Method and device for predicting injury of front-row passengers in frontal collision

Abstract

The invention relates to the technical field of automobile passive safety, in particular to a front-row passenger injury prediction method and device in a frontal collision, and the method comprises the steps: obtaining a collision waveform of a vehicle; determining constraint system parameters of the vehicle; inputting the collision waveform and the constraint system parameter into a pre-constructed front-row member prediction network model of the frontal collision, taking the constraint system parameter as a conditional variable to be fused with the collision waveform, and capturing a nonlinear relationship between the dynamic impact and the constraint system parameter, and the damage response and the damage index of the multiple parts of the passenger along with the time change can be predicted. Therefore, the problem that the application of deep learning in engineering design and safety system optimization is limited due to the fact that the influence of mechanical and physical characteristics of a constraint system is ignored in the related technology is solved.

Description

Technical Field

[0001] This application relates to the field of automotive passive safety technology, and in particular to a method and device for predicting frontal occupant injuries in a frontal collision. Background Technology

[0002] In terms of automotive collision safety, the design and optimization of occupant restraint systems play a crucial role in mitigating injuries to occupants. Among these, accurately predicting occupant injury responses in collision accidents is key to achieving this goal.

[0003] In related technologies, occupant injury prediction typically uses simulation techniques, such as finite element method (FEM) simulation and multibody dynamics simulation. However, FEM simulation requires precise mechanical parameters, has a complex simulation modeling process, and struggles to make accurate predictions for complex real-world collision conditions. While multibody dynamics simulation is more computationally efficient than the FEM method, it still requires significant computational resources for parametric design and optimization, severely limiting design iteration efficiency. Therefore, deep learning, as a key component of artificial intelligence, is gradually becoming a core technology in fields such as intelligent driving, collision prediction, and human-computer interaction due to its powerful capabilities in automatic feature extraction, nonlinear modeling, and large-scale data processing.

[0004] However, current research on deep learning in automotive collision safety focuses primarily on real-time safety decision-making or injury classification, neglecting the influence of the mechanical and physical characteristics of the constraint system, thus limiting its application in engineering design and safety system optimization. Summary of the Invention

[0005] This application provides a method and apparatus for predicting front-seat occupant injuries in a frontal collision, in order to solve the problem that the application of deep learning in engineering design and safety system optimization is limited due to the neglect of the influence of the mechanical and physical characteristics of the constraint system in related technologies.

[0006] The first aspect of this application provides a method for predicting front-seat occupant injuries in a frontal collision, comprising the following steps: acquiring a collision waveform of a vehicle; determining the restraint system parameters of the vehicle; inputting the collision waveform and the restraint system parameters into a pre-constructed front-seat occupant prediction network model, using the restraint system parameters as conditional variables to fuse with the collision waveform, capturing the nonlinear relationship between dynamic impact and restraint system parameters, so as to predict the injury response and injury indicators of multiple parts of the occupant over time.

[0007] Through the above-mentioned technical means, the embodiments of this application can use the constraint system parameters as condition variables to fuse with the collision waveform based on a pre-constructed front-row member prediction network model, thereby capturing the nonlinear relationship between dynamic impact and constraint system parameters, fully considering the influence of the mechanical and physical characteristics of the constraint system, and thus improving the generalization ability of deep learning in different technical scenarios.

[0008] Optionally, in one embodiment of this application, before inputting the collision waveform and the constraint system parameters into the pre-built frontal collision front occupant prediction network model, the method further includes: building a finite element model of a frontal collision scenario based on preset evaluation criteria; inputting the constraint system parameters into the finite element model to calculate occupant injury response and occupant injury index; generating a frontal collision occupant collision response system dataset based on the constraint system parameters, the vehicle collision waveform, the occupant injury response, and the occupant injury index; and constructing the frontal collision front occupant prediction network model using the frontal collision occupant collision response system dataset and a temporal convolutional network.

[0009] Through the above-mentioned technical means, the embodiments of this application can build a frontal collision finite element model by setting preset evaluation criteria, input constraint system parameters to calculate occupant injury-related data and generate corresponding datasets, and combine temporal convolutional networks to construct a front-row occupant prediction network model, so that the model can integrate constraint system parameter data, thereby facilitating the accurate prediction of occupant collision response in frontal collision scenarios.

[0010] Optionally, in one embodiment of this application, after generating the frontal collision occupant collision response system dataset, the method further includes: preprocessing the training data of the frontal collision occupant collision response system dataset to remove data that meets preset abnormal conditions, thereby obtaining preprocessed data; and normalizing the preprocessed data to obtain planned data that meets preset deep learning conditions.

[0011] Through the above-mentioned technical means, the embodiments of this application can perform abnormal data removal and normalization processing on the training data of the frontal collision occupant collision response system dataset to obtain standardized data that meets the requirements of deep learning. This effectively improves data quality, reduces abnormal interference, and provides accurate and suitable high-quality data support for the subsequent training of the prediction network model, ensuring the model training effect and prediction accuracy.

[0012] Optionally, in one embodiment of this application, the occupant injury response includes a head injury response and a chest injury response.

[0013] Through the above-mentioned technical means, the embodiments of this application can clearly define the occupant injury response as the head and chest injury response, thereby focusing on the key parts most vulnerable to serious injury in a collision, and improving the pertinence of safety protection and the value of data application.

[0014] Optionally, in one embodiment of this application, the formula for calculating the head injury response is: , Where t1 and t2 represent the two moments of the collision, a represents the triaxial resultant acceleration of the head's center of gravity, and g is the gravitational acceleration.

[0015] Through the above-mentioned technical means, the embodiments of this application can evaluate the head injury response based on the collision time parameters and the triaxial composite acceleration of the head center, thereby enabling rapid prediction of front-seat occupant injuries in a frontal collision of an occupant-occupied vehicle based on the head injury response.

[0016] Optionally, in one embodiment of this application, before inputting the constraint system parameters into the finite element model, the method further includes: obtaining the airbag deployment time, airbag vent diameter, seat belt force limit, and steering column crush curve scaling factor; and generating the constraint system parameters based on the airbag deployment time, airbag vent diameter, seat belt force limit, and steering column crush curve scaling factor.

[0017] Through the above-mentioned technical means, the embodiments of this application can generate restraint system parameters by integrating the airbag deployment time, vent diameter, seat belt force limiter, and steering column crush curve scaling factor. This achieves the systematic integration and precise definition of the core key parameters of the restraint system, thereby providing standardized and comprehensive parameter input support for the calculation of occupant injury response and the construction of prediction models under subsequent collision conditions, ensuring the accuracy and reliability of related analysis and modeling.

[0018] A second aspect of this application provides a frontal collision front occupant injury prediction device, comprising: an acquisition module for acquiring a vehicle collision waveform; a determination module for determining the vehicle's restraint system parameters; and a prediction module for inputting the collision waveform and restraint system parameters into a pre-constructed frontal collision front occupant prediction network module, using the restraint system parameters as condition variables to fuse with the collision waveform, capturing the nonlinear relationship between dynamic impact and restraint system parameters, and predicting the injury response and injury indicators of multiple occupant parts over time.

[0019] Through the above-mentioned technical means, the embodiments of this application can use the constraint system parameters as condition variables to fuse with the collision waveform based on a pre-constructed front-row member prediction network model, thereby capturing the nonlinear relationship between dynamic impact and constraint system parameters, fully considering the influence of the mechanical and physical characteristics of the constraint system, and thus improving the generalization ability of deep learning in different technical scenarios.

[0020] Optionally, in one embodiment of this application, the frontal collision front occupant injury prediction device further includes: a construction module, used to construct a finite element model of a vehicle frontal collision condition based on preset evaluation criteria before inputting the collision waveform and the constraint system parameters into the pre-constructed frontal collision front occupant prediction network model; a calculation module, used to input the constraint system parameters into the finite element model to calculate the occupant injury response and occupant injury index; a generation module, used to generate a frontal collision occupant collision response system dataset based on the constraint system parameters, the vehicle collision waveform, the occupant injury response, and the occupant injury index; and a modeling module, used to construct the frontal collision front occupant prediction network model using the frontal collision occupant collision response system dataset and a temporal convolutional network.

[0021] Through the above-mentioned technical means, the embodiments of this application can build a frontal collision finite element model by setting preset evaluation criteria, input constraint system parameters to calculate occupant injury-related data and generate corresponding datasets, and combine temporal convolutional networks to construct a front-row occupant prediction network model, so that the model can integrate constraint system parameter data, thereby facilitating the accurate prediction of occupant collision response in frontal collision scenarios.

[0022] Optionally, in one embodiment of this application, the frontal collision front occupant injury prediction device further includes: a first processing module, configured to preprocess the training data of the frontal collision occupant collision response system dataset after generating the dataset, remove data that meets preset abnormal conditions, and obtain preprocessed data; and a second processing module, configured to normalize the preprocessed data to obtain planned data that meets preset deep learning conditions.

[0023] Through the above-mentioned technical means, the embodiments of this application can perform abnormal data removal and normalization processing on the training data of the frontal collision occupant collision response system dataset to obtain standardized data that meets the requirements of deep learning. This effectively improves data quality, reduces abnormal interference, and provides accurate and suitable high-quality data support for the subsequent training of the prediction network model, ensuring the model training effect and prediction accuracy.

[0024] Optionally, in one embodiment of this application, the occupant injury response includes a head injury response and a chest injury response.

[0025] Through the above-mentioned technical means, the embodiments of this application can clearly define the occupant injury response as the head and chest injury response, thereby focusing on the key parts most vulnerable to serious injury in a collision, and improving the pertinence of safety protection and the value of data application.

[0026] Optionally, in one embodiment of this application, the formula for calculating the head injury response is: , Where t1 and t2 represent the two moments of the collision, a represents the triaxial resultant acceleration of the head's center of gravity, and g is the gravitational acceleration.

[0027] Through the above-mentioned technical means, the embodiments of this application can evaluate the head injury response based on the collision time parameters and the triaxial composite acceleration of the head center, thereby enabling rapid prediction of front-seat occupant injuries in a frontal collision of an occupant-occupied vehicle based on the head injury response.

[0028] Optionally, in one embodiment of this application, the frontal collision occupant injury prediction device is further configured to: obtain the airbag deployment time, airbag vent diameter, seat belt force limit, and steering column crush curve scaling factor before inputting the constraint system parameters into the finite element model; and generate the constraint system parameters based on the airbag deployment time, airbag vent diameter, seat belt force limit, and steering column crush curve scaling factor.

[0029] Through the above-mentioned technical means, the embodiments of this application can generate restraint system parameters by integrating the airbag deployment time, vent diameter, seat belt force limiter, and steering column crush curve scaling factor. This achieves the systematic integration and precise definition of the core key parameters of the restraint system, thereby providing standardized and comprehensive parameter input support for the calculation of occupant injury response and the construction of prediction models under subsequent collision conditions, ensuring the accuracy and reliability of related analysis and modeling.

[0030] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the frontal collision occupant injury prediction method as described in the above embodiments.

[0031] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting frontal occupant injuries in a frontal collision.

[0032] A fifth aspect of this application provides a computer program product that stores a computer program that, when executed by a processor, implements the above-described method for predicting frontal occupant injuries in a frontal collision.

[0033] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0034] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a frontal collision occupant injury prediction method provided according to an embodiment of this application; Figure 2 This is a schematic diagram of the architecture of an occupant injury prediction model according to a specific embodiment of this application; Figure 3 This is a schematic diagram of a multi-rigid-body simulation model according to a specific embodiment of this application; Figure 4 This is a schematic diagram of a dataset according to a specific embodiment of this application; Figure 5 This is a schematic diagram of abnormal data according to a specific embodiment of this application; Figure 6 This is a schematic diagram illustrating the training process of an occupant injury prediction model according to a specific embodiment of this application; Figure 7 This is a schematic diagram of the head acceleration curve prediction result according to a specific embodiment of this application; Figure 8 This is a schematic diagram of the chest compression curve prediction result according to a specific embodiment of this application; Figure 9 This is a schematic diagram of the frontal collision occupant injury prediction device according to an embodiment of this application; Figure 10 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0035] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0036] To address the shortcomings of existing technologies, this invention proposes an occupant injury prediction model based on Temporal Convolutional Network (TCN). This model achieves rapid and accurate prediction of injury responses at multiple locations by simultaneously inputting collision waveforms and constraint system parameters.

[0037] The following describes a method and apparatus for predicting front-seat occupant injuries in a frontal collision, based on embodiments of the present application, with reference to the accompanying drawings. Addressing the problem mentioned in the background section of the related technologies where the neglect of the mechanical and physical characteristics of the constraint system limits the application of deep learning in engineering design and safety system optimization, this application provides a method for predicting front-seat occupant injuries in a frontal collision. In this method, the vehicle collision waveform and constraint system parameters are defined as inputs to the model in time-series and numerical forms, respectively. The outputs are time-series occupant injury responses and injury indices. Leveraging the advantages of deep learning, features are extracted from the input data through training and combined with frontal collision front-seat occupant injury simulation data for prediction, thereby significantly reducing time costs. This solves the problem in the related technologies where the neglect of the mechanical and physical characteristics of the constraint system limits the application of deep learning in engineering design and safety system optimization.

[0038] Specifically, Figure 1 This is a flowchart illustrating a frontal collision occupant injury prediction method provided in an embodiment of this application.

[0039] like Figure 1 As shown, the method for predicting frontal occupant injuries in a frontal collision includes the following steps: In step S101, the collision waveform of the vehicle is acquired.

[0040] Among them, the vehicle collision waveform can be understood as the curve formed by the change of key physical parameters such as acceleration, force, and strain of the vehicle structure (such as longitudinal beams and subframes) or occupant-related components (such as dummies) over time during the collision process. It is mainly used to quantify the dynamic response and energy transfer process during the collision.

[0041] In actual implementation, vehicle collision waveforms can be obtained in the following three ways: First, real vehicle / dashboard collision tests, where sensors are deployed on key parts of the vehicle or on dummies, and data such as force and acceleration are collected by high-speed cameras to generate the waveforms, which has the highest accuracy; second, accident vehicle data extraction, which reconstructs the collision moment data by reading EDR or dashcam data; and third, computer simulation, which uses software such as LS-DYNA to build a finite element model, inputs working condition parameters, and simulates the waveforms to meet the needs of rapid verification in the early stages of R&D. No specific limitations are made here.

[0042] In step S102, the vehicle's constraint system parameters are determined.

[0043] Vehicle restraint system parameters are key design and performance parameters of the restraint system (core components such as seat belts, airbags, and seats) used to protect occupants during a car collision. They directly affect the force, displacement, and injury risk of occupants during a collision and are the core basis for the matching optimization of the restraint system and the achievement of safety performance standards (such as meeting C-NCAP requirements).

[0044] In this embodiment, the restraint system parameters can be selected as airbag ignition time, airbag vent diameter, seat belt force limiting level, steering column crush curve scaling factor, etc.

[0045] In step S103, the collision waveform and constraint system parameters are input into a pre-built frontal collision front occupant prediction network model. The constraint system parameters are used as conditional variables to be fused with the collision waveform to capture the nonlinear relationship between dynamic impact and constraint system parameters, so as to predict the injury response and injury index of multiple parts of the occupant over time.

[0046] In actual implementation, this application embodiment can choose to construct a frontal collision front occupant prediction network model based on a temporal convolutional neural network (TCN). The input of this model is the car collision waveform (temporal data) and constraint system parameters (numerical data), and the output is the occupant response (temporal data). For example, the input is the car collision waveform and constraint system parameters, the former being 1200-step temporal data and the latter consisting of four numerical values, and the output is the occupant injury response, which is also 1200-step temporal data. A temporal convolutional neural network (TCN) is built as an encoder to extract features from the collision waveform data. A fully connected layer is established to process the constraint system parameters and fuse them with the car collision waveform as conditional variables. The decoder adopts a dual-branch architecture, with each branch responsible for reconstructing the dynamic occupant response of different body parts.

[0047] Specifically, the constructed frontal collision front-seat occupant prediction network model has two input layers, used to extract features from the car collision waveform and constraint system parameters, respectively. A TCN network is built and used as one input layer, with parameters such as input dimension, number of network layers, number of network units per layer, and convolutional kernels adjusted according to the input car collision waveform. A fully connected layer is established as the second input layer, mapping the input constraint system parameters to the same dimension as the TCN output features and activating them using the ReLU function. The features extracted from the two input layers are concatenated and fused, incorporating the constraint system parameters as conditional variables into the model to capture their interaction, and then the dimension of the concatenated features is compressed. The output layer adopts a dual-branch structure, where each branch independently predicts the occupant injury response for different body parts. The output layer consists of two independent fully connected layers, each restoring the data dimension. Finally, the Sigmoid function is used for normalization, facilitating inverse normalization and comparison with real data during validation. The specific model architecture is as follows: Figure 2 As shown.

[0048] Optionally, in one embodiment of this application, before inputting the collision waveform and constraint system parameters into the pre-built frontal collision front occupant prediction network model, the method further includes: building a finite element model of the vehicle frontal collision condition based on preset evaluation criteria; inputting the constraint system parameters into the finite element model to calculate the occupant injury response and occupant injury index; generating a frontal collision occupant collision response system dataset based on the constraint system parameters, vehicle collision waveform, occupant injury response, and occupant injury index; and constructing a frontal collision front occupant prediction network model using the frontal collision occupant collision response system dataset and a temporal convolutional network.

[0049] This application embodiment can build a frontal collision finite element model by setting preset evaluation criteria, input constraint system parameters to calculate occupant injury-related data and generate a corresponding dataset, and combine a temporal convolutional network to construct a front-row occupant prediction network model, so that the model can integrate constraint system parameter data, thereby facilitating the accurate prediction of occupant collision response in frontal collision scenarios.

[0050] Regarding the construction of the finite element model: The embodiments of this application can build a finite element model of a car frontal collision condition according to the relevant C-NCAP standards.

[0051] China New Car Assessment Program (C-NCAP) 2024 adopts The impact intensity and injury risk of the occupant's head during the collision are assessed by the head injury criterion and the cumulative 3ms composite acceleration value. The chest injury of the dummy is evaluated by measuring the chest compression deformation and the chest viscosity criterion (VC). At the same time, the head acceleration curve and the chest compression curve can visualize the occupant's motion and injury situation, which can facilitate the observation of dynamic response characteristics in the collision and help identify potential injury risk factors.

[0052] For example, embodiments of this application can utilize multi-rigid-body simulation software to establish a simulation model of a frontal collision under C-NCAP standards, such as... Figure 3 As shown, the occupant restraint system in this model mainly consists of seat belts, airbags, seats, steering system, and dashboard. A Hybrid_III_50th dummy is used, and the collision speed is 50 km / h. The model has been calibrated and its occupant position and posture have been determined with reference to the real-vehicle crash test reports of benchmark models, and the response trends of the occupant's head, neck, chest, and other parts are consistent with the test results.

[0053] In one embodiment of this application, before inputting the constraint system parameters into the finite element model, the method further includes: obtaining the airbag deployment time, airbag vent diameter, seat belt force limit, and steering column crush curve scaling factor; and generating constraint system parameters based on the airbag deployment time, airbag vent diameter, seat belt force limit, and steering column crush curve scaling factor.

[0054] This application embodiment can generate constraint system parameters by integrating airbag deployment time, vent diameter, seat belt force limit, and steering column crush curve scaling factor. This achieves systematic integration and precise definition of the core key parameters of the constraint system, thereby providing standardized and comprehensive parameter input support for subsequent occupant injury response calculation and prediction model construction under collision conditions, ensuring the accuracy and reliability of related analysis and modeling.

[0055] Furthermore, regarding the generation of the dataset: The occupant collision response system involves numerous parameters, and including all of them would increase the workload and interfere with model training. In this embodiment, four sensitive parameters can be included in the dataset: airbag deployment time, airbag vent diameter, seat belt force limit, and steering column crush curve scaling factor.

[0056] The vehicle collision waveform can comprehensively reflect key parameters such as initial collision velocity, curb weight, and front-end structural characteristics, which are closely related to the damage results. Therefore, it is included in the dataset as the main load input.

[0057] Meanwhile, to ensure the accuracy of the prediction results, the established frontal collision occupant collision response system dataset should include occupant injury responses, which may include head injury responses and chest injury responses.

[0058] The data used above can be obtained through multi-rigid-body simulation. A specific range of values ​​is selected for the constraint system parameters, and these values ​​are input into multi-rigid-body simulation software to calculate the occupant injury response and injury indices (injury assessment indices). Then, the required data is extracted from multiple calculation result files and saved to create a dataset. In one embodiment of this application, after generating the frontal collision occupant collision response system dataset, the method further includes: preprocessing the training data of the frontal collision occupant collision response system dataset to remove data that meets preset abnormal conditions, thereby obtaining preprocessed data; and normalizing the preprocessed data to obtain standardized data that meets preset deep learning conditions.

[0059] In actual implementation, after generating the frontal collision occupant collision response system dataset, the data in the dataset can be further preprocessed to remove outliers and perform normalization and other operations to transform the data into the normalized data required by the deep learning model.

[0060] This application embodiment can remove and normalize abnormal data from the training data of the frontal collision occupant collision response system dataset to obtain standardized data that meets the requirements of deep learning. This effectively improves data quality, reduces abnormal interference, and provides accurate and suitable high-quality data support for the subsequent training of the prediction network model, ensuring the model training effect and prediction accuracy.

[0061] For example, in the case of vehicle collision waveforms, the collision pulses of 55 different vehicles obtained from the frontal full-width rigid barrier crash test (FRB) can be selected as the input vehicle collision waveforms in this application embodiment.

[0062] To ensure broad and uniform coverage of the multidimensional parameter space while maintaining statistical representativeness, Latin hypercube sampling can be used to generate 750 unique input combinations for the constrained system parameters.

[0063] Furthermore, by using the above two parts as inputs to the aforementioned multi-rigid-body simulation model, the corresponding occupant injury response output can be calculated. The occupant injury response and its corresponding inputs are then extracted and saved to a file, named according to the constraint system parameters, to construct a dataset, such as... Figure 4 As shown.

[0064] In this dataset, the constraint system parameters are represented numerically, while the collision waveforms and occupant response curves are formatted as time series with 1200 data steps. A small number of samples exhibiting abnormal occupant response data due to unreasonable combinations of constraint system parameters are removed. Figure 5 Data mutations or missing data are removed.

[0065] Furthermore, a program was written to extract constraint system parameters, occupant injury response data, and vehicle collision waveforms, normalize them, and then convert them into tensor form so that a deep learning network can process them. Multiple constraint system parameters are treated as a whole and simultaneously input into the deep learning network.

[0066] The dataset established in this application embodiment contains data and the range of values ​​for constraint system parameters are shown in Table 1. The specific range of values ​​for constraint system parameters can be determined by combining literature and actual engineering research.

[0067] Table 1

[0068] Occupant injury response can include head acceleration curves and chest compression curves, which can visualize occupant motion and injury status in the form of curves, making it easier to observe dynamic response characteristics during a collision and help identify potential injury risk factors.

[0069] Optionally, in one embodiment of this application, the formula for calculating the head injury response is: , In the formula, t1 and t2 represent the two moments of the collision, with a time interval of no more than 15ms; a represents the triaxial composite acceleration of the head's center of gravity (m / s2); and g is the gravitational acceleration.

[0070] The embodiments of this application can assess head injury response based on collision time parameters and triaxial composite acceleration at the head center, thereby enabling rapid prediction of front-seat occupant injuries in frontal collisions of passenger vehicles based on head injury response.

[0071] Furthermore, embodiments of this application can use the established dataset and initial model hyperparameters for training and validation.

[0072] For example, in this embodiment of the application, the established dataset can be randomly divided, with 80% of the data used as the training set and 20% as the test set. The minimum batch size is set to 1, the number of iterations is set to 2000, the learning rate is 0.0001, and the Adam optimization algorithm is used. Then, the occupant injury prediction model based on a temporal neural network (TCN) is trained. The training process is as follows: Figure 6 As shown, the loss function gradually decreases during the training process, indicating that the model recognizes the features of the input data and converges.

[0073] After the model is trained, the test set is used to validate the model, and the prediction results are as follows: Figure 7 , Figure 8As shown, to demonstrate the robustness of the model, 5-fold cross-validation was performed on the dataset for training. The design of each fold of the cross-validation differed, but the similarity scores of the prediction curves all exceeded 0.86, and the accuracy of the selected damage assessment metrics all exceeded 0.80. This result indicates that the dataset affects the model; however, the maximum difference in results is not significant, suggesting that the model exhibits high prediction accuracy and stability across different datasets.

[0074] This application's embodiments are based on a temporal neural network (TCN) to build a frontal collision occupant injury prediction model. It designs a dual-input layer fusion architecture, using the TCN network as the feature encoder for the vehicle collision temporal data, and a fully connected layer to process the constraint system parameters. The two are then spliced ​​and fused, so that the constraint system is included in the model as a conditional variable, reflecting the influence of the mechanical characteristics of the constraint system on occupant injury. Furthermore, a multi-branch decoder design is adopted, with the output layer using an independent dual-branch structure to predict the head acceleration curve and the chest compression curve, respectively.

[0075] This model uses vehicle collision waveforms and constraint system parameters as inputs to quickly predict head and chest injuries of occupants. After cross-validation, the model mentioned in this invention has strong stability and generalization ability, with a similarity score of prediction curves exceeding 0.86 and an accuracy of selected injury assessment indicators exceeding 0.80. It can accurately predict injuries to front-seat occupants in vehicle collisions, providing important support for the parameter design and optimization of constraint systems.

[0076] Compared with the prior art, the beneficial effects of the injury prediction method proposed in the embodiments of this application can be summarized as follows: (1) A front-collision front-row member prediction network model is established based on the temporal neural network (TCN) to achieve the purpose of rapid prediction, thereby saving a lot of time costs; (2) Combining simulation technology with deep learning, using multi-rigid-body simulation to acquire and expand the dataset, and obtaining occupant injury response and occupant injury index without simulation assistance after establishing the prediction model; (3) It can simultaneously process vehicle collision waveforms and constraint system parameters as inputs, and fuse the constraint system parameters as conditional variables with the waveform to effectively capture the nonlinear relationship between dynamic impact and constraint system parameters; (4) Establish a multi-branch output architecture during output to accurately predict the damage response of multiple parts of the crew over time.

[0077] The frontal collision occupant injury prediction method proposed in this application defines the vehicle collision waveform and constraint system parameters as inputs to the model in time-series and numerical forms, respectively. The outputs are time-series occupant injury responses and injury indices. Leveraging the advantages of deep learning, features are extracted from the input data through training and combined with frontal collision occupant injury simulation data for prediction, thereby significantly reducing time costs. This solves the problem in related technologies where the neglect of the mechanical and physical characteristics of the constraint system limits the application of deep learning in engineering design and safety system optimization.

[0078] Next, refer to the appendix. Figure 9 This application describes a frontal collision front occupant injury prediction device according to an embodiment of the present application.

[0079] Figure 9 This is a block diagram of a frontal collision front occupant injury prediction device according to an embodiment of this application.

[0080] like Figure 9 As shown, the frontal collision front occupant injury prediction device 10 includes: an acquisition module 100, a determination module 200, and a prediction module 300.

[0081] The acquisition module 100 is used to acquire the collision waveform of the vehicle.

[0082] The determination module 200 is used to determine the vehicle's constraint system parameters.

[0083] The prediction module 300 is used to input the collision waveform and constraint system parameters into a pre-built frontal collision front occupant prediction network module, and to use the constraint system parameters as conditional variables to fuse with the collision waveform, thereby capturing the nonlinear relationship between dynamic impact and constraint system parameters, in order to predict the injury response and injury index of multiple parts of the occupant over time.

[0084] Optionally, in one embodiment of this application, the frontal collision front occupant injury prediction device 10 further includes: a construction module, a calculation module, a generation module, and a modeling module; wherein, the construction module is used to construct a finite element model of the vehicle frontal collision condition based on preset evaluation criteria before inputting the collision waveform and constraint system parameters into the pre-constructed frontal collision front occupant prediction network model; the calculation module is used to input the constraint system parameters into the finite element model to calculate the occupant injury response and occupant injury index; the generation module is used to generate a frontal collision occupant collision response system dataset based on the constraint system parameters, vehicle collision waveform, occupant injury response, and occupant injury index; and the modeling module is used to construct a frontal collision front occupant prediction network model using the frontal collision occupant collision response system dataset and a temporal convolutional network.

[0085] Optionally, in one embodiment of this application, the frontal collision front occupant injury prediction device 10 further includes: a first processing module and a second processing module; wherein, the first processing module is used to preprocess the training data of the frontal collision occupant collision response system dataset after generating the frontal collision occupant collision response system dataset, remove data that meets preset abnormal conditions, and obtain preprocessed data; the second processing module is used to normalize the preprocessed data to obtain planned data that meets preset deep learning conditions.

[0086] Optionally, in one embodiment of this application, the occupant injury response includes a head injury response and a chest injury response.

[0087] Optionally, in one embodiment of this application, the formula for calculating the head injury response is: , In the formula, t1 and t2 represent the two moments of the collision, with a time interval of no more than 15ms; a represents the triaxial composite acceleration of the head's center of gravity (m / s2); and g is the gravitational acceleration.

[0088] Optionally, in one embodiment of this application, the frontal collision occupant injury prediction device 10 is further configured to: obtain the airbag deployment time, airbag vent diameter, seat belt force limit, and steering column crush curve scaling factor before inputting the constraint system parameters into the finite element model; and generate constraint system parameters based on the airbag deployment time, airbag vent diameter, seat belt force limit, and steering column crush curve scaling factor.

[0089] It should be noted that the foregoing explanation of the frontal collision front occupant injury prediction method embodiment also applies to the frontal collision front occupant injury prediction device of this embodiment, and will not be repeated here.

[0090] The frontal collision occupant injury prediction device proposed in this application can define the vehicle collision waveform and constraint system parameters as inputs to the model in time-series and numerical forms, respectively. The output is the occupant injury response and injury index in time-series form. Utilizing the advantages of deep learning, features of the input data are extracted through training and combined with frontal collision occupant injury simulation data for prediction, thereby greatly reducing time costs. This solves the problem in related technologies where the neglect of the mechanical and physical characteristics of the constraint system limits the application of deep learning in engineering design and safety system optimization.

[0091] Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 1001, the processor 1002, and the computer program stored on the memory 1001 and capable of running on the processor 1002.

[0092] When the processor 1002 executes the program, it implements the frontal collision front occupant injury prediction method provided in the above embodiments.

[0093] Furthermore, electronic devices also include: Communication interface 1003 is used for communication between memory 1001 and processor 1002.

[0094] The memory 1001 is used to store computer programs that can run on the processor 1002.

[0095] The memory 1001 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0096] If the memory 1001, processor 1002, and communication interface 1003 are implemented independently, then the communication interface 1003, memory 1001, and processor 1002 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized into address buses, data buses, control buses, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0097] Optionally, in a specific implementation, if the memory 1001, processor 1002, and communication interface 1003 are integrated on a single chip, then the memory 1001, processor 1002, and communication interface 1003 can communicate with each other through an internal interface.

[0098] The processor 1002 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0099] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting frontal occupant injuries in a frontal collision.

[0100] This application also provides a computer program product storing a computer program that, when executed by a processor, implements the above-described method for predicting frontal occupant injuries in a frontal collision.

[0101] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0102] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0103] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0104] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0105] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0106] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0107] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0108] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method of predicting injury to a front passenger occupant in a frontal collision, characterized by, The method comprises the following steps: acquiring a crash waveform of a vehicle; determining restraint system parameters of the vehicle; inputting the crash waveform and the restraint system parameters into a pre-constructed front-impact passenger prediction network model, taking the restraint system parameters as conditional variables to fuse with the crash waveform, capturing a nonlinear relationship between dynamic impact and the restraint system parameters, to predict injury responses and injury indicators of multiple parts of a passenger over time.

2. The method of claim 1, wherein, Before inputting the crash waveform and the restraint system parameters into the pre-constructed front-impact passenger prediction network model, the method further comprises: building a finite element model of an automobile front-impact working condition based on preset evaluation-related standards; inputting the restraint system parameters into the finite element model to calculate passenger injury responses and passenger injury indicators; generating a front-impact passenger crash response system data set based on the restraint system parameters, the automobile crash waveform, the passenger injury responses, and the passenger injury indicators; constructing the front-impact passenger prediction network model using the front-impact passenger crash response system data set and a time-series convolution network.

3. The method of claim 2, wherein, After generating the front-impact passenger crash response system data set, the method further comprises: preprocessing training data of the front-impact passenger crash response system data set to remove data satisfying preset abnormal conditions, to obtain preprocessed data; performing normalization processing on the preprocessed data to obtain planning data satisfying preset deep learning conditions.

4. The method of claim 2, wherein, The passenger injury responses include head injury responses and chest injury responses.

5. The method of claim 4, wherein, A calculation formula of the head injury responses is: ， where t1 and t2 represent two time points of a crash, a represents a three-axis combined acceleration of a head gravity center, and g is a gravitational acceleration.

6. The method of claim 2, wherein, Before inputting the restraint system parameters into the finite element model, the method further comprises: acquiring airbag point explosion time, airbag vent hole diameter, seat belt force limiter, and steering column crush curve scaling factor; generating the restraint system parameters based on the airbag point explosion time, the airbag vent hole diameter, the seat belt force limiter, and the steering column crush curve scaling factor.

7. A frontal collision front occupant injury prediction device, characterized in that, The method comprises: an acquisition module configured to acquire a crash waveform of a vehicle; a determination module configured to determine restraint system parameters of the vehicle; a prediction module configured to input the crash waveform and the restraint system parameters into a pre-constructed front-impact passenger prediction network model, taking the restraint system parameters as conditional variables to fuse with the crash waveform, capturing a nonlinear relationship between dynamic impact and the restraint system parameters, to predict injury responses and injury indicators of multiple parts of a passenger over time.

8. An electronic device, comprising: The method comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the front-impact passenger injury prediction method according to any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the front-impact passenger injury prediction method according to any one of claims 1-6.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the front-impact passenger injury prediction method according to any one of claims 1-6.

Images