Reconstruction finite element model of cars

The experimental method used in a frontal crash of cars costs much time and expense. Therefore, numerical simulation in crashworthiness is widely applied in the world. The completed car models contain a lot of parts which provided complicated structure, especially the rear of car models do not contributetobehavioroffrontalcrashwhichusuallyevaluatesinjuriesofpedestrianormotorcyclist.Inordertosavetimeandresources,asimplificationofthecarmodelsforresearchsimulationsisessentialwiththegoalofreducingapproximately50%ofcarmodelelementsandnodes.Thisstudyaimstoconstructthefiniteelementmodelsoffrontstructuresofvehiclebasedontheoriginalfiniteelementmodels.Thosenewcarmodelsmustbemaintainedimportantvaluessuchasmassandcenterofgravityposition.Byusingconditionboundaries,inertiamomentiskeptunchangedonnewmodel.Theoriginalcarmodels,whichareprovidedbytheNationalCrashAnalysisCenter (NCAC), validated by using results from experimental crash tests. The modified (simplistic) vehicle FE models are validated by comparing simulation results with experimental data and simulation results of the original vehicle finite element models. LS-Dyna software provides convenient tools and very strong to modify finite element model. There are six car models reconstructed in this research, including 1 Pick-up, 2 SUV and 3 Sedan. Because car models were not the main object to evaluate in a crash, energy and behavior of frontal part have the most important role. As a result, six simplified car models gave reasonable outcomes and reduced significantly the number of nodes and elements. Therefore, the simulation time is also reduced a lot. Simplified car models can be applied to the upcoming frontal simulations.


INTRODUCTION
The frontal car crash is one of the most well-known tests in the automotive safety industry and the finite element method (FEM) is also widely used to simulate this kind of test. The simulation, or virtual test, is useful not only in fastening the development process but also in helping to reduce expenditure. In this simulation, the numerical model of a vehicle is given an initial velocity to bump into a constrained solid wall. In the frontal car crash test, only the properties of the frontal part of the car are attractive to researchers as the other parts seem to be unaffected by the impact. Therefore, the rear parts of the vehicle can be removed to reduce the overall number of parts and elements, which then results in less time and hardware resources to run the simulation. The modified model, however, must show consistency with the full model in terms of both kinematics and dynamics. According to a study by Mathias Stein et al. 1 , the cars model was assessed at three different energy levels in the form of pedestrian crashes, low and high energy crashes against obstacles and other vehicles. Therefore, three highly parametric simplified models were established to identify the variables with high impact on the self and partner protection, pedestrian safety and insurance classification tests. Finally, the three models can be merged together into on unified parametric car model. In the research by Mathias Stein, SFE CON-CEPT was used to reduce unimportant details. Moreover, MATLAB was also used to process output files and LS-Dyna was used to solve calculations. In 2005, Y. Liu 2 published a research regarding to developing of simplified model for crashworthiness analysis. This research represented a modified method based on the existing collapse theories but the researcher developed a new collapse theory required to predict the crash behavior for the thin-walled channel section beams. All the theory and modeling method developed in this research are applied for creating simplified models. Both the simplified and detailed models are used for crashworthiness analyses, results show that the errors caused by the simplified models are fewer than 10% and the simplified models only take less than 10% of the computer time of the corresponding detailed models. Another research regarding to modify FE vehicle model of H. Al-Thairy and Y.C. Wang 3 . The main objective of this study is to present and validate a simplified numerical vehicle model that can be used to simulate the effects of vehicle frontal impact on steel columns by using the commercial finite element code ABAQUS/Explicit. The simplified numerical vehicle model treats the vehicle as a springmass system. The proposed model consists of three parts: an undeformable body representing the total vehicle mass; a spring or connector with nonlinear force-deformation relationship to represent the dynamic stiffness of the vehicle; and a rigid but weightless plate to generate the contact between the springmass system and the impacted column. The dynamic load-deformation characteristic of the spring is assumed to be bilinear: the initial linear elastic part simulating the vehicle deformation until it has reached the vehicle engine box, followed by a near rigid relationship. This concept has been validated by comparison against simulation results of steel columns under different impact velocities, axial load ratios, boundary conditions, and slenderness ratios using the fullscale vehicle model and using the proposed simplified spring-mass model. Having validated the proposed model, this study presents the derivations and validations of an equation to predict the equivalent linear stiffness of the vehicle that can be used either in a future numerical simulation model or in an energy based analytical model. Because of the complexity and time consuming of the previous method, this study will present the reduction method using only LS-Dyna software but still ensure relative accuracy with the original model. To achieve that, the modified model must have the same mass and the same position of center of gravity (C.G). The FEM car models built by The National Crash Analysis Center (NCAC) are complicated. Because of their use for engineering analysis, it is not easy to modify the geometry and topology of vehicle structures. The creation of a new FEM model is all based on flexible tools provided by the LS-DYNA software. The geometrical structure is simplified with a significantly reduced number of elements and parts while still creating constraints among the parts and the added mass to ensure accuracy for the new FEM model. The result is a newly created model that solves the problem of time and resource consumption during research simulation. Thus, the purpose of this paper is to develop the FE models of vehicle front structures based on available FE models of a sedan, a pickup, a neon, a Camry, and an SUV.

METHODOLOGY
With a large amount of cost and insufficient facilities, the experimental method in Vietnam is minimal.
Hence, the methodology in this project will be based on numerical methods. Furthermore, there is much software on the market that supports numerical computation. In particular, the finite element method, which is a popular, convenient method that saves a lot of time and money. The software can be mentioned as: ANSYS, ABAQUS, SOLID WORK and LS -DYNA. Among these packages, LS-DYNA is widely used in automobile industry for simulating crash tests, and it provides a large number of dummy models as well as car models that are compatible with LS-DYNA solvers. Therefore, LS-DYNA will be the software used in this project.

Constrained Nodal Rigid Body
Following guideline of FEA Information Inc. Global News & Industry Information 4 , Constrained Nodal Rigid Bodies (CNRB) are treated internally in LS-DYNA like a rigid body part, which uses the MAT_RIGID material model. A set of nodes is defined for each nodal rigid body definition with a minimum number of 2 nodes. Nodal rigid bodies with one node are deleted. The most common usage of the NODAL_RIGID_BODY definition is to model rigid, i.e., non/breakable, connections between structural parts. It is also common practice to model spot welds and others weld types using this definition. The *CONSTRAINED_NODE_SET option in LS-DYNA eliminates all rotational degree-offreedom within the set and should be used cautiously. In this study, the node with added mass is connected to the body by means of Nodal Rigid Body constraints (CNRB). These constraints are also used to hold the rear boundary edge to compensate for their reduction in stiffness.

Finite element car models.
The finite element (FE) models were developed through the process of reverse engineering at the National Crash Analysis Center (NCAC) of The George Washington University (GWU). This paper focuses to 06 FEM car models as shown in Fig. 1 including 01 Pickup model (Chevrolet C2500 5 , 02 SUV car models (Toyota Rav 4 6 , Ford Explorer 7 ) and 03 sedan car models (Yaris 8 , Camry 9 , and Dodge Neon 10 ). Each vehicle model has been verified with the experimental test. The NCAC provides data for each vehicle model including simulation method and experiment method. This data comes with a complete car model. These detailed FE models were constructed to include full functional capabilities of the suspension and steering subsystems, so the FE models are required to have a simplistic method to change up original FE models. In this study, simplistic algorithm is introduced below and it comprises three principal steps: -Deleting unnecessary parts.
-Conserve volume and position of central -Adding boundary conditions. In the following sections, three above steps will be discussed in more detail.

Deleting unnecessary part.
Considering the important part of vehicle in crash test simulation, the frontal structure is kept while the rear parts are unnecessary, so they will be deleted. Results after deletion are shown in Modified models part of Fig. 1.

Conserve volume and position of central.
The mass and the position of central will be change through the deleting process. So, the mass and the position of central need to balance. In other to add extra mass, a node is created and added with extra mass. Furthermore, the modified model's C.G has to the same as the original model's C.G. Therefore, the extra mass is not enough. The coordinates of this node need to be calculated and refined. Finally, a node with added mass is connected to the body by means of Nodal Rigid Body constraints (CNRB). Here, the formula: Where m 1 , m 2 are the mass of original and modified models while x 1 , x 2 are position in the x direction of original and the modified models, respectively. Repeat those for y and z direction.

Adding boundary conditions.
Although the mass and position have been preserved, due to the majority loss of the rear parts, a change in moment of inertia occurs. The rear of the car is still affected by external forces, which include gravity and lift at the rear wheels. To ignore the effects of unnecessary parts, some boundary conditions need to be added to the modified model. Position of boundary condition is shown in Fig. 2. The boundary conditions are applied to rearmost elements of the new model and the wheel housings which have only one degree of freedom in the direction car move straight. The axis of the wheel has 2 degrees of freedom which are in the straight direction and car's high direction. Thus, the modified model will ensure that there is no external force impacting the back of the vehicle so that the vehicle will be erected. It is noticed that the modified model is used to investigate the behavior of frontal collision. Therefore, energy and momentum of the modified model must be similar to the full models.

Type of element
Each model is composed of many types of elements. Depending on each part of the model, a different type of element is used. For example, element_mass (3D structural mass element) for mass node while ele-ment_shell (three, four, six, and eight node 2D thinshell elements) for windsheld, plate structure...

Simulation set up
All the modified models in this study are set up to contact with NCAP wall at 56.3 km/h as demonstration in Fig. 3. The simulation problems in the research is all frontal contacts. The CONTACT AUTOMATIC SURFACE TO SURFACE keyword was used between modified car models and NCAP wall. Velocity, acceleration, displacement, force and energy data values are considered.

Validation of mass and position of C.G
The specification of comparing of original vehicle models and the modified vehicle models is shown below from Table 1 to Table 6. All of modified models reduce almost 50% of the total number of nodes and elements except the Pickup model, the mass and location of C.G of modified vehicle models are similar to the original vehicle models.

Verification of modified vehicle models
The FE models are set to have an initial velocity of 56.3 km/h and bump into a rigid wall created by 4N-Shell element. The simulation results of the full model impacting an analytical wall downloaded from CCSA website is used for benchmarking.

Pickup-1994 Chevrolet C2500
Deformation of Pickup-1994 Chevrolet C2500 is described typically at 30 ms and 80 ms in Figure 4. The velocity of left and right seat crossmember are shown in Figure 5 and Figure 6. The velocity curve of modified model agrees well with results in 5 . The acceleration of left and right seat cross member are shown in Figure 7 and Figure 8. There is fluctuation but the tendency of all acceleration curves are similar. In particular, the acceleration curve of Modified model and NCAP Test 1741 show a good result. The rigid body displacement is shown in Figure 9 while the total wall force is represented in Figure 10. The rigid body displacement curve of Modified model is higher than that of Full model from 0.06s to 0.15s. However, the tendency of them are good. The results         The energy balance and the percentage error of total energy are shown in Figure 11 and Figure 12, respec-tively. The total kinetic energy and internal energy are lost due to non-physical energies. The average percentage error of total energy is 5.5%.  The velocity of engine bottom and engine top are shown in Figure 16 and Figure 17, respectively. They match very well. The total wall force and vehicle displacement are shown in Figure 18 and Figure 19. The curve of Full model and Modified model in Figure 18 stick together closer than others while Modified model curve is closer to NCAP test 2496 than others in Figure 19. The cause lies in the change of inertia.

SUV-2002 Ford Explorer
Deformation of SUV-2002 Ford Explorer is described typically at 30 ms and 80 ms in Figure 22.

Sedan-2010 Toyota Yaris
Deformation of Sedan-2010 Toyota Yaris is described typically at 30 ms and 80 ms in Figure 32. The accel-  The total wall force and force-displacement are shown in Figure 35 and Figure 36, respectively. All curves have similar tendency in Figure 35    Deformation of Sedan-2012 Toyota Camry is described typically at 30 ms and 80 ms in Figure 42. The acceleration of engine top and engine bottom are shown in Figure 43 and Figure 44. The acceleration curve of Modified model and Full model stick together. There is small difference between NCAP Test with the two others but insignificant in case of Engine top acceleration, Figure 43. In general, they are matched very well.  The energy balance and the percentage error of total energy are shown in Figure 48 and Figure 49, respectively. The energy balance graph shows an excellent result. The average percentage error of total energy of modified model compare to full model is 4.8%

Findig the reduction of simulation time
Modified models gives a good results when reducing a large amount of resources used in computational simulation.

Pickup-1994 Chevrolet C2500
The Elapsed time decrease 0.22% when run with modified model. Detail of the results is described in Table 7.

SUV-1997 Toyota Rav4
The Elapsed time decrease 69% when run with modified model. Detail of the results is described in Table 8.

SUV-2002 Ford Explorer
The Elapsed time decrease 28% when run with modified model. Detail of the results is described in Table 9.

Sedan-2010 Toyota Yaris
The Elapsed time decrease 58% when run with modified model. Detail of the results is described in Table 10.

Sedan-2012 Toyota Camry
The Elapsed time decrease 62% when run with modified model. Detail of the results is described in Table 11.

Sedan-1996 Dodge Neon
The Elapsed time decrease 37% when run with modified model. Detail of the results is described in Table 12.

CONCLUSION
The results show that the FE models of vehicle front structures can replace original models in a frontal crash to reduce time operation and memory resources significantly. Conservation of vehicle front structures' CG makes sure that the results of the modified