Abstract
Objective: In minicars, the survival space between the side structure and occupant is smaller than in conventional cars. This is an issue in side collisions. Therefore, in this article a solution is studied in which a lateral seat movement is imposed in the precrash phase. It generates a pre-acceleration and an initial velocity of the occupant, thus reducing the loads due to the side impact.
Methods: The assessment of the potential is done by numerical simulations and a full-vehicle crash test. The optimal parameters of the restraint system including the precrash movement, time-to-fire of head and side airbag, etc., are found using metamodel-based optimization methods by minimizing occupant loads according to European New Car Assessment Programme (Euro NCAP).
Results: The metamodel-based optimization approach is able to tune the restraint system parameters. The numerical simulations show a significant averaged reduction of 22.3% in occupant loads.
Conclusion: The results show that the lateral precrash occupant movement offers better occupant protection in side collisions.
Introduction
Recently, the number of electric minivehicles in the European market, which look like small city cars (e.g., smart for 2) but technically classed as heavy quadricycles (L7e category, curb weight < 400 kg without battery) is growing steadily. The European New Car Assessment Programme (Euro NCAP) tested vehicles of this L7e class in front and side impact, all of which revealed critical safety problems (Euro NCAP Citation2014).
In the research project Visio.M, the Technische Universität München in collaboration with Daimler, Autoliv, and IAV developed an integral safety concept for the Visio.M car, which is classified as an L7e category vehicle.
During the project, a study of potential accident situations of electric mini vehicles based on the German In-Depth Accident Study was carried out (Unselt et al. Citation2014). The analysis points out that electric minivehicles would be involved in fewer front and rear collisions and would be more subject to lateral collisions due to the increase of intersection accidents. Side pole collisions have been shown to play a less significant role.
There are 2 big challenges in side collisions of minicars: (1) the survival space between the side structure and occupant is smaller compared to conventional vehicles and (2) the incompatibility regarding the small mass and dimensions of minivehicles in relation to cars of normal size (Egertz Citation2011).
Current passive safety systems for side collisions like side airbags, safety belts, and vehicle structures start to work milliseconds after the impact. With the detection of the vehicle surroundings by means of sensors, it is possible to assess unavoidable accidents a few milliseconds before they happen (Tandler Citation2008). This time can be used to prepare the occupant and the car for the crash. For side collisions, a lateral movement of the occupant toward the middle of the vehicle can be used once the side impact is detected. As a result of this pre-acceleration, the crash pulse acting on the occupant can be reduced and the distance between the occupant and side structure is increased.
Luzan-Narro et al. (2012) implemented the occupant displacement by a rotating seat around the longitudinal axis of the vehicle. Pyrotechnical actuators start to rotate the seat 80 ms before the crash and increase the gap between side structure and occupant. Mercedes-Benz (Mellinghoff et al. Citation2009) tested the precrash system Pre-Safe-Pulse by which cushions propel the upper body of the occupant up to 50 mm toward the center of the vehicle by inflating suddenly. Numerical simulations with this system showed an average 30% reduction in rib intrusion of the occupant.
This article presents the load reduction potential of a precrash lateral occupant movement in side collisions of minicars. The entire seat is used for the occupant movement. Thus, all relevant body parts are moved in the precrash phase of a side collision.
Methods
The assessment of the potential is mainly done by simulation and a full-vehicle crash test serves as verification. The optimal parameters of the restraint system, including the precrash movement, are found using metamodel-based optimization methods by minimizing occupant loads.
Finite Element Simulation Model
The model consists of body-in-white, exterior, seat, and restraint system of Visio.M, as well as ES-2 dummy and ECE-R 95 barrier (). The assembled Visio.M simulation vehicle model has a mass of 700 kg, which represents the vehicle weight plus 2 occupants and an additional payload. The simulation was carried out with the finite element software LS-Dyna (Livermore Software Technology Corporation (LSTC), Livermore, CA).
Body-in-White Model
A multimaterial mix is used for the Visio.M body-in-white. It consists of doors and monocoque made of carbon fiber reinforced plastic (CFRP), an aluminum roof, and front and rear ends.
Seat Model
Visio.M utilizes a fixed-eye-point design and therefore a new seat concept is used. There is no seat adjustment along the vehicle longitudinal axis but an adjustment in height. The seat consists of 3 parts: seating surface, lumbar support, and backrest. All parts have the same lateral movement during the precrash phase. The seat contour is smaller compared to conventional seats; therefore, the occupant is not kept in the seat as well as in normal seats during the lateral movement.
Restraint System Model
The restraint system consists of a 3-point belt with pretensioner and 2 side airbags. Two different setups are examined. The first setup is made of a seat-mounted head–thorax airbag and pelvis airbag. It is not linked with the later described metamodel-based optimization. The second setup is made of a door-mounted pelvis–thorax airbag and a roof-mounted head airbag. Because the head of the occupant is always at the same position (fixed-eye-point design), the head airbag can be designed to be very small.
Dummy Model
The ES-2 side impact dummy from Dynamore (Germany) is used for simulation. Previously conducted tests with a lateral moving seat during the Visio.M project showed similar behavior between subjects and an ES-2 dummy in the thorax area in the first 150 ms of the lateral occupant movement (Unselt et al. Citation2014). The head of the ES-2 dummy, however, behaves differently. Additionally, the head of the ES-2 simulation model behaves differently compared to the real dummy. This has to be taken into account during the tuning of the restraint system.
The dummy movement in the precrash phase before t0 (start of crash) in the simulation was validated with experiments. Therefore, the seat was moved with a belted ES-2 dummy and the displacement and velocity were captured with a high-speed camera. A friction coefficient of 0.5 between the seat and dummy showed good alignment of simulation and experiment.
Barrier Model
The mobile deformable barrier from LSTC is used and represents an ECE-R95 side impact barrier, also used in the Euro NCAP car-to-car side impact test.
Metamodel-Based Optimization
Two independent metamodel-based optimizations are carried out, one with the lateral precrash occupant movement (precrash optimization) and one in which the seat is fixed (in-crash optimization). As a result, it is possible to compare the best design of each optimization: one with and one without lateral precrash occupant movement. The design of experiments is performed with the program Datalysor, an in-house program developed and used at Autoliv. A D-optimal plan was chosen for both optimizations, which leads to 6 independent parameters for the in-crash optimization (50 simulations) and 8 parameters for the precrash optimization (76 simulations). After the finite element simulation results are extracted, response surfaces are generated and occupant loads optimized.
Load Case
The load case is based on the Euro NCAP side collision protocol. A deformable barrier of 950 kg moves perpendicular to the vehicle at 50 km/h and struck in the driver's door.
Parameters
A group of experts from the working group chose the parameters and the associated levels for both optimizations, which are shown in together with their description. Negative time values correspond to the precrash phase—events that happen prior to the crash. The time-to-fire for the head airbag is coupled in all simulations to the value of the side airbag.
Table 1. Parameters for in-crasha and precrashb metamodel-based optimization
Two different side airbag shapes (SAB_SHAPE) are used. Shape 1 expands more in the x- and y-directions, and shape 2 expands more in the z-direction. The belt pretension (BELT_TTF) starts either initially with the seat movement or at t0. The seat movement parameter (SEAT_TTF) has 3 levels and starts 150, 100, and 50 ms before t0. The generic movement remains the same in all 3 levels and is based on experimental data. The seat movement in the experiments was slightly slower than the generic movement used in this study. A pneumatic muscle from the company Festo (Germany) in combination with a gas generator was used as an actuator. In the present study a faster seat movement was used, allowing us to achieve greater dummy displacement and velocity. Using a stronger gas generator, this faster seat movement could be easily achieved in reality. and show the lateral displacement and velocity of the seat and the ES-2 dummy head, chest, abdomen, and pelvis in a simulation without belt pretension and no crash. This is equivalent to the precrash phase in the full-vehicle simulation. As can be seen, the seat moves much earlier than the dummy and the displacement and velocity of the dummy change with time.
System Responses
The occupant load evaluation is based on the current Euro NCAP side impact protocol (Euro NCAP Citation2013). Its occupant load parameters act as system responses: head injury criterion (HIC), resultant acceleration 3 ms exceedance (a3ms), rib intrusion (RibIn), rib viscous criterion (RibVC), T12 moment (T12Mx), T12 force (T12Fy), backplate loading (BpF), total abdominal force (AbdF), and the pubic symphysis force (PubF).
Optimization
A polynomial response surface model is created for each response. In the Euro NCAP assessment, each occupant load has a higher performance limit (hpl) and a lower performance limit (lpl), whereas it is aspired to be under the higher performance limit. The limits are as follows: HIC (hpl 650/lpl 1000), a3ms (72/88), RibIn (22/42), RibVC (0.32/1.0), BpF (1/4), T12Fy (1.5/2.0), T12Mx (150/200), AbdF (1.0/2.5), PubF (3.0/6.0). The maximum score is 16 points. In some cases it is possible to reach the maximum score of 16 points with several parameter sets. Therefore, a 2-stage optimization is performed. If the score is below 16 points, the objective function is based on the Euro NCAP score calculation and tries to maximize the score. If the score reaches 16 points, the objective function tries to maximize the normalized distance d from the hpl of every load, which means that the most critical value is minimized as long as there is no lower normalized value. An example for the distance calculation for the HIC is found in EquationEq. (1)(1) :
(1)
Thus, the possible values of the normalized distance lie between 0 and 1 in case of a 16-point assessment. Because some occupant loads are valued twice as high as others in the Euro NCAP assessment (so-called modifiers), the objective function weighs some parameters differently in the case of 16 points. With the example from EquationEq. (1)(1) , the optimization problem is given by EquationEq. (2)
(2) :
(2)
The response surface models are generated in Datalysor and transferred to MATLAB (The Mathworks Inc., USA). A standard genetic algorithm together with the fmincon algorithm from the MATLAB Toolbox is used to minimize the occupant loads. The predicted values from the metamodel are verified with further finite element simulations.
Results
Prior to the second restraint system setup, a setup with a head–thorax airbag and pelvis airbag (first setup) was examined (Hierlinger et al. Citation2014). This setup was not sufficient in combination with the lateral precrash occupant movement. Induced by the seat movement, there is a large geometric variability and a robust positioning of the relatively large head–thorax airbag mounted in the backrest is difficult. In addition, the occupant pushes during the seat movement temporarily in the deployment path of the head–thorax airbag due to the slim seat contour and the airbag deploys in the back of the occupant. This can also be seen in , where the seat already traveled 100 mm after 50 ms and the farthest part of the dummy just 27 mm.
The restraint system setup with pelvis–thorax airbag and head airbag is much more robust than the previous one. With the first setup, many simulations collapsed due to out-of-position problems, but with the second setup all simulations proceeded successfully.
The coefficient of determination (R2) represents the ability of the response surface to identify the variability of the design response and has a range from 0 (bad) to 1 (good). For the in-crash model, it lies between 0.9 and 0.98. For the precrash model, it ranges from 0.71 (RibVC) to 0.99. Nevertheless, 0.71 can be considered an outlier because the next higher value is 0.88.
The predicted values from the metamodel deliver good results confirmed by additional numerical verification simulations. For the in-crash simulation, there is an average error of 3.4% for the occupant loads. The smallest distance d to the hpl is predicted for BpF (dBpF = 0.047). However, in the simulation, the smallest distance is at the T12Fy load (dT12Fy = -0.02), which is worse than in the prediction. For the precrash simulation, the average error for the occupant loads compared to the prediction is 4.9%. The smallest predicted distance is at AbdF (dAbdF = 0.195), though in the simulation the T12Fy value (dT12Fy = 0.113) is the most critical. In both cases the most critical predicted values are reduced in the verification simulations, whereas less critical values become the most critical values in the simulations due to larger prediction errors.
presents the normalized distances from the optimization for the best in-crash simulation and the best precrash simulation. The larger the normalized distance, the lower the occupant loads. Additionally, the percentage changes of the occupant loads from in-crash to precrash simulation are displayed. The simulations show a significant reduction in occupant loads due to the lateral precrash occupant movement. The in-crash optimization obtains 15.8 points, whereas the precrash optimization reaches 16 points. The most critical load is T12Fy. This load is reduced by 13.1%. In total, all but one load can be reduced. The load reduction averages 22.3% for the minicar Visio.M with the lateral precrash occupant movement. The average normalized distance of the in-crash simulation is 0.392 in contrast to the 0.468 of the precrash simulation.
The parameter values for the in-crash/precrash simulation are as follows: SAB_INFL (102/113), SAB_TTF (0.0/4.0), SAB_VENT (115/96), SAB_SHAPE (1/1), HAB_INFL (115/115), HAB_VENT (85/85), SEAT_TTF (—/50.0), BELT_TTF (—/0). In comparison, it can be seen that there are some differences in the parameters of the side airbag. The mass flow in the in-crash simulation is lower as well as the time-to-fire. In addition, the venting area is larger. Together this leads to a softer side airbag than in the precrash simulation. The parameters for the head airbag are the same in both simulations. The best results for the precrash optimization are achieved with a short precrash phase and the belt pretension at the time of the collision (t0). In this configuration, the occupant loads stay in an uncritical range during the seat movement. However, with the belt pretension at the start of the seat movement, the load on the ES-2 dummy backplate BpF is slightly over the hpl.
The occupant loads are a result of the chosen parameters. The influence on each occupant load might depend on different parameters. Using the analysis of variance and global sensitivity analysis (GSA)/SOBOL approach (Ryberg et al. Citation2012), a sensitivity analysis for the precrash optimization was performed (). The arrow in each parameter indicates whether the values must be reduced or increased in order to diminish the specific occupant load. For example, if the arrow points downwards at HIC for SEAT_TTF, it means that a short precrash phase will lead to lower HIC values. The left-most parameters have the highest influence on the occupant load. It can be seen that some parameters are competitive such as SAB_VENT for T12Mx and AbdF.
Table 2. Sensitivity analysis of the design variables from the precrash optimization
Additional to the simulations, a full-vehicle crash test with the lateral precrash occupant movement was performed. Due to many prototype components, the optimal parameter set found by optimization could not be used. In addition, the seat movement curve in the precrash phase differed from the curve used here. A result of 13.7 points according to Euro NCAP dummy assessment was achieved. The most critical occupant load was T12Mx, which exceeded the lpl. In addition, AbdF was slightly over the hpl. All other values were under their hpl.
Discussion
The restraint system setup with pelvis–thorax airbag and head airbag works very well in the examined configurations. The described occupant movement is only possible if there is enough space for the seat movement. Therefore, the application is predestined for electric vehicles without a transmission tunnel and front wheel-drive cars, whose transmission tunnel is removed. An additional interaction protection mechanism between occupants might be necessary as well. Tests conducted with subjects showed a dependency of the seat velocity in the precrash phase with the occupant weight. Adaptation to different occupant weights can be done by changing the time-to-fire of the seat (SEAT_TTF). A more optimally formed seat contour will help to better retain the occupant into the seat.
The metamodel-based optimization approach is able to tune the restraint system. The prediction errors are quite low, yet the largest errors are at the most critical values. In addition to the polynomial metamodel, 2 further metamodels (kriging, neural network) were applied to optimize the restraint system parameters. All metamodels worked well, whereas due to the D-optimal point selection the optimization and evaluation was done with the polynomial model. In future optimizations, the kriging metamodel combined with a space-filling point selection might lead to even better occupant load predictions but with a higher computational cost.
The minimum occupant loads were achieved with a short precrash time (50 ms) and the belt pretension at t0. The coupling between occupant and restraint system (especially side airbag) begins approximately 20 ms after t0, which is equivalent to 70 ms after the seat movement starts. As can be seen in and , at this time the chest and abdomen have their maximum velocity and the pelvis velocity already starts decreasing. In contrast, these parts just moved one third of the possible distance. Though the head is almost not moving at this time, its HIC value is less critical, and therefore a precrash preparation of this body part is not necessary. With a precrash time of 100 ms (SEAT_TTF = 100), the occupant loads are still better than without lateral precrash movement (most critical normalized distance = 0.073, average normalized distance = 0.4). A precrash time of 150 ms (SEAT_TTF = 150) also leads to 16 points in Euro NCAP and the most critical normalized distance = 0.08 is slightly higher than with a 100 ms precrash time, although the average normalized distance = 0.373 is lower. At this time, the dummy velocity already decreases at the pelvis, abdomen, and chest, which are the critical parts in case of occupant loads. Only the head still increases in velocity, but it is not a critical part. In return, the distance covered by the dummy almost equals the distance of the seat, maximizing the space between side structure and dummy. One reason for the poorer results with a long precrash phase is the limited optimization design space. Both SAB_INFL and SAB_VENT are at the border of the design space and therefore the side airbag is too soft. Additionally, the side airbag shape is not optimal for the large space between the side structure and dummy. The lateral extension should be larger to fill the whole space optimally. Taking this knowledge into account, an even better occupant load reduction could be possible if no fallback solution is necessary.
In all 3 precrash simulations, the belt pretension starts at t0. This is because of the BpF value, which is slightly over the hpl, making it impossible to score 16 points. In general, it is important to limit the seat velocity in the precrash phase in a way that no serious injuries occur. In a test with an ES-2 dummy and a similar seat movement curve to the generic curve used here, the load BpF remained in an uncritical range. Nevertheless, even if the 16 points are reached (no point deduction), it will be better to start the belt pretension at t0. The space between side structure and dummy would become even larger and the belt tension after t0 would be too low.
Unfortunately, it is not possible to have the same restraint system parameter set for both optimal in-crash and precrash case. A simulation with the optimal parameter set from the in-crash optimization and additional lateral precrash occupant movement (SEAT_TTF = 50, BELT_TTF = t0) results in 14 points in Euro NCAP assessment. In this case, the T12Mx load is over the lpl. All other occupant loads are under the hpl. Another simulation with a changed SAB_TTF = 4.0 comes to 14.6 points in Euro NCAP assessment. Still, the T12Mx load is over the hpl.
In the full-vehicle crash test, the Euro NCAP score was 13.7. This is mainly because of the higher side structure intrusion compared to the simulation model, which is not accurate enough. New simulations with higher intrusions have to be carried out. Nevertheless, the potential of the lateral precrash occupant movement demonstrated in this article is still valid because the intrusion for both in-crash and precrash optimization was the same, only lower as in reality. It also shows that for minicars a normal state-of-the-art restraint systems might not be sufficient.
In case of multiple impacts, the system should only be activated if the first impact is a side impact. If the occupant is not well seated (for example, after a first frontal impact) and there is no contact with the backrest, the movement of the occupant thorax is prevented and only the pelvis region is moved. This leads to a tilted occupant. In case of a rollover crash, the system could also help to reduce occupant loads.
This article showed the potential of occupant load reduction by means of lateral precrash seat movement in case of a side impact. A significant average reduction of 22.3% in occupant loads could be reached in simulations with a precrash movement. Especially for small and light vehicles, the potential of this approach is very high due to the small survival space and the normally higher intrusions.
Funding
Participants in the Visio.M consortium are, in addition to the automotive companies BMW AG (lead manager) and Daimler AG, the Technische Universitaet Muenchen as a scientific partner and Autoliv BV & Co. KG, the Federal Highway Research Institute (BAST), Continental Automotive GmbH, Finepower GmbH, Hyve AG, IAV GmbH, InnoZ GmbH, Intermap Technologies GmbH, LION Smart GmbH, Amtek Tekfor Holding GmbH, Siemens AG, Texas Instruments Germany GmbH, and TÜV SÜD AG as industrial partners. The project is funded under the priority program “Key Technologies for Electric Mobility–STROM” of the Federal Ministry for Education and Research (BMBF).
References
- Egertz D. Novel Safety Requirements and Crash Test Standards for Light-Weight Urban Vehicles. Stockholm, Sweden: KTH Royal Institute of Technology; 2011.
- Euro NCAP. European New Car Assessment Programme: Assessment Protocol—Adult Occupant Protection. Belgium: Author; 2013. Technical report, Version 6.0.
- Euro NCAP. Euro NCAP's spotlight falls on heavy quadricycles [press release]. 2014. http://preview.thenewsmarket.com/Previews/NCAP/DocumentAssets/332832.pdf. Accessed November 11, 2014.
- Hierlinger T, Unger J, Khazami M, Netzker C, Lienkamp M. Presentation of innovative protection systems of a side-impact protection concept for micro cars with analysis based on FE simulations. Paper presented at: crash.tech; April 8–9, 2014; Munich, Germany.
- Luzan-Narro J, Arregui-Dalmases C, Hernande LM, Core E, Narbona A, Selgas C. Innovative passive and active countermeasures for near side crash safety. Paper presented at: International Crashworthiness Conference (ICRASH)July18–202012; Milan, Italy.
- Mellinghoff U, Breitling T, Schöneburg R, Metzler HG. The Mercedes-Benz experimental safety vehicle ESF 2009. ATZ Worldwide. 2009;111:34–41.
- Ryberg AB, Domeij Bäckryd R, Nilsson L. Metamodel-Based Multidisciplinary Design Optimization for Automotive Applications. Linköping, Sweden: Linköping University; 2012. Technical Report.
- Tandler J. Crash safety through side impact protection—Aprosys Project for Environmental Sensorics. ATZ Worldwide. 2008;11:4–11.
- Unselt T, Unger J, Quedenbaum L, Hierlinger T. The integrated safety concept of the ultra-compact electric vehicle visio.m. Paper presented at: 12th International Symposium & Exhibition on Sophisticated Car Occupant Safety Systems (Airbag 2000+)December 1–3, 2014; Karlsruhe, Germany.