Automated Leg Placement Method for Collision Simulation Dummy

Abstract

This invention relates to the field of automotive safety testing technology and discloses an automated method for placing the legs of a collision simulation dummy. The method includes the following steps: The testing team, combining experience and habits of manual placement, programs the dummy leg placement process in CAE preprocessing software. Through secondary development programming, an execution program is obtained to automatically perform the translation, rotation, and joint adjustments of the dummy legs. The input and output of the execution program are defined: the input is the adjustment parameters required for dummy leg placement, and the output is the coordinates of the dummy leg feature points. The output of the execution program is correlated with the target seating posture provided by the client. Using the target seating posture as the optimization target, the input of the execution program is iteratively optimized to obtain the optimal output. This automated method for placing the legs of a collision simulation dummy minimizes the randomness and variability caused by manual placement of the dummy legs, improving the standardization, accuracy, and consistency of dummy placement.

Description

Technical Field

[0001] This invention relates to the field of automotive safety testing technology, specifically to an automated method for placing the legs of a collision simulation dummy. Background Technology

[0002] In automotive crash safety simulation analysis, the placement of the dummy's legs is a crucial step that often significantly impacts the simulation results. In practice, due to differences in CAE solvers and dummy versions used by project participants, as well as CAE model confidentiality considerations, communication regarding dummy leg placement is typically achieved through the coordinates of feature points on the dummy's legs, such as the ankle, knee, and H-point.

[0003] A common scenario is when a client needs to provide the dummy's leg pose to a testing team. They typically provide the feature point coordinates of the dummy's legs (which could be a CAD dummy or a multi-rigid-body simulation dummy). The testing team then positions the legs according to their own simulation dummy model, matching the input conditions. Often, due to differences in the dummy models used, the dimensions of various parts of the dummy's legs will vary. If the dummy's legs are placed manually, repeated adjustments are frequent, which is time-consuming and may not be optimal. Furthermore, different engineers may produce different results. Therefore, this manual method of dummy leg placement inevitably leads to randomness and variability.

[0004] Current manual adjustment methods for dummy leg positions are easily affected by human factors, inevitably causing discrepancies and difficulties for those involved in vehicle development. To ensure the simulation and testing are as close as possible, the dummy's leg placement should be precise and consistent during both testing and simulation.

[0005] However, many factors influence the dummy's leg score, and these factors are interconnected and coupled. Therefore, in actual development, a certain degree of randomness is often found in the dummy's leg score. Even with similar vehicle conditions and under the same collision conditions, the dummy's leg score may differ. One important reason for this phenomenon is the randomness of the dummy's leg placement. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of the above-mentioned technologies by providing an automated method for placing the legs of a collision simulation dummy, thereby minimizing the randomness and variability caused by manual placement of the dummy's legs and improving the standardization, accuracy, and consistency of dummy placement.

[0007] To achieve the above objectives, the automated leg placement method for a collision simulation dummy according to the present invention includes the following steps:

[0008] A) The experimenters combined their experience and habits in manual placement with the programmatic placement of the dummy's legs in the CAE preprocessing software. Through secondary development programming, they obtained the execution program to automatically execute the translation, rotation, and joint adjustment of the dummy's legs.

[0009] B) Define the input and output of the execution program obtained in step A), where the input is the adjustment parameters required for the placement of the dummy's legs, and the output is the coordinates of the feature points of the dummy's legs;

[0010] C) Associate the output of the execution program with the target sitting posture provided by the client, and use the target sitting posture as the optimization target to iteratively optimize the input of the execution program to obtain the best output, namely the best coordinates of the dummy's leg feature points.

[0011] Preferably, in step A), the executable program is edited and modified to delete graphical operation commands and irrelevant operation commands.

[0012] Preferably, in step C), the output of the execution program is correlated with the target sitting posture provided by the demander using a root mean square error function to evaluate system deviation and serve as an optimization target.

[0013] Preferably, in step C), during iterative optimization, the optimization algorithm used to solve the parameter identification problem is based on the surrogate model and adopts a sequential optimization strategy with domain reduction.

[0014] Preferably, the surrogate model can be support vector regression, radial basis function network, kriging, feedforward neural network, and polynomial response surface.

[0015] Compared with the prior art, the present invention has the following advantages: it minimizes the randomness and variability caused by the manual placement of the mannequin legs, and improves the standardization, accuracy and consistency of the placement of the mannequins. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0017] An automated method for positioning the legs of a collision simulation dummy includes the following steps:

[0018] A) The experimenters combined their experience and habits in manual placement with the programmatic placement of the dummy's legs in the CAE preprocessing software. Through secondary development programming, they obtained the execution program to automatically execute the translation, rotation, and joint adjustment of the dummy's legs.

[0019] B) Define the input and output of the execution program obtained in step A). ​​The input is the adjustment parameters required for the placement of the dummy's legs, and the output is the coordinates of the feature points of the dummy's legs.

[0020] C) Associate the output of the execution program with the target sitting posture provided by the client, and use the target sitting posture as the optimization target to iteratively optimize the input of the execution program to obtain the best output, namely the best coordinates of the dummy's leg feature points.

[0021] In step A), the executable program is edited and modified, deleting graphical and irrelevant operation commands. In step C), the output of the executable program is correlated with the target seating posture provided by the client using the root mean square error function to evaluate system bias and use it as the optimization objective. In step C), during iterative optimization, the optimization algorithm used to solve the parameter identification problem is based on surrogate model optimization and employs a sequential optimization strategy with domain reduction. The surrogate model can be support vector regression, radial basis function network, kriging, feedforward neural network, or multinomial response surface.

[0022] Taking the H III 50% Lower Body mannequin developed by LS_Dyna as an example, the preprocessing software used for adjusting the mannequin's sitting posture is LS_PrePost. The following describes how to achieve automatic leg placement in LS_PrePost.

[0023] First, record a program in LS-PrePost that manually positions the dummy's legs. These graphical user interface (GUI) interactive programs are written to a file named lspost.cfile. After exiting LS-PrePost, this file can be renamed, for example, to lspost_dummypose.cfile.

[0024] Then, the generated executable program is edited and modified, mainly by deleting graphical operation commands and irrelevant operation commands, and adding comments to make the program code more readable and the execution more concise and faster.

[0025] Next, the values ​​for adjusting the dummy's legs in the execution program are parameterized, and a separate command code file is created to manage these adjustment parameters. Finally, the batch processing command ("LS-PrePost c=commandfile-nographics") can be used to automatically position the dummy's legs according to the input adjustment parameters without opening the LS-PrePost GUI interface, and export the adjusted dummy model K file.

[0026] Next, iterative optimization is performed in the software. First, in the Stage window, on the Setup page, the above-mentioned program for automatically positioning the dummy's legs is used as the program for identifying system parameters. Then, in the Problem global setup window, the adjustment parameters for leg positioning are used as inputs. These parameters can change the dummy's leg positioning posture. Next, in the Stage window, on the Responses page, the coordinates of the dummy's leg feature points after the program execution are used as outputs. Finally, in the Composite window, the root mean square error (Sqrt MSE) function is used to correlate the output with the target sitting posture to evaluate system bias and use it as the optimization target.

[0027] In this embodiment, the optimization algorithm for solving the parameter identification problem is based on a surrogate model and employs a sequential optimization strategy with domain reduction. This embodiment sets up 5 iterations in the optimization software LS-OPT, with each iteration using 14 samples (i.e., the dummy leg placement program is executed 14 times in the background). The surrogate model constructed from the sample points calculates the minimum systematic bias of the current iteration. Then, it determines whether to terminate convergence. If convergence fails, it continues iterating and modifying the parameters to minimize the systematic bias value. By minimizing the systematic bias value, the dummy leg adjustment parameters are calibrated to the desired result, thus obtaining the target dummy leg placement posture.

[0028] This embodiment also compares the convergence process and time consumed by different surrogate models (using an 8-core InterXeon CPU E5-2620). Five surrogate models were selected: Support Vector Regression (SVR), Radial Basis Function Network (RBFN), Kriging, Feedforward Neural Network (FFNN), and Polynomial Response Surface (POLY).

[0029] In terms of convergence speed, except for the first-order polynomial response surface surrogate model, the other four surrogate models can achieve good dummy placement after three iterations. Regarding convergence accuracy after the fifth iteration, the Kriging surrogate model has the highest accuracy, followed by feedforward neural networks, first-order polynomial response surfaces, support vector regression, and radial basis function networks. Of course, with more iterations, the other surrogate models will also achieve similarly high accuracy. In terms of time consumption, as shown in the table below, the radial basis function networks and first-order polynomial response surfaces require less time, mainly because their surrogate models are less expensive to build than the other surrogate models.

[0030] Proxy Model Time (seconds) SVR 492 RBFN 394 Kriging 508 FFNN 523 POLY 397

[0031] The present invention provides an automated method for placing the legs of a collision simulation dummy, which minimizes the randomness and variability caused by manual placement of the dummy's legs, and improves the standardization, accuracy, and consistency of dummy placement.

Claims

1. A method for automated placement of crash test dummy legs, the method comprising: The method comprises the following steps: ​ A) The test party combines the experience and habit of manual placement, and processes the placement process of the dummy leg in the CAE pre-processing software to obtain an execution program for automatically performing the translation, rotation and joint adjustment of the dummy leg through secondary development programming; B) Defining the input and output of the execution program obtained in step A), the input being the adjustment parameters required for the placement of the dummy leg, and the output being the characteristic point coordinates of the dummy leg; C) Using a root mean square error function to associate the output of the execution program with the target placement posture provided by the demand party to evaluate the system deviation, taking the target placement posture as the optimization target, and iteratively optimizing the input of the execution program, wherein the optimization algorithm for solving the parameter identification problem is based on the optimization of the proxy model, a sequential optimization strategy with domain reduction is adopted, and the best output, i.e., the best characteristic point coordinates of the dummy leg, is obtained.

2. The automated placement method of a crash simulation dummy leg according to claim 1, wherein: In step A), the execution program is edited and modified to delete the graphic operation commands and irrelevant operation commands therein.

3. The automated leg placement method for crash test dummies according to claim 1, wherein: The proxy model is one of support vector regression, radial basis function network, Kriging, feedforward neural network and polynomial response surface.