring the safety and quality of the construction of subway station underground complexes and reducing construction risks.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an early warning method and system for compartmentalized analysis and construction of a large-scale foundation pit for bilateral expansion of a subway. Background Art

[0002] As critical urban infrastructure, the 3D modeling and construction management of underground subway station complexes have long been a key research focus in the engineering field. Traditional subway station modeling methods rely primarily on 2D drawings and a single data source, making it difficult to fully reflect the complexity of underground spaces. In recent years, with the development of technologies such as 3D laser scanning and geological exploration, multi-source data fusion has become a new trend in underground space modeling. Some researchers have proposed BIM (Building Information Modeling)-based subway station modeling methods, integrating multiple types of information to improve model accuracy and integrity. Furthermore, in construction management, some scholars have begun exploring the application of artificial intelligence and big data analysis technologies to optimize and monitor the construction process. This is particularly true when constructing large foundation pits on both sides of a subway station and dividing them into compartments. This places stringent requirements on the dynamic response capabilities of subway station modeling and construction management.

[0003] However, existing technologies still have some shortcomings. First, the fusion processing of multi-source heterogeneous data still faces challenges, especially when processing data of different scales and formats, which makes it difficult to achieve seamless integration. Second, traditional modeling methods often ignore the dynamic changes in the geological environment and cannot accurately reflect the long-term evolution of underground space. Furthermore, existing construction management methods mostly focus on static planning and lack the ability to respond in real time to dynamic changes during the construction process. Finally, in terms of anomaly detection and early warning, the sensitivity and accuracy of existing methods need to be improved, making it difficult to detect and address potential construction risks in a timely manner. Summary of the Invention

[0004] In view of this, an embodiment of the present invention provides an early warning method and system for compartment analysis and construction of large-scale foundation pits for subway double-sided construction, which is used to improve the efficiency and accuracy of compartment analysis and construction early warning of large-scale foundation pits for subway double-sided construction.

[0005] The present invention provides an early warning method for analyzing the construction of a large-scale foundation pit on both sides of a subway, comprising:

[0006] Perform multi-dimensional fusion processing on the pre-collected multi-source heterogeneous data of the subway station underground complex to obtain a comprehensive dataset;

[0007] Performing spatial interpolation and dynamic characteristic simulation on the comprehensive data set using a Kriging interpolation algorithm to obtain a three-dimensional geological structure that changes dynamically with construction;

[0008] Performing parametric multi-scale modeling on the pre-collected two-dimensional engineering drawings and three-dimensional scanning data and the three-dimensional geological structure to obtain three-dimensional representation data of the subway station structure containing multi-system information;

[0009] A multi-objective optimization algorithm is used to perform a three-dimensional finite element numerical simulation of the adjacent station expansion pit on the three-dimensional representation data of the subway station structure to obtain a target compartment scheme;

[0010] Performing a construction process simulation on the target compartment scheme and the three-dimensional representation data of the subway station structure to obtain digital representation data of the construction process including time and cost information;

[0011] Time series analysis and multi-dimensional feature extraction are performed on the digital representation data of the construction process and the real-time construction data collected in real time to obtain a construction process feature vector set, and trend prediction and anomaly detection are performed on the construction process feature vector set using a support vector regression algorithm to obtain construction process prediction results and anomaly alarm data.

[0012] The present invention also provides an early warning system for analyzing and constructing large-scale foundation pits on both sides of a subway, including:

[0013] The fusion module is used to perform multi-dimensional fusion processing on the pre-collected multi-source heterogeneous data of the subway station underground complex to obtain a comprehensive data set;

[0014] A simulation module, configured to perform spatial interpolation and dynamic characteristic simulation on the comprehensive data set using a Kriging interpolation algorithm to obtain a three-dimensional geological structure that changes dynamically with construction;

[0015] A modeling module is used to perform parametric multi-scale modeling on the pre-collected two-dimensional engineering drawings and three-dimensional scanning data and the three-dimensional geological structure to obtain three-dimensional representation data of the subway station structure containing multi-system information;

[0016] A compartmentalization module is used to perform compartmentalization analysis on the three-dimensional representation data of the subway station structure using a multi-objective optimization algorithm to perform three-dimensional finite element numerical simulation of the foundation pit for adjacent stations to obtain a target compartmentalization solution;

[0017] A simulation module is used to simulate the construction process of the target compartment scheme and the three-dimensional representation data of the subway station structure to obtain digital representation data of the construction process including time and cost information;

[0018] The detection module is used to perform time series analysis and multidimensional feature extraction on the digital representation data of the construction process and the real-time construction data collected in real time to obtain a construction process feature vector set, and perform trend prediction and anomaly detection on the construction process feature vector set through a support vector regression algorithm to obtain construction process prediction results and abnormal alarm data.

[0019] The technical solution provided by this invention utilizes multi-dimensional fusion processing of pre-collected multi-source, heterogeneous data from the subway station's underground complex to generate a comprehensive dataset. This effectively integrates data from different sources and formats, providing a comprehensive and unified data foundation for subsequent analysis. This fusion process not only improves data integrity and consistency, but also significantly reduces data redundancy and enhances data processing efficiency. Secondly, a Kriging interpolation algorithm is used to spatially interpolate and dynamically simulate the comprehensive dataset, generating a three-dimensional geological structure that reflects dynamic changes during construction. This significantly improves the accuracy and predictive power of the geological model. The Kriging interpolation algorithm accounts for spatial correlation, enabling more accurate estimation of geological parameters at unknown points, while the dynamic simulation captures the temporal evolution of the geological structure, providing more reliable geological information for engineering design and construction. Next, parametric multi-scale modeling is performed on the pre-collected two-dimensional engineering drawings, three-dimensional scanned data, and three-dimensional geological structure to generate a three-dimensional representation of the subway station structure containing information from multiple systems. This innovative modeling approach achieves seamless transition from two-dimensional to three-dimensional, organically integrating geological information with engineering structural information, significantly enhancing the model's integrity and practicality. The multi-scale modeling feature enables the model to adapt to varying accuracy requirements, reflecting both the overall structure and local details. A multi-objective optimization algorithm was used to perform a bin-by-bin analysis of the 3D representation of the subway station structure using a 3D finite element numerical simulation of the adjacent station excavation pit. This resulted in a target bin-by-bin solution, significantly improving the scientific and rational nature of construction planning. Intelligent bin-by-bin solution considers multiple objective functions, such as construction efficiency, cost control, and safety, and can find the optimal balance between multiple key indicators, providing strong support for project management. A construction process simulation was performed using the target bin-by-bin solution and the 3D representation of the subway station structure to generate a digital representation of the construction process, including time and cost information. This simulation not only foresees potential construction problems but also optimizes construction schedules and resource allocation, significantly improving efficiency and reducing risks. Finally, time series analysis and multidimensional feature extraction were performed on the digital representation of the construction process and real-time construction data collected in real time to generate a construction process feature vector set. Trend prediction and anomaly detection were performed on this feature vector set using a support vector regression algorithm, resulting in construction process prediction results and anomaly alert data. It realizes real-time monitoring and early warning of the construction process, can detect abnormal situations in time, prevent safety accidents and ensure construction quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 This is a flow chart of an early warning method for compartmentalized analysis and construction of large-scale foundation pits for bilateral expansion of subways in an embodiment of the present invention.

[0022] Figure 2 Schematic diagram of an early warning system for compartmentalized analysis and construction of large-scale foundation pits for subway double-sided expansion in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0024] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0025] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0026] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , Figure 1 This is a flow chart of an early warning method for analyzing and constructing a large-scale foundation pit on both sides of a subway according to an embodiment of the present invention. Figure 1 As shown, the following steps are included:

[0027] S101, performing multi-dimensional fusion processing on pre-collected multi-source heterogeneous data of the subway station underground complex to obtain a comprehensive data set;

[0028] S102, performing spatial interpolation and dynamic characteristic simulation on the comprehensive data set using a Kriging interpolation algorithm to obtain a three-dimensional geological structure that includes dynamic changes with construction;

[0029] S103, performing parametric multi-scale modeling on the pre-collected two-dimensional engineering drawings and three-dimensional scanning data and three-dimensional geological structure to obtain three-dimensional representation data of the subway station structure containing multi-system information;

[0030] S104, using a multi-objective optimization algorithm to perform a three-dimensional finite element numerical simulation of the adjacent station expansion pit on the three-dimensional representation data of the subway station structure, and obtain a target compartment scheme;

[0031] S105: Perform construction process simulation on the target compartment scheme and the three-dimensional representation data of the subway station structure to obtain digital representation data of the construction process including time and cost information;

[0032] S106. Perform time series analysis and multi-dimensional feature extraction on the digital representation data of the construction process and the real-time construction data collected in real time to obtain a construction process feature vector set, and perform trend prediction and anomaly detection on the construction process feature vector set using a support vector regression algorithm to obtain construction process prediction results and anomaly alarm data.

[0033] Specifically, multi-dimensional fusion processing is performed on pre-collected, heterogeneous data from multiple sources. This data includes geological survey reports, construction drawings, and historical monitoring data. Through data cleaning, standardization, and feature extraction, the data from different sources is unified in format and coordinate system. For example, for geological drilling data, it may be necessary to remove outliers, standardize depth units, and extract key stratigraphic information. Principal component analysis is used for dimensionality reduction to reduce data redundancy, while tensor decomposition algorithms are employed to explore potential relationships between the data. The resulting comprehensive dataset contains geological, structural, and environmental information. Next, the Kriging interpolation algorithm is used to perform spatial interpolation and dynamic property simulation on the comprehensive dataset. Kriging interpolation is a geostatistical method that accounts for spatial correlation and is suitable for interpolating geological data. First, a variogram analysis is performed on the data to calculate spatial correlation, and this information is then used to perform interpolation estimation. Furthermore, the temporal trends of geological characteristics are considered to construct dynamic property functions to predict geological states at future points. This not only fills data gaps between sampling points but also reflects the dynamic characteristics of geological structures.

[0034] Subsequently, parametric multiscale modeling was performed on the pre-collected 2D engineering drawings, 3D scanning data, and the 3D geological structure obtained in the previous step. First, the 2D drawings were vectorized to extract key structural lines and feature points. For the 3D scanning data, point cloud noise reduction and filtering were performed to extract key structural features. This data was then integrated with the 3D geological structure to construct a parametric geometric model. Multiscale modeling allows for the representation of structural details at different scales. For example, critical structural nodes can be represented with centimeter-level accuracy, while non-critical areas can be represented with meter-level accuracy. This approach ensures model accuracy while improving modeling efficiency. Next, a multi-objective optimization algorithm was used to perform a compartmentalized analysis of the 3D representation of the subway station structure, using a 3D finite element numerical simulation of the foundation pit for adjacent station expansion. Compartmentalization involves dividing the entire project into multiple relatively independent construction units. This approach considers multiple objectives, such as structural complexity, construction difficulty, and resource allocation. The 3D representation data was first spatially partitioned, and the structural complexity and construction difficulty of each sub-area were evaluated. A multi-objective genetic algorithm was used to optimize the compartmentalization scheme, balancing factors such as construction efficiency, cost, and safety.

[0035] Next, a construction process simulation was conducted based on the target compartmentalization scheme and the 3D representation data of the subway station structure. Construction steps were first broken down to develop a preliminary construction schedule. Resource allocation and cost estimation were then performed, and the entire construction process was simulated using a discrete event simulation algorithm. This simulation takes into account factors such as equipment operation and personnel operations, reflecting the various situations likely to be encountered in actual construction. Finally, time series analysis and multidimensional feature extraction were performed on the digital representation of the construction process and the real-time construction data collected in real time. Data was first synchronized and aligned, followed by time window partitioning and feature engineering. Principal component analysis was used for dimensionality reduction and key feature extraction. A predictive model was then constructed using the support vector regression algorithm to predict future construction status. Furthermore, anomaly thresholds and judgment rules were set to identify potential anomalies and generate anomaly alert data.

[0036] By executing the above steps, a comprehensive dataset was generated through multi-dimensional fusion of pre-collected, heterogeneous data from multiple sources within the subway station's underground complex. This effectively integrates data from different sources and formats, providing a comprehensive and unified data foundation for subsequent analysis. This fusion process not only improves data integrity and consistency, but also significantly reduces data redundancy and enhances data processing efficiency. Secondly, a Kriging interpolation algorithm was used to spatially interpolate and dynamically simulate the comprehensive dataset, generating a 3D geological structure that incorporates dynamic changes during construction. This significantly enhances the accuracy and predictive power of the geological model. The Kriging interpolation algorithm accounts for spatial correlation, enabling more accurate estimation of geological parameters at unknown points, while the dynamic simulation captures temporal variations in the geological structure, providing more reliable geological information for engineering design and construction. Next, parametric multi-scale modeling was performed on the pre-collected 2D engineering drawings, 3D scan data, and 3D geological structure, generating a 3D representation of the subway station structure that incorporates information from multiple systems. This innovative modeling approach enables seamless transition from 2D to 3D and organically integrates geological and engineering structural information, significantly improving the model's integrity and practicality. The multi-scale modeling feature enables the model to adapt to varying accuracy requirements, reflecting both the overall structure and local details. A multi-objective optimization algorithm was used to perform a bin-by-bin analysis of the 3D representation of the subway station structure using a 3D finite element numerical simulation of the adjacent station excavation pit. This resulted in a target bin-by-bin solution, significantly improving the scientific and rational nature of construction planning. Intelligent bin-by-bin solution considers multiple objective functions, such as construction efficiency, cost control, and safety, and can find the optimal balance between multiple key indicators, providing strong support for project management. A construction process simulation was performed using the target bin-by-bin solution and the 3D representation of the subway station structure to generate a digital representation of the construction process, including time and cost information. This simulation not only foresees potential construction problems but also optimizes construction schedules and resource allocation, significantly improving efficiency and reducing risks. Finally, time series analysis and multidimensional feature extraction were performed on the digital representation of the construction process and real-time construction data collected in real time to generate a construction process feature vector set. Trend prediction and anomaly detection were performed on this feature vector set using a support vector regression algorithm, resulting in construction process prediction results and anomaly alert data. It realizes real-time monitoring and early warning of the construction process, can detect abnormal situations in time, prevent safety accidents and ensure construction quality.

[0037] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0038] (1) Data cleaning is performed on the pre-collected multi-source heterogeneous data of the subway station underground complex to remove outliers and noise data to obtain preliminarily processed multi-source data;

[0039] (2) Perform data standardization on the multi-source data after preliminary processing, unify the dimensions and units of different data sources, and obtain standardized multi-source data;

[0040] (3) Perform feature extraction on the standardized multi-source data, extract the key features of each data source, and obtain multi-source feature data;

[0041] (4) Perform dimensionality reduction processing on multi-source feature data through principal component analysis algorithm to obtain feature data after dimensionality reduction;

[0042] (5) Perform data alignment on the feature data after dimensionality reduction to obtain aligned feature data;

[0043] (6) Perform multi-dimensional correlation analysis on the aligned feature data through the tensor decomposition algorithm to obtain the correlation feature data;

[0044] (7) Calculate the data fusion weight of the associated feature data to obtain the data fusion weight coefficient;

[0045] (8) Perform weighted fusion on the associated feature data according to the data fusion weight coefficient to obtain preliminary fusion data, and perform data consistency check on the preliminary fusion data to obtain verified fusion data;

[0046] (9) Perform spatiotemporal interpolation on the verified fused data to obtain complete fused data, and then perform data format unification on the complete fused data to obtain a comprehensive data set.

[0047] Specifically, data cleaning is performed to remove outliers and noise data. Outliers are data points that deviate significantly from the normal range, such as extreme readings caused by equipment failure. Noise data is inaccurate data caused by random fluctuations or interference. By setting appropriate thresholds and using statistical methods such as box plots or Z-score methods to identify and remove these outliers, we obtain preliminarily processed multi-source data. Next, the preliminarily processed multi-source data is normalized to unify the dimensions and units of different data sources. This is crucial because data from different sources may use different units or scales. For example, geological data may be measured in meters, while structural data may be measured in centimeters. Through normalization, all data are converted to the same scale range, typically [0, 1] or [-1, 1]. Common normalization methods include minimum-maximum normalization and Z-score normalization. This processed data facilitates subsequent analysis and comparison.

[0048] Feature extraction is then performed on the standardized multi-source data to extract the key features of each data source. Feature extraction aims to extract information from the raw data that best represents the essential characteristics of the data. For geological data, possible features to be extracted include stratum thickness, lithology, porosity, etc. For structural data, these may include key node coordinates, component dimensions, etc. By reducing the data dimension, the efficiency of subsequent processing is improved. The principal component analysis (PCA) algorithm is then used to reduce the dimensionality of the multi-source feature data. PCA is a commonly used unsupervised learning method that projects the original high-dimensional data into a new low-dimensional space through linear transformation while retaining the main information of the data. PCA first calculates the covariance matrix of the data, then solves the eigenvalues and eigenvectors, and selects the eigenvectors corresponding to the largest eigenvalues as the new basis. This can greatly reduce the data dimension while retaining the main variability of the data.

[0049] The next key step is to align the feature data after dimensionality reduction. Data alignment ensures that data from different sources are consistent in time and space. For example, data collected at different time points are mapped to the same time axis, or spatial data in different coordinate systems are converted to a unified reference system. This lays the foundation for subsequent multidimensional correlation analysis. Subsequently, multidimensional correlation analysis is performed on the aligned feature data using a tensor decomposition algorithm. Tensor decomposition is an effective tool for processing multidimensional data and can reveal hidden multidimensional relationships in the data. Common tensor decomposition methods include Tucker decomposition and CP decomposition. They can capture potential correlations between different data sources and provide an important basis for subsequent data fusion.

[0050] Next, data fusion weights are calculated for the associated feature data. This weight calculation takes into account the reliability, importance, and relevance of each data source. Common methods include expert scoring, entropy weighting, and the analytic hierarchy process. The calculated weight coefficients reflect the contribution of each data source to the final fusion result. The associated feature data are then weighted and fused based on the data fusion weight coefficients to produce preliminary fused data. The weighted fusion process can be understood as integrating information from different data sources to obtain a more comprehensive and reliable description. The preliminary fused data is then subjected to a data consistency check to ensure the reliability and coherence of the fusion results. This consistency check can be achieved through cross-validation or comparison with known reliable reference data.

[0051] Finally, the validated fused data undergoes spatiotemporal interpolation to fill in data gaps. Spatiotemporal interpolation takes into account the temporal and spatial distribution characteristics of the data. Common methods include kriging interpolation and inverse distance weighting. The resulting complete fused data covers the entire study area and timeframe. The final step is to standardize the format of the complete fused data to ensure that all data conform to predefined formatting standards for subsequent application and analysis. For example, in a subway station expansion project, the initial data included geotechnical parameters from 100 geological drilling points, 3D laser scanning point cloud data covering a 1,000-square-meter area, and 10 sets of 2D construction drawings. The data cleaning phase identified and removed 5% of outliers, such as significantly deviated stratum thickness data. Normalization unified data in different units (such as meters, centimeters, and megapascals) to the [0,1] range. The feature extraction phase extracted 20 key features from the geological data, including the average thickness and strength parameters of each stratum. Principal component analysis reduced the original 100-dimensional feature space to 30 dimensions, retaining 95% of the information. Data alignment unified data collected at different times onto a timeline based on the project's start date. Tensor decomposition revealed correlation patterns between geological conditions, structural loads, and settlement deformation. Weighting was performed to assign a weight of 0.4 to geological data, 0.35 to structural data, and 0.25 to monitoring data. After weighted fusion, a data consistency check identified and corrected 2% of data inconsistencies. The final spatiotemporal interpolation expanded the previously discrete data points into a 100 m x 100 m x 50 m three-dimensional grid, forming a continuous digital representation of the underground space.

[0052] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0053] (1) Partition the comprehensive dataset to obtain a partitioned dataset, perform variogram analysis on each sub-region in the partitioned dataset, calculate spatial correlation, and obtain variogram parameters;

[0054] (2) Perform Kriging weight calculation on each sub-region according to the variation function parameters to obtain the interpolation weight coefficient, and normalize the interpolation weight coefficient to obtain the normalized interpolation weight;

[0055] (3) Perform local Kriging interpolation on each sub-region in the partitioned dataset, estimate the unknown points by normalizing the interpolation weights, and obtain preliminary interpolation results;

[0056] (4) Perform boundary processing on the preliminary interpolation results to obtain continuous interpolation data;

[0057] (5) Perform time series analysis on the continuous interpolation data to identify the temporal variation trend of geological characteristics and obtain temporal variation characteristic data. Based on the temporal variation characteristic data, a dynamic characteristic function is constructed to describe the temporal variation of geological parameters and obtain the dynamic characteristic function.

[0058] (6) Using dynamic characteristic functions to extrapolate the continuous interpolation data, the geological state at future time points is predicted to obtain predicted geological data;

[0059] (7) Conduct uncertainty analysis on the predicted geological data, evaluate the reliability of the prediction results, obtain confidence interval data, and revise the predicted geological data based on the confidence interval data to obtain revised geological prediction data;

[0060] (8) The revised geological prediction data is reconstructed in three-dimensional space to generate a three-dimensional geological structure that changes dynamically with construction.

[0061] Specifically, the comprehensive dataset was partitioned to obtain a partitioned dataset. The purpose of data partitioning is to divide the entire study area into several sub-regions in order to more accurately capture local spatial variability. For each sub-region, a variogram analysis was performed to calculate spatial correlation and obtain variogram parameters. The variogram is a function that describes the correlation between spatial variables and is expressed as:

[0062]

[0063] in, is the variance function value, is the distance between sampling points, The interval is The number of sample logarithms, and The locations are and The sampling value at .

[0064] Kriging weight calculation is performed on each sub-region according to the variation function parameters to obtain the interpolation weight coefficient. Kriging interpolation is the optimal linear unbiased estimation method, and its interpolation formula is:

[0065]

[0066] in, is the predicted value of the point to be estimated, is the weight of the i-th known point, is the number of known points involved in the estimation.

[0067] The interpolation weight coefficients are normalized to obtain normalized interpolation weights. Normalization ensures that the sum of all weights is 1, which ensures the unbiasedness of the estimate. Then, local Kriging interpolation is performed on each sub-region in the partitioned data set, and the unknown points are estimated by the normalized interpolation weights to obtain preliminary interpolation results. Next, the preliminary interpolation results are subjected to boundary processing to obtain continuous interpolation data. Boundary processing solves the possible discontinuity problem between sub-regions and ensures a smooth transition of the interpolation results of the entire study area. Time series analysis is performed on the continuous interpolation data to identify the time-varying trends of geological characteristics and obtain time-varying characteristic data. Time series analysis methods include moving average, exponential smoothing, etc. A dynamic characteristic function is constructed based on the time-varying characteristic data to describe the changing law of geological parameters over time. The dynamic characteristic function can be expressed as:

[0068]

[0069] in, is the geological parameter over time The change function of , a, b, c are unknown coefficients, is a random error term.

[0070] Dynamic characteristic functions are used to extrapolate continuous interpolated data over time to predict geological conditions at future points in time, generating predicted geological data. Uncertainty analysis is performed on the predicted geological data to assess the reliability of the prediction results and generate confidence intervals. This uncertainty analysis considers factors such as model errors and parameter uncertainties and typically utilizes Monte Carlo simulations. The predicted geological data are then corrected based on the confidence intervals to generate corrected geological prediction data, improving the reliability of the prediction results. Finally, the corrected geological prediction data is reconstructed in three dimensions to generate a 3D geological structure that reflects the dynamic changes occurring during construction. During this reconstruction, a three-dimensional interpolation algorithm (such as trilinear interpolation) is used to convert the discrete predicted data points into a continuous three-dimensional representation.

[0071] For example, in a subway station expansion construction project, the study area is a spatial range of 100m×100m×50m, with a total of 200 geological drilling sampling points. First, the study area is divided into 10 sub-areas, with about 20 sampling points in each sub-area. Perform a variance function analysis on each sub-area to obtain the variance function parameters. For example, the variance function of the first sub-area is a spherical model with a base value of 0.8 and a range of 15m. The Kriging weights are calculated based on the variance function parameters. For the estimated point (50, 50, 25), the 5 nearest known points are selected for interpolation, and the normalized interpolation weights obtained are 0.3, 0.25, 0.2, 0.15, and 0.1, respectively. These weights are used to estimate the unknown points and obtain the geological parameter values of the point. Repeat this process for all sub-areas, and perform boundary processing to obtain continuous interpolation data for the entire study area. Time series analysis shows that the stratum settlement rate changes as a quadratic function, and the dynamic characteristic function is obtained by fitting. (Unit: mm / day). This function is used to predict the geological conditions for the next three months. Uncertainty analysis shows that the 95% confidence interval of the prediction result is ±2 mm.

[0072] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0073] (1) Vectorize the pre-collected two-dimensional engineering drawings, extract key structural lines and feature points, obtain vectorized drawing data, and perform coordinate system conversion on the vectorized drawing data to obtain vector data in a unified coordinate system;

[0074] (2) Perform point cloud denoising and filtering on the 3D scanning data to obtain simplified point cloud data, and perform feature recognition on the simplified point cloud data through feature extraction algorithms to extract key structural features and obtain a set of structural feature points;

[0075] (3) Data fusion is performed on the vector data and structural feature point set of the unified coordinate system to establish the correspondence between the two-dimensional drawing and the three-dimensional scanning data to obtain preliminary fusion data. Based on the preliminary fusion data, a parametric geometric model is constructed to define the structural parameters and constraints to obtain a parametric geometric expression.

[0076] (4) Perform multi-scale decomposition on the parameterized geometric expression to obtain multi-scale structural hierarchical data, and perform detail enhancement on the multi-scale structural hierarchical data to obtain enhanced multi-scale data;

[0077] (5) Spatial registration of the enhanced multi-scale data and the three-dimensional geological structure is performed to establish the spatial relationship between the geological environment and the engineering structure, and to obtain the three-dimensional representation data of the subway station structure containing multi-system information.

[0078] Specifically, two-dimensional engineering drawings are vectorized, converting the geometric information contained in them into a vector format that can be recognized and processed by computers. Vectorization involves steps such as edge detection, line tracing, and polygon recognition. These steps extract key structural lines and feature points, generating vectorized drawing data. Subsequently, the vectorized drawing data undergoes coordinate system conversion, unifying the local coordinate systems used by different drawings into a global coordinate system to generate vector data in the unified coordinate system. Next, point cloud denoising and filtering are performed on the three-dimensional scan data. Point cloud data consists of a large number of discrete three-dimensional coordinate points and often contains noise and redundant information. The denoising process uses a statistical outlier removal algorithm to remove points that significantly deviate from the main body. Filtering utilizes a voxel grid method, dividing the space into a grid of equal-sized cubes. Points within each grid are replaced by their centroids, resulting in a reduced point cloud data set. Feature extraction algorithms are then used to identify key structural features in the reduced point cloud data. Common feature extraction methods include normal vector estimation, curvature calculation, and local feature descriptors, which are used to generate a set of structural feature points. The next important step is to fuse the vector data and the structural feature point set in a unified coordinate system. To establish the correspondence between the 2D drawing and the 3D scan data, the iterative closest point (ICP) algorithm is typically used for registration. The ICP algorithm iteratively minimizes the distance between the two data sets, achieving precise alignment. This fusion generates preliminary fused data, based on which a parametric geometric model is constructed. Parametric modeling defines structural parameters and constraints, allowing the model to be flexibly changed in shape and size by adjusting the parameters, resulting in a parametric geometric representation.

[0079] The purpose of multi-scale decomposition of the parametric geometric expression is to capture structural details at different scales. Multi-scale decomposition usually uses wavelet transform or Laplace pyramid algorithm to decompose the original model into a hierarchical structure of different resolutions to obtain multi-scale structural hierarchical data. Subsequently, the multi-scale structural hierarchical data is enhanced in detail, and the detailed performance of the model at each scale is improved through techniques such as high-frequency information compensation or super-resolution reconstruction, thereby obtaining enhanced multi-scale data. Finally, the enhanced multi-scale data and the three-dimensional geological structure are spatially aligned to establish the spatial relationship between the geological environment and the engineering structure. Using spatial alignment algorithms, such as feature-based alignment or intensity-based alignment methods, the engineering structure model is accurately positioned in the geological environment, thereby obtaining three-dimensional representation data of the subway station structure containing multi-system information.

[0080] For example, in a subway station expansion project, the initial data included 10 sets of 2D CAD engineering drawings and 3D laser scanning point cloud data covering 1,000 square meters. After vectorization of the 2D drawings, 5,000 key structural lines and 2,000 feature points were extracted. Coordinate system conversion unified all data to the National Geodetic Coordinate System 2000. The 3D point cloud data originally contained 500 million points, which were reduced to 20 million points through noise reduction and filtering, preserving key structural information. A feature extraction algorithm identified 500,000 structural feature points from the reduced point cloud, including key components such as walls, columns, and pipes. The data fusion process employed the ICP algorithm through 20 iterations, ultimately achieving precise alignment of the 2D and 3D data with a fusion error of less than 5 mm. The parametric geometric model defined 200 key parameters and 500 constraints, enabling flexible adjustment of key dimensions such as platform width and column spacing. Multi-scale decomposition divided the model into five scale levels, ranging from centimeter-level accuracy to meter-level accuracy. After detail enhancement, the highest resolution reached millimeter-level accuracy. Finally, by aligning with the three-dimensional geological structure, the engineering structure was accurately matched with the surrounding geological environment within a depth of 50 meters, forming a comprehensive three-dimensional representation dataset containing multi-system information such as structure, geology, and equipment.

[0081] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0082] (1) Spatially divide the three-dimensional representation data of the subway station structure, obtain an initial zoning scheme based on the acquired adjacent station expansion scope, and optimize the boundaries of the initial zoning scheme to obtain an optimized zoning scheme;

[0083] (2) Evaluate the structural complexity of each sub-region in the optimized zoning scheme. Calculate the structural density and connectivity indicators, including neighboring stations, using a three-dimensional finite element algorithm to obtain complexity score data. Based on the complexity score data, perform a preliminary bin division of the sub-regions to obtain candidate bin division schemes.

[0084] (3) Analyze the construction difficulty of the candidate warehouse division plan to obtain the construction difficulty index, and adjust the candidate warehouse division plan according to the construction difficulty index to obtain the adjusted warehouse division plan;

[0085] (4) Conduct resource allocation simulation on the adjusted warehouse allocation plan to obtain resource demand data, and balance and optimize the warehouse allocation plan based on the resource demand data to obtain a balanced warehouse allocation plan;

[0086] (5) The balanced partitioning scheme is optimized through a multi-objective optimization algorithm to obtain the target partitioning scheme.

[0087] Specifically, the intelligent bin analysis of the three-dimensional representation data of the subway station structure is a complex process. First, the three-dimensional representation data of the subway station structure is spatially divided to obtain an initial partitioning scheme. Using the octree segmentation algorithm, the entire three-dimensional space is recursively divided into several subspaces according to the obtained neighboring station expansion range, until the complexity of each subspace meets the preset threshold. Subsequently, the boundaries of the initial partitioning scheme are optimized, and the sub-region boundaries are adjusted using the Markov random field model to better adapt to the structural characteristics, thereby obtaining an optimized partitioning scheme. The structural complexity of each sub-region in the optimized partitioning scheme is evaluated by three-dimensional finite element algorithm calculation, and the structural density and connectivity indicators are calculated. Structural density index The calculation formula is:

[0088]

[0089] in, is the volume of the structural member within the subregion, is the total volume of the subregion. Connectivity index The calculation formula is:

[0090]

[0091] in, is the number of components connected to other sub-areas, is the total number of components in the sub-area. Based on these two indicators, we obtain the complexity score data, perform a preliminary division of the sub-area, and obtain candidate division schemes.

[0092] Next, we analyze the construction difficulty of the candidate warehouse solutions and obtain the construction difficulty index. The calculation formula is:

[0093]

[0094] in, , , , is the weight coefficient, is the accessibility indicator, is the spatial constraint indicator. The candidate partitioning schemes are adjusted based on the construction difficulty indicator to obtain the adjusted partitioning scheme.

[0095] Perform resource allocation simulation on the adjusted warehouse allocation plan, use the linear programming model to calculate the resource requirements of each warehouse, and obtain resource requirement data. The objective function of the resource allocation model is:

[0096]

[0097] in, Is to use resources Allocate to sub-warehouses The cost, is the allocation decision variable. The partitioning scheme is balanced and optimized according to the resource demand data to obtain the balanced partitioning scheme. Finally, the balanced partitioning scheme is optimized by a multi-objective genetic algorithm. The mathematical expression of the multi-objective optimization problem is:

[0098]

[0099] in, is the objective function vector, including construction efficiency, cost, safety and other objectives, and These are inequality and equality constraints, respectively. This optimization problem is solved using a multi-objective genetic algorithm, ultimately yielding the target warehouse allocation solution.

[0100] For example, the three-dimensional representation data of a subway station expansion project contains 1 million grid cells. The initial space partitioning uses the octree algorithm to divide the space into 64 sub-regions. After boundary optimization, the number of sub-regions is adjusted to 58. In the structural complexity assessment, the structural density index and connectivity index of each sub-region are calculated. For example, the structural density of a sub-region , connectivity The initial division merged 58 sub-areas into 20 candidate sub-areas. In the construction difficulty analysis, the weight coefficient is assumed to be , , , , the construction difficulty index of a sub-warehouse The resource allocation simulation considered three types of resources: manpower, equipment, and materials. The linear programming model included 60 decision variables and 100 constraints. The multi-objective genetic algorithm was set to a population size of 100 and an evolutionary number of 500. The resulting target partitioning scheme included 15 partitions, with an average construction difficulty index of 0.65 for each partition, improving resource utilization by 15%.

[0101] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0102] (1) Decompose the construction process of the target warehouse plan to obtain a detailed process list, and formulate a preliminary construction schedule based on the detailed process list to obtain an initial schedule;

[0103] (2) Allocate resources to the initial schedule to obtain a resource allocation plan, and adjust the initial schedule based on the resource allocation plan to obtain an optimized schedule;

[0104] (3) Conduct construction cost estimation for the optimized schedule, calculate the direct and indirect costs of each process, obtain cost estimation data, and optimize the schedule based on the cost estimation data to obtain a cost-optimized schedule;

[0105] (4) Using discrete event simulation algorithms to simulate the construction process of the cost-optimized schedule, obtain construction process simulation data, and perform statistical analysis on the construction process simulation data to obtain construction performance evaluation results;

[0106] (5) Based on the results of the construction performance evaluation, the construction plan is optimized and adjusted, the schedule and resource allocation are updated, and digital representation data of the construction process containing time and cost information is obtained.

[0107] Specifically, the target sub-division plan is decomposed into construction steps. Using the Work Breakdown Structure (WBS) technique, the construction tasks for each sub-division are broken down into specific steps. This step decomposition takes into account construction logic, technical requirements, and resource constraints, resulting in a detailed step list. Subsequently, a preliminary construction schedule is developed based on the detailed step list. The Critical Path Method (CPM) is used to determine the dependencies and timings between steps, resulting in an initial schedule. Next, resources are allocated to the initial schedule. This resource allocation process utilizes a resource balancing algorithm, taking into account the availability and efficiency of resources such as manpower, equipment, and materials. The algorithm first identifies resource conflicts and then resolves them by adjusting the start times of non-critical steps, resulting in a resource allocation plan. Based on the resource allocation plan, the initial schedule is adjusted, which may involve reordering the steps or modifying their durations, ultimately resulting in an optimized schedule.

[0108] Estimating construction costs for the optimized schedule is the next important step. Cost estimation involves calculating direct and indirect costs. Direct costs include material, labor, and equipment costs and are derived through bill of quantities and unit price analysis. Indirect costs, such as management fees and taxes, are calculated using a ratio or quota method. These calculations generate cost estimates, which are then used to optimize the schedule using cost-time trade-off analysis (TCTO) techniques. TCTO analysis adjusts construction duration and resource allocation to find the lowest total cost solution, resulting in a cost-optimized schedule. The construction process of this cost-optimized schedule is then simulated using a discrete event simulation algorithm. Discrete event simulation treats the construction process as a series of discrete events, simulating factors such as the construction environment, equipment operation, and personnel operations. During the simulation, Monte Carlo methods are used to generate random variables to account for construction uncertainties, such as weather effects and equipment failures. The simulation results provide construction process data, including the actual start and end times of each process, resource utilization, and cost consumption.

[0109] Statistical analysis of construction process simulation data is performed to calculate key performance indicators, such as schedule achievement rate, cost control rate, and resource utilization efficiency. These indicators are used to generate construction performance evaluation results, providing a basis for subsequent optimization. Finally, the construction plan is optimized and adjusted based on the construction performance evaluation results. The optimization process uses heuristic algorithms, such as simulated annealing or genetic algorithms, to adjust process scheduling, resource allocation, and construction methods while satisfying constraints to improve overall construction efficiency. The optimization results are used to update the schedule and resource allocation, ultimately generating a digital representation of the construction process that includes time and cost information.

[0110] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0111] (1) Synchronize and align the digital representation data of the construction process and the real-time construction data collected in real time to obtain aligned time series data, and clean the aligned time series data to obtain cleaned time series data;

[0112] (2) Divide the cleaned time series data into time windows, set the appropriate time window size, obtain the segmented time series data set, and perform feature engineering on the segmented time series data set to extract time domain and frequency domain features to obtain the preliminary feature set;

[0113] (3) The preliminary feature set is reduced in dimension by principal component analysis to obtain the feature data after dimensionality reduction. The feature importance of the feature data after dimensionality reduction is evaluated to obtain the feature vector set of the construction process;

[0114] (4) Using the support vector regression algorithm to train the construction process feature vector set, a prediction model is constructed to obtain the trained prediction model, and the trained prediction model is used to predict the construction status at future time points to obtain the construction process prediction results;

[0115] (5) Perform anomaly detection on the predicted results of the construction process, set anomaly thresholds and judgment rules, identify potential anomalies, and obtain anomaly alarm data.

[0116] It is important to note that these two types of data are synchronized and aligned to ensure timestamp consistency. Using the dynamic time warping (DTW) algorithm, data with different sampling frequencies and time scales are aligned onto the same timeline, resulting in aligned time series data. Subsequently, the aligned time series data undergoes data cleaning, including outlier removal, missing value filling, and noise reduction. Outlier detection utilizes the local outlier factor (LOF) algorithm, missing value filling utilizes multiple interpolation, and noise reduction utilizes wavelet transform denoising, ultimately resulting in cleaned time series data. Next, the cleaned time series data is partitioned into time windows. The size of the time window is determined based on the data characteristics and analysis objectives, typically using a sliding window technique, with window sizes ranging from a few hours to several days. This partitioning results in a segmented time series dataset, with each segment containing complete data within a specific time range. Feature engineering is then performed on the segmented time series dataset to extract time and frequency domain features. Time domain features include statistical quantities such as mean, standard deviation, kurtosis, and skewness, while frequency domain features are extracted through fast Fourier transform (FFT), including main frequency components, power spectrum density, etc. These features constitute the preliminary feature set.

[0117] Subsequently, principal component analysis (PCA) was used to reduce the dimensionality of the preliminary feature set. PCA calculates the feature covariance matrix to identify the main eigenvectors, projecting the high-dimensional data into a low-dimensional space while preserving the main variability in the data. This dimensionality reduction not only reduces the data volume but also removes redundant information. Feature importance was then assessed on the reduced feature data. A random forest algorithm was used to calculate the importance score of each feature, and features with high scores were selected to form the construction process feature vector set. Next, a model was trained on the construction process feature vector set using the support vector regression (SVR) algorithm. SVR is a powerful regression algorithm capable of handling nonlinear relationships. During the training process, grid search and cross-validation were used to determine the optimal kernel function and hyperparameters. After training, a predictive model was obtained that captured the complex patterns and trends in the construction process. The trained predictive model was then used to predict the construction status at future points in time, resulting in construction process prediction results.

[0118] Finally, anomaly detection is performed on the construction process prediction results. This anomaly detection uses a density-based local anomaly detection algorithm (LOF), with defined anomaly thresholds and judgment rules. The LOF algorithm calculates the local density of each data point relative to its neighborhood. Outliers typically have lower local density. By comparing the LOF score with the preset threshold, potential outliers are identified and anomaly alert data is generated.

[0119] For example, the digital representation of the construction process for a subway station expansion project included 1,000 monitoring points, with data collected hourly for 30 days. Real-time construction data was collected every 15 minutes. Data synchronization and alignment unified the two types of data to an hourly scale, resulting in 72,000 data points. The data cleaning phase identified and addressed approximately 2% of outliers and missing values. Time windows were partitioned using a 6-hour sliding window, resulting in 120 time windows. Feature engineering extracted 20 time-domain features and 10 frequency-domain features from each window. PCA dimensionality reduction reduced the 30-dimensional features to 10 dimensions, retaining 95% of the information. Feature importance assessment selected the seven most important features to form the feature vector set. The SVR model was trained using 80% of the data as a training set and 20% as a test set. The model achieved a mean absolute error of 2.5% on the test set. The prediction model predicted the construction status for the next seven days. Anomaly detection sets the LOF threshold to 1.5 and identifies 15 potential anomalies, 10 of which are confirmed to be true anomalies requiring further attention and processing.

[0120] In a specific embodiment, the process of performing anomaly detection on the construction process prediction results, setting anomaly thresholds and judgment rules, identifying potential anomalies, and obtaining anomaly alarm data may specifically include the following steps:

[0121] (1) Perform data standardization on the construction process prediction results, convert different indicators to the same scale, obtain standardized prediction data, and conduct historical data comparison and analysis on the standardized prediction data, calculate the deviation from the historical normal value, and obtain the deviation data;

[0122] (2) Statistically analyze the deviation data to obtain statistical characteristic values, and set preliminary abnormal thresholds based on the statistical characteristic values to obtain preliminary threshold settings;

[0123] (3) fine-tuning the initial threshold setting to obtain an adjusted threshold setting, and formulating anomaly judgment rules based on the adjusted threshold setting to obtain an anomaly judgment rule set;

[0124] (4) Dynamic threshold calculation is performed on the standardized prediction data through the sliding window method. According to the time-varying characteristics of the data, a dynamic threshold sequence is obtained. The dynamic threshold sequence is then integrated with the anomaly judgment rule set to obtain an anomaly detection scheme.

[0125] (5) According to the anomaly detection scheme, abnormal data points are identified on the standardized prediction data to obtain abnormal alarm data.

[0126] It should be noted that the construction process prediction results are standardized to convert different indicators to the same scale. The standardization process uses the Z-score method to obtain standardized prediction data, which is then compared and analyzed with historical data to calculate the deviation from the historical normal value. Then, the deviation data is statistically analyzed to calculate the mean. and standard deviation , these statistical characteristic values are used to set the preliminary anomaly threshold. The preliminary anomaly threshold is usually set as:

[0127]

[0128] in, k It is an adjustable parameter, usually set to 2 or 3. This gives the initial threshold setting. is the preliminary anomaly threshold, which indicates the limit value for judging whether a data point is abnormal. The mean of the deviation data reflects the average difference between the forecast value and the historical normal value. is the standard deviation of the deviation data, which indicates the degree of dispersion of the deviation.

[0129] Subsequently, the initial threshold setting is fine-tuned. The fine-tuning process takes into account expert experience and historical abnormal data, and the adjustment formula is:

[0130]

[0131] in, is the threshold adjustment value, The adjustment amount is determined based on expert knowledge. Based on the adjusted threshold setting, anomaly judgment rules are formulated to form an anomaly judgment rule set.

[0132] Next, the dynamic threshold calculation is performed on the standardized prediction data using the sliding window method. The local statistical features are calculated within each window, and the dynamic threshold calculation formula is:

[0133]

[0134] in, and The time The mean and standard deviation in the window. Get the dynamic threshold sequence .

[0135] Finally, the dynamic threshold sequence is fused with the anomaly judgment rule set to form an anomaly detection solution. The fusion process uses the weighted average method:

[0136]

[0137] in, is the weight coefficient. It is the final anomaly detection threshold. According to this anomaly detection scheme, abnormal data points are identified on the standardized prediction data to obtain abnormal alarm data.

[0138] The embodiment of the present invention also provides an early warning system for analyzing the construction of large-scale foundation pits on both sides of the subway. Figure 2 As shown in the figure, the early warning system for the analysis and construction of the large foundation pit on both sides of the subway specifically includes:

[0139] The fusion module 201 is used to perform multi-dimensional fusion processing on the pre-collected multi-source heterogeneous data of the subway station underground complex to obtain a comprehensive data set;

[0140] A simulation module 202 is used to perform spatial interpolation and dynamic characteristic simulation on the comprehensive data set using a Kriging interpolation algorithm to obtain a three-dimensional geological structure that changes dynamically with construction;

[0141] Modeling module 203, configured to perform parametric multi-scale modeling on the pre-collected two-dimensional engineering drawings and three-dimensional scanning data and the three-dimensional geological structure to obtain three-dimensional representation data of the subway station structure containing multi-system information;

[0142] The compartment division module 204 is configured to perform a compartment division analysis of the three-dimensional representation data of the subway station structure by a three-dimensional finite element numerical simulation of the adjacent station expansion pit using a multi-objective optimization algorithm to obtain a target compartment division scheme;

[0143] A simulation module 205 is configured to simulate the construction process based on the target compartment scheme and the three-dimensional representation data of the subway station structure to obtain digital representation data of the construction process including time and cost information;

[0144] The detection module 206 is used to perform time series analysis and multi-dimensional feature extraction on the digital representation data of the construction process and the real-time construction data collected in real time to obtain a construction process feature vector set, and perform trend prediction and anomaly detection on the construction process feature vector set through a support vector regression algorithm to obtain construction process prediction results and abnormal alarm data.

[0145] Through the collaborative work of these modules, a comprehensive dataset is generated by multi-dimensionally fusion processing of pre-collected, heterogeneous data from multiple sources within the subway station's underground complex. This effectively integrates data from diverse sources and formats, providing a comprehensive and unified data foundation for subsequent analysis. This fusion process not only improves data integrity and consistency, but also significantly reduces data redundancy and enhances data processing efficiency. Secondly, a Kriging interpolation algorithm is used to spatially interpolate and dynamically simulate the comprehensive dataset, generating a 3D geological structure that incorporates dynamic changes during construction. This significantly enhances the accuracy and predictive power of the geological model. The Kriging interpolation algorithm accounts for spatial correlation, enabling more accurate estimation of geological parameters at unknown points, while the dynamic simulation captures temporal variations in the geological structure, providing more reliable geological information for engineering design and construction. Next, parametric multi-scale modeling is performed on pre-collected 2D engineering drawings, 3D scan data, and 3D geological structures, resulting in a 3D representation of the subway station structure that incorporates information from multiple systems. This innovative modeling approach enables seamless transition from 2D to 3D and organically integrates geological and engineering structural information, significantly improving the model's integrity and practicality. The multi-scale modeling feature enables the model to adapt to varying accuracy requirements, reflecting both the overall structure and local details. A multi-objective optimization algorithm was used to perform a bin-by-bin analysis of the 3D representation of the subway station structure using a 3D finite element numerical simulation of the adjacent station excavation pit. This resulted in a target bin-by-bin solution, significantly improving the scientific and rational nature of construction planning. Intelligent bin-by-bin solution considers multiple objective functions, such as construction efficiency, cost control, and safety, and can find the optimal balance between multiple key indicators, providing strong support for project management. A construction process simulation was performed using the target bin-by-bin solution and the 3D representation of the subway station structure to generate a digital representation of the construction process, including time and cost information. This simulation not only foresees potential construction problems but also optimizes construction schedules and resource allocation, significantly improving efficiency and reducing risks. Finally, time series analysis and multidimensional feature extraction were performed on the digital representation of the construction process and real-time construction data collected in real time to generate a construction process feature vector set. Trend prediction and anomaly detection were performed on this feature vector set using a support vector regression algorithm, resulting in construction process prediction results and anomaly alert data. It realizes real-time monitoring and early warning of the construction process, can detect abnormal situations in time, prevent safety accidents and ensure construction quality.

[0146] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. An early warning method for analyzing and constructing large foundation pits on both sides of a subway, characterized by: include: Perform multi-dimensional fusion processing on the pre-collected multi-source heterogeneous data of the subway station underground complex to obtain a comprehensive dataset; Performing spatial interpolation and dynamic characteristic simulation on the comprehensive data set using a Kriging interpolation algorithm to obtain a three-dimensional geological structure that changes dynamically with construction; Performing parametric multi-scale modeling on the pre-collected two-dimensional engineering drawings and three-dimensional scanning data and the three-dimensional geological structure to obtain three-dimensional representation data of the subway station structure containing multi-system information; A multi-objective optimization algorithm is used to perform a three-dimensional finite element numerical simulation of the adjacent station expansion pit on the three-dimensional representation data of the subway station structure to obtain a target compartment scheme; The step of performing a three-dimensional finite element numerical simulation and compartmental analysis of the adjacent station expansion pit on the three-dimensional representation data of the subway station structure by a multi-objective optimization algorithm to obtain a target compartmentalization scheme includes: The three-dimensional representation data of the subway station structure is spatially divided, and an initial zoning scheme is obtained according to the obtained expansion scope of the neighboring stations, and the boundaries of the initial zoning scheme are optimized to obtain an optimized zoning scheme; a structural complexity assessment is performed on each sub-area in the optimized zoning scheme, and the structural density and connectivity index including the neighboring stations are calculated by a three-dimensional finite element algorithm to obtain complexity score data, and the sub-areas are preliminarily divided into warehouses according to the complexity score data to obtain candidate warehouse schemes; a construction difficulty analysis is performed on the candidate warehouse schemes to obtain a construction difficulty index, and the candidate warehouse schemes are adjusted according to the construction difficulty index to obtain an adjusted warehouse scheme; a resource allocation simulation is performed on the adjusted warehouse scheme to obtain resource demand data, and the warehouse scheme is balanced and optimized according to the resource demand data to obtain a balanced warehouse scheme; a multi-objective optimization algorithm is used to perform multi-objective optimization on the balanced warehouse scheme to obtain a target warehouse scheme; Performing a construction process simulation on the target compartment scheme and the three-dimensional representation data of the subway station structure to obtain digital representation data of the construction process including time and cost information; Time series analysis and multi-dimensional feature extraction are performed on the digital representation data of the construction process and the real-time construction data collected in real time to obtain a construction process feature vector set, and trend prediction and anomaly detection are performed on the construction process feature vector set using a support vector regression algorithm to obtain construction process prediction results and anomaly alarm data.

2. The early warning method for analyzing and constructing a large-scale foundation pit on both sides of a subway according to claim 1 is characterized in that: The step of performing multi-dimensional fusion processing on the pre-collected multi-source heterogeneous data of the subway station underground complex to obtain a comprehensive data set includes: Data cleaning is performed on the pre-collected multi-source heterogeneous data of the subway station underground complex to remove outliers and noise data to obtain preliminarily processed multi-source data; Performing data standardization on the preliminarily processed multi-source data to unify the dimensions and units of different data sources to obtain standardized multi-source data; Performing feature extraction on the standardized multi-source data to extract key features of each data source to obtain multi-source feature data; Performing dimensionality reduction processing on the multi-source feature data by using a principal component analysis algorithm to obtain feature data after dimensionality reduction; Performing data alignment on the feature data after dimensionality reduction to obtain aligned feature data; Performing multi-dimensional correlation analysis on the aligned feature data using a tensor decomposition algorithm to obtain correlation feature data; Performing data fusion weight calculation on the associated feature data to obtain a data fusion weight coefficient; Performing weighted fusion on the associated feature data according to the data fusion weight coefficient to obtain preliminary fusion data, and performing data consistency check on the preliminary fusion data to obtain verified fusion data; The verified fused data is subjected to spatiotemporal interpolation to obtain complete fused data, and the complete fused data is subjected to data format unification processing to obtain the comprehensive data set.

3. The early warning method for analyzing and constructing a large-scale foundation pit on both sides of a subway according to claim 1 is characterized in that: The step of performing spatial interpolation and dynamic characteristic simulation on the comprehensive data set by using the Kriging interpolation algorithm to obtain a three-dimensional geological structure that changes dynamically with construction includes: Partitioning the comprehensive data set to obtain a partitioned data set, performing a variogram analysis on each sub-region in the partitioned data set, calculating spatial correlation, and obtaining variogram parameters; Performing Kriging weight calculation on each sub-region according to the variation function parameters to obtain an interpolation weight coefficient, and normalizing the interpolation weight coefficient to obtain a normalized interpolation weight; Performing local Kriging interpolation on each sub-region in the partitioned dataset, estimating unknown points using the normalized interpolation weights, and obtaining preliminary interpolation results; Performing boundary processing on the preliminary interpolation result to obtain continuous interpolation data; Performing time series analysis on the continuous interpolation data to identify the temporal variation trend of the geological characteristics to obtain temporal variation characteristic data, and constructing a dynamic characteristic function based on the temporal variation characteristic data to describe the temporal variation pattern of the geological parameters to obtain the dynamic characteristic function; Performing time extrapolation on the continuous interpolation data using the dynamic characteristic function to predict the geological state at a future time point to obtain predicted geological data; performing uncertainty analysis on the predicted geological data, evaluating the reliability of the prediction results, obtaining confidence interval data, and revising the predicted geological data based on the confidence interval data to obtain revised geological prediction data; The revised geological prediction data is reconstructed in three-dimensional space to generate a three-dimensional geological structure that changes dynamically with construction.

4. The early warning method for analyzing and constructing a large-scale foundation pit on both sides of a subway according to claim 1 is characterized in that: The step of performing parametric multi-scale modeling on the pre-collected two-dimensional engineering drawings and three-dimensional scanning data and the three-dimensional geological structure to obtain three-dimensional representation data of the subway station structure containing multi-system information includes: Performing vectorization processing on the pre-collected two-dimensional engineering drawings, extracting key structural lines and feature points to obtain vectorized drawing data, and performing coordinate system conversion on the vectorized drawing data to obtain vector data of a unified coordinate system; Performing point cloud denoising and filtering on the three-dimensional scanning data to obtain simplified point cloud data, and performing feature recognition on the simplified point cloud data using a feature extraction algorithm to extract key structural features to obtain a structural feature point set; Performing data fusion on the vector data of the unified coordinate system and the set of structural feature points, establishing a correspondence between the two-dimensional drawing and the three-dimensional scan data, obtaining preliminary fused data, and constructing a parametric geometric model based on the preliminary fused data, defining structural parameters and constraints, and obtaining a parametric geometric expression; Performing multi-scale decomposition on the parameterized geometric expression to obtain multi-scale structural hierarchical data, and performing detail enhancement on the multi-scale structural hierarchical data to obtain enhanced multi-scale data; The enhanced multi-scale data and the three-dimensional geological structure are spatially aligned to establish a spatial relationship between the geological environment and the engineering structure, thereby obtaining three-dimensional representation data of the subway station structure containing multi-system information.

5. The early warning method for analyzing and constructing a large-scale foundation pit on both sides of a subway according to claim 1 is characterized in that: The step of simulating the construction process of the target compartment scheme and the three-dimensional representation data of the subway station structure to obtain digital representation data of the construction process containing time and cost information includes: Decomposing the target warehouse plan into construction procedures to obtain a detailed procedure list, and formulating a preliminary construction schedule based on the detailed procedure list to obtain an initial schedule; Allocating resources to the initial schedule to obtain a resource allocation plan, and adjusting the initial schedule according to the resource allocation plan to obtain an optimized schedule plan; Performing construction cost estimation on the optimized schedule, calculating direct cost and indirect cost of each process to obtain cost estimation data, and performing cost optimization on the schedule based on the cost estimation data to obtain a cost-optimized schedule; Performing a construction process simulation on the cost-optimized schedule using a discrete event simulation algorithm to obtain construction process simulation data, and performing statistical analysis on the construction process simulation data to obtain a construction performance evaluation result; The construction plan is optimized and adjusted according to the construction performance evaluation results, the schedule and resource allocation are updated, and digital representation data of the construction process including time and cost information is obtained.

6. The early warning method for analyzing and constructing a large-scale foundation pit on both sides of a subway according to claim 5 is characterized in that: The steps of performing time series analysis and multi-dimensional feature extraction on the digital representation data of the construction process and the real-time construction data collected in real time to obtain a construction process feature vector set, and performing trend prediction and anomaly detection on the construction process feature vector set using a support vector regression algorithm to obtain a construction process prediction result and anomaly alarm data include: Performing data synchronization and alignment on the digital representation data of the construction process and the real-time construction data collected in real time to obtain aligned time series data, and performing data cleaning on the aligned time series data to obtain cleaned time series data; Divide the cleaned time series data into time windows, set an appropriate time window size, obtain a segmented time series data set, and perform feature engineering on the segmented time series data set to extract time domain and frequency domain features to obtain a preliminary feature set; Performing dimensionality reduction processing on the preliminary feature set through principal component analysis to obtain feature data after dimensionality reduction, and performing feature importance evaluation on the feature data after dimensionality reduction to obtain a construction process feature vector set; Performing model training on the construction process feature vector set using a support vector regression algorithm to construct a prediction model to obtain a trained prediction model, and using the trained prediction model to predict the construction status at a future time point to obtain a construction process prediction result; Anomaly detection is performed on the construction process prediction results, anomaly thresholds and judgment rules are set, potential anomalies are identified, and anomaly alarm data is obtained.

7. The early warning method for analyzing and constructing a large-scale foundation pit on both sides of a subway according to claim 6 is characterized in that: The steps of performing anomaly detection on the construction process prediction results, setting anomaly thresholds and judgment rules, identifying potential anomalies, and obtaining anomaly alarm data include: Performing data standardization on the construction process prediction results, converting different indicators to the same scale to obtain standardized prediction data, and performing historical data comparative analysis on the standardized prediction data, calculating the deviation from historical normal values, and obtaining deviation data; Performing statistical analysis on the deviation data to obtain statistical characteristic values, and setting preliminary abnormality thresholds according to the statistical characteristic values to obtain preliminary threshold settings; Fine-tuning the preliminary threshold setting to obtain an adjusted threshold setting, and formulating an anomaly determination rule based on the adjusted threshold setting to obtain an anomaly determination rule set; Performing dynamic threshold calculation on the standardized prediction data using a sliding window method, obtaining a dynamic threshold sequence based on the time-varying characteristics of the data, and fusing the dynamic threshold sequence with the anomaly judgment rule set to obtain an anomaly detection solution; Abnormal data points are identified on the standardized prediction data according to the abnormality detection scheme to obtain abnormal alarm data.

8. An early warning system for analyzing and constructing large-scale foundation pits on both sides of a subway, for executing the early warning method for analyzing and constructing large-scale foundation pits on both sides of a subway as claimed in any one of claims 1 to 7, characterized in that: include: The fusion module is used to perform multi-dimensional fusion processing on the pre-collected multi-source heterogeneous data of the subway station underground complex to obtain a comprehensive data set; A simulation module, configured to perform spatial interpolation and dynamic characteristic simulation on the comprehensive data set using a Kriging interpolation algorithm to obtain a three-dimensional geological structure that changes dynamically with construction; A modeling module is used to perform parametric multi-scale modeling on the pre-collected two-dimensional engineering drawings and three-dimensional scanning data and the three-dimensional geological structure to obtain three-dimensional representation data of the subway station structure containing multi-system information; A compartmentalization module is used to perform compartmentalization analysis on the three-dimensional representation data of the subway station structure using a multi-objective optimization algorithm to perform three-dimensional finite element numerical simulation of the foundation pit for adjacent stations to obtain a target compartmentalization solution; The partitioning module is specifically used to: spatially divide the three-dimensional representation data of the subway station structure, obtain an initial partitioning scheme based on the acquired neighboring station expansion range, and perform boundary optimization on the initial partitioning scheme to obtain an optimized partitioning scheme; perform structural complexity evaluation on each sub-area in the optimized partitioning scheme, calculate the structural density and connectivity index including the neighboring stations through a three-dimensional finite element algorithm to obtain complexity scoring data, and perform preliminary partitioning of the sub-areas based on the complexity scoring data to obtain candidate partitioning schemes; perform construction difficulty analysis on the candidate partitioning schemes to obtain construction difficulty indicators, and adjust the candidate partitioning schemes based on the construction difficulty indicators to obtain adjusted partitioning schemes; perform resource allocation simulation on the adjusted partitioning scheme to obtain resource demand data, and perform balance optimization on the partitioning scheme based on the resource demand data to obtain a balanced partitioning scheme; perform multi-objective optimization on the balanced partitioning scheme through a multi-objective optimization algorithm to obtain a target partitioning scheme; A simulation module is used to simulate the construction process of the target compartment scheme and the three-dimensional representation data of the subway station structure to obtain digital representation data of the construction process including time and cost information; The detection module is used to perform time series analysis and multidimensional feature extraction on the digital representation data of the construction process and the real-time construction data collected in real time to obtain a construction process feature vector set, and perform trend prediction and anomaly detection on the construction process feature vector set through a support vector regression algorithm to obtain construction process prediction results and abnormal alarm data.

Images