Track deformation analysis method and system combining BIM and finite elements

Abstract

The application discloses a kind of track deformation analysis method and system combined with BIM and finite element, it is related to data processing technical field.The method includes: collecting the BIM model parameter and multivariate environmental characteristics of sample track section, obtain sample track BIM parameter and sample environmental characteristic parameter set, collect the track deformation parameter of sample track section, obtain sample track deformation parameter;Processing obtains track deformation wavelength, configure finite element quantity;Randomly select environmental characteristic combination in multivariate environmental characteristics, and divide environmental characteristic combination parameter set in sample environmental characteristic parameter set;Based on sample track BIM parameter, environmental characteristic combination parameter set and sample track deformation parameter, carry out integrated track deformation analysis intelligent agent training and selection optimization, obtain optimized finite element combination and optimized track deformation analysis intelligent agent combination, carry out track deformation analysis.The application effectively improves the accuracy of track deformation analysis under complex environment.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for track deformation analysis that combines BIM and finite element methods. Background Technology

[0002] As rail transit engineering develops towards high speed and networking, the long-term service stability of track structures has become a core concern in the industry. Building Information Modeling (BIM) technology, with its advantages of digital and parametric modeling, as well as the precise mechanical calculation capabilities of finite element analysis, is widely used in the field of track deformation analysis. Existing technologies mostly use BIM technology to construct a three-dimensional model of the track, and then rely on the finite element method to carry out mechanical simulation analysis, thereby achieving a preliminary assessment of track deformation.

[0003] However, high-speed rail track deformation is affected by multiple factors, including temperature, geology, and load, and environmental conditions vary significantly across different regions. The complex nonlinear coupling relationships between these factors make it difficult for traditional methods to accurately capture the track deformation patterns under the combined effects of these factors, and to adapt to complex working conditions. Consequently, the accuracy of track deformation analysis results falls short of practical application requirements. Summary of the Invention

[0004] This invention provides a method and system for track deformation analysis that combines BIM and finite element methods, aiming to solve the technical problem of insufficient accuracy in track deformation analysis under complex environments in existing technologies.

[0005] In view of the above problems, the present invention provides a method and system for track deformation analysis that combines BIM and finite element method.

[0006] In a first aspect, the present invention provides a method for track deformation analysis combining BIM and finite element methods, comprising: Collect BIM model parameters and multi-environmental features of sample track segments to obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets, and collect track deformation parameters of sample track segments to obtain multiple sample track deformation parameters. Based on the multiple sample orbit deformation parameters, the orbit deformation wavelength is obtained, and the number of finite element elements is configured. According to the number of finite elements, multiple combinations of environmental features are randomly selected within the multivariate environmental features, and multiple sets of environmental feature combination parameters are divided within the multiple sets of sample environmental feature parameters. Based on the multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, an integrated track deformation analysis agent is trained and optimized to obtain an optimized finite element combination and an optimized track deformation analysis agent combination for track deformation analysis.

[0007] Secondly, the present invention provides a track deformation analysis system combining BIM and finite element methods, comprising: The track parameter acquisition module is used to acquire BIM model parameters and multi-dimensional environmental features of sample track segments, obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets, and acquire track deformation parameters of sample track segments to obtain multiple sample track deformation parameters. The finite element configuration module is used to process and obtain the orbit deformation wavelength based on the multiple sample orbit deformation parameters, and to configure the number of finite elements. The feature combination partitioning module is used to randomly select multiple environmental feature combinations within the multivariate environmental features according to the number of finite elements, and to partition multiple environmental feature combination parameter sets within the multiple sample environmental feature parameter sets. The intelligent agent training and optimization module is used to train and select an integrated track deformation analysis intelligent agent based on the multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, to obtain an optimized finite element combination and an optimized track deformation analysis intelligent agent combination for track deformation analysis.

[0008] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention provides a method and system for track deformation analysis combining BIM and finite element method. By simultaneously collecting parameters of sample track BIM models, multi-dimensional environmental features, and track deformation parameters, a comprehensive and closely related basic data system is constructed, providing data support for adapting to complex working conditions with multiple coupled factors. Based on track deformation parameters, deformation wavelengths are obtained and the number of finite elements is scientifically configured to achieve precise matching between finite elements and track deformation features, avoiding accuracy deviations caused by empirical configuration and improving the relevance of the analysis. The multi-dimensional environmental feature combination parameter set is divided according to the number of finite elements, accurately covering feature coupling scenarios in different regional environments, further adapting to the diversity of complex working conditions. Relying on an integrated intelligent agent training and optimization mechanism, dual screening of optimized finite element combinations and intelligent agent combinations is achieved, improving the accuracy of track deformation analysis results and ensuring that the analysis conclusions can meet the practical application needs of high-speed rail track safety operation and maintenance, providing an efficient and accurate technical means for long-term stability assessment of track structures. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1A flowchart illustrating a track deformation analysis method combining BIM and finite element analysis, provided as an embodiment of the present invention; Figure 2 A schematic diagram of a track deformation analysis system combining BIM and finite element method provided for an embodiment of the present invention; The components represented by each number in the attached diagram are explained below: The module includes: orbital parameter acquisition module 11, finite element configuration module 12, feature combination and partitioning module 13, and agent training and optimization module 14. Detailed Implementation

[0011] This invention provides a method and system for track deformation analysis that combines BIM and finite element methods, which is used to address the technical problem of insufficient accuracy in track deformation analysis under complex environments in existing technologies.

[0012] Example 1, as Figure 1 As shown, this invention provides a method for track deformation analysis combining BIM and finite element methods, the method comprising: S100: Collect BIM model parameters and multi-dimensional environmental features of sample track segments to obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets, and collect track deformation parameters of sample track segments to obtain multiple sample track deformation parameters.

[0013] In this embodiment of the invention, BIM model parameters and multivariate environmental features of sample track segments are collected to obtain multiple sets of sample track BIM parameters and multiple sets of sample environmental feature parameters. Track deformation parameters of the sample track segments are also collected to obtain multiple sets of sample track deformation parameters. High-speed rail track deformation is the result of multidimensional coupling effects of soil properties, environmental factors, and the track structure itself. Single or fragmented data cannot reflect the essential laws of deformation. Different regions exhibit significant differences in soil types, moisture content, and environmental factors such as temperature fluctuations and rainfall intensity. Furthermore, the material properties of structures such as the track slab, support layer, and subgrade directly affect the track's load-bearing stability, all of which require comprehensive data collection to cover the coupled scenarios. The overall track deformation is formed by the superposition of local deformations in each section; collecting only overall track data will miss local deformation features, leading to biases in subsequent analysis. Therefore, it is necessary to scientifically divide track sections and collect BIM model parameters, multivariate environmental features, and track deformation parameters to provide complete and accurate basic data for subsequent wavelength analysis, finite element configuration, and agent training.

[0014] Step S100 in the method provided in this embodiment of the invention includes: The sample track segments are divided to obtain multiple sample track intervals; Collect BIM model parameters and multi-dimensional environmental features within the multiple sample track intervals to obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets; Track deformation parameters were collected from multiple sample track intervals to obtain multiple sample track deformation parameters.

[0015] First, the sample track segments are divided into multiple sample track intervals. Sample track segments refer to selected railway track sections covering different types of typical working conditions, used as data acquisition carriers. Sample track intervals are local track segments formed based on mileage markers, geological survey report stratigraphic boundaries, and bridge, tunnel, and roadbed design boundary points. Each interval must have uniform environmental conditions such as geology and structural types, and clearly distinguishable characteristics between different intervals to facilitate accurate capture of track deformation correlation data under different working conditions. Based on the line's mileage markers, geological survey reports, and design boundary points for different structures such as bridges, tunnels, and roadbeds, the target high-speed railway line is divided. The division principle is to ensure relatively uniform environmental conditions within each interval, while maintaining distinguishable characteristics between different intervals.

[0016] For example, a section of a high-speed railway from K100+000 to K100+500 is selected as a sample track section. K100+000 indicates that this point is located at the 100-kilometer mark of the line, that is, the position where the cumulative mileage reaches 100 kilometers from the starting point of the line; +500 indicates that it extends 500 meters from the K100 kilometer mark towards the end of the line. Within this section, K100+000 to K100+080 is a silty clay subgrade section, K100+080 to K100+130 is a simply supported beam bridge section, with silty clay underneath, and K100+210 is the geological boundary between silty clay and sandy loam. For example, the section is divided into 1000 sample track sections Q1-Q1000 at intervals of 50 meters per section, covering alternating working conditions such as silty clay subgrade, simply supported beam bridge, and sandy loam subgrade. The environmental conditions are uniform within each section, but there are significant differences in structural or geological characteristics between sections.

[0017] Secondly, BIM model parameters and multi-dimensional environmental features were collected from multiple sample track sections to obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets. BIM model parameters are digital parameters of the track structure constructed based on Building Information Modeling (BIM) technology, covering data such as track geometry and structural material properties. Multi-dimensional environmental features refer to multi-dimensional parameters affecting track deformation, including soil properties, meteorological parameters, and track support system material properties. Sample environmental feature parameter sets refer to datasets formed by integrating all environmental feature parameters for a single sample track section. The BIM models of the sample track sections were retrieved, and parameters such as track gauge, rail type, track slab concrete strength, support layer mortar strength, and subgrade filler type were extracted according to section number. For each section, parameters such as soil moisture content, porosity, and compression modulus were obtained through on-site testing; meteorological parameters such as monthly average temperature and annual rainfall were retrieved from historical data; and support system material property parameters such as track slab compressive strength and support layer elastic modulus were obtained through sampling and testing. The data were organized according to section number, with each section corresponding to one set of sample track BIM parameters and one set of sample environmental feature parameter sets.

[0018] For example, data was collected for the intervals Q1-Q1000 respectively: Q1: BIM parameters: track gauge 1435mm, rail type U71Mn, track bed slab C40 concrete; environmental characteristic parameter set: silty clay moisture content 28%, void ratio 0.85, compression modulus 12MPa, average monthly temperature 15℃, annual rainfall 800mm, extreme rainfall intensity 50mm / 24h, track bed slab compressive strength 42.5MPa, track bed slab elastic modulus 35GPa, support layer mortar strength M15, support layer elastic modulus 30GPa, subgrade filling compression modulus 15MPa, subgrade filling graded crushed stone. The same method was used for Q2-Q1000, ultimately obtaining 1000 sets of sample track BIM parameters and 1000 sets of sample environmental characteristic parameter sets.

[0019] Finally, track deformation parameters were collected from multiple sample track sections to obtain multiple sample track deformation parameters. Track deformation parameters refer to the displacement dimensions of the track during service, with the index being vertical settlement, which directly reflects the degree of track deformation. Using a precision level, five evenly distributed measurement points were set on the top surface of the rail in each sample track section; the vertical settlement of each point relative to the initial track laying position was measured; the average settlement of the five measurement points within a single section was calculated as the sample track deformation parameter for that section.

[0020] For example, settlement measurements were conducted sequentially across the Q1-Q1000 range: Settlement at measurement point Q1: 1.8mm, 2.1mm, 1.9mm, 2.0mm, 2.2mm, with an average of 2.0mm, meaning the deformation parameter for Q1 is 2.0mm. Settlement at measurement point Q2: 1.3mm, 1.5mm, 1.4mm, 1.6mm, 1.7mm, with an average of 1.5mm, meaning the deformation parameter for Q2 is 1.5mm. The same method was applied to the remaining ranges, ultimately yielding 1000 sample track deformation parameters, with settlement ranging from 1.2mm to 2.8mm, covering different deformation levels.

[0021] In this embodiment of the invention, by dividing the track into sections, uniform working conditions within a single section and distinguishable features between sections are achieved, avoiding data mixing across working conditions and laying the foundation for subsequent targeted parameter collection. By collecting BIM model parameters and diverse environmental features in sections, a complete working condition and parameter association system is constructed, covering key influencing factors such as structure, geology, and environment, thus solving the problem of traditional data fragmentation. By accurately collecting track deformation parameters, and ensuring that the deformation parameters correspond one-to-one with BIM parameters and environmental feature parameters, a standardized and accurate basic dataset is formed, providing reliable data support for subsequent deformation wavelength analysis, finite element quantity configuration, and agent training.

[0022] S200: Based on the multiple sample track deformation parameters, process and obtain the track deformation wavelength, and configure the number of finite element elements.

[0023] In this embodiment of the invention, the orbit deformation wavelength is obtained by processing the multiple sample orbit deformation parameters, and the number of finite element methods (FEMs) is configured. The distribution pattern of orbit deformation exhibits a length-periodic characteristic; the period length corresponding to different deformation parameters directly reflects the complexity of the orbit deformation. The smaller the deformation wavelength, the more frequent the local deformation fluctuations of the orbit, and the higher the complexity of the deformation, requiring a higher density of FEMs to capture subtle deformation features. Conversely, the larger the deformation wavelength, the smoother the orbit deformation trend, and the FEM density can be appropriately reduced. Simultaneously, the multi-dimensional environmental features have high dimensionality; if the number of FEMs is too large, it will lead to data dimensional redundancy in subsequent agent training, causing problems such as difficulty in model convergence and decreased accuracy. Therefore, wavelength analysis based on sample orbit deformation parameters is necessary to quantify the deformation complexity, and then scientifically configure the number of FEMs to achieve a balance between analytical accuracy and computational efficiency, providing a reasonable basis for subsequent environmental feature combination and partitioning.

[0024] Step S200 in the method provided in this embodiment of the invention includes: Based on the orbit deformation parameters of multiple samples, wavelength analysis of different deformation parameters is performed to obtain the wavelengths of multiple deformation parameters, and the orbit deformation wavelength is obtained through processing. The number of finite element elements is configured according to the wavelength of the orbital deformation.

[0025] First, based on the deformation parameters of multiple sample tracks, wavelength analysis of different deformation parameters is performed to obtain the wavelengths of multiple deformation parameters, and then the track deformation wavelengths are obtained through processing.

[0026] Specifically, based on multiple sample orbit deformation parameters, wavelength analysis of different deformation parameters is performed to obtain multiple deformation parameter wavelengths. These wavelengths are then processed to obtain the orbit deformation wavelengths, including: The multiple sample track deformation parameters are equally divided into clusters to obtain multiple clustered sample track deformation parameter sets, and the cluster centers of the multiple clustered sample track deformation parameter sets are extracted as multiple representative deformation parameters. Within multiple sample orbit deformation parameters, the average distance length of repeated occurrences of sample orbit deformation parameters that are greater than or equal to the first representative deformation parameter is extracted and used as the wavelength of the first deformation parameter. Further wavelength analysis was performed based on multiple representative deformation parameters to obtain the wavelengths of multiple deformation parameters; The orbital deformation wavelength is calculated based on multiple deformation parameter wavelengths.

[0027] First, the multiple sample track deformation parameters are equally divided into clusters to obtain multiple clustered sample track deformation parameter sets. The cluster centers of these clusters are then extracted as representative deformation parameters. Equal division clustering, which divides the multiple sample track deformation parameters into equally wide intervals, is an unsupervised clustering method that uniformly divides sample deformation parameters into predetermined categories based on their numerical values, used to distinguish different levels of deformation. A clustered sample track deformation parameter set refers to the collection of all sample track deformation parameters within the same category. The cluster center is the average parameter value of each clustered sample track deformation parameter set, representing the typical characteristics of that category of deformation parameters. The representative deformation parameters are the values ​​corresponding to the cluster centers, reflecting different levels of deformation. The numerical range of all sample track deformation parameters is statistically analyzed to determine the maximum and minimum values. Based on the distribution characteristics of the deformation parameters, the number of clusters is preset, which must be less than the number of samples. The value range of the deformation parameters is evenly divided into a preset number of intervals according to their numerical values, with each interval corresponding to a cluster. Each sample track deformation parameter is assigned to its corresponding numerical interval, forming multiple clustered sample track deformation parameter sets. The average value of all parameters in each clustered sample track deformation parameter set is calculated and used as the cluster center of that category, representing the deformation parameter.

[0028] For example, given 1000 sets of sample track deformation parameters obtained in S100, the settlement range is 1.2mm-2.8mm. The numerical range is determined as follows: minimum 1.2mm, maximum 2.8mm, with a range of 1.6mm. The preset number of clusters is 3, and the ranges are evenly divided according to numerical value: first range 1.2mm-1.8mm, second range 1.8mm-2.4mm, and third range 2.4mm-2.8mm. Classification parameters: the first range contains 300 samples, forming the first cluster sample track deformation parameter set; the second range contains 500 samples, forming the second cluster sample track deformation parameter set; and the third range contains 200 samples, forming the third cluster sample track deformation parameter set. Cluster centers were calculated as follows: the mean of the first cluster = the sum of deformation parameters of 300 samples / 300 ≈ 1.5 mm; the mean of the second cluster = the sum of deformation parameters of 500 samples / 500 ≈ 2.1 mm; and the mean of the third cluster = the sum of deformation parameters of 200 samples / 200 ≈ 2.6 mm. Finally, three representative deformation parameters were obtained, which were 1.5 mm, 2.1 mm, and 2.6 mm, respectively.

[0029] Secondly, within multiple sample track deformation parameters, the average distance length of repeated occurrences of sample track deformation parameters greater than or equal to the first representative deformation parameter is extracted as the wavelength of the first deformation parameter. The first representative deformation parameter refers to the first representative deformation parameter sorted from smallest to largest value, corresponding to a low level of deformation. The average distance length of repeated occurrences refers to the average distance between adjacent positions of sample track deformation parameters greater than or equal to the target representative deformation parameter on the track segment, i.e., the spatial period length of the deformation parameter. The wavelength of the first deformation parameter is the spatial period length corresponding to the first representative deformation parameter, reflecting the distribution pattern of low-level deformation. The representative deformation parameters are sorted from smallest to largest value to determine the first representative deformation parameter; all sample track deformation parameters greater than or equal to the first representative deformation parameter are screened, and their corresponding track interval mileage positions are recorded; the distance between the mileage positions corresponding to two adjacent parameters is calculated, and the values ​​of all distances are counted; the average value of the distances is calculated as the wavelength of the first deformation parameter.

[0030] For example, the representative deformation parameters are 1.5mm, 2.1mm, and 2.6mm. After sorting the values ​​from smallest to largest, the first representative deformation parameter is 1.5mm. All sample track deformation parameters greater than or equal to 1.5mm are filtered, totaling 950 groups. The corresponding track interval mileage position is recorded for each group. The distance between the corresponding mileage positions of two adjacent samples is calculated. Because the samples are collected evenly in intervals, the distance between adjacent samples meeting the conditions is mostly 50 meters. Some samples, due to local deformation <1.5mm, show a distance of one or two intervals, resulting in 949 distances. All distance values ​​are statistically analyzed, and the average distance is approximately 65.7m. Rounded to 66m, the wavelength of the first deformation parameter is determined to be 66 meters.

[0031] Furthermore, wavelength analysis is performed based on multiple representative deformation parameters to obtain the wavelengths of these deformation parameters. These wavelengths correspond to the spatial period length of each representative deformation parameter, reflecting the distribution patterns of different levels of deformation. The remaining representative deformation parameters are selected sequentially in ascending order of value. For each selected representative deformation parameter, all sample track deformation parameters greater than or equal to that parameter are selected, and their corresponding mileage positions are recorded. The distance between the mileage positions corresponding to adjacent parameters is calculated, and the average value is taken as the wavelength of that representative deformation parameter. The above steps are repeated until wavelength analysis is completed for all representative deformation parameters, yielding multiple deformation parameter wavelengths.

[0032] For example, continuing to select the remaining representative deformation parameter of 2.1 mm: filter all sample track deformation parameters greater than or equal to 2.1 mm, totaling 320 groups, and record their corresponding mileage positions one by one; calculate the mileage interval between two adjacent samples: 180 intervals of 50 meters, 120 intervals of 100 meters, and 19 intervals of 150 meters, resulting in a total of 319 intervals; the average interval is approximately 74.8 m, rounded to 75 m; determine that the wavelength of the deformation parameter corresponding to this representative deformation parameter is 75 m. Similarly, calculate that the wavelength of the deformation parameter corresponding to the representative deformation parameter of 2.6 mm is 146 m. ​​Finally, obtain three deformation parameter wavelengths: 66 m, 75 m, and 146 m.

[0033] Finally, the orbital deformation wavelength is calculated based on multiple deformation parameter wavelengths. The orbital deformation wavelength is a weighted average of all deformation parameter wavelengths, reflecting the spatial periodic characteristics of the overall deformation of the sample orbital segment and serving as the core basis for configuring the number of finite element elements. The weight corresponding to each deformation parameter wavelength is determined, with the weight value consistent with the sample size proportion of the corresponding clustered sample orbital deformation parameter set. The orbital deformation wavelength is calculated using the weighted average formula: Orbital deformation wavelength = Σ (Deformation parameter wavelength × Corresponding weight). The calculated orbital deformation wavelength value is then output.

[0034] For example, the three deformation parameters have wavelengths of 66m, 75m, and 146m, respectively. 66m corresponds to 300 samples in the first cluster, 75m to 500 samples in the second cluster, and 146m to 200 samples in the third cluster, for a total of 1000 samples. The weights are calculated as follows: weight for wavelength 66m = 300 / 1000 = 0.3, weight for 75m = 500 / 1000 = 0.5, and weight for 146m = 200 / 1000 = 0.2. The orbital deformation wavelength is calculated as 66 × 0.3 + 75 × 0.5 + 146 × 0.2 = 86.5m, rounded to 87m.

[0035] Secondly, the number of finite element elements is configured according to the wavelength of the orbital deformation.

[0036] The number of finite element elements is configured according to the orbital deformation wavelength, including: Obtain the average deformation wavelength of other orbital segments tested over a historical period as the reference deformation wavelength; Obtain the preset number of finite elements; Based on the ratio of the reference deformation wavelength to the orbit deformation wavelength, the preset number of finite elements is adjusted and calculated to obtain the number of finite elements, wherein the number of finite elements is less than the number of features of the multi-element environmental features.

[0037] First, the average deformation wavelength of other track segments tested over a historical period is obtained as the baseline deformation wavelength. The baseline deformation wavelength refers to the average deformation wavelength of similar track segments in history, serving as a reference for measuring the deformation complexity of the current sample track segment. Deformation wavelength data of other track segments with similar structural types and operating conditions to the sample track segment are retrieved from the historical test database; the average value of the historical data is calculated and determined as the baseline deformation wavelength. For example, deformation wavelength data of a mixed high-speed railway subgrade-bridge section of the same type are retrieved from the historical database, and the calculated average deformation wavelength is 240m, thus determining the baseline deformation wavelength as 240m.

[0038] Secondly, the preset number of finite element samples (FEMs) is determined. The preset FEMs are initial FEMs set based on historical engineering experience for similar track segments, serving as the foundation for subsequent adjustments and calculations. Referring to engineering cases of similar track deformation analyses, and considering the total number of multi-dimensional environmental features, the preset FEMs are determined. The preset FEMs must be less than the total number of multi-dimensional environmental features to avoid data redundancy later. For example, if the total number of multi-dimensional environmental features for a sample track segment is 12, including 3 soil properties, 3 meteorological parameters, and 6 support system material properties, and referring to similar cases, the preset FEMs are 4. Since 4 < 12, this meets the requirement.

[0039] Finally, based on the ratio of the reference deformation wavelength to the track deformation wavelength, the preset number of finite elements is adjusted to obtain the total number of finite elements, where the number of finite elements is less than the number of features in the multi-element environmental features. The number of finite elements refers to the final determined number of finite element divisions used for track deformation analysis, corresponding to the number of environmental feature combinations selected subsequently. The number of finite elements is calculated using the adjustment formula: Number of finite elements = Preset number of finite elements × (Reference deformation wavelength / Track deformation wavelength); the smaller the deformation wavelength, the larger the ratio, and the larger the number of finite elements, to accommodate more complex deformation features; the calculation result is verified to be less than the total number of multi-element environmental features, and if not, the largest integer less than the total number is taken.

[0040] For example, given a reference deformation wavelength of 240m, an orbital deformation wavelength of 87m, a preset finite element quantity of 4, and a total of 12 multi-element environmental features. Reference deformation wavelength / orbital deformation wavelength = 240 / 87 ≈ 2.759, finite element quantity = 4 × 2.795 ≈ 11.03, rounded down to 11; verification: 11 < 12, meeting the requirement; finally, the finite element quantity is determined to be 11.

[0041] In this embodiment of the invention, equal-division clustering and wavelength analysis are used to achieve quantitative classification of track deformation parameters, accurately extract spatial periodic features of different levels of deformation, and determine the track deformation wavelength to objectively reflect the overall deformation complexity of the track segment, providing a scientific basis for the configuration of the number of finite element elements and avoiding the blindness of empirical configuration. The number of finite element elements is adjusted based on the ratio of the benchmark deformation wavelength to the track deformation wavelength to achieve a match between the deformation complexity and the finite element density. The more complex the deformation, the larger the number of finite element elements, improving the analysis accuracy; the smoother the deformation, the more appropriate the number of finite element elements, reducing the computational cost. The number of finite element elements is strictly controlled to be less than the total number of multi-dimensional environmental features, effectively avoiding the dimensional redundancy problem in subsequent agent training, ensuring model convergence and analysis accuracy, and laying a reasonable quantitative foundation for the next step of environmental feature combination and division.

[0042] S300: According to the number of finite elements, randomly select multiple combinations of environmental features within the multivariate environmental features, and divide multiple sets of environmental feature combination parameters within the multiple sets of sample environmental feature parameters.

[0043] In this embodiment of the invention, multiple combinations of environmental features are randomly selected from the multivariate environmental features according to the number of finite element samples, and multiple sets of environmental feature combination parameters are divided within the multiple sets of sample environmental feature parameters. Multivariate environmental features have high dimensionality; directly using full feature analysis would easily lead to data redundancy, a surge in computational costs, and a mismatch with the previously determined number of finite element samples. Selecting environmental feature combinations according to the number of finite element samples allows the feature dimensionality to be controlled within a reasonable range, adapting to the convergence requirements of subsequent agent training. Simultaneously, by randomly selecting multiple feature combinations, coupling scenarios of different environmental factors can be covered, avoiding the omission of key related features by a single combination, ensuring that the divided parameter sets are both targeted and comprehensive, providing diverse data support for subsequent integrated agent training.

[0044] Step S300 in the method provided in this embodiment of the invention includes: According to the number of finite elements, environmental features with the number of finite elements are randomly selected from the multi-environmental features to form the first combination of environmental features. Continue to randomly select multiple combinations of environmental features; Based on multiple combinations of environmental features, environmental feature parameters are divided within multiple sample environmental feature parameter sets to obtain multiple sets of combined environmental feature parameters.

[0045] First, based on the stated number of finite element units, environmental features of that number are randomly selected from the multi-dimensional environmental features to form the first environmental feature combination. The first environmental feature combination refers to the first randomly selected set containing environmental features of that number of finite element units. The complete list of multi-dimensional environmental features and the number of finite element units are clearly defined. A random sampling algorithm is used to select features of that number of finite element units without replacement from the complete list of features. The selected features are then organized to form the first environmental feature combination, ensuring that there are no duplicate features within the combination.

[0046] For example, the multi-environmental characteristics include a total of 12 environmental characteristics: moisture content, porosity, compressive modulus, monthly average temperature, annual rainfall, extreme rainfall intensity, ballast slab compressive strength, support layer elastic modulus, subgrade filler compressive modulus, bridge bearing elastic modulus, ballast slab elastic modulus, and subgrade filler gradation. Combining the determined number of finite element samples (11), 11 characteristics are randomly selected to form the first environmental characteristic combination: moisture content, porosity, compressive modulus, monthly average temperature, annual rainfall, ballast slab compressive strength, support layer elastic modulus, subgrade filler compressive modulus, bridge bearing elastic modulus, ballast slab elastic modulus, and subgrade filler gradation.

[0047] Secondly, multiple environmental feature combinations are randomly selected. Multiple environmental feature combinations refer to multiple feature sets, including the first environmental feature combination, each containing finite element quantity features. The features between groups differ to cover multi-coupling scenarios. Using the random sampling algorithm, the above selection logic is repeated, continuing to select finite element quantity features from the full set of features to form new combinations; controlling the differences between groups ensures that the number of overlapping features between each combination and the previously selected combinations does not exceed 6, while avoiding completely duplicated combinations; a preset number of groups are selected as needed to form multiple environmental feature combinations.

[0048] For example, two more sets of feature combinations are selected, each containing 11 features, consistent with the number of finite element samples, to differentiate from the first combination: Second environmental feature combination: moisture content, porosity, compressive modulus, monthly average temperature, extreme rainfall intensity, ballast slab compressive strength, support layer elastic modulus, subgrade filler compressive modulus, bridge bearing elastic modulus, ballast slab elastic modulus, and subgrade filler gradation; Third environmental feature combination: moisture content, porosity, compressive modulus, annual rainfall, extreme rainfall intensity, ballast slab compressive strength, support layer elastic modulus, bridge bearing elastic modulus, ballast slab elastic modulus, subgrade filler compressive modulus, and subgrade filler gradation; Finally, three environmental feature combinations are obtained, each with 11 features, maintaining coverage of core features while differentiating by removing different individual features, adapting to diverse coupled scenarios.

[0049] Finally, based on multiple combinations of environmental features, environmental feature parameters are divided within multiple sample environmental feature parameter sets to obtain multiple sets of environmental feature combination parameters. The sample environmental feature parameter set refers to the multiple sets of interval environmental feature parameters collected in S100. The environmental feature combination parameter set refers to a dedicated dataset formed by extracting the parameters corresponding to each feature from each sample parameter set for a single environmental feature combination. The feature list for each environmental feature combination is retrieved one by one; for each sample environmental feature parameter set, the corresponding parameters are extracted according to the combination feature list, and feature parameters outside the list are removed; the extracted parameters of all samples under the same combination are organized and categorized to form the environmental feature combination parameter set corresponding to that combination; the above steps are repeated to generate a dedicated combination parameter set for each environmental feature combination.

[0050] For example, based on three combinations of environmental features, the parameter set of 1000 samples is divided as follows: For the first combination, 11 corresponding feature parameters are extracted from the parameter sets of Q1-Q1000 respectively, and these are organized to form the first environmental feature combination parameter set; For the second and third combinations, similarly, 11 corresponding feature parameters are extracted from the parameter sets of Q1-Q1000 respectively, forming the second and third environmental feature combination parameter sets; Finally, three environmental feature combination parameter sets are obtained, each set containing 11 corresponding feature parameters for 1000 samples, which correspond one-to-one with the three environmental feature combinations.

[0051] In this embodiment of the invention, feature combinations are selected according to the number of finite elements, and the feature dimensions are strictly controlled to meet the dimensional constraints mentioned above, effectively avoiding data redundancy and convergence problems in subsequent intelligent agent training. Multiple differentiated feature combinations cover diverse coupled scenarios of soil, meteorology, and support systems, avoiding the analytical limitations of a single combination and improving the comprehensiveness of the data. The divided combination parameter set corresponds one-to-one with the samples and is accurately matched with the feature combinations, providing targeted and scenario-rich input data for subsequent integrated intelligent agent training, ensuring training effectiveness and analytical generalization ability.

[0052] S400: Based on the multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, perform integrated track deformation analysis agent training and selection optimization to obtain optimized finite element combination and optimized track deformation analysis agent combination, and perform track deformation analysis.

[0053] In this embodiment of the invention, based on the multiple sample track BIM parameters, multiple sets of environmental feature combinations, and multiple sample track deformation parameters, an integrated track deformation analysis agent is trained and optimized to obtain an optimized finite element combination and an optimized track deformation analysis agent combination for track deformation analysis. A single track deformation analysis model is difficult to adapt to complex working conditions under the coupling of multiple environmental features, and the model accuracy varies for different combinations of environmental features. Through integrated agent training, differentiated models can be constructed for different feature combinations; combined with a weight adjustment mechanism, the training intensity of erroneous samples can be strengthened, improving the model's fitting ability to complex deformation scenarios. Subsequently, through accuracy testing and similarity screening, the most suitable feature combination and agent model can be selected, achieving a dual improvement in track deformation analysis accuracy and generalization ability, providing a reliable analysis tool for practical engineering applications.

[0054] Step S400 in the method provided in this embodiment of the invention includes: Using multiple sample track BIM parameters, a first set of environmental feature combination parameters from multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, a basic track deformation analysis agent is trained to obtain a first track deformation analysis agent set. Continue training to obtain a set of multiple orbit deformation analysis agents; The accuracy rates of the multiple sets of track deformation analysis agents are tested, and finite element selection optimization and track deformation analysis agent selection optimization are performed to obtain optimized finite element combinations and optimized track deformation analysis agent combinations for track deformation analysis.

[0055] First, a basic track deformation analysis agent is trained using multiple sample track BIM parameters, a first environmental feature combination parameter set from multiple environmental feature combination parameter sets, and multiple sample track deformation parameters to obtain a first track deformation analysis agent set.

[0056] Specifically, multiple sample track BIM parameters, a first set of environmental feature combination parameters from multiple environmental feature combination parameter sets, and multiple sample track deformation parameters are used to train a basic track deformation analysis agent, resulting in a first track deformation analysis agent set, including: Multiple sets of first deformation analysis training data are obtained by combining multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, and multiple first initial weights are assigned. Based on machine learning, the architecture of the first first track deformation analysis intelligent agent is constructed. According to multiple first initial weights, training resources of multiple sets of first deformation analysis training data are allocated to supervise the training of the first first track deformation analysis intelligent agent. The training input of the first track deformation analysis intelligent agent is the sample track BIM parameters and the first environmental feature combination parameters, and the output label is the sample track deformation parameters. After the first first-track deformation analysis agent is trained under supervision, the accuracy of multiple sets of first deformation analysis training data is tested to obtain multiple first test results, which include correct or incorrect results. Based on multiple first test results, multiple first weight adjustment coefficients are set, and multiple first initial weights are adjusted and calculated to obtain multiple first updated weights. Among them, the first weight adjustment coefficient corresponding to the correct first test result is less than 1, and the first weight adjustment coefficient corresponding to the incorrect first test result is greater than 1. Based on multiple first update weights, training resources for multiple sets of first deformation analysis training data are allocated, and a second first orbit deformation analysis agent is continued to be constructed and trained. Continue training multiple first-orbit deformation analysis agents to obtain a set of first-orbit deformation analysis agents.

[0057] First, multiple sets of sample track BIM parameters, multiple sets of environmental feature combination parameters, and multiple sample track deformation parameters are combined to obtain multiple sets of first deformation analysis training data, and multiple initial weights are assigned. The first deformation analysis training data refers to the training dataset formed by mapping the sample track BIM parameters, the first set of environmental feature combination parameters, and the sample track deformation parameters one-to-one; it is the basic input for agent training. The initial weights are the initial weight values ​​assigned to each set of training data, used to adjust the resource allocation of different data during training. Initially, all data have equal weights, and the sum of the weights is 1. The sample track BIM parameters obtained in S100, the first set of environmental feature combination parameters obtained in S300, and the sample track deformation parameters are mapped one-to-one according to interval numbers to form the first deformation analysis training data; each set of training data is assigned an equal initial weight, calculated as initial weight = 1 / number of training data sets, ensuring that the sum of the weights of all data is 1.

[0058] For example, the 1000 sets of sample track BIM parameters from Q1 to Q1000, the 11 parameters of the first environmental feature combination, and the 1000 sample track deformation parameters are combined one by one according to the interval number to obtain 1000 sets of first deformation analysis training data: D1, D2, ..., D1000; the initial weights are assigned as follows: the initial weight of each set of training data = 1 / 1000 = 0.001, and the total weight of all data is 1.

[0059] Secondly, based on machine learning, the architecture of the first intelligent agent for track deformation analysis was constructed. Training resources for multiple sets of first-stage deformation analysis training data were allocated according to multiple initial weights to conduct supervised training on the first intelligent agent. The training input of the intelligent agent consisted of sample track BIM parameters and combined parameters of first-stage environmental features, and the output label was the sample track deformation parameters. The track deformation analysis intelligent agent is a data analysis model built based on machine learning algorithms. Its input is the sample track BIM parameters and combined parameters of environmental features, and its output is the predicted value of the track deformation parameters. Supervised training refers to a training method that uses real track deformation parameters as labels and iteratively optimizes the model parameters by comparing the differences between the model's predicted values ​​and the real values.

[0060] Specifically, a machine learning model architecture is constructed, for example, using a random forest regression model: the model parameters are set as follows: 100 decision trees, a maximum depth of 8 layers per decision tree, random sampling for features, and mean squared error (MSE) as the node splitting criterion; the input layer dimension is the sum of the BIM parameter dimension and the environmental feature combination parameter dimension, and the output layer dimension is the track deformation parameter dimension, which is 1-dimensional settlement in this embodiment; according to the first initial weight, corresponding computing resources are allocated to each group of training data, and the dataset with higher weight participates in the model parameter iteration more frequently; the training data is input into the model for supervised training, and batch gradient descent is used for iterative optimization, with mean squared error (MSE) selected as the loss function. The model parameters are optimized by iteratively calculating the error between the predicted value and the true value until the prediction error converges to a preset threshold. The convergence condition is: the mean squared error (MSE) of 5 consecutive iterations ≤ 0.05 mm, and the error fluctuation amplitude ≤ 0.02 mm, thus completing the training of the first track deformation analysis agent.

[0061] For example, a model architecture based on random forest is constructed, with core parameters set as 100 decision trees, a maximum depth of 8 layers per tree, a feature sampling ratio of 80%, and mean squared error as the node splitting criterion. The input layer consists of 14-dimensional data, including 3-dimensional BIM parameters and 11-dimensional first environmental feature combination parameters. The output layer is a 1-dimensional settlement amount. The training dataset consists of BIM parameters collected by S100 in the Q1-Q1000 interval, the first environmental feature combination parameter set generated by S303, and the actual track settlement parameters in the corresponding interval, which are integrated into 1000 sets of training data D1-D1000. Resources are allocated according to the initial weights, that is, the sampling weight of each set of data during decision tree training is 0.001. The first deformation analysis training data D1-D1000 is input into the model, and batch gradient descent is used for training. The loss function is mean squared error (MSE). Training is stopped when the mean squared error (MSE) of 5 consecutive iterations is less than 0.05 mm and the error fluctuation amplitude is ≤0.02 mm, thus obtaining the first first track deformation analysis agent.

[0062] Furthermore, after the first deformation analysis agent for the first track has undergone supervised training, the accuracy of multiple sets of first deformation analysis training data is tested to obtain multiple first test results, which include correct or incorrect results. A first test result refers to inputting multiple sets of first deformation analysis training data into the trained agent, comparing the difference between the predicted value and the actual value, and determining whether the prediction result is correct or incorrect. For example, an error ≤ 0.1mm is considered correct, and otherwise incorrect. Multiple sets of first deformation analysis training data are input into the first agent to obtain predicted values, and the test results for each set of data are obtained by comparing them with the actual deformation parameters. For example, if all 1000 sets of first deformation analysis training data D1-D1000 are input into the first first track deformation analysis agent, and the error between the predicted value and the actual deformation parameter is ≤ 0.1mm as the criterion for correctness, the final test results are 720 correct and 280 incorrect, covering samples from different working conditions, which conforms to the accuracy distribution pattern of actual deformation analysis.

[0063] Subsequently, based on multiple first test results, multiple first weight adjustment coefficients are set, and multiple first initial weights are adjusted and calculated to obtain multiple first updated weights. The first weight adjustment coefficient corresponding to a correct first test result is less than 1, while the first weight adjustment coefficient corresponding to an incorrect first test result is greater than 1. The weight adjustment coefficient is used to adjust the weights of the training data. A coefficient less than 1 for correct samples is used to decrease the weight, while a coefficient greater than 1 for incorrect samples is used to increase the weight, thus achieving reinforcement training on incorrect samples. The weight adjustment coefficients are set, and the updated weight for each group of data is calculated as: Updated Weight = Initial Weight × Adjustment Coefficient. The updated weights are then normalized to ensure that the sum of all data weights is 1.

[0064] For example, set the weight adjustment coefficients: 0.5 for correct samples and 1.5 for incorrect samples. Calculate the update weight for a single group: update weight for any correct sample = 0.001 × 0.5 = 0.0005, update weight for any incorrect sample = 0.001 × 1.5 = 0.0015. Calculate the total weight sum: total weight for 720 correct samples = 720 × 0.0005 = 0.36, total weight for 280 incorrect samples = 280 × 0.0015 = 0.42, total weight sum = 0.36 + 0.42 = 0.78. Normalize: final weight for correct samples = 0.0005 / 0.78 ≈ 0.000641, final weight for incorrect samples = 0.0015 / 0.78 ≈ 0.001923. In the final normalized weight set, 720 correct samples each account for approximately 0.000641, 280 incorrect samples each account for approximately 0.001923, and the total weight of all data is 1.

[0065] Then, training resources for multiple sets of first deformation analysis training data are allocated according to multiple first update weights to continue building and training a second first orbit deformation analysis agent. For example, using the above-mentioned normalized update weights, the training and construction steps of the first first orbit deformation analysis agent are repeated to build and train a second first orbit deformation analysis agent.

[0066] Finally, multiple first-track deformation analysis agents are trained to obtain a set of first-track deformation analysis agents. This set of first-track deformation analysis agents is a collection of differentiated agents obtained through multiple weight adjustments and model training based on the first environmental feature combination parameter set. Based on the test data after the second first-track deformation analysis agent is trained, the weights are adjusted again, and so on, iterative training is conducted according to the preset number of agents. The trained agents are then integrated to form the first-track deformation analysis agent set. For example, the second agent is trained with normalized weights corresponding to 720 correct and 280 incorrect samples. After testing, the distribution of correct and incorrect samples is recalculated and adjusted to obtain a new set of normalized weights. This iterative training continues until five first-track deformation analysis agents are trained. These five differentiated agents are then integrated to form the first-track deformation analysis agent set.

[0067] Similarly, training continues to obtain multiple sets of orbit deformation analysis agents. Multiple sets of orbit deformation analysis agents refer to the sets of agents obtained for each set of environmental feature combination parameters obtained from S300, with the number of sets matching the number of environmental feature combinations. The remaining sets of environmental feature combination parameters obtained from S300 are retrieved; for each set of feature combination parameters, the previous steps are repeated to construct the corresponding set of deformation analysis agents; finally, the number of agent sets is the same as the number of environmental feature combinations. For example, for the second and third sets of environmental feature combination parameters, five agents are trained respectively, forming the second and third agent sets, ultimately resulting in three sets of agents, each containing five agents.

[0068] Based on this, the accuracies of the multiple sets of track deformation analysis agents are tested, and finite element selection optimization and track deformation analysis agent selection optimization are performed to obtain optimized finite element combinations and optimized track deformation analysis agent combinations for track deformation analysis.

[0069] This includes testing the accuracy of multiple sets of track deformation analysis agents, performing finite element selection optimization and track deformation analysis agent selection optimization to obtain optimized finite element combinations and optimized track deformation analysis agent combinations, including: The accuracy of the multiple sets of track deformation analysis agents is tested to obtain multiple agent accuracy sets, and the mean is calculated to obtain multiple accuracy rates; The combination of environmental features corresponding to the set of intelligent agents for track deformation analysis with the highest accuracy is selected as the optimal finite element combination. Calculate the similarity between the combination of environmental features corresponding to each track deformation analysis agent and the optimized finite element combination to obtain multiple finite element similarities; Based on the agent accuracy and finite element similarity of each orbit deformation analysis agent, an agent score is calculated, and the several orbit deformation analysis agents with the highest scores are selected as the optimal combination of orbit deformation analysis agents.

[0070] First, the accuracy of the multiple orbit deformation analysis agent sets is tested to obtain multiple agent accuracy sets, and the average is calculated to obtain multiple accuracy rates. Agent accuracy refers to the proportion of samples where a single agent's prediction is correct out of the total number of samples. The average accuracy of an agent set is the arithmetic mean of the accuracy rates of all agents within a set, used to measure the overall analytical capability of the set. A validation dataset similar to the training data is selected, and the validation data is input into each agent. The accuracy rate of an individual agent is calculated, and the average accuracy rate of all agents within each set is calculated to obtain the average accuracy rate of each set. For example, selecting 200 sets of new data for validation, the average accuracy rate of the first agent set is calculated to be 85%, the average accuracy rate of the second agent set is 78%, and the average accuracy rate of the third agent set is 82%.

[0071] Secondly, the environmental feature combination corresponding to the set of agents with the highest accuracy in track deformation analysis is selected as the optimal finite element combination. The optimal finite element combination refers to the environmental feature combination corresponding to the set of agents with the highest average accuracy; it represents the most adaptable feature combination scheme. By comparing the average accuracy of all agent sets, the set of agents with the highest average accuracy is selected, and the environmental feature combination corresponding to this set is the optimal finite element combination. For example, comparing the average accuracy of three sets: the first set has an average accuracy of 85% > the third set has an average accuracy of 82% > the second set has an average accuracy of 78%, therefore, the first environmental feature combination corresponding to the first set of agents is the optimal finite element combination.

[0072] Further, the similarity between the environmental feature combination corresponding to each trajectory deformation analysis agent and the optimized finite element combination is calculated to obtain multiple finite element similarities. Finite element similarity refers to the proportion of overlapping features between any environmental feature combination and the optimized finite element combination to the total number of features in the optimized finite element combination, used to measure the degree of similarity between different combinations. The feature list of the optimized finite element combination is extracted, and the total number of features is counted; the feature list of each other environmental feature combination is extracted, and the number of overlapping features with the optimized combination is counted; the similarity is calculated as: Similarity = Number of overlapping features / Total number of features in the optimized finite element combination × 100%.

[0073] For example, the second environmental feature combination overlaps with the optimized finite element combination by 7 items, and the third environmental feature combination overlaps with the optimized finite element combination by 8 items; the similarity is calculated as follows: the similarity between the second environmental feature combination and the optimized finite element combination is 7 / 11×100%≈63.6%, and the similarity between the third environmental feature combination and the optimized finite element combination is 8 / 11×100%≈72.7%.

[0074] Subsequently, based on the agent accuracy and finite element similarity of each orbit deformation analysis agent, an agent score is calculated. The agents with the highest scores are selected as the optimal orbit deformation analysis agent combination. The agent score is a comprehensive score calculated by combining the accuracy of a single agent with the similarity of corresponding feature combinations, used to measure the agent's priority. The optimal orbit deformation analysis agent combination refers to the set of agents with the highest scores selected for the final orbit deformation analysis. The scoring calculation formula is set as follows: Agent Score = Agent Accuracy × Finite Element Similarity; the scores of all agents are calculated; the scores are sorted from highest to lowest, and the agents with the highest scores are selected to form the optimal orbit deformation analysis agent combination. For example, the scores of all 15 agents are calculated, and the top 3 are selected after sorting by score to form the optimal orbit deformation analysis agent combination.

[0075] Finally, track deformation analysis is performed. BIM model parameters of the track segment to be analyzed and environmental characteristic parameters within the optimized finite element combination are collected. These parameters are then input into the optimized track deformation analysis agent combination. The average prediction results of all agents within the combination are taken as the deformation analysis result for the track segment to be analyzed. For example, 11 environmental characteristic parameters of a certain track segment to be analyzed, obtained from the BIM parameters and optimized finite element combination, are input into three optimized agents. The predicted settlement amounts are 2.1 mm, 2.0 mm, and 2.2 mm, respectively. The average predicted settlement amount of 2.1 mm is taken as the final track deformation analysis result.

[0076] In this embodiment of the invention, a weight adjustment mechanism is used to strengthen the training intensity of erroneous samples, improve the fitting ability of the agent to complex working conditions, and solve the problem of insufficient generalization ability of a single model. Based on the dual screening of accuracy and similarity, the optimal finite element combination and agent combination are selected, which takes into account both analysis accuracy and scenario adaptability, and avoids the interference of invalid features on the analysis results. The final track deformation analysis process can directly carry out parameter collection and analysis based on the optimized combination, reduce on-site detection and calculation costs, and provide efficient and accurate technical support for track safety operation and maintenance.

[0077] Through the specific implementation methods described above, the embodiments of the present invention achieve the following technical effects: This invention provides a method and system for track deformation analysis combining BIM and finite element method (FEM). By dividing the track into sections, it simultaneously collects BIM model parameters, multi-dimensional environmental features, and track deformation parameters to construct a standardized dataset, laying a solid foundation for accurate analysis. Wavelength analysis is performed based on sample deformation parameters to quantify the spatial periodic characteristics of deformation. The number of finite element samples is adjusted in conjunction with historical benchmark wavelengths to achieve a precise match between finite element density and deformation complexity, while keeping the number below the total dimension of environmental features to avoid data redundancy. Multiple sets of differentiated environmental feature combinations are selected based on the number of finite elements, adapting to model convergence requirements and covering diverse coupled scenarios, thus improving the analysis's generalization ability. Iterative weight adjustment strengthens training on erroneous samples, and a dual screening process based on accuracy and feature similarity selects the optimal finite element combination and agent combination, addressing the limitations of a single model. Ultimately, this achieves simultaneous improvement in analysis accuracy and efficiency, adapting to complex railway track conditions, reducing detection and computational costs, and providing efficient and reliable technical support for track safety operation and stability assessment.

[0078] Example 2, as Figure 2 As shown, this invention provides a track deformation analysis system combining BIM and finite element methods, the system comprising: The track parameter acquisition module 11 is used to acquire BIM model parameters and multi-dimensional environmental features of sample track segments, obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets, and acquire track deformation parameters of sample track segments to obtain multiple sample track deformation parameters. Finite element configuration module 12 is used to process and obtain the orbit deformation wavelength based on the multiple sample orbit deformation parameters, and to configure the number of finite elements; The feature combination partitioning module 13 is used to randomly select multiple environmental feature combinations within the multi-dimensional environmental features according to the number of finite elements, and to partition multiple environmental feature combination parameter sets within the multiple sample environmental feature parameter sets. The intelligent agent training and optimization module 14 is used to train and select an integrated track deformation analysis intelligent agent based on the multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, to obtain an optimized finite element combination and an optimized track deformation analysis intelligent agent combination for track deformation analysis.

[0079] In one embodiment, the orbit parameter acquisition module 11 is further configured to: The sample track segments are divided to obtain multiple sample track intervals; Collect BIM model parameters and multi-dimensional environmental features within the multiple sample track intervals to obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets; Track deformation parameters were collected from multiple sample track intervals to obtain multiple sample track deformation parameters.

[0080] In one embodiment, the finite element configuration module 12 is further configured to: Based on the orbit deformation parameters of multiple samples, wavelength analysis of different deformation parameters is performed to obtain the wavelengths of multiple deformation parameters, and the orbit deformation wavelength is obtained through processing. The number of finite element elements is configured according to the wavelength of the orbital deformation.

[0081] Specifically, based on multiple sample orbit deformation parameters, wavelength analysis of different deformation parameters is performed to obtain multiple deformation parameter wavelengths. These wavelengths are then processed to obtain the orbit deformation wavelengths, including: The multiple sample track deformation parameters are equally divided into clusters to obtain multiple clustered sample track deformation parameter sets, and the cluster centers of the multiple clustered sample track deformation parameter sets are extracted as multiple representative deformation parameters. Within multiple sample orbit deformation parameters, the average distance length of repeated occurrences of sample orbit deformation parameters that are greater than or equal to the first representative deformation parameter is extracted and used as the wavelength of the first deformation parameter. Further wavelength analysis was performed based on multiple representative deformation parameters to obtain the wavelengths of multiple deformation parameters; The orbital deformation wavelength is calculated based on multiple deformation parameter wavelengths.

[0082] The number of finite element elements is configured according to the orbital deformation wavelength, including: Obtain the average deformation wavelength of other orbital segments tested over a historical period as the reference deformation wavelength; Obtain the preset number of finite elements; Based on the ratio of the reference deformation wavelength to the orbit deformation wavelength, the preset number of finite elements is adjusted and calculated to obtain the number of finite elements, wherein the number of finite elements is less than the number of features of the multi-element environmental features.

[0083] In one embodiment, the feature combination segmentation module 13 is further configured to: According to the number of finite elements, environmental features with the number of finite elements are randomly selected from the multi-environmental features to form the first combination of environmental features. Continue to randomly select multiple combinations of environmental features; Based on multiple combinations of environmental features, environmental feature parameters are divided within multiple sample environmental feature parameter sets to obtain multiple sets of combined environmental feature parameters.

[0084] In one embodiment, the agent training optimization module 14 is further configured to: Using multiple sample track BIM parameters, a first set of environmental feature combination parameters from multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, a basic track deformation analysis agent is trained to obtain a first track deformation analysis agent set. Continue training to obtain a set of multiple orbit deformation analysis agents; The accuracy rates of the multiple sets of track deformation analysis agents are tested, and finite element selection optimization and track deformation analysis agent selection optimization are performed to obtain optimized finite element combinations and optimized track deformation analysis agent combinations for track deformation analysis.

[0085] Specifically, multiple sample track BIM parameters, a first set of environmental feature combination parameters from multiple environmental feature combination parameter sets, and multiple sample track deformation parameters are used to train a basic track deformation analysis agent, resulting in a first track deformation analysis agent set, including: Multiple sets of first deformation analysis training data are obtained by combining multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, and multiple first initial weights are assigned. Based on machine learning, the architecture of the first first track deformation analysis intelligent agent is constructed. According to multiple first initial weights, training resources of multiple sets of first deformation analysis training data are allocated to supervise the training of the first first track deformation analysis intelligent agent. The training input of the first track deformation analysis intelligent agent is the sample track BIM parameters and the first environmental feature combination parameters, and the output label is the sample track deformation parameters. After the first first-track deformation analysis agent is trained under supervision, the accuracy of multiple sets of first deformation analysis training data is tested to obtain multiple first test results, which include correct or incorrect results. Based on multiple first test results, multiple first weight adjustment coefficients are set, and multiple first initial weights are adjusted and calculated to obtain multiple first updated weights. Among them, the first weight adjustment coefficient corresponding to the correct first test result is less than 1, and the first weight adjustment coefficient corresponding to the incorrect first test result is greater than 1. Based on multiple first update weights, training resources for multiple sets of first deformation analysis training data are allocated, and a second first orbit deformation analysis agent is continued to be constructed and trained. Continue training multiple first-orbit deformation analysis agents to obtain a set of first-orbit deformation analysis agents.

[0086] This includes testing the accuracy of multiple sets of track deformation analysis agents, performing finite element selection optimization and track deformation analysis agent selection optimization to obtain optimized finite element combinations and optimized track deformation analysis agent combinations, including: The accuracy of the multiple sets of track deformation analysis agents is tested to obtain multiple agent accuracy sets, and the mean is calculated to obtain multiple accuracy rates; The combination of environmental features corresponding to the set of intelligent agents for track deformation analysis with the highest accuracy is selected as the optimal finite element combination. Calculate the similarity between the combination of environmental features corresponding to each track deformation analysis agent and the optimized finite element combination to obtain multiple finite element similarities; Based on the agent accuracy and finite element similarity of each orbit deformation analysis agent, an agent score is calculated, and the several orbit deformation analysis agents with the highest scores are selected as the optimal combination of orbit deformation analysis agents.

[0087] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for analyzing track deformation combining BIM and finite element methods, characterized in that, The method includes: Collect BIM model parameters and multi-environmental features of sample track segments to obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets, and collect track deformation parameters of sample track segments to obtain multiple sample track deformation parameters. Based on the multiple sample orbit deformation parameters, the orbit deformation wavelength is obtained, and the number of finite element elements is configured. According to the number of finite elements, multiple combinations of environmental features are randomly selected within the multivariate environmental features, and multiple sets of environmental feature combination parameters are divided within the multiple sets of sample environmental feature parameters. Based on the multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, an integrated track deformation analysis agent is trained and optimized to obtain an optimized finite element combination and an optimized track deformation analysis agent combination for track deformation analysis.

2. The track deformation analysis method combining BIM and finite element method according to claim 1, characterized in that, The BIM model parameters and multivariate environmental features of the sample track segments were collected to obtain multiple sets of sample track BIM parameters and multiple sets of sample environmental feature parameters. Track deformation parameters of the sample track segments were also collected to obtain multiple sets of sample track deformation parameters, including: The sample track segments are divided to obtain multiple sample track intervals; Collect BIM model parameters and multi-dimensional environmental features within the multiple sample track intervals to obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets; Track deformation parameters were collected from multiple sample track intervals to obtain multiple sample track deformation parameters.

3. The track deformation analysis method combining BIM and finite element analysis according to claim 1, characterized in that, Based on the multiple sample orbit deformation parameters, the orbit deformation wavelength is obtained through processing, and the number of finite element methods is configured, including: Based on the orbit deformation parameters of multiple samples, wavelength analysis of different deformation parameters is performed to obtain the wavelengths of multiple deformation parameters, and the orbit deformation wavelength is obtained through processing. The number of finite element elements is configured according to the wavelength of the orbital deformation.

4. The track deformation analysis method combining BIM and finite element method according to claim 3, characterized in that, Based on multiple sample orbit deformation parameters, wavelength analysis of different deformation parameters is performed to obtain multiple deformation parameter wavelengths. These wavelengths are then processed to obtain the orbit deformation wavelengths, including: The multiple sample track deformation parameters are equally divided into clusters to obtain multiple clustered sample track deformation parameter sets, and the cluster centers of the multiple clustered sample track deformation parameter sets are extracted as multiple representative deformation parameters. Within multiple sample orbit deformation parameters, the average distance length of repeated occurrences of sample orbit deformation parameters that are greater than or equal to the first representative deformation parameter is extracted and used as the wavelength of the first deformation parameter. Further wavelength analysis was performed based on multiple representative deformation parameters to obtain the wavelengths of multiple deformation parameters; The orbital deformation wavelength is calculated based on multiple deformation parameter wavelengths.

5. The track deformation analysis method combining BIM and finite element method according to claim 3, characterized in that, Based on the wavelength of the orbital deformation, the number of finite element elements is configured, including: Obtain the average deformation wavelength of other orbital segments tested over a historical period as the reference deformation wavelength; Obtain the preset number of finite elements; Based on the ratio of the reference deformation wavelength to the orbit deformation wavelength, the preset number of finite elements is adjusted and calculated to obtain the number of finite elements, wherein the number of finite elements is less than the number of features of the multi-element environmental features.

6. The track deformation analysis method combining BIM and finite element method according to claim 3, characterized in that, Based on the number of finite element samples, multiple combinations of environmental features are randomly selected from the multivariate environmental features, and multiple sets of environmental feature combination parameters are divided within the multiple sets of sample environmental feature parameters, including: According to the number of finite elements, environmental features with the number of finite elements are randomly selected from the multi-environmental features to form the first combination of environmental features. Continue to randomly select multiple combinations of environmental features; Based on multiple combinations of environmental features, environmental feature parameters are divided within multiple sample environmental feature parameter sets to obtain multiple sets of combined environmental feature parameters.

7. The track deformation analysis method combining BIM and finite element analysis according to claim 1, characterized in that, Based on the multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, an integrated track deformation analysis agent is trained and optimized to obtain an optimized finite element combination and an optimized track deformation analysis agent combination. Track deformation analysis is then performed, including: Using multiple sample track BIM parameters, a first set of environmental feature combination parameters from multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, a basic track deformation analysis agent is trained to obtain a first track deformation analysis agent set. Continue training to obtain a set of multiple orbit deformation analysis agents; The accuracy rates of the multiple sets of track deformation analysis agents are tested, and finite element selection optimization and track deformation analysis agent selection optimization are performed to obtain optimized finite element combinations and optimized track deformation analysis agent combinations for track deformation analysis.

8. The track deformation analysis method combining BIM and finite element method according to claim 7, characterized in that, Using multiple sample track BIM parameters, a first set of environmental feature combination parameters from multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, a basic track deformation analysis agent is trained to obtain a first track deformation analysis agent set, including: Multiple sets of first deformation analysis training data are obtained by combining multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, and multiple first initial weights are assigned. Based on machine learning, the architecture of the first first track deformation analysis intelligent agent is constructed. According to multiple first initial weights, training resources of multiple sets of first deformation analysis training data are allocated to supervise the training of the first first track deformation analysis intelligent agent. The training input of the first track deformation analysis intelligent agent is the sample track BIM parameters and the first environmental feature combination parameters, and the output label is the sample track deformation parameters. After the first first-track deformation analysis agent is trained under supervision, the accuracy of multiple sets of first deformation analysis training data is tested to obtain multiple first test results, which include correct or incorrect results. Based on multiple first test results, multiple first weight adjustment coefficients are set, and multiple first initial weights are adjusted and calculated to obtain multiple first updated weights. Among them, the first weight adjustment coefficient corresponding to the correct first test result is less than 1, and the first weight adjustment coefficient corresponding to the incorrect first test result is greater than 1. Based on multiple first update weights, training resources for multiple sets of first deformation analysis training data are allocated, and a second first orbit deformation analysis agent is continued to be constructed and trained. Continue training multiple first-orbit deformation analysis agents to obtain a set of first-orbit deformation analysis agents.

9. The track deformation analysis method combining BIM and finite element method according to claim 7, characterized in that, The accuracies of the multiple orbit deformation analysis agent assemblies are tested, and finite element selection optimization and orbit deformation analysis agent selection optimization are performed to obtain optimized finite element combinations and optimized orbit deformation analysis agent combinations, including: The accuracy of the multiple sets of track deformation analysis agents is tested to obtain multiple agent accuracy sets, and the mean is calculated to obtain multiple accuracy rates; The combination of environmental features corresponding to the set of intelligent agents for track deformation analysis with the highest accuracy is selected as the optimal finite element combination. Calculate the similarity between the combination of environmental features corresponding to each track deformation analysis agent and the optimized finite element combination to obtain multiple finite element similarities; Based on the agent accuracy and finite element similarity of each orbit deformation analysis agent, an agent score is calculated, and the several agents with the highest scores are selected as the optimal combination of orbit deformation analysis agents.

10. A track deformation analysis system combining BIM and finite element method, characterized in that, The system is used to implement the track deformation analysis method combining BIM and finite element analysis as described in any one of claims 1-9, the system comprising: The track parameter acquisition module is used to acquire BIM model parameters and multi-dimensional environmental features of sample track segments, obtain multiple sample track BIM parameter sets and multiple sample environmental feature parameter sets, and acquire track deformation parameters of sample track segments to obtain multiple sample track deformation parameters. The finite element configuration module is used to process and obtain the orbit deformation wavelength based on the multiple sample orbit deformation parameters, and to configure the number of finite elements. The feature combination partitioning module is used to randomly select multiple environmental feature combinations within the multivariate environmental features according to the number of finite elements, and to partition multiple environmental feature combination parameter sets within the multiple sample environmental feature parameter sets. The intelligent agent training and optimization module is used to train and select an integrated track deformation analysis intelligent agent based on the multiple sample track BIM parameters, multiple environmental feature combination parameter sets, and multiple sample track deformation parameters, to obtain an optimized finite element combination and an optimized track deformation analysis intelligent agent combination for track deformation analysis.

Images