A method and device for constructing a car collision dummy neck calibration model

By constructing a neck calibration model for a car crash dummy using the PSO-TabNet neural network, the problem of low modeling accuracy was solved, high-precision neck performance prediction was achieved, and the accuracy of dummy calibration was improved.

CN118533511BActive Publication Date: 2024-11-26CHINA AUTOMOTIVE TECH & RES CENT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410995470.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-26
Estimated Expiration
2044-07-24

AI Technical Summary

Technical Problem

The existing technology for car crash dummy neck calibration models has low modeling accuracy and cannot effectively reflect the real human body's movement posture and injury situation.

Method used

A neck calibration model was constructed using a PSO-TabNet neural network. By acquiring multiple sets of experimental data, initial features were constructed and incorporated into the initial conditional data to establish a high-precision prediction model for the head and neck D-plane angle and occipital moment.

Benefits of technology

The modeling accuracy of the dummy neck calibration model has been improved, which can better reflect the neck performance characteristics of the collision dummy and improve the dynamic performance and fitting effect of the model output.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118533511B_ABST
    Figure CN118533511B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of data processing, and in particular to a method and device for constructing a calibration model for the neck of a car collision dummy. The method comprises: obtaining multiple sets of collision dummy neck calibration test data; constructing initial features according to initial condition data; wherein the initial condition data comprises time, a first velocity when a pendulum hits honeycomb aluminum, and a first acceleration of the pendulum after the collision occurs; the initial features comprise a first velocity feature, a first acceleration feature, and a second acceleration feature; the first velocity feature is the change of the velocity of the pendulum after it hits the honeycomb aluminum with the time, the first acceleration feature is the time difference between the two inflection points of the first acceleration curve, and the second acceleration feature is the average acceleration between the two inflection points; constructing a head and neck D-plane angle prediction model and an occipital moment prediction model based on a PSO‑TabNet neural network. The present invention can solve the problem of low modeling accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of data processing, and in particular to a method and device for constructing a car collision dummy neck calibration model. Background Art

[0002] The automobile collision dummy is a key test equipment for evaluating the safety performance of vehicles. In order to ensure that the dummy can restore the movement posture and injury of the real human body to the greatest extent during the collision test, it is necessary to calibrate each part of the dummy according to the regulations to judge the bionic performance of the dummy. Domestic patent CN202310405885.2 "Dummy calibration system, rear impact calibration equipment and calibration method thereof" discloses a dummy calibration system, rear impact calibration equipment and calibration method for vehicle collision test. The calibration method in this patent only mentions that the relevant data meets the expected requirements, that is, the calibration is judged to be qualified, but it does not clearly mention how to process the relevant data and what the expected requirements are. There is no practical guidance on how to effectively apply the calibration data to achieve the determination of the calibration results. Domestic patent CN202311090640.1 "Automobile collision dummy calibration result determination method, equipment and medium" designs the RMS indicator to compare the difference between the average mechanical response curve of the collision dummy and the average mechanical response curve of the real human body, and then judges the collision dummy calibration result. The average method is the simplest curve modeling method with time as the single input variable. The average mechanical response curve constructed using the average method only retains the average change relationship of the mechanical response index over time, and cannot well show the dynamic performance of the curve. In actual operation, the initial conditions of each calibration test are often different. Only by incorporating the initial conditions into the modeling input can the modeling accuracy of the mechanical response curve be guaranteed.

[0003] In view of this, the present invention is proposed. Summary of the invention

[0004] The purpose of the present invention is to provide a method and device for constructing a car collision dummy neck calibration model, so as to solve the problem of low modeling accuracy in the prior art.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present invention provides a method for constructing a car collision dummy neck calibration model, comprising:

[0007] Acquire multiple sets of automobile collision dummy neck calibration test data; the neck calibration test data includes pendulum acceleration, pendulum velocity, head and neck D-plane angle, occipital moment and time characteristics;

[0008] According to the initial condition data in the neck calibration test data, an initial feature is constructed; wherein the initial condition data includes time, a first speed when the pendulum hits the honeycomb aluminum, and a first acceleration of the pendulum after the impact occurs; the initial feature includes a first speed feature, a first acceleration feature, and a second acceleration feature; the first speed feature is the change of the speed of the pendulum after it hits the honeycomb aluminum with the time, the first acceleration feature is the time difference between two inflection points of the first acceleration curve, and the second acceleration feature is the average acceleration between the two inflection points;

[0009] Taking the initial features, time features and occipital moment as input, a head and neck D-plane angle prediction model is constructed based on the PSO-TabNet neural network;

[0010] Taking the initial features, time features and head and neck D-plane angle as input, an occipital moment prediction model is constructed based on the PSO-TabNet neural network.

[0011] As a further preferred technical solution, the first acceleration feature is constructed in the following manner:

[0012] ;in is the first acceleration feature, is the time corresponding to the inflection point of the rising section of the first acceleration curve; is the time corresponding to the inflection point of the descending section of the first acceleration curve.

[0013] As a further preferred technical solution, the second acceleration feature is constructed in the following manner:

[0014] ;in is the second acceleration characteristic, For the first acceleration curve The acceleration corresponding to the moment.

[0015] As a further preferred technical solution, the head and neck D-plane angle prediction model based on the PSO-TabNet neural network includes:

[0016] The training set data is fed into the TabNet model for fitting to obtain the initial model of the head and neck D-plane angle;

[0017] Use the initial model to predict the validation set data;

[0018] Use loss function to evaluate the fitting effect;

[0019] According to the fitting effect, the PSO algorithm is used to optimize the initial model hyperparameters;

[0020] When the number of iterations reaches the first set value, the model with the lowest loss function is selected as the head and neck D-plane angle prediction model.

[0021] As a further preferred technical solution, the loss function is:

[0022] Loss= ,in is the model output, vector is the normalized model input feature, y(i) is the calibrated test value of the model output variable, and n is the number of sample points.

[0023] As a further preferred technical solution, after the head and neck D-plane angle prediction model is constructed based on the PSO-TabNet neural network, it also includes:

[0024] The test set data is sent to the head and neck D-plane angle prediction model to calculate the fitting residual;

[0025] Determining an evaluation index according to the fitting residual;

[0026] If the evaluation index is greater than or equal to the second set value, the head and neck D-plane angle prediction model is successfully constructed;

[0027] If the evaluation index is less than the second set value, the step of constructing a head and neck D-plane angle prediction model based on the PSO-TabNet neural network is continued until the evaluation index is greater than or equal to the second set value.

[0028] As a further preferred technical solution, the fitting residual is given by the following formula:

[0029] ,in is the residual sequence, i is the sample point number, k is the test set number, y k (i) is the calibration test value of the model output variable, is the predicted value of the model;

[0030] The evaluation indicators are as follows:

[0031] , where De is the evaluation index, num is the number of test sets, is the average of the calibrated test values ​​of the sample point output variable, and n is the number of sample points.

[0032] In a second aspect, the present invention provides a device for constructing a calibration model of a neck of a car collision dummy, comprising:

[0033] An acquisition module is used to acquire multiple sets of automobile collision dummy neck calibration test data; the neck calibration test data includes pendulum acceleration, pendulum velocity, head and neck D-plane angle, occipital moment and time characteristics;

[0034] An initial feature construction module is used to construct an initial feature according to the initial condition data in the neck calibration test data; wherein the initial condition data includes time, a first speed when the pendulum hits the honeycomb aluminum, and a first acceleration of the pendulum after the impact occurs; the initial feature includes a first speed feature, a first acceleration feature, and a second acceleration feature; the first speed feature is the change of the speed of the pendulum after it hits the honeycomb aluminum with the time, the first acceleration feature is the time difference between two inflection points of the first acceleration curve, and the second acceleration feature is the average acceleration between the two inflection points;

[0035] The prediction model construction module is used to construct a head and neck D-plane angle prediction model based on the PSO-TabNet neural network with the initial features, time features and occipital moment as inputs; and to construct an occipital moment prediction model based on the PSO-TabNet neural network with the initial features, time features and head and neck D-plane angles as inputs.

[0036] In a third aspect, the present invention provides an electronic device, comprising:

[0037] at least one processor, and a memory communicatively coupled to at least one of the processors;

[0038] The memory stores instructions that can be executed by at least one of the processors, and the instructions are executed by at least one of the processors so that at least one of the processors can perform the above method.

[0039] In a fourth aspect, the present invention provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to enable the computer to execute the above method.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The method for constructing a car collision dummy neck calibration model provided by the present invention first obtains multiple groups of car collision dummy neck calibration test data, then constructs initial features according to initial condition data in the neck calibration test data, the initial features including a first velocity, a first acceleration feature and a second acceleration feature, finally takes the initial features, time features and occipital moment as input, and constructs a head and neck D-plane angle prediction model based on a PSO-TabNet neural network, and takes the initial features, time features and head and neck D-plane angle as input, and constructs an occipital moment prediction model based on a PSO-TabNet neural network.

[0042] The present invention performs regression fitting on the performance of the collision dummy neck calibration test through the PSO-TabNet neural network model, establishes a high-precision dummy performance model, and solves the problem of insufficient modeling accuracy of the existing averaging method.

[0043] Aiming at the collision dummy neck calibration test, the present invention divides the test data into initial condition data and test performance data (other data except the initial condition data), constructs initial features according to the initial condition data, and incorporates the initial features into the neural network input, so that the neural network model can better learn the performance characteristics of the collision dummy's neck, thereby solving the problem of insufficient modeling accuracy caused by the existing regression model only considering the experimental performance data and ignoring the influence of the initial conditions on the calibration test results.

[0044] Furthermore, the present invention uses the Loss function as the objective function of PSO parameter optimization. This exponential evaluation index increases the weight of the outlier fitting accuracy during hyperparameter optimization, making the dynamic performance of the model output better, and thus making the model output better fit the main performance indicators of the neck calibration test such as maximum value, maximum value corresponding time, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 It is a flow chart of the method for constructing a car collision dummy neck calibration model provided by the present invention;

[0047] Figure 2 is a schematic diagram of the head and neck calibration test;

[0048] Figure 3 is a pendulum acceleration curve obtained by the present invention;

[0049] Figure 4 It is a schematic diagram of the D-plane angle curve of the head and neck obtained by the present invention;

[0050] Figure 5 is a schematic diagram of the occipital moment curve obtained by the present invention;

[0051] Figure 6 It is a schematic diagram of the head and neck D-plane angle prediction model test constructed in an embodiment of the present invention;

[0052] Figure 7 is a schematic diagram of the test of the occipital moment prediction model constructed in an embodiment of the present invention;

[0053] Figure 8 1 is a schematic diagram of the residual error of the head and neck D-plane angle prediction model test constructed in an embodiment of the present invention;

[0054] Fig. 9 is a schematic diagram of the residual error of the occipital moment prediction model test constructed in an embodiment of the present invention;

[0055] Fig.10 It is a structural schematic diagram of a vehicle collision dummy neck calibration model building device provided by the present invention;

[0056] Fig.11 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0057] The following is a description of exemplary embodiments of the present application in conjunction with the accompanying drawings, including various details of the embodiments of the present application to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0058] As mentioned in the background technology, the prior art has the problem of low modeling accuracy. In this regard, the present invention adopts a method of incorporating initial conditions into the modeling process and constructs a performance prediction model based on a PSO-TabNet neural network. The present invention is further described in detail below in conjunction with an embodiment.

[0059] Example 1

[0060] Figure 1 This is a flow chart of a method for constructing a car collision dummy neck calibration model provided by this embodiment. This embodiment is suitable for constructing a model after obtaining the collision dummy neck calibration data, and provides support for dummy calibration tests and repeatability and reproducibility tests. This method can be executed by a car collision dummy neck calibration model construction device, which can be composed of software and / or hardware and is generally integrated in an electronic device, which can be a computer or other mobile terminal (such as a smart phone, tablet computer, etc.). For ease of understanding, each step in the control method of this embodiment is executed by a computer.

[0061] like Figure 1 As shown, this embodiment provides a method for constructing a car collision dummy neck calibration model, comprising the following steps:

[0062] S110, obtaining multiple sets of automobile collision dummy neck calibration test data; the neck calibration test data includes pendulum acceleration, pendulum velocity, head and neck D-plane angle, occipital moment and time characteristics.

[0063] The above-mentioned “pendulum acceleration” is acquired by using an acceleration sensor installed on the pendulum. Figure 2 As shown in the figure, the “D-plane angle of the head and neck” is the angle formed by the plane where the upper neck force sensor of the dummy is located (i.e., the D-plane) and the bottom surface of the lower neck (the plane shown by the dotted line at the lower neck). The “occipital moment” is the equivalent torque in the y direction acting on the dummy’s occipital position, which is calculated from the neck force collected by the upper neck force sensor. The “time feature” is the time collected in the calibration test. There is a corresponding relationship between time and the pendulum acceleration, pendulum velocity, D-plane angle of the head and neck, and occipital moment. That is to say, combined with the time feature, we can obtain the curves of the pendulum acceleration, pendulum velocity, D-plane angle of the head and neck, and occipital moment changing with time, as shown in the figure. Figure 3-Figure 5 Shown are the obtained pendulum acceleration curve, head and neck D-plane angle curve, and occipital moment curve.

[0064] Regarding the automobile crash dummy neck calibration test data, this embodiment uses 5 crash dummy head and neck assemblies, and conducts 5 tests on each assembly to obtain a total of 25 sets of data. In fact, the more data sets, the better, to ensure that there is enough data to build a performance model.

[0065] This embodiment does not impose any special restrictions on the device and test method for the head and neck calibration test, and any method that can be implemented in this technical field can be used. The head and neck calibration device uses a pendulum mounted on a stand to impact the honeycomb aluminum, giving the dummy head and neck assembly an acceleration, and measures the dummy head and neck related data through sensors to evaluate the mechanical properties of the head and neck assembly. The head and neck assembly includes: skull, scalp, occipital bone, occipital skin, upper neck force sensor, neck body, and lower neck mounting bracket.

[0066] S120. Constructing initial features according to the initial condition data in the neck calibration test data; wherein the initial condition data includes time, a first velocity when the pendulum hits the honeycomb aluminum, and a first acceleration of the pendulum after the impact occurs; the initial features include a first velocity feature, a first acceleration feature, and a second acceleration feature; the first velocity feature is a change in the velocity of the pendulum after it hits the honeycomb aluminum with the time, the first acceleration feature is a time difference between two inflection points of the first acceleration curve, and the second acceleration feature is an average acceleration between the two inflection points.

[0067] It should be understood that the above-mentioned “first acceleration curve” refers to a curve of the first acceleration changing with time.

[0068] In the neck calibration, the first speed represents the total energy when the pendulum hits, which is mainly related to the state of the calibration device. The front part of the first acceleration curve represents the acceleration of the calibration device acting on the lower end of the dummy's neck, which is mainly related to the state of the honeycomb aluminum. The above-mentioned first speed and the above-mentioned first acceleration curve will affect the neck calibration result, so the present invention uses these two characteristics as initial condition data. The first acceleration curve can be filtered using CFC60.

[0069] Furthermore, the first speed feature is constructed in such a way that a column is added to the test data to store the first speed information of this calibration test.

[0070] ,in It is A variable in a vector used to store speed information, It is The time corresponding to the vector is is the first speed measured at time 0 of this calibration test (i.e., the speed of the pendulum when it hits the honeycomb aluminum). Vector It is all the data collected at a certain moment in the calibration test, and the vector at least includes: time, occipital moment, head and neck D-plane angle and the first acceleration.

[0071] Furthermore, the first acceleration feature is constructed in the following manner:

[0072] ;in is the first acceleration feature, is the time corresponding to the inflection point of the rising section of the first acceleration curve, that is, the time corresponding to the part with the maximum jerk of the curve; It is the time corresponding to the inflection point of the descending section of the first acceleration curve, that is, the time corresponding to the position of the minimum jerk of the curve.

[0073] Furthermore, the second acceleration feature is constructed in the following manner:

[0074] ;in is the second acceleration characteristic, For the first acceleration curve The acceleration corresponding to the moment.

[0075] S130, using the initial features, time features and occipital moment as input, constructing a head and neck D-plane angle prediction model based on the PSO-TabNet neural network.

[0076] In the process of building the prediction model, the number of models that need to be established and the corresponding input variables and output variables are determined according to the type of experiment, and the initial condition characteristics are added to the input variables. The number of models and output variables is determined by the number of variables involved in the evaluation indicators of the dummy component calibration test. The function of the model is to fit the output variables using the input variables, and the model corresponds to the output variables one by one. In the neck bending test, whether the dummy neck is qualified is determined by the indicators calculated from the two variables of the head and neck D-plane angle and the occipital moment. Therefore, it is necessary to establish regression models with the head and neck D-plane angle and the occipital moment as outputs, respectively, to describe the performance of the dummy neck.

[0077] Each of the models corresponds to an output variable and a plurality of input variables, and the input variables include the initial features and the output variables corresponding to other models. In step S130, in the head and neck D-plane angle prediction model, the input variables include the first velocity feature, the first acceleration feature, the second acceleration feature, the time feature, and the occipital moment. In step S140, in the occipital moment prediction model, the input variables include the first velocity feature, the first acceleration feature, the second acceleration feature, the time feature, and the head and neck D-plane angle.

[0078] Furthermore, the construction of the head and neck D-plane angle prediction model based on the PSO-TabNet neural network includes:

[0079] The training set data is fed into the TabNet model for fitting to obtain the initial model of the head and neck D-plane angle;

[0080] Use the initial model to predict the validation set data;

[0081] Use loss function to evaluate the fitting effect;

[0082] According to the fitting effect, the PSO algorithm is used to optimize the initial model hyperparameters;

[0083] When the number of iterations reaches the first set value, the model with the lowest loss function is selected as the head and neck D-plane angle prediction model.

[0084] Before building the model, the data obtained by S110 needs to be normalized and then divided into three parts: training set, validation set, and test set. The normalization operation formula is given by the following formula:

[0085] ,in x is a parameter in the input variable, x max The maximum value obtained for the parameter; x minis the minimum value of the parameter. x new is the normalized value.

[0086] As described in step S110, this embodiment obtains a total of 25 sets of experimental data, 19 sets of data are randomly selected as training sets, 3 sets of data are selected as validation sets, and the remaining 3 sets of data are selected as test sets. The training set data is used to train the neural network model to fit the output curve in the neck calibration test. The validation set data is used to verify the fitting effect, and the model hyperparameters are adjusted based on the loss function so that the model can conform to the real calibration test curve to the greatest extent. The test set data is used to test the final fitting effect of the evaluation model.

[0087] The TabNet neural network model is a neural network model based on the self-attention mechanism, which proposes a sequence attention mechanism to determine the feature subset with the most semantic value and process it in each round of decision-making. The TabNet model combines the characteristics of the tree model and the neural network model, and has obvious advantages in processing tabular data such as collision dummy calibration test curves.

[0088] The PSO (Particle Swarm Optimization) particle swarm optimization algorithm is used to optimize the hyperparameters of the TabNet network. The PSO algorithm abstracts the feasible solution of each function to be optimized into particles, and each particle has the following elements: speed, position, and the optimal position it has experienced. The movement of particles is constrained by both speed and position. Here, an adaptive PSO algorithm that adaptively changes the weight according to the number of iterations is used to achieve parameter optimization. The objective function of the parameter optimization is the loss function of the model fitting.

[0089] Furthermore, the loss function is:

[0090] Loss= ,in is the model output, vector is the normalized model input feature, y(i) is the calibrated test value of the model output variable, and n is the number of sample points.

[0091] The evaluation of the bionic performance of the collision dummy mainly focuses on the dynamic performance of the model output, such as the maximum value and the time corresponding to the maximum value. Therefore, in order to ensure the fitting performance of the neural network model in dynamic performance, the loss function uses an exponential form to measure the difference between the predicted value and the actual value. This method enables the model parameter optimization to focus more on the fitting effect of outliers, thereby improving the modeling accuracy.

[0092] The TahNet hyperparameters optimized by the PSO algorithm include at least the following six: N_d number of decision layers; N_a number of attention layers; N_steps number of steps; learningrate learning rate; gamma factor; Optimizer_fn internal optimizer.

[0093] Further, after constructing the head and neck D-plane angle prediction model based on the PSO-TabNet neural network, the method further includes:

[0094] The test set data is sent to the head and neck D-plane angle prediction model to calculate the fitting residual;

[0095] Determining an evaluation index according to the fitting residual;

[0096] If the evaluation index is greater than or equal to the second set value, the head and neck D-plane angle prediction model is successfully constructed;

[0097] If the evaluation index is less than the second set value, the step of constructing a head and neck D-plane angle prediction model based on the PSO-TabNet neural network is continued until the evaluation index is greater than or equal to the second set value.

[0098] Furthermore, the fitting residual is given by the following formula:

[0099] ,in is the residual sequence, i is the sample point number, k is the test set number, y k (i) is the calibration test value of the model output variable, is the predicted value of the model;

[0100] The evaluation indicators are as follows:

[0101] , where De is the evaluation index, num is the number of test sets, is the average of the calibrated test values ​​of the sample point output variable, and n is the number of sample points.

[0102] The value range of the evaluation index is [0, num], and the higher the value of the evaluation index is, the better the model fitting effect is. In the present invention, the second set value is 0.99 num.

[0103] In this embodiment, the number of samples in the test set is 3, so the second set value can be set to 2.97.

[0104] like Figure 6The figure shows a schematic diagram of the test of the head and neck D-plane angle prediction model constructed in this embodiment. The actual curve in the figure is the head and neck D-plane angle curve obtained by the calibration test, the TabNet fitting is the curve obtained by the prediction model constructed in this embodiment, and the average curve is the curve obtained by the average method mentioned in the background technology of the present invention. It can be seen that the curve predicted by the model constructed by the method of the present invention is much more consistent with the actual curve than the average method. Figure 8 It can also be seen that, compared with the average method, the model residual of this embodiment fluctuates slightly around 0, and the consistency is high.

[0105] S140. Using the initial features and the data on the change of the head and neck D-plane angle over time as input, an occipital moment prediction model is constructed based on the PSO-TabNet neural network.

[0106] The specific implementation method of constructing the occipital moment prediction model based on the PSO-TabNet neural network is similar to the construction method of the head and neck D-plane angle prediction model in S130, except that the model input and output are different, which has been mentioned before and will not be repeated here.

[0107] like Figure 7 The figure shows a test diagram of the occipital moment prediction model constructed in this embodiment. The actual curve in the figure is the occipital moment curve obtained by the calibration test, the TabNet fitting is the curve obtained by the prediction model constructed in this embodiment, and the average curve is the curve obtained by the average method mentioned in the background technology of the present invention. It can be seen that the curve predicted by the model constructed by the method of the present invention is much more consistent with the actual curve than the average method. Fig. 9 It can also be seen that, compared with the average method, the model residual of this embodiment fluctuates around 0, with a smaller fluctuation range and higher consistency.

[0108] It should be noted that this embodiment only shows the construction order of a head and neck D-plane angle prediction model and a neural network to construct an occipital moment prediction model, that is, S130 is executed before S140. In fact, the head and neck D-plane angle prediction model can also be executed after or at the same time as the neural network constructs the occipital moment prediction model.

[0109] The method for constructing a car collision dummy neck calibration model provided in the present embodiment first obtains multiple groups of car collision dummy neck calibration test data, then constructs initial features according to initial condition data in the neck calibration test data, the initial features including a first velocity, a first acceleration feature and a second acceleration feature, finally constructs a head and neck D-plane angle prediction model based on a PSO-TabNet neural network with the initial features, time features and occipital moment as inputs, and constructs an occipital moment prediction model based on a PSO-TabNet neural network with the initial features, time features and head and neck D-plane angles as inputs.

[0110] This embodiment uses the PSO-TabNet neural network model to perform regression fitting on the performance of the collision dummy neck calibration test, establishes a high-precision dummy performance model, and solves the problem of insufficient modeling accuracy of the existing averaging method.

[0111] This embodiment is aimed at the collision dummy neck calibration test, divides the test data into initial condition data and test performance data (other data except the initial condition data), constructs initial features according to the initial condition data, and incorporates the initial features into the neural network input, so that the neural network model can better learn the performance characteristics of the collision dummy's neck, and solves the problem of insufficient modeling accuracy caused by the existing regression model only considering the experimental performance data and ignoring the influence of the initial conditions on the calibration test results.

[0112] Furthermore, this embodiment uses the Loss loss function as the objective function of PSO parameter optimization. This exponential evaluation index increases the weight of the outlier fitting accuracy during hyperparameter optimization, making the dynamic performance of the model output better, and thus making the model output better fit the main performance indicators of the neck calibration test such as maximum value, maximum value corresponding time, etc.

[0113] Example 2

[0114] like Fig.10 As shown, this embodiment provides a device for constructing a car collision dummy neck calibration model, comprising:

[0115] The acquisition module 201 is used to acquire multiple sets of automobile collision dummy neck calibration test data; the neck calibration test data includes pendulum acceleration, pendulum velocity, head and neck D-plane angle, occipital moment and time characteristics;

[0116] The initial feature construction module 202 is used to construct the initial feature according to the initial condition data in the neck calibration test data; wherein the initial condition data includes time, a first speed when the pendulum hits the honeycomb aluminum, and a first acceleration of the pendulum after the impact occurs; the initial feature includes a first speed feature, a first acceleration feature, and a second acceleration feature; the first speed feature is the change of the speed of the pendulum after it hits the honeycomb aluminum with the time, the first acceleration feature is the time difference between the two inflection points of the first acceleration curve, and the second acceleration feature is the average acceleration between the two inflection points;

[0117] The prediction model construction module 203 is used to construct a head and neck D-plane angle prediction model based on the PSO-TabNet neural network with the initial features, time features and occipital moment as inputs; and to construct an occipital moment prediction model based on the PSO-TabNet neural network with the initial features, time features and head and neck D-plane angles as inputs.

[0118] The device is used to execute the above method, and thus has at least functional modules and beneficial effects corresponding to the above method.

[0119] Example 3

[0120] like Fig.11 As shown, this embodiment provides an electronic device, including:

[0121] at least one processor; and

[0122] a memory communicatively connected to at least one of the processors; wherein,

[0123] The memory stores instructions executable by at least one of the processors, and the instructions are executed by at least one of the processors to enable at least one of the processors to perform the above method. At least one processor in the electronic device can perform the above method, and thus has at least the same advantages as the above method.

[0124] Optionally, the electronic device also includes interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the electronic device, including instructions stored in or on a memory to display graphical information of a GUI (Graphical User Interface) on an external input / output device (such as a display device coupled to an interface). In other embodiments, if necessary, multiple processors can be used together with multiple memories, and / or multiple buses can be used together with multiple memories. Similarly, multiple electronic devices can be connected (for example, as a server array, a group of blade servers, or a multi-processor system), and each device provides some necessary operations. Fig.11 A processor 301 is taken as an example.

[0125] The memory 302, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the method for building a car crash dummy neck calibration model in the embodiment of the present invention (for example, the acquisition module, initial feature construction module, and prediction model construction module in the car crash dummy neck calibration model construction device). The processor 301 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 302, that is, realizing the above-mentioned car crash dummy neck calibration model construction method.

[0126] The memory 302 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 302 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 302 may further include a memory remotely arranged relative to the processor 301, and these remote memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0127] The electronic device may further include: an input device 303 and an output device 304. The processor 301, the memory 302, the input device 303 and the output device 304 may be connected via a bus or other means. Fig.11 The example of connecting through bus is taken in the following.

[0128] The input device 303 can receive input digital or character information, and the output device 304 can include a display device, an auxiliary lighting device (e.g., LED), and a tactile feedback device (e.g., a vibration motor), etc. The display device can include, but is not limited to, a liquid crystal display (LCD), a light emitting diode (LED) display, and a plasma display. In some embodiments, the display device can be a touch screen.

[0129] Example 4

[0130] This embodiment provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to enable the computer to execute the above method. The computer instructions on the computer-readable storage medium are used to enable the computer to execute the above method, and thus have at least the same advantages as the above method.

[0131] The medium in the present invention may adopt any combination of one or more computer-readable media. The medium may be a computer-readable signal medium or a computer-readable storage medium. The medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples of the medium (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, the medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device, or device.

[0132] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0133] The program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, RF (Radio Frequency), etc., or any suitable combination of the above.

[0134] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0135] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution disclosed in this application can be achieved, and this document is not limited here.

[0136] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A method for constructing a car collision dummy neck calibration model, characterized in that: include: Obtain multiple sets of car crash dummy neck calibration test data; The neck calibration test data includes pendulum acceleration, pendulum velocity, head and neck D-plane angle, occipital moment and time characteristics; According to the initial condition data in the neck calibration test data, an initial feature is constructed; wherein the initial condition data includes time, a first speed when the pendulum hits the honeycomb aluminum, and a first acceleration of the pendulum after the impact occurs; the initial feature includes a first speed feature, a first acceleration feature, and a second acceleration feature; the first speed feature is the change of the speed of the pendulum after it hits the honeycomb aluminum with the time, the first acceleration feature is the time difference between two inflection points of the first acceleration curve, and the second acceleration feature is the average acceleration between the two inflection points; Taking the initial features, time features and occipital moment as input, a head and neck D-plane angle prediction model is constructed based on the PSO-TabNet neural network; Taking the initial features, time features and head and neck D-plane angle as input, an occipital moment prediction model is constructed based on the PSO-TabNet neural network.

2. The method for constructing a car collision dummy neck calibration model according to claim 1, characterized in that: The first acceleration feature is constructed as follows: ;in is the first acceleration feature, is the time corresponding to the inflection point of the rising section of the first acceleration curve; is the time corresponding to the inflection point of the descending section of the first acceleration curve.

3. The method for constructing a car collision dummy neck calibration model according to claim 2, characterized in that: The second acceleration characteristic is constructed as follows: ;in is the second acceleration characteristic, For the first acceleration curve The acceleration corresponding to the moment.

4. The method for constructing a car collision dummy neck calibration model according to claim 1, characterized in that: The head and neck D-plane angle prediction model based on the PSO-TabNet neural network includes: The training set data is fed into the TabNet model for fitting to obtain the initial model of the head and neck D-plane angle; Use the initial model to predict the validation set data; Use loss function to evaluate the fitting effect; According to the fitting effect, the PSO algorithm is used to optimize the initial model hyperparameters; When the number of iterations reaches the first set value, the model with the lowest loss function is selected as the head and neck D-plane angle prediction model.

5. The method for constructing a car collision dummy neck calibration model according to claim 4, characterized in that: The loss function is: Loss= ,in is the model output, vector is the normalized model input feature, y(i) is the calibrated test value of the model output variable, and n is the number of sample points.

6. The method for constructing a car collision dummy neck calibration model according to claim 1, characterized in that: After the head and neck D-plane angle prediction model is constructed based on the PSO-TabNet neural network, the method further includes: The test set data is sent to the head and neck D-plane angle prediction model to calculate the fitting residual; Determining an evaluation index according to the fitting residual; If the evaluation index is greater than or equal to the second set value, the head and neck D-plane angle prediction model is successfully constructed; If the evaluation index is less than the second set value, the step of constructing a head and neck D-plane angle prediction model based on the PSO-TabNet neural network is continued until the evaluation index is greater than or equal to the second set value.

7. The method for constructing a car collision dummy neck calibration model according to claim 6, characterized in that: The fit residual is given by: ,in is the residual sequence, i is the sample point number, k is the test set number, y k (i) is the calibration test value of the model output variable, is the predicted value of the model; The evaluation indicators are as follows: , where De is the evaluation index, num is the number of test sets, is the average of the calibrated test values ​​of the sample point output variable, and n is the number of sample points.

8. A device for constructing a car collision dummy neck calibration model, characterized in that: include: An acquisition module is used to acquire multiple sets of automobile collision dummy neck calibration test data; The neck calibration test data includes pendulum acceleration, pendulum velocity, head and neck D-plane angle, occipital moment and time characteristics; An initial feature construction module is used to construct an initial feature according to the initial condition data in the neck calibration test data; wherein the initial condition data includes time, a first speed when the pendulum hits the honeycomb aluminum, and a first acceleration of the pendulum after the impact occurs; the initial feature includes a first speed feature, a first acceleration feature, and a second acceleration feature; the first speed feature is the change of the speed of the pendulum after it hits the honeycomb aluminum with the time, the first acceleration feature is the time difference between two inflection points of the first acceleration curve, and the second acceleration feature is the average acceleration between the two inflection points; The prediction model construction module is used to construct a head and neck D-plane angle prediction model based on the PSO-TabNet neural network with the initial features, time features and occipital moment as inputs; and to construct an occipital moment prediction model based on the PSO-TabNet neural network with the initial features, time features and head and neck D-plane angles as inputs.

9. An electronic device, characterized in that: include: at least one processor, and a memory communicatively coupled to at least one of the processors; The memory stores instructions executable by at least one of the processors, and the instructions are executed by at least one of the processors so that the at least one processor can execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method according to any one of claims 1 to 7.