A dummy injury evaluation method based on a simplified model deep learning algorithm

Abstract

The present application relates to the technical field of automobile safety evaluation, and in particular to a dummy injury evaluation method based on a simplified model deep learning algorithm. The method comprises the following steps: constructing a database of one-dimensional simplified models, including: collecting vehicle speed-time curve data, using a calibrated one-dimensional simplified model to replace the whole vehicle model for vehicle speed-time curve data sampling; based on the calibrated one-dimensional simplified model, generating a vehicle speed-time curve sample set by discretizing the load distribution of the main force transmission path; performing finite element analysis by substituting the constraint system CAE model, extracting the injury curve data of each part of the dummy, and arranging the extracted dummy injury curve data into a database form; using a CNN-LSTM architecture to build a deep learning model, setting the parameters and connection relationship of each layer of the model; used for predicting the dummy injury curve; and training the CNN-LSTM prediction model. The technical solution can improve the efficiency and accuracy of dummy injury curve prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automobile safety evaluation, and particularly relates to a dummy injury evaluation method based on a simplified model deep learning algorithm. BACKGROUND

[0002] Based on mature architecture, platform and vehicle model, vehicle modification development is one of the main processes of current new product development of each automobile factory. The development cycle and cost of the modified vehicle are usually related to the range of changes. For vehicle crash safety, when the range of changes is limited to the external shape, engine compartment layout and local structural components of the main force transmission path, the project team usually expects to use the interior, restraint system configuration and parameters of the base vehicle to achieve the purpose of minimizing the development cycle and cost. In order to ensure that the above strategy is achieved, in the product development process, in addition to using necessary CAE means to verify the crashworthiness indicators of the whole vehicle structure, the performance of the restraint system of the modified vehicle must be evaluated in real time to ensure that the dummy injury indicators meet the safety regulations and star level development requirements.

[0003] REFERENCE Figure 1 The traditional dummy injury evaluation method usually relies on detailed whole vehicle crash simulation analysis, and the "acceleration-time" or "speed-time" curve of the vehicle body is input into the simulation model of the restraint system of the base vehicle to obtain the injury value of each part of the dummy, so as to determine whether the design scheme of the modified vehicle meets the requirements. This method relies on high-performance servers, finite element solvers and high-precision dummy models. In complex and variable crash conditions, the establishment and adjustment of the model often requires a large amount of time and resources. However, in the product development process, the scheme changes frequently. On the one hand, this evaluation process needs to consume a large amount of manpower and software and hardware resources, and on the other hand, due to the long CAE calculation time, the restraint system dummy injury evaluation is delayed, which seriously affects the product development efficiency, making it difficult for automobile manufacturers to quickly respond to market changes and launch modified vehicle models that meet safety performance and cost effectiveness.

[0004] Although with the rapid development of artificial intelligence (AI) technology, automobile manufacturers have increased their investment in AI technology in the research and development stage, and the application research of AI technology in the field of automobile safety has also increased, for example, in the field of automobile restraint system performance, some research has applied long short-term memory network (LSTM) model to the prediction of whipping performance, and some research has used time domain reduction model algorithm to realize the prediction of dummy injury curve, and the prediction accuracy of various deep learning algorithms in occupant injury evaluation has been studied. However, these studies have not effectively solved the problem of rapid evaluation of restraint system dummy injury indicators in the development of modified vehicle models. The current AI application research has not yet integrated different models and algorithms to meet the dual needs of efficiency and accuracy of restraint system dummy injury evaluation in the development of modified vehicles. SUMMARY

[0005] The present application aims to provide a dummy injury evaluation method based on a simplified model deep learning algorithm, which can improve the efficiency and accuracy of dummy injury curve prediction.

[0006] To achieve the above-mentioned purpose, the technical scheme provides a dummy injury evaluation method based on a simplified model deep learning algorithm, comprising:

[0007] A database of one-dimensional simplified models is constructed, comprising:

[0008] The vehicle body speed-time curve data is collected, and the calibrated one-dimensional simplified model is used to replace the whole vehicle model to sample the vehicle body speed-time curve data; based on the calibrated one-dimensional simplified model, the load distribution of the discrete main force transmission path is generated to generate a vehicle body speed-time curve sample set for the input of the constraint system CAE analysis;

[0009] The generated vehicle body speed-time curve sample set is substituted into the constraint system CAE model for finite element analysis, and the injury curve data of each part of the dummy is extracted, and the extracted dummy injury curve data is arranged in the form of a database;

[0010] A deep learning model is built using a CNN-LSTM architecture, and the parameters and connection relationships of each layer of the model are set; and the dummy injury curve is predicted;

[0011] The training of the CNN-LSTM prediction model is performed.

[0012] The beneficial effects of the basic scheme are: the technical scheme uses a one-dimensional simplified model to replace a complex whole vehicle model to sample the vehicle body speed-time curve data, which greatly reduces the calculation amount and time cost. In the vehicle design and safety evaluation stage, key data can be obtained more quickly, thereby accelerating the entire evaluation process.

[0013] After the one-dimensional simplified model is calibrated, it can more accurately reflect the dynamic behavior of the vehicle in actual collision. Combined with the discrete main force transmission path load distribution, the generated vehicle body speed-time curve sample set is closer to the actual situation, providing a reliable basis for subsequent CAE analysis.

[0014] By substituting the generated vehicle body speed-time curve sample set into the constraint system CAE model for finite element analysis, detailed injury curve data of each part of the dummy can be extracted.

[0015] The deep learning model built with the CNN-LSTM architecture combines the spatial feature extraction capability of CNN and the time sequence dependence capturing capability of LSTM, can consider the multi-modal characteristics, spatial features and time dependencies of the data, and can fully utilize the historical data and feature information to perform high-precision prediction on the dummy injury curve. When processing time series data related to vehicle collision, it can more accurately mine potential information in the data, has better prediction accuracy than traditional models, and can provide more reliable results for dummy injury evaluation.

[0016] Compared with traditional physical tests, this method avoids a large number of physical tests and labor costs. Through the prediction of the deep learning model, the injury of the dummy can be accurately evaluated without actual collision tests, thereby reducing the overall evaluation cost.

[0017] The application of this method helps to promote the intelligent development of automobile safety evaluation. By continuously training and optimizing the deep learning model, the prediction accuracy and generalization ability can be further improved, providing strong support for vehicle design and safety performance improvement.

[0018] In summary, the technical scheme has significant beneficial effects in improving evaluation efficiency, enhancing data accuracy, building rich database resources, improving prediction ability, reducing evaluation cost, and promoting the improvement of intelligence and safety.

[0019] As an implementable preferred solution, the one-dimensional simplified model includes a plurality of one-dimensional spring units, each one-dimensional unit representing a main force transmission path or component of the vehicle, and the dynamic response of the whole vehicle during the collision process is simulated by setting the mechanical properties of each one-dimensional unit.

[0020] As an implementable preferred solution, the restraint system CAE model includes a vehicle body structure, interior parts, seat belts, airbags, and a dummy model, and is simulated and analyzed using finite element software; and based on the specified signal processing requirements, the CAE results are batch-processed, and the dummy head acceleration curve and chest compression curve are extracted as input data for the deep learning model.

[0021] As an implementable preferred solution, the mechanical properties of each one-dimensional unit are set to simulate the dynamic response of the whole vehicle during the collision process, including the following contents:

[0022] The acceleration curve of the vehicle body is predicted by the load characteristics of the discrete main force transmission path, or the main force transmission path load value is reversely designed according to the target acceleration curve, and when used in a modified vehicle, the main force transmission path load must be reversely calibrated based on the calculation results of the detailed finite element model of the whole vehicle; the force-displacement curve of the cross section of the main force transmission path in the collision direction is extracted from the calculation results of the detailed finite element model, and the load definition and fine tuning of the one-dimensional spring element of the one-dimensional simplified model of the main force transmission path simplified model are carried out, and the acceleration and speed time curves of the whole vehicle are matched.

[0023] As an implementable preferred solution, a deep learning model is built using a CNN-LSTM architecture, including the following:

[0024] Each data in the data set is serialized and regularized according to the RNN data rules, and the regularized calculation formula is as follows:

[0025]

[0026] In the formula, is the original data; is the normalized data, whose value range is [0, 1]; and are the maximum and minimum values of the data on the time curve in all sample data, respectively.

[0027] As an implementable preferred solution, when building the prediction model, the CNN-LSTM algorithm combines the spatial feature extraction capability of CNN and the time sequence dependence capture capability of LSTM, realizing the consideration of the multi-modal characteristics, spatial features and time dependence of the data; no pooling is set in the CNN network and no full connection step is set.

[0028] As an implementable preferred solution, the training of the CNN-LSTM prediction model includes the following:

[0029] The data set is divided into a training set and a test set by using the random sampling method without replacement, and a CNN-LSTM architecture is built based on the open source Python deep learning framework, and an Adam optimizer and a mean square error function MSE are set as training parameters for model training.

[0030] As an implementable preferred solution, the Adam optimizer in the Pytorch tool is selected for model training, including the following:

[0031] The training parameters are set, including the learning rate of the Adam optimizer and the loss function MSE; the calculation formula is as follows:

[0032]

[0033] In the formula: is the number of samples, is the true value, is the neural network model prediction value;

[0034] The training set data is input into the model for iterative training until the loss function converges or reaches a preset number of iterations.

[0035] As an implementable preferred solution, the accuracy of the CNN-LSTM prediction model is also evaluated, including the following contents:

[0036] The accuracy of the predicted curve is evaluated from the aspects of channel, phase, amplitude and slope, and the mathematical expression is as follows:

[0037]

[0038] In the formula, R is the accuracy value, , , and are the channel correlation, phase correlation, amplitude correlation and slope correlation of the reference curve and the predicted curve respectively.

[0039] As an implementable preferred solution, the accuracy of the CNN-LSTM prediction model is also evaluated, including the following contents:

[0040] The prediction accuracy R value of the test set dummy head acceleration curve and chest compression amount curve; compare the test set prediction results with the CAE calculation results. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 is a schematic diagram of the structure of a traditional restraint system CAE model.

[0042] Figure 2 is a logic diagram of a dummy injury evaluation method based on a simplified model deep learning algorithm.

[0043] Figure 3 is a schematic diagram of one-dimensional spring element distribution.

[0044] Figure 4 is a schematic diagram of a one-dimensional (1D) simplified model structure.

[0045] Figure 5 is a schematic diagram of the comparison of the results of the one-dimensional (1D) simplified model and the basic model of the acceleration-time curve.

[0046] Figure 6 is a schematic diagram of the comparison of the results of the one-dimensional (1D) simplified model and the basic model of the velocity-time curve.

[0047] Figure 7It is a speed-time curve sample distribution schematic diagram.

[0048] Figure 8 It is a CNN-LSTM model architecture schematic diagram.

[0049] Figure 9 It is a head acceleration model CNN-LSTM loss function convergence curve schematic diagram.

[0050] Figure 10 It is a chest compression amount model CNN-LSTM loss function convergence curve schematic diagram.

[0051] Figure 11 It is a head acceleration model pre-test sample prediction accuracy R value comparison schematic diagram.

[0052] Figure 12 It is a chest compression amount test sample prediction accuracy R value comparison schematic diagram.

[0053] Figure 13 It is a head acceleration model prediction result and CAE calculation result comparison curve schematic diagram.

[0054] Figure 14 It is a chest compression amount prediction result and CAE calculation result comparison curve schematic diagram.

[0055] Figure 15 It is a structure schematic diagram of an electronic device of an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the technical solutions of the present application and the advantages thereof clearer, the technical solutions of the present application will be further described in detail below with reference to the drawings. It can be understood that the specific embodiments described herein are only part of the embodiments of the present application, and are only used to explain the present application, but not to limit the present application. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered in isolation, and they can be combined with each other to achieve better technical effects. The same reference numerals appearing in the drawings of the following embodiments represent the same features or components, which can be applied to different embodiments.

[0057] In addition, unless otherwise defined, the technical terms or scientific terms used in the description of the present application should have the usual meanings understood by those of ordinary skill in the art to which the present application belongs.

[0058] The electronic device 500 includes a processor 501, a communication interface 502, a memory 503, and a bus 504.

[0059] The present application will be further described in detail below with reference to the drawings:

[0060] Reference Figure 2A dummy injury assessment method based on a simplified model deep learning algorithm, comprising:

[0061] A dummy injury assessment method based on a simplified model deep learning algorithm, the method comprising the following steps:

[0062] Step S100, constructing a one-dimensional (1D) simplified model database, comprising:

[0063] Step S101, collecting vehicle body speed-time curve data, and using a calibrated one-dimensional (1D) simplified model to replace the whole vehicle model to sample the vehicle body speed-time curve data. The one-dimensional (1D) simplified model has the advantage of high calculation efficiency, and the single calculation time is much lower than that of the whole vehicle collision model, which can significantly reduce the consumption of calculation resources.

[0064] The whole vehicle collision model is simplified into a one-dimensional (1D) simplified model, referring to Figure 3 , comprising a plurality of one-dimensional (1D) spring units, rigid parts, tires and beam placeholder parts, and each one-dimensional unit represents a main force transmission path or component of the vehicle, such as the front longitudinal beam, the engine compartment, the passenger compartment, etc. By setting the mechanical properties (such as stiffness, damping, etc.) of each one-dimensional unit, the dynamic response of the whole vehicle during the collision process is simulated. Compared with the detailed collision model of the whole vehicle, the single calculation time of the 1D simplified model under 8-core solving resources is less than 1 min, which has extremely high calculation efficiency.

[0065] Figure 4 The one-dimensional (1D) simplified model shown is mainly applied to the vehicle frontal 100% rigid wall collision condition, and the load characteristics of the main force transmission path are discretized to predict the vehicle body acceleration curve or to design the main force transmission path load value according to the target acceleration curve. Therefore, when used in a modified vehicle, the main force transmission path load must be reverse calibrated based on the calculation results of the detailed finite element model of the whole vehicle. In this process, the load definition and fine-tuning of the one-dimensional (1D) simplified model main force transmission path one-dimensional spring unit can be performed by extracting the force-displacement curve of the main force transmission path cross section along the collision direction of the detailed finite element model calculation results, matching the whole vehicle acceleration and speed time curve, and completing the calibration. In addition, under the premise of certain mass distribution, there is a strict theoretical correlation between the vehicle acceleration and the load, and the one-dimensional (1D) simplified model has universality in passenger cars.

[0066] Specifically, based on the finite element calculation results of a certain passenger car 56km / h frontal 100% overlap rigid wall collision, the 1D model is simplified and calibrated. The input (such as collision speed, collision angle, etc.) of the 1D simplified model is consistent with the basic model, and the collision simulation calculation is performed. Then, the acceleration curves (referring to Figure 5 ), and the speed-time curves (referring to Figure 6) and other key indicators, so that the 1D model acceleration curve, speed curve and the original detailed model of the vehicle are consistent, and the error is within the acceptable range.

[0067] Occupant Load Carrier (OLC, passenger load factor) as an index to evaluate the acceleration of the vehicle body, is often used to evaluate the difficulty of the development of the restraint system performance. After integrating the vehicle body acceleration curve and solving OLC using CAE post-processing software Meta, the OLC values of the 1D simplified model and the detailed model of the vehicle body acceleration curve are 32.1g and 31.5g respectively, and the relative error is only 1.9%.

[0068] Based on the calibrated 1D simplified model, the vehicle speed-time curve sample set for the input of the restraint system CAE analysis is generated by discretizing the load distribution of the main force transmission path. Specifically, the load distribution on the main force transmission path is discretized to obtain a series of vehicle speed-time curve samples under different load distributions. In this embodiment, under the condition of 56km / h frontal 100% overlap rigid wall collision, by discretizing the load distribution of the main force transmission path of the 1D model under the frontal collision condition, the Lsdyna finite element software is used for calculation, and 200 groups of vehicle "speed-time" curves for the input of the restraint system CAE analysis are generated by extracting the output of the rigid passenger compartment B column position node, as shown in Figure 7 .

[0069] Step S102, the vehicle speed-time curve sample set generated in step S11 is substituted into the restraint system CAE model for finite element analysis. The restraint system CAE model includes the vehicle body structure, interior trim parts, restraint system (such as seat belt, airbag, etc.) and dummy model, etc. In this embodiment, finite element software (such as LS-DYNA) is used for simulation analysis.

[0070] The CAE results are batched and post-processed based on the signal processing requirements specified in the C-NCAP management rules (2024 version) Appendix A frontal 100% overlap rigid wall collision test procedure. The processing process includes extracting the injury curve data of each part of the dummy (such as head, chest, etc.), and performing filtering, smoothing and other processing to obtain more accurate and reliable injury curve data. In this embodiment, the dummy head acceleration curve and the chest compression curve are extracted as the input data of the deep learning model.

[0071] The extracted dummy head acceleration curve and chest compression curve data are arranged in the form of a database to facilitate the training and testing of the subsequent deep learning model. The database includes two parts of input variables (vehicle speed-time curve) and output variables (dummy head acceleration curve and chest compression curve).

[0072] Step S200, constructing a prediction model, comprising:

[0073] Referring to Figure 8 , an open-source Python deep learning framework Pytorch is used to build a convolutional neural network-long short-term memory network (CNN-LSTM) architecture. The CNN-LSTM algorithm combines the spatial feature extraction capability of CNN and the time sequence dependence capture capability of LSTM. The long short-term memory network (LSTM) introduces a forgetting gate, an input gate and an output gate, and uses an adaptive method to control information flow, thereby capturing long-term dependent information. The CNN-LSTM algorithm increases the convolution operation on one-dimensional input data on the basis of the traditional LSTM algorithm, and can consider the multi-modal characteristics, spatial features and time dependencies of data, thereby having good prediction accuracy.

[0074] The CNN-LSTM structure parameters are shown in Table 1.

[0075] Table 1

[0076]

[0077] Since the sequence data is processed in the embodiment, the time sequence value is the same as the size of the CNN convolution kernel, and no edge data padding is performed in the CNN network. The output data of a single convolution kernel after convolution is a scalar, so no pooling is set in the CNN network. Secondly, in the model, only the output data is subjected to a splicing expansion and a negative value elimination operation based on the Relu function after CNN convolution, so there is no full connection step in the CNN network architecture.

[0078] Each data in the data set is serialized and regularized according to the RNN data rule. In the embodiment, the time range of the time domain data is 0-100 ms, the data interval is 0.1 ms, the time sequence parameter is 3, and the regularization calculation formula is as follows:

[0079]

[0080] In the formula, is the original data; is the regularization data, and the value range is between 0-1; and are the maximum and minimum values of the data on the time domain curve in all sample data, respectively.

[0081] Step S300, after the database is constructed and the prediction model is selected, the CNN-LSTM prediction model is trained. The training process includes data division, model building, parameter setting and training iteration.

[0082] The original data set is divided into a training set and a test set according to a certain proportion. In the present application, 200 groups of data are divided into 150 groups of training set and 50 groups of test set by using the random sampling method without replacement. The same method is used to select 10 groups, 50 groups, 100 groups and 150 groups of samples from the training data set for model training.

[0083] The CNN-LSTM architecture is built based on the open source Python deep learning framework Pytorch. The model layer parameters and connection relationship are set according to the training parameters defined in the foregoing.

[0084] The training parameters of the Adam optimizer, such as the learning rate and the loss function MSE, are set. The Adam optimizer in the Pytorch tool is selected for model training, and the learning rate has adaptive characteristics. The initial learning rate is set to 0.001, and the loss function is the mean square error function MSE. Generally, the smaller the MSE is during training, the more excellent the model generalization ability is, and the calculation formula is as follows:

[0085]

[0086] In the formula: is the number of samples, is the true value, is the predicted value of the neural network model.

[0087] The training set data is input into the model for iterative training until the loss function converges or the preset number of iterations is reached. During the training process, the loss value of each iteration is recorded to evaluate the model performance.

[0088] The loss function MSE value curve during the training process of the dummy head acceleration and chest compression amount model is shown in Figure 9 and Figure 10 It can be seen that there is a slight oscillation phenomenon of MSE in the later stage of the training process.

[0089] Step S400, the accuracy analysis of the CNN-LSTM prediction model, comprising:

[0090] After training, the accuracy of the CNN-LSTM prediction model needs to be evaluated. The evaluation method includes comparing the test set prediction results with the CAE calculation results, calculating the prediction accuracy index, etc.

[0091] In this embodiment, the ISO / TS 18571 standard is used to evaluate the accuracy of the predicted curve from four aspects of channel, phase, amplitude and slope, and the mathematical expression is as follows:

[0092]

[0093] In the formula, R is the accuracy value, , , and Channel correlation, phase correlation, amplitude correlation and slope correlation of the reference curve and the prediction curve, respectively.

[0094] The prediction accuracy R values of the test set dummy head acceleration curve and chest compression amount curve are shown in Figure 11 and Figure 12 The closer the R value is to 1, the higher the prediction accuracy. The average R values of each prediction model are shown in Table 2.

[0095] Table 2: Average R values of each model

[0096]

[0097] By comparing the prediction results of the test set with the CAE calculation results, it is found that the R values of the CNN-LSTM prediction model on the two responses (head acceleration and chest compression amount) are at a relatively high level except for sample point 2. Among them, the average R value of each model for the head acceleration curve is greater than 0.85, and the average R value for the chest compression amount curve is greater than 0.89. This indicates that the CNN-LSTM prediction model has high prediction accuracy and generalization ability.

[0098] In addition, the influence of the training sample size on the generalization performance of the model is also studied. By comparing the prediction accuracy and loss function convergence of the model under different training sample sizes, it is found that when the training sample size reaches a certain size (such as more than 100 groups), continuing to increase the training sample size has limited effect on improving the generalization ability of the model. Therefore, in actual application, appropriate training sample size can be selected according to specific circumstances to balance the calculation resources and prediction accuracy.

[0099] To better show the influence of the training sample size on the prediction effect of the model, further analysis is carried out in combination with the prediction curve. A certain sample point in the test sample data set is selected, and the curve prediction effect is shown in Figure 13 and Figure 14 It can be seen that the consistency of the model prediction results and the CAE calculation results of the two curve responses is good. After evaluation, the trained models above have basically met the use requirements of the improved development based on the vehicle model.

[0100] The embodiments of the present disclosure also provide a dummy injury evaluation system based on a simplified model deep learning algorithm, which uses a dummy injury evaluation method based on a simplified model deep learning algorithm.

[0101] The embodiments of the present disclosure also provide a storage medium having a computer program stored therein, wherein the computer program, when executed by a processor, can implement all steps of the dummy injury evaluation method based on a simplified model deep learning algorithm.

[0102] Those skilled in the art can understand that all or part of the process of the dummy injury evaluation method based on the simplified model deep learning algorithm can be completed by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer readable storage medium, and when the program is executed, the process of each embodiment of the dummy injury evaluation method based on the simplified model deep learning algorithm can be included. Any reference to memory, storage, database or other medium used in each embodiment provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0103] The embodiments of the present application also provide an electronic device, which includes a memory, a processor and a computer program stored in the memory and executable on the processor. The processor implements the steps of the dummy injury evaluation method based on the simplified model deep learning algorithm when executing the program. In the embodiments of the present application, the processor is the control center of the computer system, which can be the processor of a physical machine or the processor of a virtual machine.

[0104] Reference Figure 15 The electronic device 500 includes at least one processor 501, at least one communication interface 502, at least one memory 503 and at least one bus 504. The bus 504 is used to realize the connection communication between the components, the communication interface 502 is used for signaling or data communication with other node devices, and the memory 503 stores machine readable instructions executable by the processor 501. When the electronic device 500 is running, the processor 501 communicates with the memory 503 through the bus 504, and the machine readable instructions are executed by the processor 501 when called. The steps of the dummy injury evaluation method based on the simplified model deep learning algorithm are as described above.

[0105] The above is only an embodiment of the present application, and common knowledge of specific structures and properties in the scheme is not described in detail, and the ordinary skilled person in the art knows all the ordinary technical knowledge in the field of the present application before the application date or the priority date, can know all the prior art in the field, and has the ability to apply conventional experimental means before that date, and the ordinary skilled person in the art can improve and implement the present scheme under the guidance of the present application, and some typical known structures or known methods should not be an obstacle for the ordinary skilled person in the art to implement the present application. It should be pointed out that for those skilled in the art, without departing from the structure of the present application, a number of modifications and improvements can be made, which should also be considered as the protection scope of the present application, which will not affect the effect and practicality of the patent. The protection scope of the present application should be subject to the content of its claims, and the specific implementation mode and the like in the specification can be used to explain the content of the claims.

Claims

1. A dummy injury assessment method based on a simplified model deep learning algorithm, characterized in that: Comprise: A database of one-dimensional simplified models is constructed, the one-dimensional simplified models comprising a plurality of one-dimensional spring units, each one-dimensional unit representing a main force transmission path or component of a vehicle, and the dynamic response of the vehicle in a collision process is simulated by setting the mechanical properties of each one-dimensional unit, including: Collecting vehicle body speed-time curve data, and using the calibrated one-dimensional simplified model to replace the vehicle model to sample the vehicle body speed-time curve data; Based on the calibrated one-dimensional simplified model, the load distribution of the main force transmission path is discretized to generate a vehicle body speed-time curve sample set for the input of the CAE analysis of the restraint system; The generated vehicle body speed-time curve sample set is substituted into the CAE model of the restraint system for finite element analysis, the injury curve data of each part of the dummy is extracted, and the extracted dummy injury curve data is arranged in the form of a database; A deep learning model is built using the CNN-LSTM architecture, and the parameters and connection relationships of each layer of the model are set; for predicting the dummy injury curve; The training of the CNN-LSTM prediction model is performed.

2. The dummy injury assessment method based on a simplified model deep learning algorithm according to claim 1, characterized in that: The restraint system CAE model includes the vehicle body structure, interior trim parts, seat belts, airbags, and dummy models, and is simulated and analyzed using finite element software; and based on the specified signal processing requirements, the CAE results are post-processed in batches, and the dummy head acceleration curve and chest compression curve are extracted as input data for the deep learning model.

3. The dummy injury assessment method based on a simplified model deep learning algorithm according to claim 1, characterized in that: The mechanical properties of each one-dimensional unit are set to simulate the dynamic response of the vehicle in the collision process, including the following contents: The load characteristics of the main force transmission path are discretized to predict the vehicle body acceleration curve or to design the main force transmission path load value in reverse according to the target acceleration curve, and when used in a modified vehicle, the main force transmission path load must be reverse calibrated based on the calculation results of the detailed finite element model of the vehicle; The load definition and fine-tuning of the one-dimensional spring unit of the main force transmission path simplified model of the one-dimensional simplified model are performed by extracting the force-displacement curve of the main force transmission path section along the collision direction from the calculation results of the detailed finite element model, and the vehicle acceleration and speed time curve are matched.

4. The dummy injury assessment method based on a simplified model deep learning algorithm according to claim 1, characterized in that: A deep learning model is built using the CNN-LSTM architecture, including the following contents: Each data in the data set is serialized and regularized according to the RNN data rules, and the regularized calculation formula is as follows: In the formula, is the original data; is the regularization data, whose value interval is [0, 1]; and are the maximum and minimum values of the data on the time-domain curve in all sample data, respectively.

5. The dummy injury assessment method based on a simplified model deep learning algorithm according to claim 1, characterized in that: When building the prediction model, the CNN-LSTM algorithm combines the spatial feature extraction capability of CNN and the time sequence dependence capture capability of LSTM to realize the consideration of the multi-modal characteristics, spatial features and time dependencies of the data; no pooling is set in the CNN network and no full connection step is performed.

6. The dummy injury assessment method based on a simplified model deep learning algorithm according to claim 1, characterized in that: The training of the CNN-LSTM prediction model is performed, including the following contents: The data set is divided into a training set and a test set using the random sampling method without replacement, and a CNN-LSTM architecture is built based on the open-source Python deep learning framework, the Adam optimizer and the mean square error function MSE are set as the training parameters for model training.

7. The dummy injury assessment method based on a simplified model deep learning algorithm according to claim 6, characterized in that: The Adam optimizer in the Pytorch tool is selected for model training, including the following contents: The training parameters are set, including the learning rate of the Adam optimizer and the loss function MSE; the calculation formula is as follows: In the formula: is the sample quantity, is the true value, is the neural network model prediction value; The training set data is input into the model for iterative training until the loss function converges or a preset number of iterations is reached. 8.The dummy injury assessment method based on a simplified model deep learning algorithm according to claim 1, characterized in that: The accuracy of the CNN-LSTM prediction model is also evaluated, including the following: The accuracy of the predicted curve is evaluated from four aspects: channel, phase, amplitude, and slope, and the mathematical expression is as follows: where R is a precision value, , , and are channel correlation, phase correlation, amplitude correlation and slope correlation of the reference curve and the prediction curve, respectively.

9. The dummy injury assessment method based on a simplified model deep learning algorithm according to claim 8, characterized in that: The accuracy of the CNN-LSTM prediction model is also evaluated, including the following: The test set dummy head acceleration curve and chest compression volume curve prediction accuracy R value; compare the test set prediction results with the CAE calculation results.

Images