A bridge deck slag retaining wall stability analysis system and method for high-speed railways

Abstract

The application provides a bridge deck slag retaining wall stability analysis system and method for high-speed railway, and relates to the technical field of high-speed railway engineering. The system comprises a data acquisition and processing unit, a structure analysis and calculation unit, and a stability evaluation and early warning unit. The data acquisition and processing unit is used to obtain original data and historical data of the geometric parameters of the slag retaining wall, material performance parameters, and environmental loads. The structure analysis and calculation unit is used to simulate the dynamic stress and displacement response of the slag retaining wall under multiple working conditions in the whole life cycle based on the original data and historical data. The stability evaluation and early warning unit is used to adaptively and intelligently judge the stability state of the slag retaining wall based on the dynamic stress and displacement response, grade the early warning of potential risks, and predict the evolution trend. The application is used in the stability analysis process of the bridge deck slag retaining wall, and solves the technical problem of lagging risk early warning in the stability analysis of the bridge deck slag retaining wall of the high-speed railway in the prior art.

Description

A Stability Analysis System and Method for Bridge Deck Retaining Walls in High-Speed ​​Railways Technical Field

[0001] This application relates to the field of high-speed railway engineering technology, and in particular to a system and method for analyzing the stability of bridge deck retaining walls in high-speed railways. Background Technology

[0002] High-speed railways, as a vital backbone of land transportation, rely heavily on the stability of their track structures for operational safety. Bridge deck retaining walls, as key auxiliary components of high-speed railway bridge decks, play an irreplaceable role in track system stability and train safety. However, with the increasing operational mileage and service life of high-speed railways, bridge deck retaining walls are constantly exposed to complex environments, bearing the coupled effects of various loads such as train dynamic loads, temperature changes, earthquakes, and water immersion. Furthermore, concrete materials experience performance degradation over time, including carbonation and steel corrosion. These factors directly impact the stability of the retaining walls. Failure to promptly identify and manage these risks can lead to ballast loss, track deviation, and even train safety accidents. Current technologies primarily employ static assessment methods for retaining wall stability analysis. However, static assessment methods are limited by specific operating conditions, making it difficult to reflect full life-cycle performance degradation and multi-load coupling effects. Moreover, the reliance on human experience in assessment results leads to delayed risk warnings. Therefore, a life-cycle stability analysis method for high-speed railway bridge deck retaining walls is urgently needed to address these issues. Summary of the Invention

[0003] This application provides a system and method for analyzing the stability of bridge deck retaining walls in high-speed railways, which solves the technical problem of delayed risk warning in the existing technology for analyzing the stability of bridge deck retaining walls in high-speed railways.

[0004] To achieve the above objectives, this application adopts the following technical solution:

[0005] Firstly, a stability analysis system for bridge deck retaining walls of high-speed railways is provided, including: a data acquisition and processing unit, a structural analysis and calculation unit, and a stability assessment and early warning unit;

[0006] The data acquisition and processing unit is used to acquire raw and historical data of the geometric parameters, material performance parameters, and environmental loads of the retaining wall; the structural analysis and calculation unit is used to simulate the dynamic stress and displacement response of the retaining wall under multiple working conditions throughout its entire life cycle based on the raw and historical data; and the stability assessment and early warning unit is used to adaptively and intelligently determine the stability state of the retaining wall based on the dynamic stress and displacement response, classify and warn of potential risks, and predict its evolution trend.

[0007] In conjunction with the first aspect mentioned above, in one possible implementation, the data acquisition and processing unit includes: a multi-source data acquisition module, a data fusion and cleaning module, and a full lifecycle database module;

[0008] The system includes a multi-source data acquisition module, which collects raw data through sensors, BIM model import, and on-site inspection. The raw data includes geometric parameters of the retaining wall, material performance parameters, and environmental load data. The environmental load data includes train load data, temperature data, and precipitation data. The data fusion and cleaning module is used to standardize and align the raw data, fill in missing values, and handle outliers to form a clean dataset with a unified spatiotemporal benchmark. The full lifecycle database module is used to store the raw data of the retaining wall from construction to operation.

[0009] In conjunction with the first aspect mentioned above, in one possible implementation, the structural analysis calculation unit includes: a damage evolution modeling module, a dynamic mechanical simulation module, and a result output and verification module;

[0010] Among them, the damage evolution modeling module is used to establish a mathematical model based on historical data to reflect the full life cycle performance degradation of materials due to aging and fatigue accumulation; the dynamic mechanics simulation module is used to construct a spatiotemporally coupled finite element model that integrates the damage evolution mechanism and real-time environmental load, perform dynamic response calculations under multiple working conditions, and obtain calculation results; the result output and verification module is used to generate and verify the dynamic stress and displacement response data of the retaining wall under various working conditions based on the mathematical model and calculation results.

[0011] In conjunction with the first aspect mentioned above, in one possible implementation, the stability assessment and early warning unit includes: a deep learning training module, a dynamic reliability analysis module, and a risk early warning and trend prediction module;

[0012] Among them, the deep learning training module is used to train the stability discrimination model of the slag retaining wall based on historical data and calculation results; the dynamic reliability analysis module is used to calculate the failure probability of the slag retaining wall at different service stages based on dynamic stress and displacement response data under various working conditions, quantify the stability safety margin, and obtain the stability assessment results; the risk warning and trend prediction module is used to generate multi-level risk warning signals based on the stability assessment results, and predict the evolution trend of potential risks through time series analysis.

[0013] Secondly, a method for stability analysis of bridge deck retaining walls in high-speed railways is provided, including: acquiring raw and historical data for stability analysis of bridge deck retaining walls; the raw data includes: raw data of retaining wall geometric parameters, raw data of material performance parameters, and raw data of environmental loads; the historical data includes: historical data of retaining wall geometric parameters, historical data of material performance parameters, and historical data of environmental loads; based on the raw and historical data, simulating the dynamic stress and displacement response of the retaining wall under multiple working conditions throughout its entire life cycle; based on the dynamic stress and displacement response, adaptively and intelligently judging the stability state of the retaining wall, classifying and warning of potential risks, and predicting its evolution trend.

[0014] In conjunction with the second aspect mentioned above, one possible implementation involves simulating the dynamic stress and displacement response of the retaining wall under multiple working conditions throughout its entire life cycle, based on raw and historical data. This includes: establishing a mathematical model reflecting the performance degradation of the material throughout its entire life cycle based on historical data, and obtaining material performance degradation parameters; constructing a spatiotemporally coupled finite element model that integrates damage evolution mechanisms and real-time environmental loads based on the material performance degradation parameters, performing dynamic response calculations under multiple working conditions, and obtaining a dynamic response feature set of the retaining wall throughout its entire life cycle; and generating and verifying the dynamic stress and displacement response data of the retaining wall under various working conditions based on the mathematical model and calculation results.

[0015] In conjunction with the second aspect mentioned above, one possible implementation involves constructing a spatiotemporally coupled finite element model that integrates damage evolution mechanisms and real-time environmental loads based on material performance degradation parameters. This model is then used to perform dynamic response calculations under multiple operating conditions, resulting in a dynamic response feature set for the entire lifecycle of the retaining wall. This includes: embedding material performance degradation parameters into the constitutive relation of the finite element model to obtain a structural mechanics model that dynamically updates with service time; establishing a spatiotemporal load field based on environmental load data using a load coupling algorithm; performing adaptive mesh generation, dynamic boundary condition loading, and transient dynamic equilibrium equation solving sequentially for various preset operating conditions based on the load field, outputting stress-time history curves and displacement contour maps under the corresponding operating conditions; and generating a dynamic response feature set for the entire lifecycle of the retaining wall by integrating the multi-condition calculation results using a condition weight allocation algorithm based on the stress-time history curves and displacement contour maps under the corresponding operating conditions.

[0016] In conjunction with the second aspect above, in one possible implementation, the load coupling algorithm satisfies the following formula:

[0017]

[0018] in, Let be the total load in the spacetime coordinate system (x,y,z,t). Let (x, y, z, t) be the spatiotemporal distribution function of the vehicle dynamic load in the spatiotemporal coordinate system (x, y, z, t). Let be the spatiotemporal distribution function of temperature load in the spatiotemporal coordinate system (x,y,z,t). Let be the spatiotemporal distribution function of seismic load in the spatiotemporal coordinate system (x,y,z,t). Let (x,y,z,t) be the spatiotemporal distribution function of the water pressure load in the spatiotemporal coordinate system (x,y,z,t). The specific three-dimensional spatial location of the load application point is represented by the spacetime coordinate system (x, y, z, t), where t represents the specific time at which the load acts on that point; the transient dynamic equilibrium equations satisfy the following formula:

[0019]

[0020] Where M is the mass matrix, C is the damping matrix, K(t) is the stiffness matrix varying with time t, and u(t) is the displacement vector. Let be the velocity vector at time t. Let be the acceleration vector at time t. Let t be the total load vector at time t.

[0021] In conjunction with the second aspect mentioned above, one possible implementation involves adaptively and intelligently determining the stability state of the retaining wall based on dynamic stress and displacement response, providing graded early warnings for potential risks, and predicting their evolution trends. This includes: training a retaining wall stability discrimination model based on historical data and calculation results; calculating the failure probability of the retaining wall at different service stages using the retaining wall stability discrimination model based on dynamic stress and displacement response data under various working conditions, quantifying the stability safety margin, and obtaining stability assessment results; generating multi-level risk early warning signals based on the stability assessment results, and predicting the evolution trend of potential risks through time series analysis.

[0022] In conjunction with the second aspect mentioned above, one possible implementation involves training a slag retaining wall stability discrimination model, including: constructing a training set and a validation set based on historical stress and displacement data; inputting the training set into a bidirectional LSTM model based on an attention mechanism for training, and outputting risk discrimination results and a preliminary trained model; the risk discrimination results include: stable, slightly risky, and heavily risky; using a transfer learning method, the parameters of the preliminary trained model are used as initial weights, and the original data of the slag retaining wall of the new line are imported for fine-tuning to obtain a trained model.

[0023] This application provides a stability analysis system and method for bridge deck retaining walls in high-speed railways. By constructing a spatiotemporally coupled finite element model that integrates the damage evolution mechanism of the retaining wall throughout its entire life cycle with real-time environmental loads, and combining multi-condition dynamic response calculation and result verification, it achieves accurate simulation of the dynamic stress and displacement response of the retaining wall under different service stages and working conditions. Furthermore, it uses a bidirectional LSTM network model based on an attention mechanism to intelligently identify the stability state of the retaining wall, quantifying the failure probability and safety margin at different service stages. Through multi-level risk warning and time-series trend prediction, it improves the intelligence level of retaining wall stability assessment and the timeliness and accuracy of risk warning, providing strong support for the safe operation and maintenance of bridge deck retaining walls in high-speed railways. This solves the technical problem of lagging risk warning in the stability analysis of bridge deck retaining walls in high-speed railways in the prior art.

[0024] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0025] Figure 1 is a system architecture diagram of a bridge deck retaining wall stability analysis system for high-speed railways provided in an embodiment of this application;

[0026] Figure 2 is a flowchart illustrating a method for analyzing the stability of a bridge deck retaining wall for high-speed railways, as provided in an embodiment of this application.

[0027] Figure 3 is a flowchart illustrating another method for analyzing the stability of a bridge deck retaining wall for high-speed railways, as provided in an embodiment of this application.

[0028] Figure 4 is a flowchart illustrating the prediction of potential risk evolution trends in a stability analysis method for bridge deck retaining walls of high-speed railways provided in an embodiment of this application. Detailed Implementation

[0029] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0030] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0031] The stability analysis method for bridge deck retaining walls of high-speed railways provided in this application embodiment can be applied to the stability analysis system for bridge deck retaining walls of high-speed railways shown in Figure 1. As shown in Figure 1, the system includes: a data acquisition and processing unit 100, a structural analysis and calculation unit 200, and a stability assessment and early warning unit 300.

[0032] The data acquisition and processing unit 100 is used to acquire the original and historical data of the geometric parameters, material performance parameters, and environmental loads of the retaining wall; the structural analysis and calculation unit 200 is used to simulate the dynamic stress and displacement response of the retaining wall under multiple working conditions throughout its entire life cycle based on the original and historical data; and the stability assessment and early warning unit 300 is used to adaptively and intelligently determine the stability state of the retaining wall based on the dynamic stress and displacement response, provide graded early warnings for potential risks, and predict their evolution trend.

[0033] In one possible implementation, the data acquisition and processing unit 100 includes: a multi-source data acquisition module 101, a data fusion and cleaning module 102, and a full lifecycle database module 103;

[0034] The multi-source data acquisition module 101 is used to collect raw data through sensors, BIM model import, and on-site inspection. The raw data includes: geometric parameters of the retaining wall, material performance parameters, and environmental load data. The environmental load data includes: train load data, temperature data, and precipitation data. The data fusion and cleaning module 102 is used to standardize and align the raw data, fill in missing values, and handle outliers to form a clean dataset with a unified spatiotemporal benchmark. The full life cycle database module 103 is used to store the raw data of the retaining wall from construction to operation.

[0035] Preferably, the multi-source data acquisition module 101 simultaneously collects geometric parameters, material performance parameters, and environmental load data such as train load, temperature, and precipitation of the retaining wall through sensors, BIM model import, and on-site detection. Sensor acquisition includes fiber optic grating sensors (monitoring stress and strain), displacement sensors (monitoring retaining wall settlement and lateral displacement), temperature and humidity sensors, and rain gauges. Geometric parameters include wall height, wall thickness, and foundation depth. The data fusion and cleaning module 102 standardizes and aligns the collected raw data, fills in missing values, and handles outliers to form a clean dataset with a unified spatiotemporal benchmark. The full lifecycle database module 103 stores historical data, real-time monitoring data, and load parameters of the retaining wall from construction to operation.

[0036] In one possible implementation, the structural analysis calculation unit 200 includes: a damage evolution modeling module 201, a dynamic mechanical simulation module 202, and a result output and verification module 203.

[0037] Among them, the damage evolution modeling module 201 is used to establish a mathematical model reflecting the full life cycle performance degradation of materials based on historical data; the dynamic mechanics simulation module 202 is used to construct a spatiotemporally coupled finite element model that integrates the damage evolution mechanism and real-time environmental load, perform dynamic response calculations under multiple working conditions, and obtain calculation results; the result output and verification module 203 is used to generate and verify the dynamic stress and displacement response data of the slag retaining wall under various working conditions based on the mathematical model and calculation results.

[0038] Preferably, the damage evolution modeling module 201 establishes a mathematical model based on historical monitoring data to reflect the full life cycle performance degradation of materials due to aging and fatigue accumulation; the dynamic mechanics simulation module 202 constructs a spatiotemporally coupled finite element model that integrates the damage evolution mechanism and real-time environmental load, performs dynamic response calculations under multiple working conditions, and the result output and verification module 203 generates and verifies the dynamic stress and displacement response data of the retaining wall under various working conditions.

[0039] In one possible implementation, the dynamic mechanics simulation module 202 constructs a spatiotemporally coupled finite element model that integrates the damage evolution mechanism and real-time environmental loads, and performs dynamic response calculations under multiple working conditions. This includes: embedding the material property degradation parameters output by the damage evolution modeling module 201 into the constitutive relation of the finite element model to generate a structural mechanics model that is dynamically updated over time; importing real-time environmental load data and establishing a spatiotemporal load field through a load coupling algorithm; sequentially performing adaptive mesh generation, dynamic loading of boundary conditions, and solving of transient dynamic equations for various preset working conditions such as normal operation, earthquake, and immersion, and outputting stress time history curves and displacement cloud maps under the corresponding working conditions; and generating a dynamic response feature set of the retaining wall throughout its entire life cycle by comprehensively analyzing the multi-working-condition calculation results through a working-condition weight allocation algorithm.

[0040] As an example, in an embodiment of this application, the constitutive relation for material property degradation satisfies the following formula:

[0041]

[0042] in, Let be the stress of the material at time t. The elastic modulus degrades over time. Let K be the initial elastic modulus, K be the degradation coefficient, and t be the service time. In response, Let be the initial stress at time t.

[0043] As an example, in an embodiment of this application, the load coupling algorithm satisfies the following formula:

[0044]

[0045] in, Let be the total load in the spacetime coordinate system (x,y,z,t). Let (x, y, z, t) be the spatiotemporal distribution function of the vehicle dynamic load in the spatiotemporal coordinate system (x, y, z, t). Let be the spatiotemporal distribution function of temperature load in the spatiotemporal coordinate system (x,y,z,t). Let be the spatiotemporal distribution function of seismic load in the spatiotemporal coordinate system (x,y,z,t). Let (x,y,z,t) be the spatiotemporal distribution function of the water pressure load in the spatiotemporal coordinate system (x,y,z,t). The specific three-dimensional spatial location of the load application point is indicated. In the spatiotemporal coordinate system (x,y,z,t), t represents the specific time when the load applies to that point.

[0046] As an example, in an embodiment of this application, the transient dynamic equilibrium satisfies the following equation:

[0047]

[0048] Where M is the mass matrix, C is the damping matrix, K(t) is the stiffness matrix varying with time t, and u(t) is the displacement vector. Let be the velocity vector at time t. Let be the acceleration vector at time t. Let t be the total load vector at time t.

[0049] As an example, in this embodiment of the application, the weighted fusion of multi-condition results satisfies the following formula:

[0050]

[0051] In the formula, The dynamic response comprehensive characteristic value throughout the entire life cycle, where n is the total number of operating conditions. For the first Weighting coefficients for different working conditions For the first The response characteristic values ​​under various operating conditions.

[0052] In one possible implementation, the stability assessment and early warning unit 300 includes: a deep learning training module 301, a dynamic reliability analysis module 302, and a risk early warning and trend prediction module 303.

[0053] Among them, the deep learning training module 301 is used to train the stability discrimination model of the slag retaining wall based on historical data and calculation results; the dynamic reliability analysis module 302 is used to calculate the failure probability of the slag retaining wall at different service stages based on dynamic stress and displacement response data under various working conditions, quantify the stability safety margin, and obtain the stability assessment results; the risk warning and trend prediction module 303 is used to generate multi-level risk warning signals based on the stability assessment results, and predict the evolution trend of potential risks through time series analysis.

[0054] Preferably, the deep learning training module 301 trains the stability discrimination model of the retaining wall based on historical monitoring data and calculation results; the dynamic reliability analysis module 302 calculates the failure probability of the retaining wall at different service stages by combining real-time stress and displacement data, and quantifies the stability safety margin; the risk warning and trend prediction module 303 generates multi-level risk warning signals based on the stability assessment results, and predicts the evolution trend of potential risks through time series analysis.

[0055] In one possible implementation, the deep learning training module 301 trains a stability discrimination model for the retaining wall based on historical monitoring data and calculation results. This includes: preprocessing the historical monitoring data and the stress and displacement results output by the structural analysis and calculation unit 200, including outlier removal, feature normalization, and proportional division of the training and validation sets; constructing a bidirectional LSTM network model based on an attention mechanism, where the input layer contains 12 key feature parameters of the retaining wall, the hidden layer has 3 layers of adaptive neurons, and the output layer outputs three discrimination results: "stable," "mild risk," and "severe risk" through a softmax function; using the cross-entropy loss function as the optimization objective, the Adam algorithm is used for model training, an initial learning rate is set, and overfitting is prevented through a learning rate decay strategy and an early stopping mechanism; and using a transfer learning method, the trained model parameters are used as initial weights, and the monitoring data of the retaining wall of the new line is imported for fine-tuning.

[0056] As an example, in this embodiment, the historical monitoring data and the stress and displacement results output by the structural analysis calculation unit 200 are preprocessed, including outlier removal, feature normalization, and proportional division of the training and validation sets. Outlier removal: In addition to the 3σ principle, outliers with stress data greater than 25MPa are removed based on the mechanical limit of the retaining wall (e.g., the compressive strength of concrete is 25MPa). Feature normalization: Min-Max standardization is used to transform the data to the [0,1] interval to avoid the impact of data magnitude differences on model training. Data set division: The training and validation sets are divided in an 8:2 ratio. The training set is used for model parameter learning, and the validation set is used to evaluate the model's generalization ability to ensure the model's discrimination accuracy on new data. A bidirectional LSTM network model based on an attention mechanism was constructed. The input layer contains 12 key feature parameters of the retaining wall, the hidden layer has 3 layers of adaptive neurons, and the output layer outputs three classification results: "stable," "mild risk," and "severe risk" through a softmax function. The 12 key feature parameters include maximum stress, maximum displacement, stress change rate, displacement change rate, concrete carbonation depth, steel corrosion rate, train passing frequency, ambient temperature, precipitation, service time, foundation settlement, and expansion joint width, covering three dimensions: structural state of the retaining wall, environmental load, and service history. The number of neurons in the hidden layer is set as follows: The first layer has 128 neurons (extracting basic features), the second layer has 64 neurons (feature dimensionality reduction and fusion), and the third layer has 32 neurons (advanced feature extraction). An attention mechanism is used to strengthen the weight allocation of key features (such as maximum stress and displacement change rate), improving the model's discrimination accuracy. The Adam algorithm is used for model training with the cross-entropy loss function as the optimization objective. An initial learning rate of 0.001 is set, and overfitting is prevented through a learning rate decay strategy and an early stopping mechanism. Transfer learning is used, with the trained model parameters as initial weights, and monitoring data from the new line's retaining wall is imported for fine-tuning. The dynamic reliability analysis module 302, combined with real-time stress and displacement data, calculates the failure probability of the retaining wall at different service stages, quantifying the stability safety margin. The failure probability calculation method uses Monte Carlo simulation, based on real-time stress and displacement data and material limits (such as the ultimate stress of concrete). ), generate 10,000 random samples, and calculate the stress greater than in the samples. The proportion is the failure probability. Safety margin quantification: safety margin ,when It was determined to have a "high safety margin". For "medium safety margin", To provide a "low safety margin" and a quantitative basis for subsequent risk warnings, the risk warning and trend prediction module 303 generates multi-level risk warning signals based on the stability assessment results and predicts the evolution trend of potential risks through time series analysis. Multi-level risk warning: Blue warning (mild risk): Monitoring frequency needs to be increased (from once per minute to once every 30 seconds); Yellow alert (moderate risk): Specialized testing and repair plans are required; Red Alert (Severe Risk): Safety Margin The line needs to be immediately blocked and emergency reinforcement implemented.

[0057] The stability analysis system for bridge deck retaining walls of high-speed railways provided in this application integrates functional modules such as data acquisition and processing, structural analysis and calculation, and stability assessment and early warning, and constructs an intelligent structural safety analysis system for the entire life cycle and all working conditions. Through multi-source data acquisition and fusion cleaning, high-precision acquisition and long-term accumulation of multi-dimensional information such as the geometric parameters, material properties, and environmental loads of the retaining wall were achieved, providing a reliable data foundation for the system. By establishing a damage evolution model and a spatiotemporally coupled finite element simulation model, the dynamic influence of factors such as material degradation, environmental effects, and train loads on the structural response of the retaining wall can be realistically reflected, thereby obtaining stress and displacement characteristics that are more consistent with reality. Based on the evaluation mechanism of deep learning and dynamic reliability analysis, the stability state of the retaining wall can be adaptively and intelligently judged, and the failure probability and safety margin can be quantitatively calculated, realizing the transformation from qualitative assessment to quantitative prediction. At the same time, the risk warning and trend prediction module can identify potential instability risks in advance through time series analysis, output multi-level warning signals, and guide operation and maintenance personnel to take timely prevention and control measures, realizing continuous monitoring, scientific evaluation, and intelligent early warning of the retaining wall from the construction to the operation stage, improving the safety, reliability, and operation and maintenance efficiency of high-speed railway bridge structures.

[0058] To address the technical problem of delayed risk warning in the stability analysis of bridge deck retaining walls for high-speed railways in existing technologies, this application provides a method for stability analysis of bridge deck retaining walls for high-speed railways. The method includes: acquiring raw and historical data for the stability analysis of the bridge deck retaining wall; the raw data includes: raw data of the retaining wall's geometric parameters, raw data of its material performance parameters, and raw data of its environmental loads; the historical data includes: historical data of the retaining wall's geometric parameters, historical data of its material performance parameters, and historical data of its environmental loads; based on the raw and historical data, simulating the dynamic stress and displacement response of the retaining wall under multiple working conditions throughout its entire life cycle; and based on the dynamic stress and displacement response, adaptively and intelligently determining the stability state of the retaining wall, classifying and warning of potential risks, and predicting its evolution trend.

[0059] Figure 2 is a flowchart illustrating the stability analysis method for bridge deck retaining walls of high-speed railways provided in this embodiment of the application. As shown in Figure 2, the method includes:

[0060] S201. Obtain the original and historical data for the stability analysis of the bridge deck retaining wall.

[0061] The original data includes: original data on the geometric parameters of the retaining wall, original data on the material performance parameters, and original data on the environmental loads; the historical data includes: historical data on the geometric parameters of the retaining wall, historical data on the material performance parameters, and historical data on the environmental loads.

[0062] In one possible implementation, sensing devices such as fiber optic grating sensors, displacement sensors, temperature and humidity sensors, and rain gauges are deployed. Combined with BIM model import and regular on-site inspections, raw data on the geometric parameters (wall height, wall thickness, foundation depth), material performance parameters, and environmental loads such as train dynamic loads, temperature, and precipitation of the retaining wall are collected simultaneously. The collected multi-source heterogeneous raw data is then standardized and aligned (unifying the spatiotemporal reference), missing value imputation (e.g., using linear interpolation), and outlier handling (e.g., removing outliers using the 3σ principle and the ultimate compressive strength of concrete) to form a clean, consistent, standardized dataset. This processed data, along with historical archive data, is stored in a full lifecycle database, which stores all relevant data from construction to operation in chronological order.

[0063] S202. Based on raw and historical data, simulate the dynamic stress and displacement response of the retaining wall under multiple working conditions throughout its entire life cycle.

[0064] The simulation objects are a spatiotemporally coupled finite element model, a damage evolution model, and time-domain or frequency-domain load inputs representing the actual operational load spectrum.

[0065] In one possible implementation, a mathematical model reflecting the full life cycle performance degradation of materials due to aging and fatigue accumulation is established based on historical data to obtain material performance degradation parameters. Based on the material performance degradation parameters, a spatiotemporally coupled finite element model integrating damage evolution mechanism and real-time environmental load is constructed to perform dynamic response calculations under multiple working conditions, thereby obtaining the dynamic response feature set of the slag retaining wall throughout its entire life cycle. Based on the mathematical model and calculation results, dynamic stress and displacement response data of the slag retaining wall under various working conditions are generated and verified.

[0066] S203. Based on dynamic stress and displacement response, the stability state of the retaining wall is adaptively and intelligently judged, potential risks are classified and warned, and their evolution trend is predicted.

[0067] Among them, adaptive intelligent discrimination refers to automatically identifying abnormal patterns by combining deep learning algorithms with physical features; graded early warning refers to outputting risk signals according to set thresholds or probability levels.

[0068] In one possible implementation, a stability discrimination model for the retaining wall is trained based on historical data and calculation results. Based on dynamic stress and displacement response data under various working conditions, the failure probability of the retaining wall at different service stages is calculated through the stability discrimination model, the stability safety margin is quantified, and the stability assessment results are obtained. Based on the stability assessment results, multi-level risk warning signals are generated, and the evolution trend of potential risks is predicted through time series analysis.

[0069] The bridge deck retaining wall stability analysis method provided in this application, through the acquisition of multi-source raw data and historical data, enables the system to comprehensively acquire high-precision information covering geometric parameters, material properties, and environmental loads, providing sufficient data support for structural simulation. Through full life-cycle multi-condition dynamic simulation, based on the damage evolution and environmental coupling mechanism, it accurately calculates the stress and displacement response of the retaining wall at different service stages, truly reflecting the structural stress characteristics and performance degradation patterns. Furthermore, by utilizing deep learning and dynamic reliability analysis methods, it adaptively assesses the stability state of the retaining wall, quantifies the safety margin, and predicts the evolution trend of potential instability risks. This achieves closed-loop intelligent analysis from data acquisition and structural analysis to risk prediction, improving the accuracy and real-time performance of bridge deck retaining wall stability assessment and solving the technical problem of delayed risk warning in existing technologies for high-speed railway bridge deck retaining wall stability analysis.

[0070] In one possible implementation of this application embodiment, referring to FIG2 and FIG3, the above-mentioned S202 can be specifically implemented by the following S301, S302 and S303, which are described in detail below:

[0071] S301. Based on historical data, establish a mathematical model that reflects the full life cycle performance degradation of materials due to aging and fatigue accumulation, and obtain material performance degradation parameters.

[0072] In one possible implementation, a mathematical model of material performance degradation is established based on historical monitoring data in a full life cycle database, reflecting the effects of concrete carbonization, steel corrosion, and fatigue accumulation. This model quantifies the performance degradation of the retaining wall over service time, thereby obtaining material performance degradation parameters.

[0073] As an example, in an embodiment of this application, the constitutive relation for material property degradation satisfies the following formula:

[0074]

[0075] in, Let be the stress of the material at time t. The elastic modulus degrades over time. Let K be the initial elastic modulus, K be the degradation coefficient, and t be the service time. In response, Let be the initial stress at time t.

[0076] S302. Based on material performance degradation parameters, a spatiotemporal coupled finite element model integrating damage evolution mechanism and real-time environmental load is constructed to perform dynamic response calculation under multiple working conditions, and obtain the dynamic response feature set of the slag retaining wall throughout its entire life cycle.

[0077] In one possible implementation, material property degradation parameters are embedded into the constitutive relation of the finite element model to obtain a structural mechanics model that is dynamically updated over service time. Based on environmental load data, a spatiotemporal load field is established through a load coupling algorithm. Based on the load field, for various preset working conditions, adaptive mesh generation, dynamic loading of boundary conditions, and solution of transient dynamic equilibrium equations are performed sequentially to output stress-time history curves and displacement contour maps under the corresponding working conditions. Based on the stress-time history curves and displacement contour maps under the corresponding working conditions, a working condition weight allocation algorithm is used to integrate the calculation results of multiple working conditions to generate a dynamic response feature set of the slag retaining wall throughout its entire life cycle.

[0078] As an example, in an embodiment of this application, the load coupling algorithm satisfies the following formula:

[0079]

[0080] in, Let be the total load in the spacetime coordinate system (x,y,z,t). Let (x, y, z, t) be the spatiotemporal distribution function of the vehicle dynamic load in the spatiotemporal coordinate system (x, y, z, t). Let be the spatiotemporal distribution function of temperature load in the spatiotemporal coordinate system (x,y,z,t). Let be the spatiotemporal distribution function of seismic load in the spatiotemporal coordinate system (x,y,z,t). Let (x,y,z,t) be the spatiotemporal distribution function of the water pressure load in the spatiotemporal coordinate system (x,y,z,t). The specific three-dimensional spatial location of the load application point is represented by the spacetime coordinate system (x,y,z,t), where t represents the specific time when the load applies to that point.

[0081] As an example, in an embodiment of this application, the transient dynamic equilibrium equation satisfies the following formula:

[0082]

[0083] Where M is the mass matrix, C is the damping matrix, K(t) is the stiffness matrix varying with time t, and u(t) is the displacement vector. Let be the velocity vector at time t. Let be the acceleration vector at time t. Let t be the total load vector at time t.

[0084] S303. Based on mathematical models and calculation results, generate and verify dynamic stress and displacement response data of the retaining wall under various working conditions.

[0085] In one possible implementation, the simulation time series and sensor measured data are registered and aligned in the time and frequency domains, and the simulation accuracy is evaluated using an error index. For areas with deviations, degradation parameters or boundary conditions are adjusted using a Bayesian calibration method, and the simulation is repeated until the error meets a preset threshold. Finally, the calibrated dynamic stress and displacement response data are compiled into a structured database. Then, using a working condition weighting algorithm, the calculation results for each working condition are weighted and fused to generate and verify the dynamic stress and displacement response data of the retaining wall under various working conditions.

[0086] As an example, in this embodiment of the application, the weighted fusion of multi-condition results satisfies the following formula:

[0087]

[0088] In the formula, The dynamic response comprehensive characteristic value throughout the entire life cycle, where n is the total number of operating conditions. For the first Weighting coefficients for different working conditions For the first The response characteristic values ​​under various operating conditions.

[0089] This application's embodiments establish a mathematical model reflecting material aging and fatigue accumulation, enabling accurate extraction of material performance degradation parameters based on historical data. This model accurately reflects the strength attenuation and damage evolution of concrete and reinforcing steel during long-term service, providing reliable input for structural calculations. By integrating the damage evolution mechanism with a spatiotemporally coupled finite element model of real-time environmental loads, dynamic simulations can be conducted under multiple conditions such as train loads, temperature changes, precipitation, and earthquakes. This comprehensively acquires the stress, displacement, and damage evolution characteristics of the retaining wall at different service stages, improving the realism and applicability of the calculation results. Furthermore, by generating and verifying the calculation results, bidirectional calibration between simulation data and field monitoring data is achieved, ensuring the accuracy and reliability of the model predictions. This enables a quantifiable expression of the retaining wall's structural performance degradation process and a high-precision restoration of its dynamic response characteristics, providing a solid physical and data foundation for subsequent stability assessments and risk warnings.

[0090] In one possible implementation of this application embodiment, referring to FIG2 and FIG4, the above-mentioned S203 can be specifically implemented by the following S401, S402 and S403, which are described in detail below:

[0091] S401. Based on historical data and calculation results, train the stability discrimination model of the retaining wall.

[0092] In one possible implementation, a training set and a validation set are constructed based on the stress and displacement results of historical data. The training set is then input into a bidirectional LSTM model based on an attention mechanism for training, and the output includes risk discrimination results and a preliminary trained model. The risk discrimination results include: stable, mild risk, and severe risk. Using a transfer learning method, the parameters of the preliminary trained model are used as initial weights and the original data of the retaining wall of the new line are imported for fine-tuning to obtain a trained model.

[0093] It should be noted that the bidirectional LSTM network model based on the attention mechanism constructed in this application is a pre-trained model, which is trained in advance before use. The training process includes: preprocessing the historical monitoring data and the stress and displacement results output by the structural analysis calculation unit, including outlier removal, feature normalization, and proportional division of the training set and validation set; constructing a bidirectional LSTM network model based on the attention mechanism, wherein the input layer contains 12 key feature parameters of the retaining wall, the hidden layer has 3 layers of adaptive neurons, and the output layer outputs three classification results of "stable", "mild risk" and "severe risk" through the softmax function; using the cross-entropy loss function as the optimization objective, the Adam algorithm is used for model training, an initial learning rate is set, and overfitting is prevented through a learning rate decay strategy and an early stopping mechanism; using the transfer learning method, the trained model parameters are used as initial weights, and the monitoring data of the retaining wall of the new line are imported for fine-tuning.

[0094] As an example, in this embodiment of the application, a bidirectional LSTM network model based on an attention mechanism is constructed. The input of the model is 12 key feature parameters covering structural state, environmental load and service history. Real-time data and calculation results are input into the model, and the model outputs a preliminary intelligent judgment result of "stable", "mild risk" or "severe risk".

[0095] S402. Based on dynamic stress and displacement response data under various working conditions, the failure probability of the retaining wall at different service stages is calculated through the retaining wall stability discrimination model, the stability safety margin is quantified, and the stability assessment results are obtained.

[0096] In one possible implementation, by combining real-time stress and displacement data and employing Monte Carlo simulation probabilistic statistical methods, the failure probability of the retaining wall at different service stages is calculated, and the safety margin is further calculated based on the failure probability.

[0097] As an example, in an embodiment of this application, the safety margin S satisfies the following formula:

[0098]

[0099] in, This represents the failure probability.

[0100] As an example, in this embodiment of the application, the Monte Carlo simulation method is used, based on real-time stress-displacement data and material limit values ​​(such as the ultimate stress of concrete). ), generate 10,000 random samples, and calculate the stress greater than in the samples. The proportion of this is the probability of failure. Safety margin ,when It was determined to have a "high safety margin". For "medium safety margin", The term "low safety margin" provides a quantitative basis for subsequent risk warnings.

[0101] S403. Based on the stability assessment results, generate multi-level risk warning signals and predict the evolution trend of potential risks through time series analysis.

[0102] In one possible implementation, based on the results of stability assessment, multi-level risk warning signals are generated using different preset levels, and the evolution trend of potential risks is predicted through time series analysis.

[0103] As an example, in this embodiment of the application, a multi-level warning is triggered by combining the intelligent discrimination result and the safety margin quantification result. The warning levels include: Blue Warning (mild risk): Monitoring frequency needs to be increased (from once per minute to once every 30 seconds); Yellow alert (moderate risk): Specialized testing and repair plans are required; Red Alert (Severe Risk): Safety Margin If necessary, the line should be immediately blocked and emergency reinforcement implemented; and based on time series analysis algorithms (such as ARIMA or LSTM prediction models), historical data on safety margins or key response parameters should be learned to predict their future trends.

[0104] This application's embodiments train a stability discrimination model for retaining walls. Based on the integration of historical data and multi-condition simulation results, an intelligent identification system with self-learning and adaptive capabilities is constructed. This system enables automatic identification of the health status of retaining walls and extraction of abnormal features, reducing reliance on manual experience and improving the objectivity and efficiency of the analysis. The discrimination model is used to calculate the failure probability and quantify the safety margin of structural response data at different service stages, forming quantifiable and comparable stability assessment results. This accurately identifies potential weak points in the structure, providing a scientific basis for safety supervision and operation and maintenance decisions. Based on the assessment results, multi-level risk warning signals are generated, and time series analysis methods are combined to predict risk evolution trends, realizing the extension from static assessment to dynamic prediction, enabling potential hazards to be identified and intervened in advance.

[0105] The above primarily describes the solutions of the embodiments of this application from the perspective of device implementation. It is understood that each device, such as the stability analysis system for bridge deck retaining walls of high-speed railways, includes at least one of the hardware structures and software modules corresponding to the execution of each function in order to achieve the above-mentioned functions. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in a hardware or computer software-driven hardware manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0106] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0107] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0108] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of this application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and modifications.

Claims

1. A stability analysis system for bridge deck retaining walls of high-speed railways, characterized in that, include: Data acquisition and processing unit, structural analysis and calculation unit, stability assessment and early warning unit; The data acquisition and processing unit is used to acquire the original and historical data of the geometric parameters, material performance parameters, and environmental loads of the retaining wall; the structural analysis and calculation unit is used to simulate the dynamic stress and displacement response of the retaining wall under multiple working conditions throughout its entire life cycle based on the original and historical data. The structural analysis and calculation unit includes: a damage evolution modeling module, a dynamic mechanics simulation module, and a result output and verification module. The damage evolution modeling module is used to establish a mathematical model reflecting the full life-cycle performance degradation of materials due to aging and fatigue accumulation based on historical data. The dynamic mechanics simulation module is used to construct a spatiotemporally coupled finite element model integrating the damage evolution mechanism and real-time environmental loads, perform dynamic response calculations under multiple working conditions, and obtain calculation results. The result output and verification module is used to generate and verify the dynamic stress and displacement response data of the retaining wall under various working conditions based on the mathematical model and the calculation results. The stability assessment and early warning unit is used to adaptively and intelligently determine the stability state of the retaining wall based on the dynamic stress and displacement response, provide graded early warnings for potential risks, and predict their evolution trends. The stability assessment and early warning unit is also used to calculate the failure probability of the retaining wall at different service stages based on the dynamic stress and displacement response data under various working conditions, and quantify the stability safety margin. The adaptive intelligent judgment of the stability state of the retaining wall based on the dynamic stress and displacement response includes: constructing a bidirectional LSTM network model based on an attention mechanism, with the input of the model being dynamic stress and displacement response data under various working conditions, and the output of the model being the preliminary intelligent judgment result of the risk; the calculation of the failure probability of the retaining wall at different service stages based on the dynamic stress and displacement response data under various working conditions, and quantification of the stability safety margin, includes: combining the real-time stress and displacement response data of the dynamic stress, using the Monte Carlo simulation probabilistic statistical method to calculate the failure probability of the retaining wall at different service stages, and further calculating the safety margin based on the failure probability to obtain the safety margin quantification result; the stability assessment and early warning unit is also used to generate multi-level risk early warning signals based on the stability assessment results, and predict the evolution trend of potential risks through time series analysis to trigger multi-level risk early warnings; the stability assessment results include: the preliminary intelligent judgment result and the safety margin quantification result.

2. The stability analysis system for bridge deck retaining walls of high-speed railways according to claim 1, characterized in that, The data acquisition and processing unit includes: a multi-source data acquisition module, a data fusion and cleaning module, and a full lifecycle database module. The multi-source data acquisition module is used to collect raw data through sensors, BIM model import, and on-site inspection. The raw data includes: geometric parameters of the retaining wall, material performance parameters, and environmental load data. The environmental load data includes: train load data, temperature data, and precipitation data. The data fusion and cleaning module is used to standardize and align the raw data, fill in missing values, and handle outliers to form a clean dataset with a unified spatiotemporal benchmark. The full lifecycle database module is used to store the raw data of the retaining wall from construction to operation.

3. A method for stability analysis of bridge deck retaining walls in high-speed railways, applied to the stability analysis system for bridge deck retaining walls in high-speed railways as described in any one of claims 1-2, characterized in that, include: Obtain raw and historical data for stability analysis of the bridge deck retaining wall; The raw data includes: raw data of the geometric parameters of the retaining wall, raw data of the material performance parameters, and raw data of the environmental load; the historical data includes: historical data of the geometric parameters of the retaining wall, historical data of the material performance parameters, and historical data of the environmental load; based on the raw data and the historical data, the dynamic stress and displacement response of the retaining wall under multiple working conditions throughout its entire life cycle are simulated; based on the dynamic stress and the displacement response, the stability state of the retaining wall is adaptively and intelligently judged, potential risks are graded and warned, and their evolution trend is predicted.

4. The method for stability analysis of bridge deck retaining walls for high-speed railways according to claim 3, characterized in that, The simulation of the dynamic stress and displacement response of the retaining wall under multiple working conditions throughout its entire life cycle, based on the original data and the historical data, includes: establishing a mathematical model reflecting the performance degradation of the material throughout its entire life cycle based on historical data, and obtaining material performance degradation parameters; constructing a spatiotemporally coupled finite element model that integrates damage evolution mechanism and real-time environmental load based on the material performance degradation parameters, performing dynamic response calculations under multiple working conditions, and obtaining a dynamic response feature set of the retaining wall throughout its entire life cycle; and generating and verifying the dynamic stress and displacement response data of the retaining wall under various working conditions based on the mathematical model and the calculation results.

5. The method for stability analysis of bridge deck retaining walls for high-speed railways according to claim 4, characterized in that, The process involves constructing a spatiotemporally coupled finite element model that integrates damage evolution mechanisms and real-time environmental loads based on the material performance degradation parameters. This model is then used to perform dynamic response calculations under multiple operating conditions, resulting in a dynamic response feature set for the entire lifecycle of the retaining wall. This includes: embedding material performance degradation parameters into the constitutive relation of the finite element model to obtain a structural mechanics model that dynamically updates with service time; establishing a spatiotemporal load field based on environmental load data using a load coupling algorithm; performing adaptive mesh generation, dynamic boundary condition loading, and transient dynamic equilibrium equation solving sequentially for various preset operating conditions based on the load field, outputting stress-time history curves and displacement contour maps under the corresponding operating conditions; and generating a dynamic response feature set for the entire lifecycle of the retaining wall by integrating the multi-condition calculation results using a condition weight allocation algorithm based on the stress-time history curves and displacement contour maps under the corresponding operating conditions.

6. The method for stability analysis of bridge deck retaining walls for high-speed railways according to claim 5, characterized in that, The load coupling algorithm satisfies the following formula: in, Let be the total load in the spacetime coordinate system (x,y,z,t). Let (x, y, z, t) be the spatiotemporal distribution function of the vehicle dynamic load in the spatiotemporal coordinate system (x, y, z, t). Let be the spatiotemporal distribution function of temperature load in the spatiotemporal coordinate system (x,y,z,t). Let be the spatiotemporal distribution function of seismic load in the spatiotemporal coordinate system (x,y,z,t). Let (x,y,z,t) be the spatiotemporal distribution function of the water pressure load in the spatiotemporal coordinate system (x,y,z,t). The specific three-dimensional spatial location of the load application point is represented by the spacetime coordinate system (x, y, z, t), where t represents the specific time at which the load acts on that point; the transient dynamic equilibrium equation satisfies the following formula: Where M is the mass matrix, C is the damping matrix, K(t) is the stiffness matrix varying with time t, and u(t) is the displacement vector at time t. Let be the velocity vector at time t. Let be the acceleration vector at time t. Let t be the total load vector at time t.

7. The method for stability analysis of bridge deck retaining walls for high-speed railways according to claim 3, characterized in that, The adaptive intelligent judgment of the stability state of the retaining wall based on the dynamic stress and displacement response, the graded early warning of potential risks, and the prediction of their evolution trend include: training a retaining wall stability judgment model based on historical data and calculation results; calculating the failure probability of the retaining wall at different service stages based on dynamic stress and displacement response data under various working conditions, quantifying the stability safety margin, and obtaining stability assessment results; generating multi-level risk early warning signals based on the stability assessment results, and predicting the evolution trend of potential risks through time series analysis.

8. The method for stability analysis of bridge deck retaining walls for high-speed railways according to claim 7, characterized in that, The training model for the stability judgment of the retaining wall includes: constructing a training set and a validation set based on the stress and displacement results of historical data; inputting the training set into a bidirectional LSTM model based on an attention mechanism for training, and outputting risk judgment results and a preliminary training model; the risk judgment results include: stable, slightly risky, and heavily risky; using a transfer learning method, the parameters of the preliminary training model are used as initial weights, and the original data of the retaining wall of the new line are imported for fine-tuning to obtain a trained model.

Images