CROSS-REFERENCE TO RELATED APPLICATION

[0001] This PCT International Patent Application claims the benefit of U.S. Provisional Patent Application Serial No. 63/309,114, filed on February 11, 2022 and titled “Statistical Method For Substructure Crash Simulation Model,” the entire disclosure of which is hereby incorporated by reference.

FIELD

[0002] The present disclosure relates generally to finite element analysis of a vehicle model. More specifically, the present disclosure relates to using a portion of a full vehicle model for a crash analysis.

BACKGROUND

[0003] Crashworthiness simulation is critical to verify vehicle safety, and it is directly involved in the development process in evaluating crashworthiness performance of a vehicle. The subsystem designs, like A pillar, B pillar, doors, and cradles, may be required to be verified for crashworthiness on a full vehicle crash setting, while subsystem design iterations are not on a full vehicle environment.

[0004] There are existing mathematical crash modelling methods to help in bridging the subsystem design iteration and full vehicle verification requirement, like reduced-order dynamic models, multi-body models, nonlinear finite element models, response surface models, crash pulse models, and hybrid models. Among these strategies, finite element model is the most informative one, and explicit nonlinear finite element analysis is the most widely acceptable modelling procedure to be shared between original equipment manufacturers (OEM) and suppliers in the automotive industry. Finite element models and settings could be directly transferred between them. A full vehicle model is expensive and time consuming to run. Computer-aided engineering (CAE) engineers seek a reduced finite element model method in the crash event by simplifying the element formulation or use a smaller portion of the full finite element model in the analysis.

(0005] Parametrizing the finite element model using a simplified element formulation gives a good result in the vehicle system-level performance, but it may require tuning element parameters, and the result may not be as informative as two-dimensional (2D) and three- dimensional (3D) elements formulated models. It is also common to truncate the finite element model, then connected by lumped mass, or to substructure the full vehicle model with interface connections. For the truncated model with lumped mass, it would add extra acceleration at the lumped mass connection interface in an explicit solver. The model size must be defined larger than necessary to overcome the extra acceleration. For the substructure vehicle model, LS-DYNA provides the user with *INTERFACE keywords by saving and reusing displacement and velocity time histories. Either truncation or substructure finite element models encounter the same barrier: there is no well-defined method to guide the truncation/substructure procedure. The scope of the reduced model definition and link of the performance from the subsystem simulation back to the full vehicle may rely on an engineers’ subjective judgment without an objective check.

SUMMARY

[0006] The present disclosure provides a method for crash simulation of a vehicle structure. The method includes: simulating a crash using a full model of the vehicle structure; identifying, based on the simulated crash of the full model, a critical crash area of the vehicle structure; generating, using the critical crash area, a substructure model of the vehicle structure smaller than the full model of the vehicle structure; and simulating the crash using the substructure model of the vehicle structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Further details, features and advantages of designs of the invention result from the following description of embodiment examples in reference to the associated drawings.

(0008] FIG. 1 shows a top-down view of a LS-DYNA setup for an Insurance Institute for Highway Safety (IIHS) lateral impact test;

(0009] FIGS. 2A-2B show graphs of Internal Energy Density (IED) distribution on the vehicle model and zoomed view by auto bins method;

(00101 FIG. 3 A shows a graph of IED distribution on the vehicle model by log;

(0011 ] FIG. 3B shows a graph of IED distribution by an auto bins method;

[0012] FIG. 4 shows a schematic block diagram of a controller;

[0013] FIG. 5 shows an Elements output from 70% confidence interval from high-density- value distribution (distribution 2);

[0014] FIG. 6 shows DBSCAN clustering results for elements in the IED distribution on the vehicle model;

[0015] FIGS. 7A-7B show overlap clustering result on the high internal energy elements;

[0016] FIG 8A shows a full vehicle model;

(0017] FIG. 8B shows a substructure model based on a 90% confidence interval;

[0018] FIG. 8C shows a substructure model based on a 80% confidence interval;

[0019] FIG. 8D shows a substructure model based on a 70% confidence interval;

[0020] FIG. 9 shows a 2011 Honda Accord test result from the IIHS lateral impact;

[0021] FIG 10A shows a full vehicle model after the IIHS lateral impact;

[0022] FIG. 10B shows a substructure model after the IIHS lateral impact, based on a 90% confidence interval; [0023] FIG. 10C shows a substructure model after the IIHS lateral impact, based on a 80% confidence interval;

[0024] FIG. 10D shows a substructure model after the IIHS lateral impact, based on a 70% confidence interval;

[0025] FIG. 11 shows a graph with overlapping plots of contact forces for the full vehicle and substructure models;

[0026] FIG. 12 shows a graph with overlapping plots of side intrusions for the full vehicle and substructure models; and

[0027] FIG. 13 shows a flow chart illustrating steps in a method for crash simulation of a vehicle structure.

DETAILED DESCRIPTION

[0028] Referring to the drawings, the present invention will be described in detail in view of following embodiments. The present disclosure provides a method to identify the critical crash zone based on energy distribution and machine learning clustering is proposed to build a confident substructure model. The results from substructure models are compared and evaluated with the full vehicle crash result, and the accuracy is proven.

Substructure Modelling Process

[0029] The full vehicle simulation method is the most straightforward one: All the design changes, no matter how small they are, would be performed on the full vehicle crash simulation. While a full vehicle simulation method requires the largest simulation time. The substructure method takes more steps but less simulation time with specified reliability: It is beneficial for the tier-one supplier who performs multiple subsystem design updates and optimizations, and these changes are required to be verified in the full vehicle environment. The identified substructure model by Step 1 from Table 1 would be reused in all the following design updates. The substructure model establishment steps are listed as follows, and details would be elaborated.

•• Step 1. Run the full vehicle model crash in LS-DYNA explicit solver;

•• Step 2. Identify the critical crash area based on energy absorption distribution;

•• Step 3. Substructure the full vehicle model, using ANSA preprocessor for FEA simulation from BETA CAE Systems; and

•• Step 4. Design iterations based on the substructure model from the previous step.

TABLE 1 Different finite element analysis methods for vehicle crash simulation in design iterations.

Run the Full Vehicle Crash Model

[0030] The full crash vehicle model used in the present disclosure is a 2011 Honda Accord, which is used as a standard lightweight vehicle (LWV 1.2). The finite element vehicle model is taken as it is. Physical test result used for comparison comes from the National Highway Traffic Safety Administration (NHTSA) open source data. The Insurance Institute for Highway Safety (IIHS) lateral moving deformable barrier (MDB) comes from LS-DYNA database, as shown in FIG. 1.

[0031] As shown in FIG. 1, a subject vehicle 20 is impacted by a MDB 22 that simulates a collision from another vehicle on a side of the subject vehicle 20 and at a reference distance d _r that of 753 millimeters behind a front axle 21 of the subject vehicle 20. In an actual test, the actual location of the impact may be different from the reference distance d _r . For example, the impact may be 759 mm behind the front axle 21 of the subject vehicle 20.

Identify the Critical Crash Area

[0032] In crash events, the most critical performance for crashworthiness is the energy absorption ability of the components: parts with different materials are deformed on a designed load path to dissipate the impact energy and protect the occupancy by means of controlled mechanisms and modes. During energy absorption, the materials need to be able to maintain a gradual deformation in the load profile. The energy absorption by the deformation is defined by the internal energy, which is the surface below the load-displacement curve. This 2011 Honda Accord is mainly composed of a 2D shell element. To normalize the internal energy on the 2D surface, internal energy density (TED) on 2D shell elements is used to identify the critical crash area for the crashworthiness. Requesting and loading plotting data from LS-DYNA binout files, shell elements from the vehicle side (excluding the MDB) are plotted for the IED distribution by the open-source Python package from LASSO GmbH. The frequency magnitude is largely sparse in different frequencies from 0 to 30,000 Hz; see FIGS. 2A-2B.

[0033] The details to identify the critical crash area based on IED distribution is challenging. It would be explained in detail below, under the heading “Identify the critical crash area based on energy absorption.” Substructure the Full Vehicle Model

[0034] LS-DYNA provides INTERFACE keywords to connect the substructure and original model. *INTERFACE_COMPONENT_OPTION1_{OPTION2} and

*INTERFACE_LINKING_NODE_OPTION are combined to create an interface for use in subsequent linking calculations. *1NTERFACE keyword provides the definition for surfaces, nodal lines, and nodal points where displacement and velocity time histories are saved at some specified frequency. The calculation is defined as follows: where Uconstrained is the nodes set constrained to follow the displacement of the interface uunked is the set of nodes followed by the constrained nodes set, ci is the displacement of the NID node in the linking file, ci is the displacement of node LNID, Q\ is the rotation into the local coordinates of the linking file, Qi is the rotation into the local coordinate system, // / is the scale factors FX, FY, and FZ. They act on the nodal displacements in the global coordinate system of the constrained calculation.

Design Iterations on the Substructure Model

[0035] After the substructure model is established, the same substructure model could be reused in the following subsystem design evaluation process to save modelling effort and simulation cost. The CAE engineer is granted more local detailed modelling capability on the subsystem level, including, but not limited to mesh refinement, material study, feature change, and optimization with a much lower computational cost compared with a full vehicle simulation. Identify the Critical Crash Area Based on Energy Absorption

[0036| The ZED plot with large frequency discrepancy from FIGS. 2A-2B is transferred to the log domain to reduce skewness with large values.

[0037| The plot bins number for FIG. 3 A is initially and arbitrarily assigned as eight (8), and it shows a skew normal distribution. To get the sufficient plot bin number, auto bin method is used to capture the distribution details.

[0038] From FIG. 3B distribution, a larger bin number (more than eight) is generated. Instead of a skew normal distribution in FIG. 3 A, FIG. 3B shows a blended combination of two distributions 24, 26 including a low-density-value distribution 24 and a high-density-value distribution 26. The high-density-value distribution 26 is the critical crash area around the impact region where the MDB 22 hit the subject vehicle 20. The Gaussian mixture method is used to decouple the two distributions. Using sklearn. mixture Python package, set the number of mixture components to be 2 and covariance type to be “FULL” each component using its own general covariance matrix. It provides the mean value as -7.8 for the low-density-value distribution 24 and 0.1 for the high-density -value distribution 26. The high-density -value distribution 26 may be assumed to be a normal distribution. Based on normal distribution mean and standard deviation value from the high-density-value distribution 26, elements from 70%, 80%, and 90% confidence levels are output for one side normal distribution.

[0039] FIG. 4 shows a schematic block diagram of a controller 50. The controller 50 includes a processor 52 coupled to a storage memory 54. The storage memory 54 stores instructions, such as program code for execution by the processor 52, in an instruction storage 56. The storage memory 54 also includes data storage 58 for holding data to be used by the processor 52. The data storage 58 may record, for example, the covariance matrices and/or the outcome of functions calculated by the processor 52.

[0040] FIG. 5 shows a shaded crash area region 28 indicating elements from distribution by 70% one side confidence interval around the front and rear doors with A, B, and C pillars included. This is the location where the MDB 22 hits the subject vehicle 20 from the side. The output could not be used directly as there are also noise elements far away from the door area. These elements need to be filtered out from the major crash area at the side of the subject vehicle 20.

[0041 ] From the DYNA simulation setup, there are no straightforward ways to prevent the noise elements. Density-based spatial clustering of applications with noise (DBSCAN) is a data clustering algorithm. To eliminate the noise elements, the skleam. cluster Python package is used, and the element position information is input in the global coordinate system. The model is an average 10 mm mesh size model with a minimal 4 mm and maximum 20 mm element size, then set s (maximum distance between two samples for one to be considered as in the neighborhood of the other) as 0.2. From FIG. 6, the DBSCAN algorithm successfully clusters the crash area in the dark colored group, labeled “CRASH AREA”.

[0042] The DBSCAN clustering result, shown as the shaded crash area region 28 on FIG. 6, is input to the element with a specific confidence interval from FIG. 5. The blue area from FIG.

7 shows the critical crash zone after filtering out the noise. For the outer panel view (FIG. 7A), the door panel for the impact area is selected. From the inner panel view (Fig. 7B), the floor extrusions for the impact load path are selected including A, B, and C pillars. These are important components in the side impact. Then other-box selection by software ANSA BETA CAE is used to create the 70%-90% confidence interval substructure models (FIGS. 8A-8D). FIG. 8A shows a fill vehicle model 30 representing an entirety of the subject vehicle 20. FIG. 8B shows a first substructure model 32 representing a substructure of the subject vehicle 20 corresponding to a confidence interval of 90%. FIG. 8C shows a second substructure model 34 representing a substructure of the subject vehicle 20 corresponding to a confidence interval of 80%. FIG. 8D third a fourth substructure model 36 representing a substructure of the subject vehicle 20 corresponding to a confidence interval of 70%.

Identify Simulation Result Comparison and Evaluation

[0043] The simulation result from the substructure model is compared to a full vehicle finite element model and physical test result with respect to after-crash deformation, contact force, and intrusion curves.

[0044] The test and simulation results show similar deformation after side impact; see FIGS. 9 and 10A-10D. FIG. 10A shows the full vehicle model 30 after the side impact. FIG. 10B shows the first substructure model 32 after the side impact. FIG. 10C shows the second substructure model 34 after the side impact. FIG. 10D shows the third substructure model 36 after the side impact.

[0045] The contact force between the MDB 22 and subject vehicle 20 is plotted in FIG.

11. FIG. 11 includes a first plot 50 for the full the full vehicle model 30, a second plot 52 for the first substructure model 32, a third plot 54 for the second substructure model 34, and a fourth plot 56 for the third substructure model 36. As shown, curves from the full vehicle model 30 and the various different substructure models 32, 34, 36 are overlapped with high correlation, over 0.99 by the Pearson correlation.

[0046] According to the contact force curves, the first maj or structure deformation happens between 0.00 second and 0.025 second. The force increases almost linearly to 200 kN, then the MDB and vehicle separate after the force peak, and then the contact force drops to zero between 0.1 second and 0.125 second.

[0047] The test data is extracted from the NHTSA test report picture. The number of simulation output data interval frequencies does not match the data extracted from the picture. To compare unsynchronized simulation data with respect to the test data, dynamic time warping is used to measure the similarity between test and full/sub structure simulation sequences (FIG. 12). FIG. 12 includes a first plot 60 for the full the full vehicle model 30, a second plot 62 for the first substructure model 32, a third plot 64 for the second substructure model 34, and a fourth plot 66 for the third substructure model 36. FIG. 12 shows, by the intrusion curves, the various different substructure models 32, 34, 36 each having a same similarity toward the test result, compared with the full vehicle model 30.

[0048] The simulation is performed on an Intel-MPI 2018 Xeon64 on LS-DYNA MPP s R11.0.0. In Table 2, the 2D number after IED distribution refers to FIG. 5. The higher the confidence interval is used, the more elements are included, from 158,000 for 70% confidence interval to 308,000 for 90% confidence interval. The 2D element number in simulation refers to FIGS. 8A-8D, the element number increases with the confidence interval increases, from 900,000 for 70% confidence interval to 943,000 for 90% confidence interval. The number of elements in simulation is much larger than the number identified is due to lacking the precise preprocessing tool to substructure the full vehicle model exactly following the clustering result boundary. In this situation, the simulation time is reduced by around 60% compared with the full vehicle simulation, from 3 hours 27 minutes to one and half hour. With respect to the Pearson correlation coefficient of the contact force between the MDB and vehicle, it is close to 1.0. With respect to the similarity between full vehicle and substructure vehicle intrusion number, it shows a similar range around

160.

TABLE 2 - Simulation time and accuracy comparison.

[0049] A method 100 of crash simulation of a vehicle structure is shown in the flow chart of FIG. 13. Some or all parts of the method 100 can be performed by the controller 50, in accordance with some embodiments of the present disclosure. As can be appreciated in light of the disclosure, the order of operation within the method is not limited to the sequential execution as illustrated in FIG. 13, but may be performed in one or more varying orders as applicable and in accordance with the present disclosure. [0050] The method 100 includes simulating a crash using a full model of the vehicle structure, at step 102. For example, the processor 52 may load and execute program instructions from the instruction storage 56 to perform step 102. In some embodiments, a finite element analysis simulation program, such as LS-DYNA may be used to simulate the crash at step 102. In some embodiments, simulating the crash using the finite element analysis simulation program may include using an explicit solver.

[0051 ] In some embodiments, step 102 may include using an Automatic Net-generation for Structural Analysis (ANSA) preprocessor, such as an ANSA preprocessor for FEA simulation program from BETA CAE Systems, to generate the substructure model of the vehicle structure.

[0052] The method 100 also includes identifying, based on the simulated crash of the full model, a critical crash area of the vehicle structure at step 104. For example, the processor 52 may load and execute program instructions from the instruction storage 56 to perform step 104.

[0053] In some embodiments, step 102 further includes determining an energy absorption distribution, and step 104 includes identifying the critical crash area of the vehicle structure based on the energy absorption distribution.

[0054] The method 100 also includes generating, using the critical crash area, a substructure model of the vehicle structure smaller than the full model of the vehicle structure at step 106. For example, the processor 52 may load and execute program instructions from the instruction storage 56 to perform step 106.

[0055] In some embodiments, step 102 further includes determining a clustering of the crash area using a data clustering algorithm, and step 106 includes using the clustering of the crash area to generate the substructure model. For example, the data clustering algorithm may include a Density-based spatial clustering of applications with noise (DBSCAN) algorithm. [0056] In some embodiments, generating the substructure model of the vehicle structure at step 106 includes computing a number of elements in the substructure model to cause the substructure model to have a predetermined confidence interval. For example, the predetermined confidence interval may be at least 70%. In another example, the predetermined confidence interval may be at least 80%. In another example, the predetermined confidence interval may be at least 90%.

[0057] The method 100 also includes simulating a crash using the substructure model of the vehicle structure at step 108. For example, the processor 52 may load and execute program instructions from the instruction storage 56 to perform step 108.

[0058] The method 100 also includes generating a revised substructure model of the vehicle structure based on a result of the simulation using the substructure model at step 110. For example, the processor 52 may load and execute program instructions from the instruction storage 56 to perform step 110.

[0059] The method 100 also includes simulating the crash using the revised substructure model of the vehicle structure at step 112. The simulation at step 112 may be performed similarly or identically to step 102, except using the revised substructure model. For example, the processor 52 may load and execute program instructions from the instruction storage 56 to perform step 112.

Conclusion

[0060] Using the method of the present disclosure, the substructure model shows the same level of accuracy compared with the full vehicle model when visually inspecting the after-impact deformation, comparing contact force between the MDB 22 and the subject vehicle 20, comparing intrusion curves from the door side. And the substructure model costs around 40% of the full vehicle simulation time. In the design change verification application for crash-related events, it could be performed on the 70% confidence interval substructure model. It provides the CAE engineer with less simulation cost and more chances of design iteration. With a more precise substructure tool, the substructure model could further refine the size.

(0061] The system, methods and/or processes described above, and steps thereof, may be realized in hardware, software or any combination of hardware and software suitable for a particular application. The hardware may include a general purpose computer and/or dedicated computing device or specific computing device or particular aspect or component of a specific computing device. The processes may be realized in one or more microprocessors, microcontrollers, embedded microcontrollers, programmable digital signal processors or other programmable device, along with internal and/or external memory. The processes may also, or alternatively, be embodied in an application specific integrated circuit, a programmable gate array, programmable array logic, or any other device or combination of devices that may be configured to process electronic signals. It will further be appreciated that one or more of the processes may be realized as a computer executable code capable of being executed on a machine readable medium.

|0062| The computer executable code may be created using a structured programming language such as C, an object oriented programming language such as C++, or any other high- level or low-level programming language (including assembly languages, hardware description languages, and database programming languages and technologies) that may be stored, compiled or interpreted to run on one of the above devices as well as heterogeneous combinations of processors processor architectures, or combinations of different hardware and software, or any other machine capable of executing program instructions. [0063] Thus, in one aspect, each method described above and combinations thereof may be embodied in computer executable code that, when executing on one or more computing devices performs the steps thereof. In another aspect, the methods may be embodied in systems that perform the steps thereof, and may be distributed across devices in a number of ways, or all of the functionality may be integrated into a dedicated, standalone device or other hardware. In another aspect, the means for performing the steps associated with the processes described above may include any of the hardware and/or software described above. All such permutations and combinations are intended to fall within the scope of the present disclosure.

[0064] The foregoing description is not intended to be exhaustive or to limit the disclosure. Individual elements or features of a particular embodiment are generally not limited to that particular embodiment, but, where applicable, are interchangeable and can be used in a selected embodiment, even if not specifically shown or described. The same may also be varied in many ways. Such variations are not to be regarded as a departure from the disclosure, and all such modifications are intended to be included within the scope of the disclosure.