Accident analysis method, system, product and medium based on digital twin

Abstract

A digital twin-based accident analysis method, system, product, and medium. The method includes: calculating the actual force value and theoretical force value by collecting guardrail deformation data and vehicle driving records; dividing the guardrail into different force areas according to the degree of deformation, and calculating the difference ratio between the theoretical and actual force values ​​in each area; combining the vehicle collision angle, collision location, and deformation characteristics to determine the reliability index of each area and establish a weight coefficient table; using the regional weighted integral method, combining the difference ratio with the reliability weight to calculate the final force compensation value, providing a more accurate basis for accident analysis. Implementation of the technical solution provided by this application improves the accuracy of highway traffic guardrail accident analysis.

Description

Technical Field

[0001] The present application relates to the field of electrical digital data processing, and in particular to an accident analysis method, system, product and medium based on digital twins. Background Art

[0002] With the rapid expansion of the global highway network and the continued growth in the number of motor vehicles, highway traffic safety protection facilities have become critical infrastructure for ensuring driving safety. Guardrails, among other things, are crucial for preventing vehicles from deviating from their routes and mitigating collision damage. Their performance evaluation and accident analysis play an irreplaceable role in determining traffic accident liability, developing safety standards, and improving road design. Accurately understanding the collision mechanisms between vehicles and guardrails is crucial for improving road safety and reducing traffic accident casualties.

[0003] Currently, highway guardrail accident analysis typically relies on a combination of on-site inspections and sensor data collection. Analysts collect guardrail deformation data, vehicle traces, and sensor-recorded impact force data from the accident scene. Then, based on Newtonian mechanics and classical collision theory, they calculate the collision force and deformation energy using a static mechanics model. This technology primarily measures the permanent deformation of the guardrail and the extent of vehicle damage. This, combined with vehicle mass parameters and estimated collision velocity, creates a linear collision model to reconstruct the accident process, thereby analyzing the guardrail's protective performance and the cause of the accident. With the rapid development of digital technology, digital twin technology has begun to emerge in the field of traffic safety. By creating a high-precision virtual model of the guardrail, combining it with real-time sensor data and historical collision information, digital twins can replicate and simulate the collision process in a virtual environment, providing a new perspective for accident analysis.

[0004] However, in related technologies, even if digital twin technology is used for simulation, in special scenarios where a vehicle collides with a guardrail at a tricky angle, the deformation of the guardrail in the digital model still does not match the actual vehicle driving conditions, resulting in deviations in the digital twin's accident restoration process, affecting the accuracy and reliability of the accident analysis results. Summary of the Invention

[0005] This application provides a digital twin-based accident analysis method, system, product, and medium for improving the accuracy of highway traffic guardrail accident analysis.

[0006] In a first aspect of the present application, a digital twin-based accident analysis method is provided, the method comprising:

[0007] The guardrail force value is calculated based on the deformation of the guardrail at the scene of the accident and the vehicle driving data is collected by the vehicle driving recorder; the vehicle driving data is input into the preset theoretical force calculation model to obtain the theoretical force value of the vehicle on the guardrail; the guardrail force value is divided according to different deformation degrees to obtain the regional force distribution of different force areas; the theoretical force value is divided according to the regional force distribution to obtain the regional theoretical force value and the regional actual force value corresponding to each force area; for each force area, the difference ratio between the regional theoretical force value and the regional actual force value is calculated to obtain a difference ratio data group; according to the vehicle collision angle and collision location information in the vehicle driving data, combined with the deformation degree and force characteristics of each force area in the deformation situation, the reliability index of each area is calculated, and a reliability weight coefficient table is constructed; based on the difference ratio data set and the reliability weight coefficient table, the final force compensation value is calculated using the regional weighted integral method; the accident analysis is performed based on the final force compensation value in combination with the preset digital twin model.

[0008] In the above-described embodiment, the force process of a guardrail collision is reconstructed. Regional segmentation overcomes the limitations of related technologies in overall averaging. The difference ratio data set accurately quantifies the difference between theoretical and actual forces. The reliability weight coefficient effectively addresses the nonlinear relationship between collision angle and deformation degree. The regional weighted integration method ensures the accuracy of force calculations in complex force scenarios. In particular, in complex scenarios such as vehicles colliding at challenging angles, this method can capture the inconsistency between localized high-intensity deformation and the overall force distribution, making accident analysis results more consistent with physical reality. This reduces modeling bias in digital twin technology for challenging angle collisions, improves the precision of virtual-reality fusion accident reconstruction, and significantly enhances the accuracy and reliability of road safety assessments.

[0009] In conjunction with some embodiments of the first aspect, in some embodiments, based on the vehicle collision angle and collision location information in the vehicle driving data, combined with the deformation degree and force characteristics of each force-bearing area in the deformation situation, the reliability index of each area is calculated, and a reliability weight coefficient table is constructed, specifically including:

[0010] According to the collision angle and collision location information in the vehicle driving data, the preset angle mapping function and position evaluation function are used to calculate the collision direction coefficient and location importance coefficient respectively; the deformation direction vector in the deformation situation is compared with the theoretical collision force direction vector to calculate the deformation direction consistency parameter; the deformation abnormal area in the deformation situation is identified; the deformation abnormal area includes the deformation discontinuity area, the deformation direction abnormal area and the deformation area beyond the elastic limit of the material; for the deformation abnormal area, the historical damage correction algorithm, the fracture area compensation algorithm or the deformation continuity reconstruction algorithm is used according to its abnormality type to generate the corrected deformation consistency parameter; the collision direction coefficient, the location importance coefficient, the deformation consistency parameter and the corrected deformation consistency parameter are comprehensively considered to calculate the reliability index of each force-bearing area, and a reliability weight coefficient table is constructed through normalization processing.

[0011] In the above embodiment, by constructing a multi-dimensional evaluation system, in-depth analysis of the consistency between deformation and theoretical force vectors, identifying and classifying abnormal deformation areas, and using targeted algorithms to correct different types of abnormalities, the data interference caused by material fracture and discontinuous deformation is effectively overcome, and the credibility of data in each area is accurately identified and quantified. In particular, in extreme cases such as material fracture and discontinuous deformation caused by strong collisions, a reasonable weight distribution can be provided.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the collision direction coefficient, the location importance coefficient, the deformation consistency parameter, and the corrected deformation consistency parameter are comprehensively considered to calculate the reliability index of each force-bearing area, and a reliability weight coefficient table is constructed through normalization processing, specifically including:

[0013] The reliability index of each stress area is calculated by combining the collision direction coefficient, the location importance coefficient, the deformation consistency parameter, the corrected deformation consistency parameter, and the first formula; wherein the first formula is: ; Among them, R is the reliability index, is the collision direction coefficient, which represents the effective component of the collision force in the normal direction. is the location importance coefficient, which indicates the structural stress characteristics. is the deformation consistency parameter, and the modified deformation consistency parameter is used for abnormal areas , 、 、 、 is the preset parameter constant; S is the special deformation adjustment factor, and the calculation formula is , 、 、 They are respectively the historical damage degree, fracture characteristic degree and discontinuity degree; by normalizing the reliability index of each stress area, a reliability weight coefficient table is constructed.

[0014] In the above-mentioned embodiment, a mathematical quantification model for the credibility of deformation data was established. This multi-nonlinear transformation mechanism makes the reliability index insensitive to small changes while highly responsive to significant differences, resolving complex situations that conventional linear weighting cannot address. This formula captures the core characteristics of collision physics, establishes a gradient-based assessment system for data credibility, and provides a scientifically sound weight distribution for subsequent regional weighted integration.

[0015] In conjunction with some embodiments of the first aspect, in some embodiments, for each force-bearing area, the difference ratio between the theoretical force value of the area and the actual force value of the area is calculated to obtain a difference ratio data group, which specifically includes:

[0016] The stress area where the actual stress value exceeds the preset maximum stress threshold is marked as an ultra-high intensity collision area; the stress response curve of the guardrail material under different deformation degrees is obtained; the stress response curve is pre-established based on the material mechanical test data; according to the stress response curve, linear transformation is used in the preset first range area, quadratic function calibration is used in the preset second range area, and exponential function calibration is used in the preset third range area to construct a piecewise calibration function; the piecewise calibration function is applied to the ultra-high intensity collision area to calculate the actual stress value of the calibration area corresponding to the ultra-high intensity collision area; for each ultra-high intensity collision area, the ratio of the theoretical stress value of the area to the actual stress value of the calibration area is calculated; for the stress area that does not belong to the ultra-high intensity collision area, the ratio of the theoretical stress value of the area to the actual stress value of the area is calculated to obtain a difference ratio data group.

[0017] In the above-mentioned examples, the nonlinear response characteristics of materials in extreme states were incorporated to reconstruct the actual stress conditions in large deformation regions. In particular, exponential function calibration demonstrated significant advantages in processing the plastic deformation region after the material yield point. This freed the force calculations in severe collision scenarios from the limitations of linear models, significantly improving the accuracy of accident analysis in extreme situations.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, after calculating, for each ultra-high intensity collision area, the ratio of the theoretical regional force value to the actual regional force value, and for force areas that do not belong to the ultra-high intensity collision area, calculating the ratio of the theoretical regional force value to the actual regional force value, and obtaining the difference ratio data set, the method further includes:

[0019] Calculate the difference ratio change rate between the actual force value of the calibration area of ​​the ultra-high intensity collision area and the actual force value of the adjacent force area in the difference ratio data group; mark the ultra-high intensity collision area whose difference ratio change rate exceeds the preset theoretical difference ratio change rate range as an area that violates the force transmission continuity principle; search the historical accident database to extract similar historical cases whose historical vehicle driving data and vehicle driving data are within the similarity range of the preset vehicle driving data; extract the historical actual regional force values ​​corresponding to the area that violates the force transmission continuity principle in the similar historical cases; determine whether the similarity between the actual force value of the calibration area of ​​the area that violates the force transmission continuity principle and the historical actual regional force value is within the preset similarity threshold range; if not, apply the preset nonlinear calibration model to recalibrate the actual force value of the calibration area.

[0020] In the above-mentioned embodiment, multiple protective barriers are established to ensure that the force analysis results conform to the basic principles of physics. In particular, in complex collision environments, this prevents physical contradictions that may be caused by purely mathematical models, enabling the system to handle special collision scenarios that conventional models struggle to address. This improves the ability to analyze accidents in extreme situations and provides a reliable basis for evaluating traffic safety protection facilities.

[0021] In conjunction with some embodiments of the first aspect, in some embodiments, a final force compensation value is calculated using a regional weighted integration method based on a difference ratio matrix and a reliability weight coefficient table, specifically including:

[0022] The difference ratio matrix and the reliability weight coefficient table are matched and aligned one-to-one through spatial coordinates to obtain a weighted difference ratio data set; the collision concentration area whose corresponding difference ratio in the difference ratio matrix is ​​greater than the preset difference ratio threshold and whose corresponding weight value in the reliability weight coefficient table is higher than the preset weight threshold is evenly divided into a preset number of refined grids; an integral calculation is performed on the refined grids, and the calculation formula is the spatial integral of the product of the refined grid difference ratio and the refined grid weight coefficient of each refined grid; and the final force compensation value is obtained through spatial integration.

[0023] In the above example, the computational resolution in key areas is specifically improved, avoiding the waste of resources required for global high-resolution calculations. This meticulous analysis of key areas ensures that the characteristics of complex collision patterns are fully captured, and the specific force characteristics of concentrated areas are specifically addressed. This balance between computational efficiency and result quality provides a strong mathematical foundation for guardrail collision accident analysis.

[0024] In conjunction with some embodiments of the first aspect, in some embodiments, after obtaining the final force compensation value through spatial integration, the method further includes:

[0025] The final force compensation value is randomly perturbed a preset number of times, the standard deviation of the compensation value change within the preset perturbation range is calculated, and a confidence interval of the compensation value is established; the final force compensation value is applied to the theoretical force calculation model to calculate the compensated force value after compensation; based on the compensated force value, a collision process simulation model is constructed to simulate the deformation state of the guardrail, and the average Euclidean distance error and the maximum Euclidean distance error between the simulated deformation and the actual detected collision trace are calculated; areas where the maximum Euclidean distance error exceeds the preset error threshold are marked as compensation abnormal areas; and a compensation reliability assessment data package is generated.

[0026] In the above example, a quantitative indicator of result reliability is provided, and theoretical calculation results are converted into observable physical deformations through simulation. This achieves closed-loop verification from numerical to real-world data, proactively identifies and compensates for abnormal areas, and provides analysts with key areas of focus. This establishes a foundation for trust in accident analysis results, elevating guardrail collision analysis from a single numerical result to a complete solution that includes confidence intervals and reliability assessments.

[0027] In a second aspect, an embodiment of the present application provides an accident analysis system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the accident analysis system to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0028] In a third aspect, an embodiment of the present application provides a computer program product comprising instructions. When the computer program product is run on an accident analysis system, the accident analysis system executes the method described in the first aspect and any possible implementation of the first aspect.

[0029] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium comprising instructions. When the instructions are executed on an accident analysis system, the accident analysis system executes the method described in the first aspect and any possible implementation of the first aspect.

[0030] It is understandable that the accident analysis system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the accident analysis method based on digital twins provided in the embodiments of this application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects of the corresponding methods and will not be repeated here.

[0031] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0032] 1. This application reconstructs the force process of a guardrail collision. By using regional division, it overcomes the limitations of related technologies in overall averaging. The difference ratio data set accurately quantifies the difference between theoretical and actual forces. The reliability weight coefficient effectively addresses the nonlinear relationship between collision angle and deformation degree. The regional weighted integration method ensures the accuracy of force calculations in complex force scenarios. In particular, in complex scenarios such as vehicles colliding at challenging angles, this method can capture the inconsistency between local high-intensity deformation and overall force distribution, making accident analysis results more consistent with physical reality. This reduces the modeling bias of digital twin technology in challenging angle collision scenarios, improves the accuracy of virtual-reality fusion accident reconstruction, and significantly enhances the accuracy and reliability of road safety assessments.

[0033] 2. This application establishes multiple protective barriers to ensure that force analysis results conform to basic principles of physics. Specifically, in complex collision environments, this prevents physical inconsistencies that can arise from purely mathematical models, enabling the system to handle special collision scenarios that conventional models struggle to address. This enhances accident analysis capabilities in extreme scenarios and provides a reliable basis for evaluating traffic safety protection facilities.

[0034] 3. This application specifically improves computational resolution in key areas, avoiding the waste of resources associated with global high-resolution calculations. Detailed analysis of key areas ensures that the characteristics of complex collision patterns are fully captured, and the specific stress characteristics of concentrated areas are prioritized. This balance between computational efficiency and result quality provides a strong mathematical foundation for guardrail collision accident analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flowchart of the accident analysis method based on digital twins in an embodiment of the present application;

[0036] Figure 2 is another flow chart of the accident analysis method based on digital twins in an embodiment of the present application;

[0037] Figure 3 It is a schematic diagram of an exemplary hardware structure of the accident analysis system in an embodiment of the present application. DETAILED DESCRIPTION

[0038] The terms used in the following examples of the present application are for the purpose of describing specific embodiments only and are not intended to limit the present application. As used in this application, the singular expressions "a," "an," "said," "above," "the," and "this" are intended to include plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in this application refers to and encompasses any and all possible combinations of one or more of the listed items.

[0039] In the following, the terms "first" and "second" are used for descriptive purposes only and should not be understood to imply or suggest relative importance or implicitly indicate the number of the technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of this application, unless otherwise specified, "plurality" means two or more.

[0040] Among related technologies, highway guardrail accident analysis has begun to integrate on-site investigations, sensor data collection, and digital twin technology for comprehensive analysis. Traditional analysis methods rely on guardrail deformation data, vehicle trace information, and impact force data to establish a static model based on mechanical principles. Digital twin technology, on the other hand, attempts to reproduce the collision process in a digital environment by constructing a high-precision virtual model of the protective facility. However, even with the help of digital twin technology, when a vehicle collides with the guardrail at an awkward angle or in an irregular manner, the deformation representation in the virtual model still does not match the actual vehicle driving data. The digital model cannot accurately reflect the dynamic force process in complex scenarios, resulting in deviations in the digital twin's reconstruction of the accident process, affecting the accuracy of the analysis results and the scientific evaluation of guardrail performance.

[0041] In the embodiment of this application, the guardrail deformation is divided into regions, and combined with the vehicle collision angle and collision location information, a reliability weight coefficient table is constructed. The difference ratio analysis of the regional theoretical force value and the actual force value is performed, and the regional weighted integral method is used to calculate the final force compensation value. The final force compensation value is used to optimize the parameters in the digital twin model. This not only allows for more accurate assessment of the mechanical properties of each stress area, but also dynamically restores the accident process in digital space, overcomes the modeling deviation caused by collisions at tricky angles, improves the accuracy and scientific nature of the digital twin accident analysis results, and provides an important basis for guardrail performance optimization and accident prevention.

[0042] Figure 1 This is a flowchart of the accident analysis method based on digital twins in the embodiment of the present application, including the following steps:

[0043] S101, a guardrail force value calculated based on the guardrail deformation at the accident scene and vehicle driving data collected by a vehicle driving recorder;

[0044] Specifically, the guardrail's deformed area is measured in three dimensions to obtain deformation geometry data. Based on the mechanical properties of the guardrail material and incorporating material mechanics theory, Hooke's law is applied in the elastic region and a nonlinear constitutive relationship is employed in the plastic region to establish a deformation-force relationship model. Finite element inverse analysis is then applied to infer the force distribution from the known deformation. Simultaneously, driving data recorded by the vehicle's dashcam is collected, including parameters such as vehicle speed, acceleration, azimuth, and attitude changes.

[0045] S102: Inputting vehicle driving data into a preset theoretical force calculation model to obtain a theoretical force value of the vehicle on the guardrail;

[0046] Specifically, the theoretical force calculation model is built based on a neural network model. The vehicle driving data obtained in the previous step is standardized and converted into an input vector acceptable to the neural network. This processed data vector is then input into the preset neural network model, and a forward propagation calculation is performed to determine the theoretical force distribution of the vehicle on the guardrail, thereby obtaining the theoretical force calculation model.

[0047] This theoretical force calculation model is a computer representation of the complex collision mechanics process. It uses a large number of historical case studies to learn the inherent relationship between vehicle dynamic parameters and collision forces. During the training phase, a large amount of driving data from known vehicle collision accidents and force values ​​calculated using traditional mechanical methods were collected to construct a training set. Using supervised learning methods, the network weights were continuously adjusted through a backpropagation algorithm to ultimately minimize the error between the model output and the known theoretical force values.

[0048] The construction of this theoretical force calculation model first accumulated a wealth of vehicle collision samples through large-scale accident databases, collision experiments and computer simulations. Each sample contains complete vehicle driving parameters (speed, acceleration, mass, angle, etc.) and a force distribution map obtained through high-precision mechanical analysis; the data is then cleaned, standardized and feature extracted to construct a high-quality training set; a neural network is used in the model architecture, which includes a specially designed time feature extraction layer, spatial feature fusion layer and mechanical constraint layer; the training process adopts a phased strategy, first using synthetic data for pre-training to establish a preliminary mapping relationship, and then fine-tuning through real collision data. At the same time, physical constraint regularization terms are introduced to ensure that the prediction results conform to the laws of mechanics; the Adam optimizer is used in combination with cosine annealing learning rate scheduling in the optimization algorithm, and K-fold cross-validation is used to ensure the model's generalization ability; the final model is verified by collision tests and continuously iterated and optimized through feedback from actual accident case analysis.

[0049] In some embodiments of this application, when vehicle data is incomplete or noisy, the model employs a robustness enhancement design. This design, based on data augmentation and noise simulation techniques, artificially introduces different types of missing data and noise during training, enabling the network to learn how to handle imperfect data.

[0050] S103, dividing the force values ​​of the guardrail according to different deformation degrees to obtain regional force distributions of different force-bearing areas;

[0051] Specifically, a three-dimensional scan is performed on the deformed area of ​​the guardrail to obtain a spatial data point cloud of the deformation; the displacement difference between the deformed area and the original state is calculated to quantify the deformation degree of each point and construct a deformation intensity field; based on the stress-strain relationship of the material, a linear mapping is used in the elastic range, and a nonlinear mapping function is used in the plastic range to establish the relationship between deformation and force, and the deformation intensity is mapped to the corresponding stress distribution; a data clustering algorithm is used to discretize the continuous deformation field into several areas with similar force characteristics based on the gradient change of the deformation degree; finally, the average force value and force direction in each divided area are calculated in combination with the structural characteristics and material parameters of the guardrail to form a regional force distribution map.

[0052] In some embodiments of the present application, when plastic fracture occurs in the guardrail, the traditional elastic-plastic deformation analysis method is no longer applicable. At this time, a supplementary analysis method of fracture mechanics is required. First, the fracture surface characteristics are identified to determine the fracture type (ductile fracture or brittle fracture); then, the critical stress intensity factor at the time of fracture is calculated based on the fracture mechanics theory; finally, the force value that causes the fracture is inferred through the energy balance equation. In the complex case of multiple collisions and superimposed deformations, it is necessary to use deformation feature separation technology. Through corrosion degree analysis, deformation direction identification and material hardening property measurement, the deformation contribution caused by collisions at different times can be distinguished, and the force distribution of each collision can be calculated separately.

[0053] S104, dividing the theoretical force value according to the regional force distribution to obtain the regional theoretical force value and the regional actual force value corresponding to each force region;

[0054] Specifically, the force-bearing area of ​​the guardrail determined based on the aforementioned deformation analysis is divided into a basic spatial reference system; then the theoretical force value calculated by the neural network model is projected in space, and the theoretical force component corresponding to each area is determined based on the relationship between the theoretically calculated force vector and the spatial position of each area; at the same time, for each divided area, a mechanical inversion calculation is performed based on the actual deformation characteristics to obtain the actual force value of the area and associate it with the corresponding spatial area coordinates; finally, a regional-level theoretical-actual force value correspondence table is established to form basic mechanical analysis data at the regional granularity.

[0055] S105, marking the stress area where the actual stress value exceeds the preset maximum stress threshold as an ultra-high intensity collision area;

[0056] Specifically, the maximum stress thresholds of the guardrail's materials and structure are first determined—the maximum force the guardrail can withstand while maintaining basic functional integrity. These thresholds are derived from design specifications, material strength theory, structural failure mode analysis, or extensive experimental data. The actual stress values ​​calculated for each area are then compared with the preset maximum stress thresholds for that area. If the stress value in a region exceeds the threshold, the system automatically marks that region as an ultra-high-intensity collision zone.

[0057] S106. Obtaining force response curves of the guardrail material under different deformation degrees;

[0058] Specifically, the various materials used in guardrails (typically high-strength steel, aluminum alloy, or composite materials) undergo mechanical property testing in advance. These tests include standard tensile, compression, bending, and impact tests, comprehensively covering the material's response characteristics under various stress conditions. During the tests, high-precision sensing equipment records the deformation of the material specimen under gradually increasing external forces in real time, establishing raw force-displacement data pairs. This raw data is then mathematically processed to calculate the material's stress-strain relationship, focusing specifically on the transition characteristics between the elastic, yield, strengthening, and fracture zones. The discrete test data is then fitted using mathematical models (such as the Ramberg-Osgood model and the Johnson-Cook model) to obtain a continuous, smooth mathematical expression, representing the force-response curve.

[0059] S107. Construct a segmented calibration function based on the force response curve by using linear transformation in a preset first range, quadratic function calibration in a preset second range, and exponential function calibration in a preset third range.

[0060] Specifically, the obtained material stress response curve is divided into ranges. Usually, based on the deformation mechanism of the material, the entire deformation process is divided into three continuous regions with different mechanical properties: the first range region dominated by elastic deformation, the second range region of elastic-plastic transition, and the third range region dominated by plastic deformation; then, for each range region, the most suitable mathematical function to describe its physical properties is selected: in the first range region, the material follows Hooke's law, and stress is proportional to strain, so the linear function y=ax+b is used for calibration; in the second range region, the material enters the yield stage, and the stress-strain relationship shows nonlinear growth but the change is relatively gentle, so the quadratic function is used. Fitting is performed; in the third range, the material strengthens or softens, the stress-strain relationship changes dramatically, and an exponential function is used Capture this rapidly changing characteristic; finally, apply continuity and smoothness constraints at the regional intersection points to ensure the continuity of functions in different regions in terms of value and derivative, forming a complete piecewise calibration function.

[0061] S108, applying the segmented calibration function to the ultra-high intensity collision area to calculate the actual force value of the calibration area corresponding to the ultra-high intensity collision area;

[0062] Specifically, based on the ultra-high intensity collision area marked in the previous steps, the deformation value distribution in each area is determined, and the spatial displacement field of the deformed surface is usually obtained through high-precision three-dimensional scanning; for the deformation amount of each scanning point, the range area (first, second or third range) in which it is located is judged, and the corresponding calibration function is selected: when the deformation amount falls in the first range, the linear function is applied, the quadratic function is applied in the second range, and the exponential function is applied in the third range; then, by substituting the measured deformation amount into the corresponding calibration function, the calibrated force value of each point is calculated; then, based on the principle of spatial integration, the force values ​​of the discrete points are integrated or weighted averaged in the area to obtain the overall calibrated force value of the entire ultra-high intensity collision area.

[0063] S109. For each ultra-high intensity collision region, calculate the ratio of the theoretical force value of the region to the actual force value of the calibration region. For the force regions that do not belong to the ultra-high intensity collision region, calculate the ratio of the theoretical force value of the region to the actual force value of the region to obtain a difference ratio data set.

[0064] Specifically, for each identified ultra-high intensity collision area, the actual force value of the calibration area calculated in the previous step is extracted, and the theoretical force value of the corresponding area is obtained from the theoretical collision model at the same time; then, the ratio is calculated by mathematical operation = theoretical force / actual calibration force, which reflects the degree of prediction deviation of the theoretical model under extreme collision conditions; then, for ordinary force areas that do not belong to the ultra-high intensity collision area, the regional theoretical force value and the regional actual force value (conventional calculated value that has not been specially calibrated) are also extracted, and the ratio is calculated = theoretical force / actual force; then the ratio data of all areas are organized according to the order of spatial position or force magnitude to form a complete difference ratio data group.

[0065] In some embodiments of the present application, the appearance of ultra-high intensity collision areas is caused by measurement errors or calculation deviations, which will lead to errors in the force prediction results. It is necessary to apply multiple historical verification and nonlinear calibration techniques to perform abnormal force value inspection and correction to ensure the scientific reliability of extreme collision mechanics analysis.

[0066] The rate of change of the difference ratio between the actual force values ​​of the calibrated area in the ultra-high-intensity collision region and the actual force values ​​of the adjacent force regions in the difference ratio data set was calculated. First, each ultra-high-intensity collision region and its adjacent regions were identified, and their spatial positional relationship and distance parameters were determined. The calculated ratio data for two types of regions were then extracted: the ratio of the theoretical force value to the calibrated actual force value in the ultra-high-intensity region and the ratio of the theoretical force value to the actual force value in the adjacent normal region. The difference between the two ratios was then calculated: the ratio of the ultra-high-intensity region minus the ratio of the normal region. This ratio difference was then divided by the standardized spatial distance between the two regions to obtain the rate of change, which represents the magnitude of the spatial variation in the force field. The calculation process is based on the concept of stress gradient in continuum mechanics theory, converting the ratio changes between discrete regions into a quasi-continuous gradient field representation.

[0067] Ultra-high-intensity collision regions where the difference ratio change rate exceeds the preset theoretical difference ratio change rate range are marked as regions that violate the principle of force transmission continuity. First, based on the material continuum mechanics theory and the structural force transmission characteristics, a reasonable theoretical difference ratio change rate range is determined. This range is usually determined through statistical analysis of a large amount of collision test data or finite element simulation results, and represents the reasonable gradient limit of force transmission between adjacent regions under normal physical conditions. The actual difference ratio change rate calculated above for each ultra-high-intensity collision region is then compared with this preset range. When the change rate of a certain region exceeds the theoretical range, the system marks it as a region that violates the principle of force transmission continuity.

[0068] The historical accident database is searched to extract historical cases whose historical vehicle driving data is similar to the vehicle driving data within the preset vehicle driving data similarity range. First, the vehicle driving data characteristics of the current collision accident are clarified, including multi-dimensional parameters such as vehicle type, mass, speed, driving direction, collision angle and collision point location. Then, the similarity judgment criteria are defined. Usually, multi-dimensional similarity calculation methods such as weighted Euclidean distance or Mahalanobis distance are used to assign importance weights to different parameters and construct a similarity quantification model. Then, a preset similarity range threshold is set. This threshold is determined based on statistical analysis and represents an acceptable similarity tolerance. The system then automatically connects to the historical accident database, which contains a large number of standardized historical collision cases and their complete parameter sets. The historical cases are traversed and compared using a similarity calculation algorithm to calculate the similarity index with the current case. Finally, a set of historical cases whose similarity index is within the preset threshold range is selected.

[0069] Extract the historical actual regional force values ​​corresponding to the areas that violated the principle of force transmission continuity in similar historical cases. First, determine the areas marked as violating the principle of force transmission continuity in the current case and record their precise spatial coordinates and geometric features on the guardrail. Then, establish a standardized spatial reference system for each retrieved similar historical case, usually based on key structural points (such as pillar positions and end positions) for coordinate alignment and scale calibration. Then, through a spatial mapping algorithm, project the coordinates of the continuity violation areas in the current case into the standardized coordinate system of the historical case to identify the corresponding areas in the historical case. Then, accurately extract the actual measured or calculated force value data of these corresponding areas from the force analysis database of the historical case. Finally, perform necessary normalization on the extracted historical force data to eliminate possible scale differences between different cases and make the data comparable.

[0070] Determine whether the similarity between the actual force values ​​of the calibration area in the area that violates the principle of continuity of force transmission and the historical actual force values ​​of the area is within the preset similarity threshold. First, the actual force values ​​of the calibration area of ​​the current case obtained in the previous step and the corresponding force values ​​of the historical case extracted are preprocessed, including unit unification and data standardization; then the similarity index of the two sets of data is calculated, usually using statistics such as relative error, correlation coefficient or root mean square deviation to quantify the degree of closeness of the two sets of data; then the calculated similarity index is compared with the preset threshold, which is usually determined based on a large number of verification test statistics and represents an acceptable normal fluctuation range; then the system automatically determines whether the similarity is within the threshold range. If it is within the range, it indicates that although the current calibration result violates the continuity principle, it is consistent with the historical data, which may reflect a special physical phenomenon; if it is outside the range, it indicates that the current calibration result violates both theoretical expectations and historical experience, and there may be calculation or measurement deviations.

[0071] If the range is exceeded, the preset nonlinear calibration model is applied to recalibrate the actual force value of the calibration area; the nonlinear calibration model is pre-established based on sensor experimental data and material extreme force test data. In the event that the similarity exceeds the preset similarity threshold range, the pre-established nonlinear calibration model is first called. This model is constructed through systematic experiments - actual guardrail material samples are placed in precision loading equipment, and the correspondence between sensor output and actual force is measured at different degrees of deformation. At the same time, combined with the special nonlinear response data of the material under the extreme state, a complex mapping relationship is established through high-order polynomial fitting, neural network algorithms, or semi-empirical formulas of physical constraints; then the deformation data and environmental parameters of the area that violates the continuity principle are input into the model; then the model automatically considers complex factors such as material strain hardening, strain rate effect, temperature influence, etc., and generates a revised force value prediction; finally, this revised value replaces the previous calibration result.

[0072] S110, calculating the reliability index of each region based on the vehicle collision angle and collision location information in the vehicle driving data, combined with the deformation degree and stress characteristics of each stress region in the deformation situation, and constructing a reliability weight coefficient table;

[0073] Specifically, first, the collision angle and collision site coordinates are obtained from the vehicle driving data, which come from the vehicle data recorder or accident reconstruction calculation; then, for each divided guardrail force area, the deformation direction vector is obtained through three-dimensional scanning or measurement technology, which reflects the dominant direction of regional deformation; then the collision direction vector is calculated, which is determined by the collision angle and geometric relationship; then the consistency index of the collision direction and the deformation direction is calculated by the vector dot product, that is, the dot product of the two vectors divided by the product of the lengths of the two vectors. The index value ranges from negative one to positive one, and the closer the value is to positive one, the higher the direction consistency; based on the consistency index, a nonlinear mapping function is applied to calculate the reliability index, usually using an S-type function to amplify the differences in the middle area; finally, a normalized weight coefficient table is constructed according to the reliability index of each area.

[0074] S111. Based on the difference ratio data set and the reliability weight coefficient table, the final force compensation value is calculated using the regional weighted integration method;

[0075] Specifically, for each divided force area of ​​the guardrail, the difference ratio data of the area (the ratio of the theoretical force value to the actual force value) and the corresponding reliability weight coefficient are extracted; then regional data weighting processing is performed, and the difference ratio of each area is multiplied by its reliability weight coefficient to obtain weighted difference data. This step ensures that the high reliability area has a greater contribution to the final result; then a spatial integration operation is performed, and the weighted differences of all areas are integrated and summed in the entire guardrail force space, usually using a discrete numerical integration method such as the rectangular method, trapezoidal method or Simpson method; finally, the integration result is divided by the weighted sum for normalization to obtain the final force compensation value.

[0076] S112. Perform accident analysis based on the final force compensation value and the preset digital twin model.

[0077] Specifically, the calculated final force compensation values ​​are first imported as parameters into the digital twin platform to adjust the material properties, structural stiffness coefficients, and energy absorption characteristics of the virtual guardrail model. During the adjustment process, the system specifically corrects the mechanical parameters of each area based on the compensation values ​​of different force-bearing areas, enabling the digital twin model to accurately reflect the nonlinear deformation behavior in actual collisions. Once the parameters are adjusted, the digital twin model will reconstruct the entire accident scenario and simulate the entire collision process through time-series deduction. This will restore key information such as the vehicle's driving trajectory, collision angle, and contact time at the time of the accident, reconstruct the accident process, and complete the accident analysis.

[0078] The digital twin model is a guardrail virtual simulation system built using high-precision 3D modeling technology. In accident analysis applications, the digital twin model implements a bidirectional data flow mechanism. It receives actual guardrail deformation data and vehicle driving parameters as input, and dynamically adjusts its own parameters based on the final force compensation value, achieving a precise mapping of the virtual and real world. The model incorporates multiple material constitutive equations and structural dynamics algorithms, enabling it to handle complex mechanical behavior in a variety of scenarios, including high-speed collisions and low-speed scrapes. In particular, it provides a detailed simulation of the strain softening and fracture characteristics of steel during the plastic deformation stage.

[0079] The digital twin model also integrates vehicle dynamics and terrain models to form a complete accident scenario reconstruction system. During collision simulation, the system calculates changes in friction between the tires and the road surface, dynamic adjustments to the vehicle's posture, and the reaction of guardrail deformation to vehicle motion. This captures the energy conversion and momentum transfer of the entire collision process over a timescale, providing a scientific basis for determining the cause of the accident and assigning responsibility.

[0080] In the above embodiment, high-precision accident reconstruction and force assessment are achieved by combining the regional division of the guardrail deformation state with the analysis of vehicle driving data. This method obtains actual and theoretical force data; identifies ultra-high-intensity collision areas and applies a piecewise calibration function; constructs a reliability weight coefficient based on the collision angle and location information; calculates the force compensation value through regional weighted integration, and optimizes the parameters of the digital twin model. Related technologies mainly rely on on-site surveys and sensor data to establish static mechanical models based on Newtonian mechanics and collision theory. However, in collision scenarios with tricky angles, there is still a mismatch between the digital model and the actual situation, which affects the accuracy of accident analysis. Compared with related technologies, the technical solution provided by this application establishes a regional analysis framework to deal with local force differences; adopts a piecewise calibration function to solve the nonlinear response of materials; integrates a reliability weight system; and applies a weighted integration method to obtain a global optimal solution, accurately restoring the dynamic process in complex collision scenarios. This effectively solves the modeling deviation of digital twin technology at special collision angles and significantly improves the accuracy of virtual accident reconstruction and the reliability of protective facility performance evaluation.

[0081] In other embodiments of this application, in traffic accidents involving complex angle collisions and high-intensity deformation, localized force assessment distortion may occur. Using the digital twin-based accident analysis method provided in this application, precise mechanical reconstruction of complex collision scenarios can be achieved through intelligent identification of abnormal deformation areas, construction of a multi-factor weighting system, and weighted integral calculation of refined grids in differential areas. This effectively addresses the analytical challenge of discrepancies between guardrail deformation and vehicle driving data caused by challenging angle collisions.

[0082] like Figure 2FIG. 1 is another flow chart of the accident analysis method based on digital twins provided in an embodiment of the present application, comprising the following steps:

[0083] S201, a guardrail force value calculated based on the deformation of the guardrail at the accident scene and vehicle driving data collected by a vehicle driving recorder;

[0084] S202: Inputting vehicle driving data into a preset theoretical force calculation model to obtain a theoretical force value of the vehicle on the guardrail;

[0085] S203, dividing the force values ​​of the guardrail according to different deformation degrees to obtain regional force distributions of different force-bearing areas;

[0086] S204, dividing the theoretical force value according to the regional force distribution to obtain the regional theoretical force value and the regional actual force value corresponding to each force region;

[0087] S205, for each stress area, respectively calculating the difference ratio between the theoretical stress value of the area and the actual stress value of the area to obtain a difference ratio data set;

[0088] S206: Calculate the collision direction coefficient and the collision location importance coefficient respectively using a preset angle mapping function and a position evaluation function based on the collision angle and collision location information in the vehicle driving data;

[0089] Specifically, first, the collision angle value (usually expressed as the angle between the longitudinal axis of the vehicle and the normal of the guardrail surface) and the collision site coordinate data (usually expressed as the position mark and height along the guardrail) are extracted from the vehicle driving data; then the extracted collision angle is input into a preset angle mapping function, which is a mathematical relationship that converts the angle value into a standardized collision direction coefficient. The design purpose is to reflect the degree of influence of collisions at different angles on the force characteristics of the guardrail; then the collision site coordinates are input into the position evaluation function, which converts the position information into a standardized site importance coefficient based on the structural characteristics and force characteristics of different positions of the guardrail; finally, two key coefficients are obtained: the collision direction coefficient (reflecting the angle influence) and the site importance coefficient (reflecting the position influence).

[0090] The angle mapping function and position evaluation function are special evaluation tools developed based on a large amount of collision test data and guardrail structural mechanics analysis. The angle mapping function is established through mechanical experiments and numerical simulations to convert the collision angle into the collision efficiency coefficient.

[0091] S207, comparing the deformation direction vector in the deformation situation with the theoretical collision force direction vector, and calculating the deformation direction consistency parameter;

[0092] Specifically, first, the deformation direction vectors of each evaluation area of ​​the guardrail are obtained. These vectors are obtained through three-dimensional scanning or multi-point measurement technology, and represent the main direction and amplitude of the deformation; then, based on the vehicle collision angle, mass distribution and dynamic characteristics, the theoretical collision force direction vector is calculated, which represents the direction in which the collision force should be transmitted in theory; then, for each evaluation area, the spatial angle between the deformation direction vector and the theoretical collision force direction vector is calculated, usually using a vector dot product operation and normalization combined with the vector length; then the angle value is converted into a consistency parameter, usually using a cosine mapping or a custom nonlinear mapping function, so that the consistency parameter is distributed between zero and one, and the closer the value is to one, the higher the directional consistency; finally, the consistency parameters of all areas are statistically analyzed to evaluate the degree of conformity of the overall deformation with theoretical expectations.

[0093] S208, identifying abnormal deformation areas in the deformation situation;

[0094] Specifically, first, the deformation discontinuity area is identified by calculating the deformation gradient between adjacent measurement points or grid units. When the gradient value exceeds the preset threshold, it is determined to be a deformation discontinuity area. This discontinuity usually manifests as a mutation, fracture or folding; then the deformation direction abnormal area is identified, and the actual deformation direction of each area is compared with the deformation direction expected based on the collision force transfer theory. When the deviation angle between the two exceeds the preset threshold, it is determined to be a direction abnormal area; then the deformation area that exceeds the elastic limit of the material is identified, and the corresponding strain value is calculated by measuring the deformation amount and combining the material characteristic parameters. When the strain value exceeds the elastic limit of the material, it is determined to be a plastic deformation area; finally, the information of the three types of abnormal areas is integrated to form a complete distribution map of deformation abnormal areas.

[0095] S209: For the abnormal deformation area, according to the abnormality type, a historical damage correction algorithm, a fracture area compensation algorithm, or a deformation continuity reconstruction algorithm is used to generate a corrected deformation consistency parameter;

[0096] Specifically, first, the abnormal deformation areas identified in the above steps are classified to clarify the abnormal type of each area; then, for abnormal areas with historical damage, a historical damage correction algorithm is applied. This algorithm separates the deformation caused by historical damage and the current collision by comparing the pre-collision state record with the current state, eliminating the interference of historical factors; then, a fracture area compensation algorithm is applied to the fracture area. This algorithm is based on the fracture mechanics theory and combines the fracture edge characteristics and material performance parameters to reconstruct the continuous deformation field and critical fracture force before fracture; then, a deformation continuity reconstruction algorithm is applied to the deformation discontinuous area. This algorithm is based on the deformation pattern of the surrounding normal area and restores the theoretical deformation state of the discontinuous area through mathematical interpolation or physical constraint model; finally, based on the corrected deformation data, the consistency parameters of the deformation direction of each area and the theoretical collision force direction are recalculated to obtain more accurate physical rationality evaluation indicators.

[0097] The historical damage correction algorithm, fracture area compensation algorithm, and deformation continuity reconstruction algorithm are pre-defined functions. The historical damage correction algorithm compares the guardrail's state records before and after a collision, identifying and filtering out deformations caused by previous accidents or environmental factors, ensuring that only deformation data generated by the current collision is analyzed. The fracture area compensation algorithm, based on the principles of material fracture mechanics, analyzes fracture edge characteristics and material parameters to reconstruct the material's mechanical state and critical fracture force before fracture, addressing the issue of missing information in the fracture area. The deformation continuity reconstruction algorithm targets regions of discontinuous deformation, utilizing the deformation gradients of adjacent normal regions and the physical constraint model. Through mathematical interpolation and boundary condition matching, it restores the continuity distribution of the deformation field and the theoretical deformation state of the damaged area.

[0098] In some embodiments of this application, standard correction algorithms require adjustments for multi-material composite structures or non-standard collision conditions. For steel-plastic composite guardrails, due to the significantly different deformation mechanisms and fracture characteristics of the two materials, a layered correction strategy is required: first, identify the different material regions and their interfaces; then, apply a correction algorithm based on elastic-plastic theory to the steel portion and a correction algorithm based on viscoelastic theory to the plastic portion; then, address the combined effects of the material interfaces, taking into account interface constraints and force transmission characteristics; and finally, integrate the correction results for each component.

[0099] S210, comprehensively calculating the collision direction coefficient, the location importance coefficient, the deformation consistency parameter, and the corrected deformation consistency parameter, calculating the reliability index of each stress-bearing area, and constructing a reliability weight coefficient table through normalization processing;

[0100] Specifically, first, four key evaluation indicators are collected: collision direction coefficient (a quantitative indicator reflecting the influence of collision angle on the force of guardrail), location importance coefficient (a quantitative indicator representing the structural characteristics and functional importance of collision location), deformation consistency parameter (an indicator measuring the degree of consistency between deformation direction and theoretical collision force direction), and corrected deformation consistency parameter (an optimized consistency indicator after being processed by abnormal area correction algorithm); then, for each divided force area, a multi-indicator comprehensive evaluation method is used to calculate the reliability index, usually using weighted average method or multi-indicator nonlinear combination function to weight and integrate the four indicators according to their relative importance to obtain a single reliability index value; then, the calculated reliability index of each area is normalized, usually using maximum value normalization or sum normalization method to ensure that the numerical range of all weight coefficients is appropriate and the sum is one; finally, a complete reliability weight coefficient table is formed, which records in detail the trust weight that should be given to each area of ​​the guardrail in the subsequent force calculation and analysis.

[0101] In some embodiments of the present application, the reliability index of each force-bearing area is calculated by comprehensively considering the collision direction coefficient, the location importance coefficient, the deformation consistency parameter, the modified deformation consistency parameter, and the first formula; wherein the first formula is: ; Among them, R is the reliability index, is the collision direction coefficient, which represents the effective component of the collision force in the normal direction. is the location importance coefficient, which indicates the structural stress characteristics. is the deformation consistency parameter, and the modified deformation consistency parameter is used for abnormal areas , 、 、 、 is the preset parameter constant; S is the special deformation adjustment factor, and the calculation formula is , 、 、 They are the degree of historical damage, the degree of fracture characteristics and the degree of discontinuity, respectively.

[0102] The first formula proposed in the embodiment of this application is a highly optimized multi-factor nonlinear comprehensive evaluation mathematical structure, which realizes the scientific integration of collision characteristics, structural characteristics and deformation characteristics. and location importance coefficient The product of is used as the basic quantity to reflect the effectiveness and position criticality of the collision force; then the deformation consistency parameter is introduced The index term , the item in It decays rapidly when it is low, reflecting the threshold constraint effect of consistency on reliability; then the reward mechanism for high-quality data is realized through the hyperbolic tangent enhancement term. Exceeding the threshold , the reliability score is improved nonlinearly; finally multiplied by the special deformation adjustment factor The "maximum value" operation ensures that any serious anomaly (historical damage, fracture characteristics or discontinuity) can be fully reflected, reflecting the principle of the short board effect in safety assessment. The preset parameter constants in this formula are 、 、 、 It is used to adjust the sensitivity and weight distribution of various factors. After optimization and calibration, it can adapt to the characteristic evaluation requirements of different guardrail systems.

[0103] First, the first formula adopts a product-type basic structure. Any deficiency in any key factor will significantly reduce the overall reliability, which conforms to the conservative principle of engineering safety assessment. Second, the introduction of exponential and hyperbolic tangent terms enables the model to exhibit different sensitivities in different parameter intervals, enabling precise distinction between boundary conditions and extreme values. Third, the special deformation adjustment factor adopts a maximum value mechanism, ensuring high sensitivity to various abnormal conditions and avoiding the problem of multiple abnormalities masking each other. Fourth, the overall formula structure achieves a scientific mapping from multidimensional evaluation indicators to a single reliability index, simplifying the subsequent analysis process. Finally, the various components of the formula have clear physical meanings and mathematical interpretations, which improves the interpretability and persuasiveness of the assessment results. Overall, this reliability index calculation method based on advanced mathematical models successfully transforms complex engineering problems into precise numerical assessments, providing more reliable and objective technical support for traffic safety facility performance analysis and accident responsibility determination. It has important practical value and theoretical innovation significance in improving road safety and optimizing the design of protective facilities.

[0104] S211, matching and aligning the difference ratio matrix and the reliability weight coefficient table in one-to-one correspondence through spatial coordinates to obtain a weighted difference ratio data set;

[0105] Specifically, two key input data are first identified: the difference ratio matrix (spatial distribution data recording the ratio of the theoretical force value to the actual force value in each area of ​​the guardrail) and the reliability weight coefficient table (spatial distribution data recording the reliability assessment results of the data in each area); then a unified spatial reference system is established, usually using a local coordinate system with the starting point of the guardrail as the origin and the direction of the guardrail as the main axis; then the two data sets are spatially aligned to ensure that the difference ratio data of each spatial position can accurately correspond to the reliability weight coefficient of the corresponding position; then, for each spatial unit, the difference ratio value is associated with the corresponding reliability weight coefficient and stored to form an enhanced data set containing the original difference, position information and weight coefficient; finally, data validity verification is performed to ensure that all spatial units have valid matching results.

[0106] S212, evenly dividing the collision concentration area whose corresponding difference ratio in the difference ratio matrix is ​​greater than a preset difference ratio threshold and whose corresponding weight value in the reliability weight coefficient table is higher than a preset weight threshold into a preset number of refined grids;

[0107] Specifically, collision concentration areas are first identified based on two key indicators: the difference ratio (an indicator indicating the degree of deviation between theoretical and actual forces) and the reliability weight coefficient (an indicator indicating the credibility of the data); a dual-threshold screening mechanism is then applied, and areas that meet both the "difference ratio greater than a preset difference ratio threshold" and the "weight value higher than a preset weight threshold" conditions are marked as collision concentration areas. This ensures that the selected areas have both significant mechanical characteristics (large difference) and reliable data support (high weight); mesh refinement is then applied to the identified collision concentration areas, evenly dividing the original coarse analysis mesh into a preset number of finer mesh cells, usually using regular subdivision strategies such as bisection, quartering, or nine-sectioning; finally, the data in the refined mesh are spatially interpolated or recalculated to ensure that the refined mesh cells have reasonable data values.

[0108] S213, performing integral calculation on the refined grids, where the calculation formula is the spatial integral of the product of the refined grid difference ratio and the refined grid weight coefficient of each refined grid;

[0109] Specifically, two key calculation quantities are first obtained: the refined grid difference ratio (defined as the ratio of the theoretical force value of the refined grid to the actual force value of the guardrail, reflecting the degree of deviation between the mechanical model prediction and the actual measurement) and the refined grid weight coefficient (a credibility factor determined based on the comprehensive determination of the vehicle collision angle, collision location information and force distribution); then, for each refined grid unit, the product of the difference ratio and the weight coefficient is calculated to form a weighted difference field; then, a spatial integration operation is performed on the entire collision concentration area, usually using numerical integration methods such as the rectangular method, trapezoidal method or Simpson method, to convert the weighted difference field on the discrete grid into a continuous integration result; finally, the integration result is associated with the area of ​​the integration region to obtain a comprehensive indicator characterizing the overall force deviation.

[0110] S214. Obtaining a final force compensation value through spatial integration;

[0111] Specifically, the optimization objective is first defined: finding an optimal compensation value such that the sum of the weighted squared errors between the theoretical force value plus the compensation value and the actual measured force value is minimized across the entire grid space. An objective function is then constructed: the square of the difference between the theoretical force value at each grid point plus the compensation value to be determined and the actual force value, multiplied by the reliability weight of each point, and integrated over the entire grid space. The objective function is then differentiated with respect to the compensation value δ and set equal to zero, yielding an analytical expression for the optimal compensation value, typically the integral of the product of the weight and the difference divided by the integral of the weight. The integral is then calculated using numerical integration methods (such as the composite trapezoidal method or the Simpson method) using the grid data obtained in the previous steps. Finally, the optimal compensation value is obtained, which ensures a global best fit between the theoretical model and the actual measurements under a given weight distribution. This method, based on the variational principle and optimization theory, achieves global optimization correction of the force analysis by systematically considering all data points within the entire analysis domain.

[0112] In some embodiments of the present application, when the collision situation is complex and there are multiple uncertainties, and the standard compensation method may not be sufficient to accurately simulate the actual deformation of all areas, the reliability of the compensation model is verified and evaluated, especially in identifying those areas that cannot accurately simulate the actual deformation even after compensation adjustment.

[0113] The final force compensation value is randomly perturbed a preset number of times, the standard deviation of the compensation value variation within the preset perturbation range is calculated, and a confidence interval for the compensation value is established. First, the basic parameters of the random perturbation are determined: the number of perturbations (usually several hundred to several thousand times) and the perturbation range (usually set to a few percent to around ten percent of the original compensation value). The calculated final force compensation value is then randomly perturbed multiple times, each time generating a random offset within the preset range, resulting in a series of perturbed compensation value samples. These perturbation samples are then statistically analyzed to calculate the sample mean (which should be close to the original compensation value) and the sample standard deviation (which reflects the stability and sensitivity of the compensation value). Based on the theory of standard normal distribution, the standard deviation is used to construct a confidence interval for the compensation value. Finally, the best estimate of the compensation value and its uncertainty range are output. This method, based on statistical sampling theory and the principles of Monte Carlo simulation, artificially introduces random perturbations to assess the model's sensitivity to changes in input parameters, quantify the degree of uncertainty in the compensation value, and provide more complete analysis results.

[0114] Apply the final force compensation value to the theoretical force calculation model to calculate the compensated force value after compensation. Apply the final force compensation value obtained in the previous step to the theoretical force calculation model to obtain a more accurate force prediction value. The theoretical force calculation model is usually based on the principles of collision mechanics, and considers factors such as vehicle mass, speed, and collision angle to calculate the theoretical force value of each point on the guardrail. However, due to model simplification and parameter uncertainty, the theoretical value often deviates from the actual value. The final force compensation value is a correction for this systematic deviation. There are several possible ways to apply this compensation value: additive compensation (theoretical value + compensation value), multiplicative compensation (theoretical value × (1 + compensation coefficient)) or more complex function compensation.

[0115] Based on the compensated force values, a collision process simulation model is constructed to simulate the guardrail deformation state. The average Euclidean distance error and the maximum Euclidean distance error between the simulated deformation and the actual detected collision traces are calculated. First, a finite element analysis model or other mechanical simulation model is established based on the compensated force values ​​and the material properties of the guardrail (such as elastic modulus, yield strength, fracture toughness, etc.). Then, a computer simulation is performed to simulate the deformation process of the guardrail under the action of force to obtain three-dimensional geometric data of the simulated deformation. Then, the guardrail deformation measurement data at the actual collision scene is obtained, usually through three-dimensional scanning or multi-point measurement methods. The Euclidean distance between the two sets of geometric data (i.e., the straight-line distance between corresponding points in three-dimensional space) is then calculated to obtain the average Euclidean distance error (reflecting the overall fitting accuracy) and the maximum Euclidean distance error (reflecting the local maximum deviation). These error indicators quantitatively characterize the degree of fit between the compensated model and the actual situation.

[0116] Regions where the maximum Euclidean distance error exceeds a preset error threshold are marked as compensation anomaly regions. First, a threshold for the maximum Euclidean distance error is set. This threshold represents the maximum acceptable model deviation and is typically determined based on engineering requirements, material properties, and measurement accuracy. The maximum Euclidean distance error calculated in the previous step is then compared with the preset threshold to identify all spatial regions where the error exceeds the threshold. Regional clustering analysis is then performed to merge adjacent outliers into continuous outlier regions. The boundaries of each identified outlier region are then defined and the coordinates recorded. Finally, these regions are marked as "compensation anomaly regions," indicating that the current compensation model fails to meet the expected accuracy requirements within these regions.

[0117] Generate a compensation reliability assessment data package; the compensation reliability assessment data package records the spatial coordinate range and compensation error value of the compensation anomaly area. First, collect all the compensation anomaly area information identified in the previous steps, including their spatial location, geometric range, and corresponding error data. Then, standardize this information and convert it into a unified data format to ensure cross-platform use and long-term storage. Then, generate a detailed attribute description for each anomaly area, including the area boundary coordinates (usually expressed in a three-dimensional coordinate system), area or volume, maximum error value, average error value, and its spatial distribution characteristics. Then, integrate all the anomaly area data and the overall assessment results into a structured data package. Finally, attach necessary metadata, such as analysis date, compensation method used, threshold setting, and other background information.

[0118] S215. Perform accident analysis based on the final force compensation value and the preset digital twin model.

[0119] Steps S201-S204, S215 and Figure 1 In the illustrated embodiment, steps S101 - S104 and S112 are similar, and the descriptions of steps S101 - S104 and S112 may be referred to, and will not be repeated here.

[0120] In the above-described embodiment, high-precision accident reconstruction in complex collision scenarios is achieved through multi-dimensional data fusion and regionalized fine-grained analysis. First, actual deformation data and theoretical calculated values ​​are obtained; the guardrail is divided into multiple force-bearing regions and the difference ratio is calculated; the collision direction and importance coefficient are constructed based on the collision angle and location; abnormal deformation areas are identified and corrected using specialized algorithms; a reliability weighting system is constructed; a refined grid is applied to key areas; and the final force compensation value is calculated through weighted spatial integration. This solves the problem of difficult angle collisions that traditional methods cannot handle. The combination of correction algorithms effectively processes information about abnormal deformation areas, while refined grids and spatial integration ensure computational accuracy. The entire process establishes a complete technical chain from raw data to reliable analysis results, making guardrail collision accident analysis more consistent with physical facts.

[0121] The following introduces an exemplary accident analysis system 300 provided in an embodiment of the present application. Figure 3 3 is a schematic diagram of an exemplary hardware structure of the accident analysis system 300 provided in an embodiment of the present application.

[0122] In some embodiments, the accident analysis system 300 is a computer device or the accident analysis system 300 includes a computer device. The computer device includes a processor, a memory and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with other external terminals or servers through a network connection. In some embodiments, the network interface can be a wired network interface, and in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, the method in the embodiment of the present application is implemented.

[0123] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0124] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0125] As used in the above embodiments, the term “when…” may be interpreted as “if…” or “after…” or “in response to determining…” or “in response to detecting…”, depending on the context. Similarly, the phrases “upon determining…” or “if (stated condition or event) is detected” may be interpreted as “if determining…” or “in response to determining…” or “upon detecting (stated condition or event)” or “in response to detecting (stated condition or event)”, depending on the context.

[0126] In the above embodiments, all or part of the embodiments can be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this application are generated in whole or in part. 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, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disk, hard disk, tape), optical media (e.g., DVD), or semiconductor media (e.g., solid-state drive).

[0127] Those skilled in the art will appreciate that all or part of the process steps in the above-described method embodiments can be implemented by a computer program instructing the relevant hardware. The program can be stored in a computer-readable storage medium, and when executed, the program can include the process steps in the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. An accident analysis method based on digital twins, characterized in that: include: The guardrail force value is calculated based on the deformation of the guardrail at the accident scene and the vehicle driving data collected by the vehicle driving recorder is obtained; Inputting the vehicle driving data into a preset theoretical force calculation model to obtain a theoretical force value of the vehicle on the guardrail; The theoretical force calculation model is obtained by pre-training historical vehicle data and corresponding historical theoretical force values ​​using a neural network as a framework; Dividing the force values ​​of the guardrail according to different deformation degrees based on the deformation condition to obtain regional force distributions of different force-bearing areas; Dividing the theoretical force value according to the regional force distribution to obtain a regional theoretical force value and a regional actual force value corresponding to each force region; For each stress area, respectively calculate the difference ratio between the theoretical stress value of the area and the actual stress value of the area to obtain a difference ratio data set; the difference ratio is the difference between the theoretical stress value of the area and the actual stress value of the area divided by the actual stress value of the area; Based on the vehicle collision angle and collision location information in the vehicle driving data, combined with the deformation degree and force characteristics of each stress-bearing area in the deformation situation, a reliability index of each area is calculated to construct a reliability weight coefficient table; the reliability index is determined by the consistency between the collision direction and the deformation direction; Based on the difference ratio data set and the reliability weight coefficient table, a regional weighted integration method is used to calculate the final force compensation value; the regional weighted integration method includes multiplying the difference ratio of each force-bearing area by the corresponding reliability weight coefficient, and performing spatial integration on all weighted regional data; Accident analysis is performed based on the final force compensation value combined with the preset digital twin model.

2. The method according to claim 1, characterized in that The reliability index of each region is calculated based on the vehicle collision angle and collision location information in the vehicle driving data, combined with the deformation degree and stress characteristics of each stress region in the deformation situation, and a reliability weight coefficient table is constructed, specifically including: According to the collision angle and collision location information in the vehicle driving data, a preset angle mapping function and a position evaluation function are used to calculate the collision direction coefficient and the location importance coefficient respectively; Comparing the deformation direction vector in the deformation situation with the theoretical collision force direction vector, and calculating the deformation direction consistency parameter; Identifying abnormal deformation areas in the deformation situation; the abnormal deformation areas include deformation discontinuity areas, deformation direction abnormal areas, and deformation areas that exceed the elastic limit of the material; For the abnormal deformation area, according to its abnormality type, a historical damage correction algorithm, a fracture area compensation algorithm or a deformation continuity reconstruction algorithm is used to generate a corrected deformation consistency parameter; The reliability index of each stress-bearing area is calculated by comprehensively considering the collision direction coefficient, the part importance coefficient, the deformation consistency parameter, and the corrected deformation consistency parameter, and a reliability weight coefficient table is constructed through normalization processing; The reliability weight coefficient table records the weight coefficients of various areas of the guardrail in the force calculation.

3. The method according to claim 2, characterized in that The reliability index of each stress-bearing area is calculated by comprehensively considering the collision direction coefficient, the location importance coefficient, the deformation consistency parameter, and the corrected deformation consistency parameter, and a reliability weight coefficient table is constructed through normalization processing, specifically including: Calculating the reliability index of each force-bearing area by combining the collision direction coefficient, the part importance coefficient, the deformation consistency parameter, the corrected deformation consistency parameter, and the first formula; Among them, the first formula is: ； in, is the reliability index, is the collision direction coefficient, which represents the effective component of the collision force in the normal direction, is the importance coefficient of the part, which indicates the stress characteristics of the structure. is the deformation consistency parameter, and the modified deformation consistency parameter is used for the abnormal area , is the preset parameter constant; S is the special deformation adjustment factor, and the calculation formula is , They are the degree of historical damage, degree of fracture characteristics and degree of discontinuity respectively; By normalizing the reliability index of each stress-bearing area, a reliability weight coefficient table is constructed.

4. The method according to claim 1, wherein For each stress area, the difference ratio between the theoretical stress value of the area and the actual stress value of the area is calculated to obtain a difference ratio data group, which specifically includes: Marking a stress area where the actual stress value of the area exceeds a preset maximum stress threshold as an ultra-high intensity collision area; Obtaining a force response curve of the guardrail material under different deformation degrees; the force response curve is pre-established based on material mechanical test data; According to the force response curve, a piecewise calibration function is constructed by adopting linear transformation in a preset first range region, adopting quadratic function calibration in a preset second range region, and adopting exponential function calibration in a preset third range region; the preset third range region is higher than the preset second range region, and the preset second range region is higher than the first range region, and the first range region, the preset second range region, and the preset third range region add up to the full range of the force response curve; Applying the segmented calibration function to the ultra-high intensity collision area to calculate an actual force value of the calibration area corresponding to the ultra-high intensity collision area; For each of the ultra-high intensity collision areas, the ratio of the theoretical force value of the area to the actual force value of the calibration area is calculated. For the force area that does not belong to the ultra-high intensity collision area, the ratio of the theoretical force value of the area to the actual force value of the area is calculated to obtain a difference ratio data group.

5. The method according to claim 4, characterized in that After calculating, for each ultra-high intensity collision area, the ratio of the theoretical force value of the area to the actual force value of the calibration area, and calculating, for a force area that does not belong to the ultra-high intensity collision area, the ratio of the theoretical force value of the area to the actual force value of the area, and obtaining a difference ratio data set, the method further includes: Calculating the difference ratio change rate between the actual force value of the calibration area of ​​the ultra-high intensity collision area and the actual force value of the adjacent force area in the difference ratio data group; Marking the ultra-high intensity collision region where the difference ratio change rate exceeds a preset theoretical difference ratio change rate range as a region violating the force transmission continuity principle; Searching a historical accident database to extract historical cases whose historical vehicle driving data is similar to the vehicle driving data within a preset vehicle driving data similarity range; Extracting historical actual regional force values ​​corresponding to the region where the principle of force transmission continuity is violated in the similar historical cases; Determining whether the similarity between the actual force value of the calibration area and the historical actual force value of the area that violates the principle of continuity of force transmission is within a preset similarity threshold range; If not, a preset nonlinear calibration model is applied to recalibrate the actual force value of the calibration area; the nonlinear calibration model is established in advance based on sensor experimental data and material limit force test data.

6. The method according to claim 1, wherein The method of calculating the final force compensation value based on the difference ratio data group and the reliability weight coefficient table by using a regional weighted integration method specifically includes: Matching and aligning the difference ratio data set with the reliability weight coefficient table in a one-to-one correspondence of spatial coordinates to obtain a weighted difference ratio data set; Evenly dividing the collision concentration area whose corresponding difference ratio in the difference ratio data group is greater than a preset difference ratio threshold and whose corresponding weight value in the reliability weight coefficient table is higher than the preset weight threshold into a preset number of refined grids; performing an integral calculation on the refined grids, the calculation formula being the spatial integral of the product of a refined grid difference ratio and a refined grid weight coefficient for each refined grid; the refined grid difference ratio being the ratio of the theoretical force value of the refined grid to the guardrail force value of the refined grid; the refined grid weight coefficient being obtained by combining the vehicle collision angle, collision location information, and force distribution in the collision concentrated area in the vehicle driving data; The final force compensation value is obtained through the spatial integration; the final force compensation value is a value that minimizes the sum of squares of deviations between the theoretical force value and the actual force value of each point in the grid space.

7. The method according to claim 6, characterized in that After obtaining the final force compensation value through the spatial integration, the method further includes: Performing a preset number of random perturbations on the final force compensation value, calculating the standard deviation of the compensation value change within the preset perturbation range, and establishing a confidence interval for the compensation value; Applying the final force compensation value to the theoretical force calculation model to calculate a compensated force value after compensation; Based on the compensated force value, a collision process simulation model is constructed to simulate and generate a deformation state of the guardrail, and an average Euclidean distance error and a maximum Euclidean distance error between the simulated deformation and the actual detected collision trace are calculated; Marking the area where the maximum Euclidean distance error exceeds a preset error threshold as a compensation abnormal area; A compensation reliability evaluation data packet is generated; the compensation reliability evaluation data packet records the spatial coordinate range and compensation error value of the compensation abnormal area.

8. An accident analysis system, characterized in that: The accident analysis system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the accident analysis system to execute the method described in any one of claims 1-7.

9. A computer program product comprising instructions, characterized in that When the computer program product is run on an accident analysis system, the accident analysis system is caused to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on the accident analysis system, the accident analysis system is caused to execute the method according to any one of claims 1 to 7.

Images