A method for establishing a passenger injury risk curve based on multi-source data fusion
Patent Information
- Application Number
- CN202610997445.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-06
- Publication Date
- 2026-09-22
AI Technical Summary
[0006]本发明意在提供一种多源数据融合的乘员损伤风险曲线建立方法，以解决PMHS样本稀缺且年龄结构严重偏向老年群体，无法建立符合真实交通事故损伤规律的乘员损伤风险曲线的技术问题
[0018]该改进的有益效果是：现有技术直接使用PMHS数据拟合风险曲线，受限于PMHS样本数量稀少且年龄结构严重偏向老年群体。本方案并非直接使用PMHS数据，而是利用S3中已标定的含年龄跨模型映射方程，将PMHS实验所测得的物理量转换为ATD等效物理量。这一转换使得稀缺的真实人体实验数据能够以ATD为统一基准融入大样本数据集，实现了“以小样本PMHS标定大样本统计”的技术突破，既保留了PMHS在真实人体损伤机理验证上的独特价值，又突破了PMHS数据稀缺对风险曲线建立的制约。
Smart Images
Figure CN122800271A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of vehicle safety technology, specifically to a method for establishing occupant injury risk curves based on multi-source data fusion. Background Technology
[0002] Passive safety testing of vehicles is a core means of evaluating vehicle crash safety performance and reducing traffic accident injuries and fatalities. In this testing and evaluation system, the injury risk curve of the Anthropomorphic Test Device (ATD, crash test dummy) is a key technical basis for determining the vehicle's safety level and evaluating the degree of human injury. The injury risk curve establishes a quantitative mapping relationship between specific physical quantities and the probability of human injury, providing an important data foundation for vehicle structure optimization, constraint system matching, and safety strategy formulation.
[0003] The injury risk curves used in existing assessment standards and procedures are mainly based on the results of Postmortem Human Subjects (PMHS) experiments. However, PMHS experiments, as a traditional data source, have significant technical limitations: First, the PMHS sample is heavily biased towards the elderly in terms of age structure, making it difficult to cover occupants of all ages widely distributed in real traffic environments. As a result, the injury risk curves established based on this sample cannot accurately represent age-related differences in injury thresholds. Second, PMHS experiments are constrained by multiple factors such as medical ethics review, scarcity of sample sources, and experimental technical conditions. The establishment and updating of risk curves are lengthy and costly, and the sustainability of data acquisition is poor. Third, existing injury risk curves based on PMHS lack direct data connections with real traffic accidents, making it difficult to reflect the injury distribution patterns under actual collision conditions.
[0004] Meanwhile, with the development of finite element simulation technology, Autoencoder Diagnostic Test (ATD) and Human Body Model (HBM) have been widely used in vehicle collision safety research. As a standardized physical testing tool, ATD's output can be directly used for regulatory evaluation; however, due to its simplified mechanical structure, ATD inherently differs from the real human body in biomechanical response. HBM possesses higher anatomical realism and biomechanical fidelity, providing more detailed information on human tissue response, but it cannot directly replace ATD as a standardized testing tool for regulatory evaluation. Currently, ATD and HBM are typically used independently, lacking the technical means to simultaneously simulate them under the same conditions and establish a quantitative mapping relationship. This prevents the biomechanical advantages of HBM from being effectively translated into damage assessment criteria applicable to ATD.
[0005] In addition, real traffic accident data contains a wealth of actual damage information, but existing technologies have failed to effectively integrate it with simulation data and PMHS experimental data, resulting in the lack of closed-loop verification of the damage risk curve based on the damage patterns of real accidents. Summary of the Invention
[0006] The present invention aims to provide a method for establishing occupant injury risk curves by fusing multi-source data, in order to solve the technical problem that PMHS samples are scarce and the age structure is seriously biased towards the elderly, making it impossible to establish occupant injury risk curves that conform to the injury patterns of real traffic accidents.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: a method for establishing an occupant injury risk curve based on multi-source data fusion, comprising: establishing a traffic accident dataset D1, wherein D1 includes collision information, vehicle information and occupant information, wherein the occupant information includes age and AIS injury level; Based on the boundary conditions of D1, a simulation matrix is established, and ATD simulation and HBM simulation are performed for the same working condition to obtain the ATD simulation dataset D2 and the HBM simulation dataset. Based on the D2 and the HBM simulation dataset, a closed equation with age parameters is used to establish the mapping relationship between ATD and HBM, and the optimal fitting parameters are determined by the least squares method. A PMHS experimental dataset D3 is established, and the PMHS experimental data in D3 is converted into ATD equivalent physical quantities using the optimal fitting parameters to form dataset D4. The target dataset D5 is formed by merging D2 and D4, and a damage risk curve containing age covariate is fitted based on D5; The actual damage data in D1 is used for verification against the predicted damage probability calculated based on the damage risk curve.
[0008] The principle and advantages of this scheme are as follows: Existing technologies typically use a single data source (such as PMHS experimental data or simulation data only) in isolation to establish injury risk curves. This results in curves that are either limited by the scarcity of PMHS samples and age bias, or lack verification of real-world human injury mechanisms. This scheme constructs a complete data chain that starts from real accidents, proceeds through dual simulation calibration and PMHS conversion, and finally reverts to real-world accident closed-loop verification, using real accident data (D1) as the demand-driven and verification benchmark, ATD as a unified metric bridge, HBM as a supplement for biomechanical realism, and PMHS as the source of real-world human calibration. Thus, even with scarce PMHS samples and an age structure biased towards the elderly, it is still possible to establish occupant injury risk curves that cover all age groups and conform to the injury patterns of real traffic accidents.
[0009] Preferably, as an improvement, the occupant information in D1 includes the occupant's seating position, age, gender, height, weight, and AIS damage level of each body part based on AIS05 coding; the collision information includes vehicle speed, collision angle, collision location, and maximum deformation; the vehicle information includes vehicle category, vehicle size, seat belt usage, bumper height, and the height and length of the front edge of the hood.
[0010] The beneficial effects of this improvement are as follows: By standardizing and formatting the collision information, vehicle information, and occupant information (including age and AIS05 coded damage level) in the accident data for storage, a standardized interface for multi-source heterogeneous data is established. This allows subsequent simulation matrices to directly call upon the boundary conditions of real accidents, ensuring a physical correspondence between the simulation conditions and real traffic accidents. This guarantees the effectiveness of the correlation between simulation output and accident damage labels, solving the problem of the disconnect between simulation data and real accident data in existing technologies.
[0011] Preferably, as an improvement, the step of performing ATD simulation and HBM simulation separately for the same operating condition includes: The boundary conditions of D1 are input into the simulation system. The boundary conditions include vehicle speed, collision angle, collision location, vehicle type, occupant seating position, seat belt usage, age, gender, height, and weight. Based on the boundary conditions, ATD simulation model and HBM simulation model were established and calculations were performed. The kinematic parameters of the ATD simulation model are extracted to form D2, and the kinematic parameters include the acceleration, collision force, compression amount and compression rate of each body part; The biomechanical and kinematic parameters of the HBM simulation model are extracted to form the HBM simulation dataset, wherein the biomechanical parameters include stress and strain.
[0012] The beneficial effect of this improvement is that existing technologies typically run ATD or HBM simulations independently, failing to obtain paired response data for both models under the same collision conditions. This approach, by simultaneously inputting the boundary conditions of D1 into both the ATD and HBM simulation models and performing calculations, obtains the ATD kinematic parameters and HBM biomechanical / kinematic parameters under the same working condition, providing a data foundation for subsequent establishment of cross-model quantitative mapping. The high-fidelity biomechanical response (stress, strain) of HBM can be transformed into standardized physical quantities applicable to ATD through subsequent mapping relationships, compensating for the biomechanical response deviations caused by the simplification of the mechanical structure in ATD.
[0013] Preferably, as an improvement, the closed equation containing the age parameter is:
[0014] in, Physical quantities measured by the ATD dummy For the age parameter of the HBM model, The physical quantities measured by the HBM model. , , These are the parameters to be fitted; The determination of the optimal fitting parameters using the least squares method includes minimizing the sum of squared residuals:
[0015] When the value of E is minimized, the optimal solution is determined. , , .
[0016] The beneficial effect of this improvement is that it introduces an age parameter as an explicit correction factor into the mapping relationship between ATD and HBM. Existing cross-model mapping techniques typically only consider the linear or nonlinear transformation of the physical quantities themselves, neglecting the differences in tissue material properties, geometry, and damage thresholds among individuals of different ages. This scheme uses a power-law form to incorporate age as an independent variable into the mapping equation, ensuring that the HBM-to-ATD conversion not only reflects the equivalence of physical quantities but also reflects age-related changes in biomechanical properties. The optimal parameters are determined by minimizing the sum of squared residuals using least squares, ensuring a statistically optimal fit to the mapping relationship.
[0017] Preferably, as an improvement, the conversion of the PMHS experimental data in D3 into ATD equivalent physical quantities using the optimal fitting parameters includes: Based on mapping relationship The physical quantities measured by the PMHS experiment Convert to ATD equivalent physical quantity ; Where Age represents the age of the PMHS sample. , , These are the optimal fitting parameters; The equivalent physical quantity of ATD is integrated with the corresponding PMHS age and AIS damage information to form D4.
[0018] The beneficial effects of this improvement are as follows: Existing technologies directly use PMHS data to fit risk curves, which is limited by the scarcity of PMHS samples and their significant age bias towards the elderly. This approach does not directly use PMHS data, but instead utilizes the age-inclusive cross-model mapping equation already calibrated in S3 to convert the physical quantities measured in PMHS experiments into equivalent physical quantities for ATD. This conversion allows scarce real-world human experimental data to be integrated into large-sample datasets using ATD as a unified benchmark, achieving a technological breakthrough of "calibrating large-sample statistics with small-sample PMHS data." This retains the unique value of PMHS in verifying real-world injury mechanisms while overcoming the constraint of PMHS data scarcity on risk curve establishment.
[0019] Preferably, as an improvement, the ATD corresponds to the physical quantity, the AIS damage level of the occupant / PMHS corresponding part, and the age; the ATD corresponds to the physical quantity, including the ATD simulation physical quantity in D2 and the ATD equivalent physical quantity in D4.
[0020] The beneficial effects of this improvement are as follows: By fusing the ATD simulation dataset D2 (large sample, e.g., 320 cases) with the PMHS transformed dataset D4 (small sample, e.g., 15 cases) under a unified ATD physical quantity benchmark, D5 simultaneously possesses statistical significance (from the large sample simulation) and representativeness of real human injury mechanisms (from the PMHS transformed data). Furthermore, D5 retains age and AIS injury grade labels, providing a data foundation with sufficient sample size, real human experimental information, and age covariate for subsequent survival model fitting, thus solving the problem of insufficient or unrealistic samples from a single data source.
[0021] Preferably, as an improvement, the injury risk curve fitted based on the D5 with an age covariate includes: The D5 values were fitted using three survival models: Weibull, Log-normal, and Log-logistic, with age as a covariate. The AIC criterion is used to judge the model fitting quality, and the model with the lowest AIC value is selected as the damage risk curve. in, is the maximum value of the model likelihood function, and k is the number of model parameters.
[0022] The beneficial effects of this improvement are as follows: Existing technologies typically use a single pre-set model (such as always using the Weibull model) to establish risk curves, which may lead to curve deviations due to mismatches between model assumptions and data distribution. This solution uses a competitive fitting process among Weibull, Log-normal, and Log-logistic models, objectively selecting the model with the best fitting quality based on the AIC criterion, thus avoiding systematic errors caused by subjective model selection. Simultaneously, by incorporating age as a covariate into the survival model, the injury risk curve can quantify the continuous impact of age on injury probability, solving the problem that traditional curves cannot characterize age-related differences in injury.
[0023] Preferably, as an improvement, the verification includes: Calculate the actual damage probability of a certain part in D1. ,in This represents the total number of passengers in D1. The number of people who suffered specific AIS injuries to a particular area; Substitute the corresponding data from D2 into the damage risk curve to calculate the predicted damage probability. ,in The probability of injury for the i-th occupant calculated using the injury risk curve; Calculate the probability difference ,when If the percentage is ≤5%, the validity of the damage risk curve is deemed to have been verified.
[0024] The beneficial effect of this improvement is that it calculates the probability difference between the actual damage probability in D1 and the predicted probability from the risk curve, and uses... Using a threshold of ≤5% to determine validity forms a closed loop that starts with real accident data, proceeds through multi-source fusion modeling, and finally verifies against real accidents. This verification step ensures that the established damage risk curve is statistically consistent with the damage patterns in real traffic accidents, solving the problem in existing technologies where risk curves lack verification with real accident data and cannot confirm their predictive effectiveness.
[0025] Preferably, as an improvement, the D3 includes experimental information and personnel information. The experimental information includes impact velocity, impact site, and relevant kinematic parameters of the personnel. The personnel information includes age, gender, height, weight, and AIS level of each body part.
[0026] The beneficial effects of this improvement are as follows: By storing PMHS experimental information (impact velocity, impact location, kinematic parameters) and personnel information (age, gender, height, weight, AIS level) using the same keyword format based on accident data keywords, the consistency of PMHS data and accident data in terms of data format and field definitions is ensured. This allows PMHS data to be seamlessly integrated into subsequent mapping transformation and data fusion processes without additional data cleaning and field alignment operations, improving the efficiency and accuracy of multi-source data fusion.
[0027] Preferably, as an improvement, when the model with the lowest AIC value is the Weibull model, the damage risk curve is:
[0028] in, The physical quantity corresponding to the target human body part in D5 is ATD, where Age is age. The intercept is... For age coefficient, For shape parameters.
[0029] The beneficial effect of this improvement is that it provides an explicit mathematical expression for the damage probability that is continuously modulated by age through a scale parameter, i.e., age through... The scaling parameter acts on the Weibull distribution. This specific form makes the impact of age on injury risk no longer a simple grouped statistical analysis, but a continuous and differentiable quantitative relationship. It can accurately reflect the gradual change in the probability of injury to occupants at different age stages from youth to old age, providing a calculable analytical tool for age-adaptive evaluation in vehicle safety design. Attached Figure Description
[0030] Figure 1 This is an overall flowchart of an embodiment of the present invention. Detailed Implementation
[0031] The following detailed description illustrates the specific implementation method: Example The basics are as follows: Figure 1 As shown, a method for establishing occupant injury risk curves through multi-source data fusion includes: S1. Create a traffic accident dataset D1, and format and store the accident data in the form of keywords for subsequent calls by various modules.
[0032] Each accident report includes at least collision information, vehicle information, and occupant information. Collision information includes vehicle speed, collision angle, collision location, and maximum deformation; vehicle information includes vehicle type, vehicle dimensions, seatbelt usage, bumper height, and the height and length of the front edge of the hood; occupant information includes occupant's seating position, age, gender, height, weight, and the AIS (Abbreviated Injury Scale) injury level for each body part, with the injury level uniformly coded using AIS05.
[0033] For example, when establishing the THOR dummy chest AIS3+ injury risk curve with chest compression as the indicator, the traffic accident dataset D1 can contain 200 cases of frontal collision accidents involving 320 occupants. The information of each occupant is recorded according to the above fields, providing realistic boundary conditions and damage labels for the subsequent establishment of the simulation matrix.
[0034] S2. Based on the boundary conditions of the accident data in S1, establish a simulation matrix, and conduct two synchronous simulations of ATD and HBM for the same working condition, and output the target simulation results to form the corresponding dataset.
[0035] Specifically, the boundary conditions from the accident data processing module are input into the simulation system to establish the simulation matrices for ATD and HBM and to perform simulation calculations. Boundary conditions include vehicle speed, collision angle, collision location, vehicle type, occupant seating position, seatbelt usage, age, gender, height, and weight.
[0036] Perform ATD simulations and extract the simulation results to form the ATD simulation dataset D2. D2 includes kinematic parameters obtained through simulation calculations, such as acceleration, impact force, compression amount, and compression rate of various body parts of the dummy, and is correlated with the age and AIS injury information of the corresponding occupants in the D1 dataset.
[0037] For example, for the 320 occupants in the aforementioned 200 frontal collision accidents, the chest compression amount of the THOR dummy can be extracted and combined with the age of the corresponding occupant and the chest AIS damage information (whether it reaches AIS3+) to form dataset D2.
[0038] Simultaneously, HBM simulations are performed, and the HBM calculation results are extracted to form an HBM simulation dataset. This dataset includes biomechanical parameters and kinematic parameters; biomechanical parameters, such as stress and strain; and kinematic parameters, such as acceleration, impact force, compression amount, and compression rate of various body parts of the human model.
[0039] For example, the chest compression amount of the HBM model can be extracted and used to establish a mapping relationship with the chest compression amount of the THOR dummy.
[0040] S3. Based on the ATD and HBM simulation datasets obtained in S2, a mapping relationship is established between the simulation results of the two datasets using a closed equation containing age parameters, and the optimal fitting parameters are determined by the least squares method.
[0041] The core of this mapping relationship lies in estimating the corresponding physical quantities of the ATD dummy for each accident case through simulation, and obtaining the optimal key parameters in the closed equation through the simulation results of ATD and HBM.
[0042] The closed equation containing the age parameter is:
[0043] in, This is the equivalent value (target parameter) of the physical quantity corresponding to ATD. This refers to the physical quantity measured in the PMHS experiment.
[0044] The specific mapping equation can be expressed as:
[0045] in, These are the physical quantities measured by the ATD dummy in the simulation; The age parameter of the HBM model in the simulation; These are the physical quantities measured by the HBM model in the simulation; , , These are the parameters to be fitted.
[0046] Minimize the sum of squared residuals using the least squares method:
[0047] In the formula, Let be the physical quantity measured by the ATD dummy in the i-th simulation; Let be the age of the HBM model in the i-th simulation; Let be the physical quantity measured by the HBM model in the i-th simulation; E is the sum of squared residuals. When E is minimized, the corresponding parameter is determined as the optimal solution, denoted as . , , .
[0048] For example, when establishing the mapping relationship between the chest compression of the THOR dummy, the calculated HBM chest compression is extracted, and the relationship between the HBM chest compression and the THOR dummy chest compression is fitted using the aforementioned power-law formula. ,in This represents the chest compression measured on the THOR dummy during the simulation. Using 320 sets of simulation data, the value was determined using the least squares error formula described above. , , The optimal parameters are those that minimize the sum of squared residuals E, which are then determined as the optimal solution. , , .
[0049] Traditional methods often use ATD or HBM alone for simulation, but ATD is based on a simplified mechanical structure, and its biomechanical response has inherent differences from that of the real human body; while HBM has higher anatomical realism, it cannot be directly used for standardized vehicle safety testing.
[0050] This step involves simultaneous simulation of ATD and HBM under the same working conditions. The biomechanical realism of HBM is used to compensate for the physical measurement limitations of ATD. At the same time, by introducing an age parameter as a correction factor, the influence of changes in human tissue material properties, geometric structure and damage threshold caused by age differences on the response results is compensated.
[0051] This age-inclusive cross-model mapping equation not only establishes a precise bridge between ATD and real human injuries, but also enables subsequent risk curves to accurately reflect the injury patterns of different age groups, solving the problem that traditional curves have a single age structure and cannot characterize differences in age-related injuries.
[0052] S4. Establish PMHS experimental dataset D3. This dataset is based on accident data keywords and uses the same keyword format to format and store PMHS experimental data. The information includes at least experimental information and personnel information.
[0053] Experimental information includes impact velocity, impact location, and relevant kinematic parameters of the personnel (such as acceleration of various body parts, impact force, compression amount, and compression rate); personnel information includes age, gender, height, weight, and AIS level of various body parts.
[0054] Using the optimal fitting parameters already determined in S3 , , A mapping relationship between PMHS and ATD was established, and PMHS experimental data were converted into equivalent physical quantities of ATD to form dataset D4.
[0055] The mapping relationship is as follows:
[0056] In the formula, The calculated equivalent physical quantity (target parameter) corresponding to the ATD. is the physical quantity measured in the PMHS experiment; Age is the age of the PMHS sample; , , The optimal fitting parameters were determined for S3. The calculated equivalent values were then integrated with the age of the corresponding PMHS and the AIS lesion information of each site in the D3 dataset to form dataset D4.
[0057] For example, the PMHS experimental dataset D3 may contain data from 15 chest impact experiments. Utilizing the optimal solution... , , Establish a mapping relationship between the THOR dummy and PMHS:
[0058] In the formula, The equivalent value of the chest compression of the THOR dummy (target parameter) is calculated. The value represents the chest compression measured in the PMHS experiment; Age represents the PMHS age. After this mapping transformation, the data from 15 PMHS experiments, together with the corresponding PMHS age and chest AIS3+ injury information, form dataset D4.
[0059] The establishment of traditional injury risk curves relies heavily on PMHS experiments, but PMHS samples are not only scarce, but also heavily biased towards the elderly population, and are subject to strict limitations in medical ethics and sample collection techniques.
[0060] This approach does not directly use PMHS data to fit curves. Instead, it utilizes a calibrated age-biased cross-model mapping equation to convert PMHS experimental data into equivalent physical quantities of ATD. This conversion allows small-sample, high-age-biased PMHS data to be integrated into large-sample datasets using ATD as a unified benchmark, significantly expanding the data foundation for establishing risk curves. Simultaneously, the unique value of PMHS data in validating real-world injury mechanisms is preserved, achieving a technological breakthrough of "calibrating large-sample statistics with small samples," effectively alleviating the constraint of PMHS data scarcity on risk curve updates.
[0061] S5. Call the ATD simulation dataset D2 obtained in S2 and the PMHS conversion dataset D4 obtained in S4, and merge them to form a new target dataset D5.
[0062] Dataset D5 includes the physical quantities corresponding to ATD, the AIS level of the corresponding occupant / PMHS location in the accident, and the age of the occupant / PMHS in the accident. The physical quantities corresponding to ATD include those in D2. With D4 .
[0063] For example, dataset D5 can be constructed by merging the data from Table 1 and Table 2 as follows: Table 1. Dataset D2 ATD Simulation Data Table
[0064] Table 2 Datasheet for D4 PMHS Conversion
[0065] Table 1 shows an example of the ATD simulation dataset D2, which contains 320 sets of THOR dummy chest compression simulation results. The column represents the measured value of the chest compression of the THOR dummy in each simulation. The column "Damage Reached AIS3+" indicates whether the chest injury of the corresponding occupant in the D1 dataset reached the AIS3+ level (0 indicates not reached, 1 indicates reached). The "Age" column represents the age of the corresponding occupant.
[0066] Table 2 shows examples of the PMHS transformed dataset D4, containing 15 cases of equivalent chest compression of THOR dummies after S4 mapping transformation. The column represents the THOR equivalent value obtained by converting the measured chest compression amount of PMHS using the optimal fitting parameters, and the meanings of the remaining fields are the same as in Table 1.
[0067] By merging Table 1 and Table 2, dataset D5 is formed, which combines large sample data (320 cases) from real accident simulation with small sample data (15 cases) from transformed PMHS under a unified ATD physical quantity benchmark, providing a target dataset that simultaneously covers simulation and real human experiment information and includes age covariates for subsequent survival model fitting.
[0068] Based on this, the D5 dataset was fitted using three survival models: Weibull, Log-normal, and Log-logistic, with age included as a covariate in the model.
[0069] The model fit quality was judged using the AIC (Akaike Information Criterion) criterion. The formula for calculating the AIC value is as follows:
[0070] In the formula, is the maximum value of the model's likelihood function; k is the number of parameters in the model. By comparing the AIC values of the three models, it is recommended to choose the function with the lowest AIC value as the final damage risk curve model.
[0071] For example, if the Weibull model has the lowest AIC value after fitting the fused dataset D5, then the objective function is determined as follows:
[0072] In the formula, The compression of the THOR dummy chest in dataset D5 (i.e., in D2) With D4 Age is the age; The intercept; This is the age coefficient (reflecting the influence of age on the scale parameter). For shape parameters.
[0073] This multi-model selection strategy ensures that the injury risk curve is best adapted to the data distribution characteristics, while the introduction of the age covariate enables the curve to quantify the impact of different ages on the probability of injury.
[0074] S6. To verify the effectiveness of the established ATD injury risk curve in predicting occupant injury, and to form a closed-loop verification that starts from real data, integrates simulation and modeling, and finally regresses to real data, the following verification steps are performed: First, calculate the actual damage probability of a certain part in the D1 dataset:
[0075] In the formula, This represents the total number of occupants in the D1 dataset; This refers to the number of people in the D1 dataset who suffered a certain type of AIS injury to a specific body part. For example, calculating the actual probability of chest AIS3+ injury among 320 occupants in the D1 dataset, i.e. ,in The number of occupants in the D1 dataset who suffered AIS3+ chest injuries.
[0076] Secondly, by substituting the corresponding data from the D2 dataset into the established ATD injury risk curve (objective function), the predicted injury probability of a certain location is calculated:
[0077] In the formula, Let be the probability of injury calculated by the ATD injury risk curve (objective function) for the i-th occupant in the simulation corresponding to the D1 dataset. For example, in the chest AIS3+ validation, denoted as . ,in This represents the probability of chest AIS3+ injury for the i-th occupant in the simulation corresponding to the D1 dataset, calculated using the injury risk curve (objective function).
[0078] Finally, calculate the probability difference:
[0079] like If the risk level is ≤5%, the effectiveness of the ATD (Advanced Treatment Detection) injury risk curve for the target human body part is considered validated. For example, if the probability of injury to the chest AIS3+ is poor... If the THOR dummy chest AIS3+ injury risk curve is valid, it is considered that the curve is consistent with the injury patterns in real traffic accidents in a statistical sense.
[0080] This solution achieves a complete data chain from real traffic accident data to simulation calibration and then to post-mortem experimental verification by merging four sources of data: accident data, ATD data, HBM data, and PMHS data at each level. Through synchronous simulation of ATD and HBM data and cross-model mapping including age, it solves the problems of deviation between traditional ATD and real human response and the lack of age factors. Through the PMHS data conversion mechanism, it overcomes the constraint of small sample PMHS data on the establishment of risk curves.
[0081] Therefore, even in the absence of large-scale PMHS experimental data, this method can still utilize ATD and HBM finite element models to establish human injury risk curves for arbitrary locations and risk levels, and combine them with real traffic accident data to make the curves more consistent with actual injury patterns. This method can be widely applied to the vehicle safety evaluation stage in vehicle passive safety testing, providing a more accurate and comprehensive tool for quantifying injury risk for vehicle safety performance assessment.
[0082] The above descriptions are merely embodiments of the present invention, and common knowledge such as specific technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solutions of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.
Claims
1. A method for establishing occupant injury risk curves through multi-source data fusion, characterized in that, include: A traffic accident dataset D1 is established, which includes collision information, vehicle information, and occupant information, including occupant age and AIS injury level. Based on the boundary conditions of D1, a simulation matrix is established, and ATD simulation and HBM simulation are performed for the same working condition to obtain the ATD simulation dataset D2 and the HBM simulation dataset. Based on the D2 and the HBM simulation dataset, a closed equation with age parameters is used to establish the mapping relationship between ATD and HBM, and the optimal fitting parameters are determined by the least squares method. A PMHS experimental dataset D3 is established, and the PMHS experimental data in D3 is converted into ATD equivalent physical quantities using the optimal fitting parameters to form dataset D4. The target dataset D5 is formed by merging D2 and D4, and a damage risk curve containing age covariate is fitted based on D5; The actual damage data in D1 is used for verification against the predicted damage probability calculated based on the damage risk curve.
2. The method for establishing an occupant injury risk curve based on multi-source data fusion according to claim 1, characterized in that: The occupant information in D1 includes the occupant's seating position, age, gender, height, weight, and AIS damage level of each body part based on AIS05 coding; the collision information includes vehicle speed, collision angle, collision location, and maximum deformation; the vehicle information includes vehicle category, vehicle size, seat belt usage, bumper height, and the height and length of the front edge of the hood.
3. The method for establishing an occupant injury risk curve based on multi-source data fusion according to claim 2, characterized in that: The process of performing ATD simulation and HBM simulation for the same operating condition includes: The boundary conditions of D1 are input into the simulation system. The boundary conditions include vehicle speed, collision angle, collision location, vehicle type, occupant seating position, seat belt usage, age, gender, height, and weight. Based on the boundary conditions, ATD simulation model and HBM simulation model were established and calculations were performed. The kinematic parameters of the ATD simulation model are extracted to form D2, and the kinematic parameters include the acceleration, collision force, compression amount and compression rate of each body part; The biomechanical and kinematic parameters of the HBM simulation model are extracted to form the HBM simulation dataset, wherein the biomechanical parameters include stress and strain.
4. The method for establishing an occupant injury risk curve based on multi-source data fusion according to claim 3, characterized in that, The closed equation containing the age parameter is: in, The physical quantities measured by the ATD dummy For the age parameter of the HBM model, The physical quantities measured by the HBM model. , , These are the parameters to be fitted; The determination of the optimal fitting parameters using the least squares method includes minimizing the sum of squared residuals: When the value of E is minimized, the optimal solution is determined. , , .
5. The method for establishing an occupant injury risk curve based on multi-source data fusion according to claim 4, characterized in that, The process of converting the PMHS experimental data in D3 into ATD equivalent physical quantities using the optimal fitting parameters includes: Based on mapping relationship The physical quantities measured by the PMHS experiment Convert to ATD equivalent physical quantity ; Where Age represents the age of the PMHS sample. , , These are the optimal fitting parameters; The equivalent physical quantity of ATD is integrated with the corresponding PMHS age and AIS damage information to form D4.
6. The method for establishing an occupant injury risk curve based on multi-source data fusion according to claim 5, characterized in that: The physical quantities corresponding to ATD, the AIS damage level and age of the corresponding part of the occupant / PMHS; the physical quantities corresponding to ATD include the simulated physical quantities of ATD in D2 and the equivalent physical quantities of ATD in D4.
7. The method for establishing an occupant injury risk curve based on multi-source data fusion according to claim 6, characterized in that, The injury risk curve fitted based on the D5 model, including the age covariate, includes: The D5 values were fitted using three survival models: Weibull, Log-normal, and Log-logistic, with age as a covariate. The AIC criterion is used to judge the model fitting quality, and the model with the lowest AIC value is selected as the damage risk curve. in, is the maximum value of the model likelihood function, and k is the number of model parameters.
8. The method for establishing an occupant injury risk curve based on multi-source data fusion according to claim 7, characterized in that, The verification includes: Calculate the actual damage probability of a certain part in D1. ,in This represents the total number of passengers in D1. The number of people who suffered specific AIS injuries to a particular area; Substitute the corresponding data from D2 into the damage risk curve to calculate the predicted damage probability. ,in The probability of injury for the i-th occupant calculated using the injury risk curve; Calculate the probability difference ,when If the percentage is ≤5%, the validity of the damage risk curve is deemed to have been verified.
9. The method for establishing an occupant injury risk curve based on multi-source data fusion according to claim 8, characterized in that: The D3 includes experimental information and personnel information. The experimental information includes impact velocity, impact location, and relevant kinematic parameters of the personnel. The personnel information includes age, gender, height, weight, and AIS level of each body part.
10. The method for establishing an occupant injury risk curve based on multi-source data fusion according to claim 9, characterized in that: When the model with the lowest AIC value is the Weibull model, the damage risk curve is: in, The physical quantity corresponding to the target human body part in D5 is ATD, where Age is age. The intercept is... For age coefficient, For shape parameters.