Anti-collision wall stress test digital twin system and method

Abstract

The present application relates to the technical field of computer simulation, in particular to a kind of anti-collision wall stress test digital twin system and method.The system includes: physical anti-collision wall module is laid with multiple sensors to collect actual data;Data acquisition module is responsible for the transmission after pre-processing sensor data;Data processing module is simulated, analyzed and optimized by constructing and calibrating parameterized finite element model in combination with artificial intelligence algorithm on the stress condition of anti-collision wall;User interaction module realizes result display and parameter input interaction;The method is based on the above-mentioned system, covers the complete process from model construction to result output and feedback.By the present application, high-precision, real-time anti-collision wall mechanical property testing and evaluation can be realized, which can effectively reduce the testing cost and improve the efficiency, provide strong support for the structural safety protection and optimization design of anti-collision wall, and is suitable for anti-collision wall structure safety detection and performance prediction in highway, bridge, tunnel and other scenes.

Description

Technical Field

[0001] This invention relates to the field of computer simulation technology, specifically to a digital twin system and method for testing the force on a crash barrier. Background Technology

[0002] As a critical protective component for road traffic, industrial plants, and important infrastructure, crash barriers' mechanical performance under accidental impacts directly affects personnel safety and structural integrity. Traditional crash barrier stress testing methods primarily rely on physical vehicle crash tests or bench tests. These methods are not only costly, time-consuming, and labor-intensive, requiring demanding facilities, but also inherently dangerous, making it difficult to comprehensively and systematically evaluate various operating conditions. Furthermore, physical tests have a limited number of sensors, making it difficult to obtain complete stress-strain distribution and damage evolution processes within the structure.

[0003] In recent years, numerical simulation techniques such as finite element analysis have been widely used in structural collision simulation. However, the accuracy of simple numerical simulations is highly dependent on the precision of model parameters, which often contain uncertainties, leading to significant deviations between simulation results and actual conditions. To improve the predictive accuracy of simulation models, model calibration is usually required, but this also faces challenges in the absence of sufficient experimental data.

[0004] Digital twin technology, acting as a bridge between the physical and virtual worlds, provides novel solutions for the monitoring, analysis, prediction, and optimization of complex systems by integrating real-time data from physical entities, simulation models, and intelligent algorithms. Digital twin technology can simulate behavior and monitor operations throughout the entire lifecycle of an object using real-time data from the physical object's sensors.

[0005] Applying digital twin technology to the stress testing of crash barriers is expected to overcome the limitations of traditional physical experiments and simple numerical simulations. By constructing a high-fidelity digital twin model of the crash barrier and calibrating the model using a small amount of targeted physical experimental data, accurate predictions of the mechanical response of the crash barrier under various virtual impact conditions can be achieved. This can not only significantly reduce testing costs and time, and improve testing efficiency, but also reveal the damage mechanism and weak points of the crash barrier, providing strong support for the optimized design and safety assessment of crash barriers. Summary of the Invention

[0006] This invention provides a digital twin system and method for testing the stress on a crash barrier. By deeply integrating physical entities and virtual models and combining artificial intelligence algorithms, it achieves high-precision simulation, real-time monitoring, and intelligent optimization of the mechanical performance of the crash barrier, solving the problems of high cost, low accuracy, and poor real-time performance in existing technologies.

[0007] A digital twin system for testing the stress of a crash barrier includes:

[0008] Physical crash barrier module: A physical crash barrier entity installed in a preset test site or actual application scenario. The physical crash barrier is equipped with a sensor network, which includes at least one selected from strain sensors, acceleration sensors, displacement sensors, and temperature sensors.

[0009] Data acquisition module: connected to the sensor network, used to acquire sensor data from the sensor network and preprocess the data;

[0010] Data processing module: Connected to the data acquisition module, used to receive, process and analyze sensor data, and drive the operation of the digital twin model;

[0011] User interaction module: Provides a graphical user interface for displaying the simulation result data and the AI-generated optimization scheme, and receives virtual impact parameters input by the user.

[0012] Preferably, the data processing module includes:

[0013] Model building and maintenance unit: builds and maintains a parameterized finite element model corresponding to the physical crash barrier, which serves as the core model of the digital twin of the physical crash barrier. The parameterized finite element model includes a macroscopic structural model and a microscopic mechanical model.

[0014] Data receiving and preprocessing unit: Receives the sensor data transmitted by the data acquisition module and performs preprocessing operations;

[0015] Model calibration unit:

[0016] Perform preliminary physical tests to obtain baseline sensor data, the preliminary physical tests including at least one of non-destructive testing, static loading testing, modal testing, or low-energy shock testing;

[0017] The operating conditions of the preliminary physical test are simulated on the core digital twin model to obtain the initial simulation response;

[0018] Compare the initial simulation response with the baseline sensor data;

[0019] An optimization algorithm is used to iteratively adjust one or more preset parameters of the digital twin core model until the difference between the subsequent simulation response of the digital twin core model and the baseline sensor data is less than a preset threshold, thereby obtaining the calibrated digital twin model.

[0020] Impact simulation unit: Based on user-defined virtual impact parameters, the calibrated digital twin model is used to simulate the mechanical response of the physical crash barrier under impact events defined by the virtual impact parameters. The virtual impact parameters include at least one of the following: impact velocity, mass, angle, and surface material properties.

[0021] Result generation and analysis unit: generates and analyzes the simulation result data of the mechanical response and extracts key performance indicators;

[0022] Artificial intelligence-assisted decision-making unit:

[0023] Based on deep learning convolutional neural network (CNN), feature extraction and pattern recognition are performed on the sensor data and simulation result data to automatically identify areas of abnormal stress and potential structural weak points.

[0024] The performance of the physical collision barrier under future operating conditions is predicted using a Long Short-Term Memory (LSTM) network.

[0025] Genetic algorithms are used to generate optimized schemes that include material replacement and structural reinforcement locations;

[0026] Multi-scale modeling and updating units:

[0027] A micromechanical model is established for the local structural weak points of the crash barrier, and updated in tandem with the macroscopic structural model;

[0028] The physical sensor data and virtual sensor node data are fused using the Kalman filter algorithm. The update time interval and weight allocation are automatically adjusted based on the data change frequency and fluctuation characteristics. The calculation formula is as follows:

[0029] ;

[0030] In the formula, for Weights are updated in real time. The standard deviation of the current data. This represents the largest standard deviation in history. For data change frequency, To preset the maximum frequency, and It is an adjustable coefficient, and .

[0031] Preferably, the artificial intelligence-assisted decision-making unit includes:

[0032] Data preprocessing layer: Normalizes and reduces noise in the raw sensor data;

[0033] Feature extraction layer: Automatically extracts spatiotemporal features of force data through a CNN network;

[0034] Classification and Recognition Layer: Based on the Softmax classifier, it identifies different types of abnormal force patterns;

[0035] Interpretive Generation Layer: The SHAP method is used to generate interpretable risk assessment reports.

[0036] A digital twin method for testing the stress of a crash barrier includes:

[0037] Construct and maintain a parametric finite element model: Construct and maintain a parametric finite element model corresponding to the physical crash barrier, which serves as the core model of the digital twin of the physical crash barrier. The physical crash barrier is equipped with a sensor network, which includes at least one selected from strain sensors, acceleration sensors, displacement sensors, and temperature sensors.

[0038] Receiving and preprocessing sensor data: Receiving sensor data from the data acquisition module connected to the sensor network and preprocessing it;

[0039] Model calibration:

[0040] The model calibration unit performs preliminary physical tests on the physical crash barrier to obtain baseline sensor data. The preliminary physical tests include at least one of non-destructive testing, static loading testing, modal testing, or low-energy impact testing.

[0041] The operating conditions of the preliminary physical test are simulated on the core digital twin model to obtain the initial simulation response;

[0042] Compare the initial simulation response with the baseline sensor data;

[0043] An optimization algorithm is used to iteratively adjust one or more preset parameters of the digital twin core model until the difference between the subsequent simulation response of the digital twin core model and the baseline sensor data is less than a preset threshold, thereby obtaining the calibrated digital twin model.

[0044] Receive user-defined virtual impact parameters: Receive virtual impact parameters input by the user, wherein the virtual impact parameters include at least one of the following: impactor velocity, mass, angle, and surface material properties;

[0045] The impact simulation process involves using the calibrated digital twin model to simulate the mechanical response of the physical crash barrier under impact events defined by the virtual impact parameters.

[0046] Artificial intelligence-assisted analysis and prediction:

[0047] The AI-assisted decision-making unit performs feature extraction and anomaly identification on the sensor data and simulation result data based on a deep learning CNN network, generating a heat map containing risk level and location.

[0048] The stress trend under future working conditions is predicted by the Long Short-Term Memory Network (LSTM), and the optimized scheme of the anti-collision wall structure is generated by combining the genetic algorithm.

[0049] Analyze and output simulation results data: Analyze and output simulation results data of the mechanical response to evaluate the stress performance of the physical crash barrier;

[0050] Results visualization and feedback: The simulation results data and the optimization scheme generated by artificial intelligence are visualized and displayed through a graphical user interface, and user feedback is received.

[0051] Preferably, the model calibration further includes a microstructure response update sub-step:

[0052] For the key connection points and material defect areas of the crash barrier, a micromechanical model was established;

[0053] Based on the macroscopic stress field distribution, the microscopic model is updated using a multi-scale calculation method.

[0054] The microstructure evolution results are fed back to the macroscopic model to correct the overall mechanical performance prediction.

[0055] Preferably, the artificial intelligence-assisted analysis and prediction further includes:

[0056] Construct a Generative Adversarial Network (GAN) to generate virtual test data with different levels of damage to expand the training samples;

[0057] By employing transfer learning, the trained model is applied to the performance evaluation of different types of crash barriers, reducing training time and data requirements.

[0058] Preferably, the digital twin method for testing the force of a crash barrier further includes the following steps: simultaneously acquiring real-time sensor data during the physical impact test of the physical crash barrier; and using the real-time sensor data to further calibrate the digital twin model.

[0059] The simulation results of the calibrated digital twin model under the same working conditions of the physical impact test are compared and verified with the real-time sensor data, and the simulation results are used to provide the internal state information of the physical crash barrier in the area not covered by the sensor.

[0060] Compared with the prior art, the advantages of this invention are:

[0061] Improve testing efficiency and accuracy: Virtual model simulation analysis reduces the scale of physical experiments, effectively lowering testing costs and shortening the testing cycle. At the same time, with the help of precise model building and calibration mechanisms, high-precision simulation and prediction of the mechanical performance of crash barriers can be achieved, providing reliable support for their structural safety assessment.

[0062] Enhance intelligent decision-making capabilities: By leveraging artificial intelligence technologies such as deep learning, the system can monitor and quickly identify abnormal stress conditions on the crash barrier in real time, and generate specific and actionable quantitative optimization solutions based on the analysis results. This effectively assists in the iterative design of the crash barrier structure, improving its overall performance and safety.

[0063] Comprehensive assessment and in-depth insights: It can simulate areas where sensors are difficult to place or physical quantities that are difficult to measure directly in physical experiments, providing a more comprehensive understanding of the failure mechanism of crash barriers. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0065] One embodiment provides a digital twin system for testing the force of a crash barrier, including a physical crash barrier module, a data acquisition module, a data processing module, and a user interaction module. The specific implementation details of each module are described in detail below:

[0066] Physical crash barrier module: In practical applications, such as crash barriers on highways, bridges, or tunnels, a physical crash barrier is installed. A carefully deployed sensor network is embedded in this physical crash barrier, including strain sensors, acceleration sensors, displacement sensors, and temperature sensors. Strain sensors can be attached to key stress-bearing areas of the crash barrier, such as corners and connection points, to measure the strain of the wall during stress. Acceleration sensors are distributed at different locations to capture acceleration changes at the moment of impact or other dynamic conditions. Displacement sensors monitor potential displacement of the wall. Temperature sensors acquire ambient temperature and the wall's own temperature changes, as temperature affects material properties and structural mechanical characteristics. These sensors work together to comprehensively collect various raw data reflecting the physical state of the crash barrier.

[0067] Data Acquisition Module: The data acquisition module connects to the sensor network on the physical collision avoidance module via wired or wireless means. After receiving raw data from various sensors in real time, this module immediately performs preprocessing operations on this data. For example, for the acquired analog sensor data, it converts it into digital signals using an analog-to-digital converter for easier subsequent transmission and processing. Simultaneously, filtering algorithms (such as mean filtering and Kalman filtering) are used to remove noise interference from the data, improving data quality and ensuring the accuracy and reliability of the data transmitted to the data processing module.

[0068] Data Processing Module: As the core hub of the entire system, the data processing module has multiple functional units. These units work together to achieve the complex functions of the system, as detailed below:

[0069] Model building and maintenance unit:

[0070] First, based on detailed parameters such as the actual geometric dimensions, material properties (e.g., concrete strength grade, steel elastic modulus), and connection methods of the physical crash barrier, a parametric finite element model corresponding to the physical crash barrier is constructed using professional finite element modeling software. This model includes a macroscopic structural model to simulate the mechanical behavior of the entire crash barrier at the macroscopic level, such as overall deformation and stress distribution; it also includes a microscopic mechanical model, focusing on the mechanical response at the microscopic structural level of the materials within the crash barrier, such as the propagation of microcracks within the concrete and the evolution of microscopic damage in the steel.

[0071] During system operation, the parametric finite element model is continuously maintained based on subsequent test data and model update requirements. This includes timely updates of material parameters, geometric dimension changes, and other information to ensure that the model accurately reflects the actual state of the physical crash barrier.

[0072] Data Receiving and Preprocessing Unit: This unit receives pre-processed sensor data from the data acquisition module via a pre-defined data interface. It then performs a further format conversion to ensure all data conforms to the format requirements for subsequent processing and analysis. For example, data from different sensors is organized according to unified timestamps and data types. Simultaneously, a data integrity check is performed again; missing or abnormal data points are filled or repaired using interpolation to ensure data continuity and availability.

[0073] Model calibration unit:

[0074] Preliminary physical testing: A series of preliminary physical tests are carried out on the physical crash barrier, including non-destructive testing, static loading testing, modal testing or low-energy impact testing, etc., to obtain baseline sensor data.

[0075] Simulated operating conditions and comparative analysis: The operating conditions of the preliminary physical tests are accurately reproduced on the constructed digital twin core model (i.e., the parametric finite element model), and the corresponding initial simulation response is obtained through finite element simulation calculation. Then, the initial simulation response is compared with the actual collected baseline sensor data to analyze the differences between the two.

[0076] Parameter Iterative Adjustment: Using optimization algorithms (such as genetic algorithms, particle swarm optimization, etc.), one or more preset parameters of the digital twin core model are iteratively adjusted based on the differences. This process is continuously repeated until the difference between the subsequent simulation response of the digital twin core model and the baseline sensor data is less than a preset threshold (such as setting the root mean square error to be less than 5%), thereby obtaining a calibrated digital twin model that can accurately simulate the actual mechanical performance of a physical crash barrier.

[0077] Impact Simulation Unit: Based on the virtual impact parameters input by the user through the user interaction module, which specifically cover key elements such as the velocity, mass, angle, and surface material properties of the impacting object, the unit uses a pre-calibrated digital twin model and the dynamic analysis function of finite element simulation software to simulate the mechanical response of the physical crash barrier under impact events defined by these virtual impact parameters. This includes, but is not limited to, changes in stress distribution, the location of strain peaks, the overall deformation morphology, and energy absorption.

[0078] The results generation and analysis unit performs in-depth analysis of the simulation results data of the mechanical response obtained from the impact simulation unit, extracting a series of key performance indicators. For example, it calculates stress distribution cloud maps through post-processing algorithms, visually displaying the stress magnitude and distribution of various parts of the crash barrier; it analyzes strain distribution to identify strain concentration areas and determine potential damage locations; it analyzes deformation modes to observe the overall and local deformation characteristics of the crash barrier under impact; it assesses the degree of damage, determining whether the crash barrier is damaged and the severity of the damage based on certain damage judgment criteria (such as strain energy density, stress intensity factor, etc.); it analyzes energy absorption characteristics to understand the crash barrier's ability to absorb and dissipate energy during impact; and it also focuses on indicators such as the penetration depth of the impacting object to comprehensively judge the protective performance of the crash barrier.

[0079] Artificial intelligence-assisted decision-making unit:

[0080] Data preprocessing layer: First, the raw sensor data and simulation results data are normalized to map the data to a specific range and eliminate the influence of differences between different data volumes. At the same time, noise reduction algorithms are used to further remove residual noise components in the data, improve data quality, and lay a good foundation for subsequent deep learning analysis.

[0081] Feature extraction layer: The preprocessed data is input into the CNN network, and the network automatically learns and extracts the spatiotemporal features of the stress data, such as stress change features and strain change patterns at different times and locations, forming a representative feature vector.

[0082] Classification and Recognition Layer: Based on the extracted feature vectors, the Softmax classifier is used to identify and classify different types of stress anomaly patterns, such as distinguishing between anomalies caused by local stress concentration or anomalies caused by internal material defects, and providing corresponding probability values ​​to facilitate accurate judgment of the type and probability of the anomaly.

[0083] Explanation and Generation Layer: Using the SHAP method, the decisions made by the artificial intelligence model are explained, generating an interpretable risk assessment report that clearly shows the contribution of each input factor to the final risk assessment result, enabling users to intuitively understand the basis and process of the model's decision-making.

[0084] Multi-scale modeling and updating units:

[0085] Microscopic Mechanical Model Establishment and Collaborative Update: For local structural weaknesses identified in the crash barrier through analysis, a microscopic mechanical model is established using micromechanical theories and methods. This model details the microstructural changes and damage evolution mechanisms of the local area under stress. Then, the microscopic mechanical model is collaboratively updated with the macroscopic structural model to achieve the interrelation and information transfer between the overall mechanical response at the macroscopic level and the local damage evolution at the microscopic level. For example, when the macroscopic stress field distribution changes, the microscopic model is updated accordingly using multi-scale calculation methods. Simultaneously, the microstructural evolution results are fed back to the macroscopic model, correcting the overall mechanical performance predicted by the macroscopic model. This ensures that the entire model system accurately reflects the actual mechanical state changes of the crash barrier from the macroscopic to the microscopic level.

[0086] Data fusion and weight adjustment: The Kalman filter algorithm is used to fuse physical sensor data and virtual sensor node data. During actual operation, the frequency of data changes (e.g., measured by the number of fluctuations per unit time) and fluctuation characteristics (e.g., the standard deviation of the data reflects the amplitude of fluctuation) are monitored in real time, and weights are adjusted according to the formula... It automatically adjusts the update interval and weight allocation. and This is an adjustable coefficient (which can be set based on practical experience or previous test results, for example...). , ), The standard deviation of the current data. This represents the largest standard deviation in history. For data change frequency, The maximum preset frequency is used. Through this dynamic weight adjustment mechanism, the model can reasonably integrate multi-source data under different operating conditions, achieving more accurate updates and simulations.

[0087] User Interaction Module: This module enables user-friendly interaction through a graphical user interface. On one hand, it clearly displays simulation results data (such as visualized charts and animations of key performance indicators) obtained from the result generation and analysis unit in the data processing module, as well as optimization schemes generated by the AI-assisted decision-making unit. This allows users to fully understand the stress performance of the crash barrier and potential improvement directions. On the other hand, it provides corresponding input interfaces to receive virtual impact parameters input by the user. Users can flexibly set different impact conditions according to actual needs, thereby driving the impact simulation unit in the data processing module to carry out corresponding simulation analysis.

[0088] In another embodiment, the method flow implementation steps corresponding to the system are provided as follows:

[0089] Constructing and maintaining a parametric finite element model: Using professional finite element modeling software, combined with detailed information such as the actual engineering drawings, material parameters, and installation environment of the physical crash barrier, a corresponding parametric finite element model is constructed. This model encompasses both a macroscopic structural model and a microscopic mechanical model. During the subsequent operation of the entire system, continuous monitoring is conducted on potential structural changes and material aging of the physical crash barrier, and the relevant parameters of the model are updated and maintained in a timely manner to ensure that the model accurately reflects the state of the physical entity.

[0090] Receiving and Preprocessing Sensor Data: A stable communication connection is established between the data acquisition module and the sensor network on the physical collision avoidance module to receive various types of raw data collected by the sensors in real time. Then, following a predetermined data preprocessing process, analog-to-digital conversion, filtering algorithms, and operations such as data format unification, integrity checks, and repairs are performed to preprocess the sensor data, ensuring the accuracy and usability of the data and providing a high-quality data foundation for subsequent model calibration and other analytical work.

[0091] Model calibration:

[0092] Preliminary physical testing was conducted: various preliminary physical testing methods, including non-destructive testing, static loading testing, modal testing, and low-energy impact testing, were carried out on the physical crash barrier in accordance with the specifications to obtain rich baseline sensor data. The various parameter settings and corresponding sensor response data during each test were recorded in detail to provide a practical reference for subsequent model calibration.

[0093] Simulated working conditions and acquisition of initial simulation response: The actual working conditions of the above preliminary physical tests are completely input into the constructed digital twin core model. The mechanical response under the corresponding working conditions is accurately simulated by finite element simulation software to obtain the initial simulation response data.

[0094] Comparative analysis and parameter adjustment: The initial simulation response data is compared with the actual collected baseline sensor data. The accuracy of the model is evaluated by calculating the difference indicators between the two (such as root mean square error, relative error, etc.). Subsequently, the key parameters in the core digital twin model are iteratively adjusted using optimization algorithms until the difference between the subsequent simulation response of the model and the baseline sensor data is less than a preset threshold. Finally, the calibrated digital twin model is obtained, which can accurately simulate the actual mechanical behavior of a physical crash barrier.

[0095] Receive user-defined virtual impact parameters: Through the graphical user interface provided by the user interaction module, the virtual impact parameters input by the user are received. These parameters cover multiple aspects such as the velocity, mass, angle, and surface material properties of the impacting object. Users can flexibly set different combinations of virtual impact parameters according to actual research needs, simulation scenarios, or safety assessment objectives, so as to comprehensively examine the mechanical performance of the crash barrier under various possible impact conditions.

[0096] The impact simulation process involves transmitting the user-inputted virtual impact parameters to the calibrated digital twin model. Using the dynamic analysis function of the finite element simulation software, the impact simulation calculation is initiated to simulate the complete mechanical response process of the physical crash barrier under the impact event defined by the corresponding virtual impact parameters. This includes detailed mechanical behavior at each stage, from the propagation of stress waves at the moment of impact to the deformation development of the entire crash barrier structure and the final energy absorption and dissipation, resulting in comprehensive and detailed mechanical response simulation data.

[0097] Artificial intelligence-assisted analysis and prediction:

[0098] Feature extraction and anomaly identification: The preprocessed sensor data and the results of the impact simulation are input into the deep learning convolutional neural network in the artificial intelligence-assisted decision-making unit. With the powerful feature extraction capability of CNN, the spatiotemporal features in the data are automatically learned and extracted. Then, based on the extracted features, the classification and recognition algorithm is used to identify and judge the areas of abnormal stress and potential structural weak points, and generate a heat map containing risk level and location to intuitively show the distribution and severity of the abnormal situation.

[0099] Performance prediction and optimization scheme generation: A long short-term memory network is used to analyze and learn from historical sensor data and current simulation results to predict the stress trends of the physical crash barrier within a certain future timeframe, thus identifying potential safety hazards in advance. Simultaneously, combined with a genetic algorithm, optimization schemes are generated with the goals of improving the mechanical performance of the crash barrier and reducing damage risk. These schemes include specific details such as material replacement suggestions and structural reinforcement locations, providing strong decision support for the subsequent maintenance, modification, and optimized design of the crash barrier.

[0100] Analyze and output simulation results data: Analyze the mechanical response simulation results data obtained from the impact simulation process, extract key performance indicators such as stress distribution, strain distribution, deformation mode, damage degree, energy absorption characteristics, and impact penetration depth, and present these indicators in an intuitive and easy-to-understand form through data visualization technology, so that users can fully understand the mechanical performance of the crash barrier under specific impact conditions.

[0101] Results Visualization and Feedback: Through the graphical user interface of the user interaction module, the simulation results data obtained from the above analysis and the optimization schemes generated by the AI-assisted decision-making unit are visualized and presented to users in the form of charts, graphs, and text descriptions. This allows users to clearly understand the stress conditions, potential risks, and improvement directions of the crash barrier. Simultaneously, a corresponding feedback channel is set up on the interface to receive user feedback such as questions, opinions, or further analysis requests regarding the displayed content. Based on user feedback, the system's parameter settings, simulation conditions, or analysis methods are adjusted in a timely manner, achieving good interaction with users and continuously optimizing the application effect of the entire crash barrier stress testing digital twin system and method.

[0102] Through the specific implementation of the above system architecture and methodology, the digital twin system and method for testing the force of crash barriers of the present invention can efficiently and accurately achieve comprehensive testing, analysis, and optimization decisions on the mechanical properties of crash barriers, providing strong technical support for the safety assurance and performance improvement of crash barriers.

[0103] In this specification, the terms "an embodiment," "example," "specific example," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0104] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A digital twin system for testing the force of a crash barrier, characterized in that, include: Physical crash barrier module: A physical crash barrier entity installed in a preset test site or actual application scenario. The physical crash barrier is equipped with a sensor network, which includes at least one selected from strain sensors, acceleration sensors, displacement sensors, and temperature sensors. Data acquisition module: connected to the sensor network, used to acquire sensor data from the sensor network and preprocess the data; Data processing module: Connected to the data acquisition module, used to receive, process and analyze sensor data, and drive the operation of the digital twin model; The data processing module includes a model building and maintenance unit, a data receiving and preprocessing unit, a model calibration unit, an impact simulation unit, and a result generation and analysis unit. The model building and maintenance unit is used to build and maintain a parameterized finite element model corresponding to the physical crash barrier, which serves as the core model of the digital twin of the physical crash barrier. The parameterized finite element model includes a macroscopic structural model and a microscopic mechanical model. The data receiving and preprocessing unit is used to receive the sensor data transmitted by the data acquisition module and perform preprocessing operations. The model calibration unit is used to adjust one or more preset parameters of the digital twin core model using an optimization algorithm to calibrate the digital twin model. The impact simulation unit is used to simulate the mechanical response of the physical crash barrier under impact events defined by the virtual impact parameters based on user-defined virtual impact parameters and using the calibrated digital twin model. The virtual impact parameters include at least one of the following: impactor velocity, mass, angle, and surface material properties. The result generation and analysis unit is used to generate and analyze the simulation result data of the mechanical response and extract key performance indicators, which include at least one of stress distribution, strain distribution, deformation mode, damage degree, energy absorption characteristics or impact penetration depth. The step of adjusting one or more preset parameters of the digital twin core model using an optimization algorithm to calibrate the digital twin model includes: Perform preliminary physical tests to obtain baseline sensor data, the preliminary physical tests including at least one of non-destructive testing, static loading testing, modal testing, or low-energy shock testing; The operating conditions of the preliminary physical test are simulated on the core digital twin model to obtain the initial simulation response; Compare the initial simulation response with the baseline sensor data; An optimization algorithm is used to iteratively adjust one or more preset parameters of the digital twin core model until the difference between the subsequent simulation response of the digital twin core model and the baseline sensor data is less than a preset threshold, thus obtaining the calibrated digital twin model. The data processing module also includes a multi-scale modeling and updating unit: A micromechanical model is established for the local structural weak points of the crash barrier, and updated in tandem with the macroscopic structural model; The processed sensor data and virtual sensor node data are fused using the Kalman filter algorithm. The update time interval and weight allocation are automatically adjusted based on the data change frequency and fluctuation characteristics. The calculation formula is as follows: ； In the formula, for Weights are updated in real time. The standard deviation of the current data. This represents the largest standard deviation in history. For data change frequency, To preset the maximum frequency, and It is an adjustable coefficient, and ; User interaction module: Provides a graphical user interface for displaying the simulation results data and AI-generated optimization schemes, and receives virtual impact parameters input by the user.

2. The digital twin system for testing the force of a crash barrier according to claim 1, characterized in that, The data processing module also includes an artificial intelligence-assisted decision-making unit: Based on deep learning convolutional neural network (CNN), feature extraction and pattern recognition are performed on the sensor data and simulation result data to automatically identify areas of abnormal stress and potential structural weak points. The performance of the physical collision barrier under future operating conditions is predicted using a Long Short-Term Memory (LSTM) network. Genetic algorithms are used to generate optimized schemes that include material replacement and structural reinforcement locations.

3. The digital twin system for testing the force of a crash barrier according to claim 2, characterized in that, The AI-assisted decision-making unit includes: Data preprocessing layer: Normalizes and reduces noise in the raw sensor data; Feature extraction layer: Automatically extracts spatiotemporal features of force data through a CNN network; Classification and Recognition Layer: Based on the Softmax classifier, it identifies different types of abnormal force patterns; Interpretive Generation Layer: The SHAP method is used to generate interpretable risk assessment reports.

4. A digital twin method for testing the force of a crash barrier, based on the digital twin system for testing the force of a crash barrier as described in any one of claims 1-3, characterized in that, include: Construct and maintain a parametric finite element model: Construct and maintain a parametric finite element model corresponding to the physical crash barrier, which serves as the core model of the digital twin of the physical crash barrier. The physical crash barrier is equipped with a sensor network, which includes at least one selected from strain sensors, acceleration sensors, displacement sensors, and temperature sensors. Receiving and preprocessing sensor data: Receiving sensor data from the data acquisition module connected to the sensor network and preprocessing it; Model calibration: The model calibration unit performs preliminary physical tests on the physical crash barrier to obtain baseline sensor data. The preliminary physical tests include at least one of non-destructive testing, static loading testing, modal testing, or low-energy impact testing. The operating conditions of the preliminary physical test are simulated on the core digital twin model to obtain the initial simulation response; Compare the initial simulation response with the baseline sensor data; An optimization algorithm is used to iteratively adjust one or more preset parameters of the digital twin core model until the difference between the subsequent simulation response of the digital twin core model and the baseline sensor data is less than a preset threshold, thereby obtaining the calibrated digital twin model. Receive user-defined virtual impact parameters: Receive virtual impact parameters input by the user, wherein the virtual impact parameters include at least one of the following: impactor velocity, mass, angle, and surface material properties; The impact simulation process involves using the calibrated digital twin model to simulate the mechanical response of the physical crash barrier under impact events defined by the virtual impact parameters. Artificial intelligence-assisted analysis and prediction: The AI-assisted decision-making unit performs feature extraction and anomaly identification on the sensor data and simulation result data based on a deep learning CNN network, generating a heat map containing risk level and location. The stress trend under future working conditions is predicted by the Long Short-Term Memory Network (LSTM), and the optimized scheme of the anti-collision wall structure is generated by combining the genetic algorithm. Analyze and output simulation results data: Analyze and output simulation results data of the mechanical response to evaluate the stress performance of the physical crash barrier; Results visualization and feedback: The simulation results data and the optimization scheme generated by artificial intelligence are visualized and displayed through a graphical user interface, and user feedback is received.

5. The digital twin method for testing the force of a crash barrier according to claim 4, characterized in that, The model calibration further includes a microstructure response update step: For the key connection points and material defect areas of the crash barrier, a micromechanical model was established; Based on the macroscopic stress field distribution, the microscopic model is updated using a multi-scale calculation method. The microstructure evolution results are fed back to the macroscopic model to correct the overall mechanical performance prediction.

6. The digital twin method for testing the force of a crash barrier according to claim 4, characterized in that, The aforementioned AI-assisted analysis and prediction further includes: Construct a Generative Adversarial Network (GAN) to generate virtual test data with different levels of damage to expand the training samples; By employing transfer learning, the trained model is applied to the performance evaluation of different types of crash barriers, reducing training time and data requirements.

7. The digital twin method for testing the force of a crash barrier according to claim 4, characterized in that, Further includes: During the physical impact test of the physical crash barrier, real-time sensor data is collected simultaneously. The real-time sensor data will be used to further calibrate the digital twin model; The simulation results of the calibrated digital twin model under the same working conditions of the physical impact test are compared and verified with the real-time sensor data, and the simulation results are used to provide the internal state information of the physical crash barrier in the area not covered by the sensor.

Images