Improved Vehicle Crashworthiness Design by Control of the Energy Absorption for Different Collision Situations

Transcription

1 Improved Vehicle Crashworthiness Design by Control of the Energy Absorption for Different Collision Situations

2 CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EINDHOVEN Witteman, Willibrordus J. Improved Vehicle Crashworthiness Design by Control of the Energy Absorption for Different Collision Situations / by Willibrordus J. Witteman. - Eindhoven : Technische Universiteit Eindhoven, Doctoral dissertation, Eindhoven University of Technology With literature list and summary in Dutch. Proefschrift. ISBN NUGI 834 Subject headings: vehicles; crashworthiness / longitudinal member / vehicle structure / numerical simulation Trefwoorden: voertuigen; botsveiligheid / langsligger / autoconstructie / numerieke simulatie Printing: Universiteitsdrukkerij TU Eindhoven Copyright 1999 W.J. Witteman All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form by any means, electronic, mechanical, photocopying, recording or otherwise, without the written permission of the author.

3 Improved Vehicle Crashworthiness Design by Control of the Energy Absorption for Different Collision Situations PROEFSCHRIFT ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de Rector Magnificus, prof.dr. M. Rem, voor een commissie aangewezen door het College voor Promoties in het openbaar te verdedigen op dinsdag 15 juni 1999 om uur door Willibrordus Jacobus Witteman geboren te Geldrop

4 Dit proefschrift is goedgekeurd door de promotoren: prof.dr.ir. R.F.C. Kriens en prof.dr.ing. G. Belingardi

5 Vehicle safety research and new safer cars are expensive. Human life is priceless.

6

7 Summary 7 Summary Increased traffic intensity, growing concern of the public and new stringent legislation, have made vehicle safety one of the major research areas in automotive engineering. Especially the unfavorable crash results (large deformation of the passenger compartment of many cars) occurring in more realistic crash tests, which deviate from the compulsory full overlap crash test against a concrete block, are reason to worry. In the case of a partial frontal overlap (offset) collision or an off-axis crash direction only part of the vehicle structure can be used for energy absorption. This leads to dangerous intrusions of the passenger compartment, because only one of the two longitudinal members is used for energy absorption. This thesis describes the design of a new frontal vehicle structure that directs the asymmetric crash load of an offset collision as an axial load to the second unloaded longitudinal member. Only by using both longitudinal members and through a progressive folding pattern, enough energy can be absorbed in the front structure to prevent a deformation of the passenger compartment. To prevent a premature bending collapse, the new longitudinal members consist of two functional components: an inside square crushing column for a normal stable axial force level and a stiff outside sliding supporting structure that gives the necessary extra bending resistance. An integrated cable system transmits the force to the other longitudinal member. With this novel design concept, a vehicle has similar energy absorption in the front structure for the entire range of collision situations (full, offset, oblique). By means of numerical crash simulations, this concept has been optimized and evaluated. Results show that for an entire range of frontal collision situations similar deceleration curves can be obtained. However, to further reduce the injury level of the occupants, optimal crash decelerations for various crash velocities are necessary. To this aim, a method is described for numerical FEM dummy simulations to obtain optimized crash pulses for different velocities. The novel concept is very suitable to adapt the structural stiffness to these new deceleration pulses. To realize the optimal deceleration during the crash for each velocity, solutions have been presented based on controllable energy absorption by additional friction or based on controllable hydraulic flow restriction. With this total design, an optimal vehicle deceleration curve is possible for each velocity over the entire frontal collision spectrum, yielding the lowest levels of the occupant injury criteria.

8 8

9 Samenvatting 9 Samenvatting Toegenomen verkeersintensiteit, meer bewustwording van de consument en strengere wetgeving, maken veiligheid van auto s tot belangrijk onderzoeksgebied in de automobielontwikkeling. Vooral de tegenvallende botsresultaten (grote vervormingen van het passagiersgedeelte van veel auto s) in op de realiteit gelijkende botstesten (die afwijken van de verplichte volledige overlap botstest tegen een betonblok), zijn zorgwekkend. In het geval van een gedeeltelijke overlap (offset) botsing of een schuine botsrichting kan slechts een gedeelte van de voertuigconstructie gebruikt worden voor de energie absorptie. Dit lijdt tot gevaarlijke intrusies van het passagierscompartiment, omdat slechts één van de twee langsliggerbalken voor energie absorptie wordt gebruikt. In dit proefschrift wordt een nieuwe frontale voertuigconstructie beschreven, die bij een offsetbotsing de krachten op slechts één getroffen langsligger doorleidt naar de andere ongetroffen langsligger en wel zo dat deze ook axiaal deformeert. Alleen op deze manier absorberen beide langsliggers met voortgaande plooivorming genoeg energie om een vervorming van het passagiersgedeelte te voorkomen. Om een vroegtijdige knik te voorkomen bestaan de nieuwe langsliggers uit twee functionele delen: een inwendige vierkante crashkoker voor een normaal stabiel axiaal krachtniveau en een uitwendige stijve ondersteunende schuifconstructie voor de noodzakelijke extra buigweerstand. Een geïntegreerd kabelsysteem zorgt voor de krachtoverbrenging naar de andere langsligger. Met dit nieuwe ontwerp heeft een auto een vergelijkbare energie absorptie in de frontale voertuigconstructie voor de hele range van botssituaties (volledig, offset, schuin). Met behulp van numerieke botssimulaties is het nieuwe concept geoptimaliseerd en beproefd. De resultaten laten zien dat vergelijkbare vertragingscurven kunnen worden verkregen voor een heel spectrum van frontale botssituaties. Om echter het letselniveau van de inzittenden verder te reduceren, zijn de optimale botsvertragingen bij verschillende botssnelheden nodig. Hiervoor is een methode beschreven waarmee met numerieke FEM dummy simulaties geoptimaliseerde vertragingscurven bij verschillende snelheden zijn verkregen. Het nieuwe concept is zeer geschikt om de constructiestijfheid aan te passen voor deze nieuwe vertragingscurven. Om de optimale vertraging tijdens een botsing voor elke snelheid te realiseren, worden constructieve oplossingen voorgesteld gebaseerd op regelbare energie absorptie door extra wrijving of gebaseerd op regelbare hydraulische stroombegrenzers. Met dit totale ontwerp is een optimale voertuig vertragingscurve

10 10 voor elke botssnelheid voor het gehele frontale botsgebied mogelijk, wat tot de laagste niveaus van de letselwaarden leidt.

11 Table of contents 11 Table of contents Summary... 7 Samenvatting General Introduction Description of the research issue Research objectives Research strategy Thesis outline The Necessity of Improved Crashworthiness Design Introduction Frontal crash parameters Collision speed Obstacle type Collision place and direction Analysis of useful crash situations Optimal crash pulses Representative crash tests as a design goal Overview of actual and expected legal test requirements Conclusions Numerical Design of Stable Energy Absorbing Longitudinal Members Introduction Simulation parameters as design requirements Research overview of crushing columns Simulation results Discussion of the simulation results... 54

12 12 Table of contents 3.6. Design of the triggering for a constant stable force level Determination of the optimal trigger position Determination of the most efficient trigger geometry Determination of the best dimension of the bead initiator Design of a longitudinal cross-section A new design concept for functionally decomposed longitudinal members Manufacturing and axial crushing of the advanced longitudinal member Conclusions Design of a Frontal Safety Structure Suited for Different Crash Situations Introduction The cable connection system for a symmetric force distribution Numerical simulation of the cable-supported longitudinal structure Building the longitudinal structure in a numeric frontal car model Numerical simulation of a full overlap crash Numerical simulation of a 40 per cent offset crash Numerical simulation of a 30 degrees crash A non-axial component test as verification is not realistic Conclusions A Structural Solution to Realize the Desired Deceleration Pulse Introduction Example of a method for optimizing the deceleration pulse Structural design specifications for different crash velocities The necessity of an adaptive structure Energy absorption by friction Future possibilities Design of a hydraulically controlled frontal car structure Conclusions

13 Table of contents Conclusions and Recommendations Overview of the research Conclusions Recommendations Appendix A Numerical Simulation Method with PAM-CRASH Appendix B Calculation of the Temperature Increase after Energy Absorption by Friction References Acknowledgements Curriculum vitae

14 14 Table of contents

15 General Introduction 15 Chapter 1 General Introduction Serious or fatal traffic accidents are considered as one of the most threatening dangers in daily life. It is an unexpected event that can change people s life radically. In the Netherlands (15 million inhabitants) about 1200 people are killed annually and more as people are injured. About half of these are attributed to car occupants. See Table 1.1 in which accident statistics are shown since Despite increased mobility, these values are slightly on the decline. Although a safe driving style minimizes accident risk, car occupants are also exposed to unexpected road conditions and risky or drunken drive behavior of other road users (like a badly timed passing maneuver or a slipping car on the wrong side of the road). Especially, frontal accidents on country roads against other cars have a high fatality rate. Due to efforts to avoid frontal collisions, the car front is generally only partly involved and not always axially. In addition, the incompatibility between different vehicles yields more fatalities. These collision situations are until now not legally tested. Only a few car manufacturers are using such collision situations as safety design goal for a longer time. Table 1.1. Number of death people of different traffic participants in The Netherlands (SWOV 1998) Car driver Car passenger Truck Delivery van Bus Motorcycle/scooter Motorbike Small motorbike Bicycle Pedestrian Other Total

16 16 Chapter 1 Increased traffic intensity, growing concern of the general public, and more stringent legislation have made vehicle safety one of the major research areas in automotive engineering. An area of particular concern is the early crashworthiness design of cars. Cars have to pass the compulsory crash test as issued by the authorities. However, this test does not guarantee that cars are safe in crash situations that deviate from the prescribed one (Witteman 1993). Hence, the entire collision spectrum within which the car must be safe must be investigated Description of the research issue The improved frontal crashworthiness of cars necessitates totally new design concepts, which take into account that the majority of collisions occur with partial frontal overlap and under off-axis load directions. Realistic crash tests with partial overlap have shown that conventional longitudinal structures are not capable of absorbing all the energy in the car front without deforming the passenger compartment. See Figure 1.1 for an offset test against a rigid barrier and Figure 1.2 in which a car is used in a full overlap and in an offset test against a deformable barrier for comparison. It is clear to see that in case of a full overlap collision there is no intrusion of the passenger compartment, while in the offset test the passenger compartment of the same car collapses. The reason for this is that the structure of the longitudinal members is specifically designed for meeting the less severe requirements of the compulsory full overlap test, in which both longitudinals are loaded axially. Figure 1.1. Example of large deformation of the passenger compartment during an offset crash with 55 km/h (Auto Motor und Sport spezial 1992, photo H.P. Seufert).

17 General Introduction 17 Figure 1.2. Example of identical cars with the same collision speed in a full (above) and in an offset (under) crash (video fragment TRRL). For improved frontal car safety it is necessary to design a structure that absorbs enough energy in each realistic crash situation. To protect the occupants, the passenger compartment should not be deformed and intrusion must be avoided too. To prevent excessive deceleration levels, the available deformation distance in front of the passenger compartment must be used completely for a predetermined crash velocity. This implies that in a given vehicle concept the structure must have a specific stiffness. Normally, the two main longitudinal members will absorb most of the crash energy with a progressive folding deformation of a steel column. The main problem is that in real car collisions these two longitudinal members often are not loaded in a synchronous fashion and also not loaded pure axially. The majority of collisions occur with partial frontal overlap, in which only one longitudinal is loaded, or under an off-axis load direction. This implies that most longitudinals fail under a premature bending collapse rather than a much more energy absorbing progressive folding pattern. This gives rise to two design conflicts. The first conflict is that the same amount of energy must be absorbed either with a single or with both

18 18 Chapter 1 longitudinals. The second conflict is that the same amount of energy must be absorbed in the case of an off-axis impact angle as in the case of a normal axial impact. These problems can not be solved by just increasing the stiffness of the longitudinals in such a way that each longitudinal is capable of absorbing all of the energy, see the following reasons. To absorb enough energy, a stiff longitudinal is needed for the offset crash in which normally only one longitudinal is loaded. The same longitudinal must be more supple in case of a full overlap crash, since both longitudinals must not exceed the desired deceleration level (Witteman 1993). In addition, a stiff longitudinal is needed to absorb enough energy in an off-axis load direction resulting in a higher bending resistance to help transform off-axis loads into axial loads and to prevent a bending collapse. The same but more supple longitudinal is needed in the case of a normal axial load to avoid overly high deceleration forces. Another issue is the crash velocity. To absorb all the kinetic energy, which is proportional with the square of the velocity, the deformable structure length must have a specific stiffness. This stiffness results in an average mean force, which multiplied with the deformation shortening gives the absorbed energy. For an acceptable injury level of the occupants, the total deceleration level must be as low as possible, using the maximum available deformation length without deforming the passenger compartment. This means that for example in a 64 km/h crash compared with a 32 km/h crash, a four times longer deformation distance is needed for the same deceleration level. See Figure 1.3 in which the relation between impact energy, deformation length, force and crash velocity is plotted for different vehicle masses, stiffnesses and average deceleration levels. In this figure the example is plotted with the dashed lines. For a crash velocity of 32 km/h respectively 64 km/h the necessary deformation length for the same constant deceleration of 20 g is 20 cm respectively 80 cm. Since a deformation length is mostly restricted to cm, it is not desirable to use only cm with a relative high deceleration level for the 32 km/h crash. Another example is plotted with the dotted lines, how to walk around. There is started with a 56 km/h crash velocity and a vehicle mass of 1100 kg. This gives an impact energy of 133 knm. In case of 80 cm available deformation length, the crash force is 166 kn and the average deceleration is 15.4 g. Although the stiffness normally increases during the crash and at higher crash speeds there is made use of the stiff engine; the only way to generate an optimal crash pulse at different collision speeds is variable structure stiffness. After detection of the crash velocity, the optimal stiffness of the longitudinal member should be realized.

19 General Introduction Figure 1.3. Walk around diagram of dynamic collision relations

20 20 Chapter Research objectives The objective of the research project presented here, was to design a concept structure that substitutes the conventional energy absorbing longitudinal members in a frontal vehicle structure and yields optimized deceleration pulses for different crash velocities and overlap percentages. To this aim the structure must have a stiffness that can be varied in accordance to the specific crash situations. The novel design presented in this thesis can cope with the following four crashworthiness problems: 1. In the case of a full overlap crash (both longitudinals and engine involved) as in the case of an offset crash (at 40 per cent overlap only one longitudinal directly involved) a similar amount of energy must be absorbed by the front structure. 2. In the case of an oblique load direction as in the case of an axial load direction a similar amount of energy must be absorbed by the front structure. 3. With a not much longer deformation length, much more energy must be absorbed at high crash velocities (resulting in less fatal injuries) and a lower injury level must be obtained at lower crash velocities. 4. A deceleration pulse must be obtained which is optimal (lowest injury level) for the concerning collision speed and the chosen dummy restraint parameters Research strategy Increased protection for the entire collision spectrum can be obtained by structures consisting of longitudinal members with an advanced geometric form, giving higher bending resistance without increasing the axial stiffness, in conjunction with a rigid connection between the front ends of these members. From several longitudinal square cross-sections, the influence of the width and thickness dimensions on the crash behavior is evaluated. The purpose of this study is to conceive an advanced geometric design for a longitudinal member optimized for a wide collision spectrum. The influence of various crash situations on the amount of energy absorbed by such longitudinal members will be discussed and representative crash tests are proposed. However, to reach a similar amount of energy absorption in case of an offset collision compared with a full overlap situation, additional measures are necessary. For this, the other longitudinal, which is not directly loaded, has to crumple axially as well. A cable system is introduced to perform axial shortening of the unloaded longitudinal with a tensile force to the rear. To prove the new concept, numerical simulations of a full overlap, a 40 per cent offset and an oblique 30 degrees crash with a complete

21 General Introduction 21 frontal car structure have been carried out. In these simulations, deceleration levels in the same order of magnitude are found. Since it has been demonstrated that it is possible to design a vehicle structure which generates a crash pulse that is almost independent of the crash direction and overlap percentage, it is useful to do an independent search for optimal pulses at several crash velocities, because the found structure-based pulses are not obviously the optimal pulses for minimal injury to the occupants. Therefore, the reverse question is answered: which crash pulse gives the lowest injury levels with an already optimized restraint system, instead of finding the optimized restraint system for a given crash pulse. For this research, a method is described in which a numeric model of an interior and a FEM dummy has been used to find the levels of the injury criteria. To compare the results of different crash pulses, an overall severity index has been used. From a described research an optimal pulse has been found after several considered pulse variations at a crash speed of 56 km/h. This pulse, used as example, deviates much from a traditional pulse, which shows normally an increasing stiffness of the structure near the end of the crash, but gives as it seems much lower injuries. During the first part of the deformation length the deceleration level can be high, then a low deceleration interval is desired, and at the end (dummy is restrained by belt and airbag) the deceleration can be high again. Also for other crash velocities, pulses are mentioned with adapted pulse characteristics for optimal results. Finally, new ideas are given how to further customize the energy absorption for different crash velocities to reach the optimal pulse. Therefore, an intelligent structure must be built which realizes such optimal pulses as closely as possible. In case of a high passenger floor (like an MPV), new design concepts are discussed. The research described in this thesis has a mean focus point on technically realizable design solutions, which are realistic but conscious not optimized initially for the weight and costs to find the highest technical potential of the proposed solutions Thesis outline The contents of this thesis are structured as follows. In Chapter 2, an inventory is given of parameters that determine useful crash situations. The influence of specific crash situations on the vehicle structure load is analyzed and arguments are mentioned which prove why at least two frontal impact tests are necessary for vehicle crashworthiness assessment. Two extreme crash situations are represented, which generally result in very different crash pulses and which cope with the just prescribed new legal test situations.

22 22 Chapter 1 To find optimal cross-section geometry of the new longitudinal, in Chapter 3 several basic forms are compared on energy absorption at different collision situations. For a square cross-section, the influence of thickness and perimeter on crashworthiness aspects is analyzed. From these results, a new concept is represented, which combines a higher bending resistance with an unchanged axial force level. In addition, the most efficient way of triggering to generate a stable folding process starting at the front part of the longitudinal is discussed. The manufacturing of the new longitudinal and a quasi-static experiment on a part as verification is described. The folding process is as expected. To absorb enough energy in offset collisions as well, an additional cable system is designed to operate with the new longitudinal structure. This system is described in Chapter 4. A numerical model of a vehicle front has been developed to simulate different crash situations. The results show that the system works (reported problems are mostly numerical) and that in each crash situation a similar deceleration curve can be obtained. In Chapter 5, a method for numerical crash simulations with a dummy is described, which can be used to find an optimal deceleration pulse. Structural designs are presented with the aid of which optimal pulses can be obtained. Especially a stronger deceleration in the first part of the crash needs an intelligent solution, in which the additional energy absorption is dependent on the crash velocity. Design concepts based on friction or based on hydraulic flow restriction are presented. Finally, in Chapter 6 the conclusions and recommendations are given.

23 The Necessity of Improved Crashworthiness Design 23 Chapter 2 The Necessity of Improved Crashworthiness Design 2.1. Introduction In the present chapter, the crash situation and the tests that are used to assess vehicle crashworthiness are reviewed. It will be shown that at least two frontal impact tests are necessary for frontal vehicle crashworthiness assessment. In recent years a large number of frontal crash tests have been reported (Auto Motor und Sport 1992, Consumentengids ) which were designed specifically to represent real crash situations. In these tests, many cars showed a crash behavior leading to unacceptable deformations of the passenger compartment yielding less survival space. Apparent reason for this is that insufficient design efforts were made to cope with the high and uneven mechanical loads occurring in such realistic collisions. The design was based on the until 1998 only compulsory full overlap test against a rigid wall. Improved frontal crashworthiness of cars necessitates additional design requirements, which take into account that the majority of collisions occur with partial frontal overlap, at oblique angles of incidence, and at velocities that deviate significantly from the regulated test speeds. In reality, the collision statistics ask for a design that is optimized for the entire range of frontal collision situations. Ideally, a vehicle's crashworthiness should be validated by multiple crash tests. Evidently, this is economically unfeasible. To solve this problem the introduction of a set of at least two compulsory tests are proposed in this chapter. If these chosen tests represent extreme cases within the frontal collision spectrum, crashworthiness in intermediate situations may then be interpolated from the respective test results. In new car development projects, the car body is optimized to comply with the compulsory crash tests as issued by the authorities. Consequently, the production car will show a sufficient level of crashworthiness in collisions that match or closely resemble those simulated in the tests. However, in collision situations that deviate from those represented in the tests, its actual performance is uncertain. Defining the proper crash tests is very important since it largely influences the crashworthiness design and, ultimately, determines the passive safety performance of the car. Although this topic also concerns side impact crashes, the discussion in this thesis will be limited to frontal crashes.

24 24 Chapter Frontal crash parameters Obviously, one must know the entire spectrum of all possible collision types in order to design a car that will be safe enough in any collision that may occur. Collision statistics reported by car manufacturers (Seiffert 1992, Justen 1993) and research institutes (Her Majesty 1991) and in the National Automotive Sampling System (NASS) (Stucki 1998) and the Fatality Analysis Reporting System (FARS) provide an extensive database of crash parameters. Important parameters are the collision speed, the obstacle type, the impact location and the impact direction. From this databases, suitable ranges can be defined which cover the collision situations most likely to occur. The following observations can be made: 1. At least 90 percent of all frontal collisions take place at speeds up to 56 km/h, see Figure Naturally, the number of different obstacle types is endless. However it appears that a major division can be made into three standard obstacle types, i.e. the rigid wall (simulating buildings or heavy trucks), the deformable barrier (simulating other vehicles), and the pole (simulating trees and pillars). 3. The majority of frontal collisions happen with frontal overlap percentages (the part of the bumper that makes contact with an obstacle) varying from 30 up to 100 per cent. 4. Frontal collisions are considered to occur in impact directions having angles of incidence with the longitudinal car axis varying from -30 degrees up to 30 degrees. How each of these crash parameters influences the proper design of the vehicle's safety structure will be briefly discussed.

25 The Necessity of Improved Crashworthiness Design 25 Cumulative frequency in % 100 % % % % Figure 2.1. Cumulative frequency of velocities in frontal collisions (source: Justen 1993). Velocity [km/h] Collision speed For optimal frontal crash behavior, all kinetic energy should be dissipated by the front structure. The lowest deceleration level of the passenger compartment is obtained, if the available deformation length in front of the car is as long as possible, see Figure 2.2. In this figure the needed deformation length at an average deceleration level is given. The upper curve is based on a completely linear relation between force and deformation length; the lower curve is based on a constant force independent of the deformation length, which gives the highest total energy absorption. In reality, the relation between deceleration level and deformation length will lie between these two curves. For a specific crash velocity, the optimal situation is achieved if the entire available deformation length is used without deforming the passenger compartment. This implies, that in a given vehicle concept the structure must have a specific stiffness which is determined by the relation between the crash energy at this velocity and the available deformation length. Higher velocities result in a higher level of kinetic energy, which cannot be fully dissipated by this front structure. Hence, the passenger

26 26 Chapter 2 compartment has to deform, which means that the necessary survival space cannot be guaranteed. Lower velocities will not use the whole available front structure causing the forces acting upon the occupant to be higher than necessary. In case of a deformable barrier, the barrier also absorbs energy and increases the total deformation length. So for a similar level of energy absorption in the vehicle structure, the crash velocity of the car against a deformable barrier must be higher as in case of a crash against a rigid wall F deceleration level [g] F S 20 S deformation at a speed of 56 km/h against a rigid wall [m] Figure 2.2. Average deceleration level as a function of deformation length Obstacle type The stiffness of the obstacle has a large influence on the crash behavior of a car. The stiff parts in the frontal structure of a car are the two main longitudinal members (possibly combined with the front wheel suspension) and the engine. They are responsible for absorbing large amounts of energy during a crash. A typical longitudinal member that collapses in a regular pattern can absorb about 25 per cent of the impact energy (De Santis 1996). If, in addition, the engine is forced to decelerate very rapidly and high

27 The Necessity of Improved Crashworthiness Design 27 forces originating from the engine deform the stiff firewall, about half of the impact energy will be absorbed by the remaining front structure (Hobbs 1991). One must note that this is only true in the case of a collision against a rigid wall, a special situation in which very high load forces can be directed into the stiff parts of the car structure. In the case of a deformable barrier, the barrier will generally not be capable of generating such high loads. Hence, the stiff parts cannot deform from the very start of the impact. This results in large deformations in the supple parts of the structure. Consequently the front structure will absorb less energy. This leads to intrusion and deformation of the passenger compartment. Despite the fact that the decelerations are lower, the amount of crash energy that must be absorbed stays the same. The so-called dynamic stiffness of the structure (resistance against a rapid change of kinetic energy) becomes lower, which causes the deformation to occur near the passenger compartment rather than at the front end of the car (Hobbs 1993). Figure 2.3 shows a rough estimation of energy absorption distributed on the frontal structure during a crash with 56 km/h against a rigid barrier (De Santis 1996, Leeuwen 1997). firewall longitudinals 5 % 10 % 10 % Second half 20 % engine 7.5 % 7.5 % First half 5 % 5 % 5 % 7.5 % 7.5 % 5 % front panel 5 % Figure 2.3. Estimated energy absorption percentages in the frontal structure Collision place and direction The overlap percentage determines which parts of the frontal structure of the car are hit and contribute to the energy absorption. In case of a full overlap against a rigid wall, the two stiff longitudinal members and the motion of the engine can absorb most of the

28 28 Chapter 2 energy. During the first half of the crash duration, mainly the longitudinal members will be loaded. In the second half, the engine will be loaded as well. In crashes with partial overlap percentages of 70 per cent and lower, not all available stiff structures are used for energy absorption (Justen 1993), see Table 2.1. Table 2.1. Relative energy absorption for several frontal crash overlaps against a rigid wall (see Figure 2.3). frontal overlap percentage stiff parts in the structure part of total energy absorption first half of crash duration part of total energy absorption second half of crash duration % 2 longitudinals + surrounding structure + engine / firewall 40-70% 1 longitudinal + surrounding structure + engine / firewall 50 % 50 % 25 % 35 % 30-40% 1 longitudinal + surrounding structure 25 % 15 % Research (Ragland 1991) has shown that in a car to car crash the percentages in the first column of Table 2.1 are even higher. Reason for this is that the car front has a nonuniform stiffness distribution, up to an overlap percentage of 50 per cent only one longitudinal member is absorbing energy, while the second member and the engine are not involved. A crash against a stiff pole can be regarded as a crash with a small overlap against a rigid wall. Only one stiff part e.g. one of the longitudinals or the engine will be hit. Changing the crash direction in a test against a rigid wall from zero to 30 degrees leads to a so called glance-off (the car grazes the wall and changes direction to move further) if no anti slide (vertical strips on the rigid wall where the car structure hooks on) is used (Justen 1993). The crash load on the longitudinal will be lower. With the use of an anti slide, the car will turn with the front towards the wall and the bending moment caused by the inertia load will become an axial load if the longitudinal has

29 The Necessity of Improved Crashworthiness Design 29 enough bending resistance. In general the 30 degrees impact presents a benign environment for a restrained occupant and did not reflect the impact conditions which lead to fatalities and serious injuries in the real world (Hackney 1985) Analysis of useful crash situations Accident analyses (Hobbs 1993) have shown that two-thirds of the collisions in which car occupants have been injured are frontal impacts, of which two-thirds occur in car to car accidents. The chances for collisions occurring with a full frontal overlap, or with an overlap percentage up to one-third, or with an overlap percentage from onethird up to two-thirds are comparable. This means that two-thirds of the frontal collisions use only part of the front structure. Also two-thirds of the collisions take place with impact directions under normal angle of incidence with the longitudinal car axis. Defining the proper design requirement for a considered crash velocity is very important. Crashes at speeds above this design velocity lead to a higher number of fatal injuries. On the other hand, crashes at lower speeds lead to a higher number of minor injuries. Setting the design speed at 56 km/h is sufficient for more than 90 per cent of the frontal collisions. This speed is compatible with current restraint systems. A test speed of 48 or 50 km/h is too low, since one-third of the offset crashes having 40 or 50 per cent overlap occur at higher speeds, see Figure 2.1. Summarizing, a suitable design crash situation will be crash at 56 km/h under normal angle of incidence against a deformable barrier with partial overlap. Accident investigations have also shown that the major cause of serious and fatal injuries in frontal car crashes is the intrusion into the passenger compartment (Hobbs 1993). High seat belt forces on the occupant lead to minor injuries, while contact with the car interior or penetrating parts leads to major injuries. As a consequence, it is better to design a car body that is a lit too stiff (leading to increased deceleration forces) than a car body that is a lit too supple (leading to decreased deceleration forces but with too much intrusions). Optimal safety implies on the one hand a sufficiently low deceleration level and on the other hand no intrusion of the passenger compartment under a wide range of crash situations. This is a design dilemma. The structure must be neither too stiff nor too supple. Referring to Table 2.1 it is clear that a full frontal overlap gives maximum energy absorption by the stiff parts. In a crash having an overlap of only per cent,

30 30 Chapter 2 merely half of the regular amount of energy is absorbed during the first half of the crash. During the second half, the amount is even worse, about 30 per cent of the regular value. Combining these percentages yields a total energy absorption of about 40 per cent of the regular value. This implies that the total stiffness of the structure available in this partial overlap crash is only 40 per cent of the stiffness regularly available in the full overlap crash. It must be remarked that a high stiffness is especially important near the passenger compartment. This 2 ½ times difference in stiffness affects the crash behavior of a car substantially. For the same available deformation length the deceleration level of a car in a per cent overlap crash is approximately 2 ½ times as low as the deceleration level during a full overlap crash. Suppose a car, designed for a full overlap crash, where the available deformation length at a specific collision speed would be fully utilized, becomes involved in a crash with a per cent overlap with the same speed and the same available deformation length. Then, 60 per cent of the crash energy will not absorbed by the front structure. This remaining energy must be absorbed by the passenger compartment. This is clearly unacceptable. Table 2.2 shows the mentioned differences. It also shows the differences for a reversed comparison: a car that is designed for a specific crash velocity in a per cent overlap crash that becomes involved in a collision with full overlap. Table 2.2. Approximate relative deceleration levels in two different crash situations with different design goals. Designed for Deceleration level at full overlap crash Deceleration level at 30-40% overlap crash Full overlap Reference = 100 % 40 % 30-40% overlap 250 % 100 % It will be clear that a 2 ½ times higher deceleration level leads to much higher forces on the occupant. In the reversed case of a deceleration level which is 2 ½ times lower than programmed, the car structure will not be able to absorb enough crash energy. This leads to high intrusion into the passenger compartment and probably to fatal casualties. An optimal crashworthiness requires a weighted design of the body structure for these two extreme cases.

31 The Necessity of Improved Crashworthiness Design 31 For the choice of a specific partial overlap percentage, the difference in consequences between a 30 and a 40 per cent overlap crash for the occupant must be considered. A 40 per cent overlap crash generates a higher deceleration level and more intrusion through the firewall in front of the occupant then a 30 per cent overlap crash. A 30 per cent overlap crash only leads to a higher intrusion level in the side structure (sill and door) and not in front of the occupant. In conclusion two extreme different design crash situations can be found, i.e. the full overlap crash and the 40 per cent overlap crash Optimal crash pulses The stiffness of the available front structure determines the deceleration pulse during a crash. This pulse should have a certain shape, ensuring minimal risk for the occupant. During a heavy collision, there are three important phases: 1. Crash initiation phase. In this phase, the sensor triggering for the belt pretensioner and the airbag must take place. For optimal sensor triggering, the front end of the car should be sufficiently stiff to generate within a short time interval a velocity change that lies above the trigger value of about 6 km/h. 2. Airbag deployment phase. In this phase, the airbag is inflated and the occupant tightens the belts while moving forwards with a relative velocity with respect to the car. This relative velocity should be sufficiently low, because in practice many injuries are the result of reaching a still inflating airbag or hitting the fully inflated airbag with a relatively high velocity. The deceleration of the car should be sufficiently low in this phase, implying that the stiffness must be relatively low. 3. Occupant contact phase. In this phase, the occupant has hit the airbag and there is a stiff contact between the occupant and the car. In this phase high decelerations may occur because the occupant will not be subjected to further shock loads caused by contacts with the interior. The frontal car structure should be stiff enough to decelerate substantially in the remaining time. Research (Brantman 1991) has shown that for optimal occupant safety in a collision with 48 km/h impact velocity, the first phase lasts between 10 and 30 ms, the second phase lasts 35 ms and the last phase fills up the remaining time to a total of maximal 90 ms. In the first and second phase, the optimal relative velocity values are 8 km/h each. Figure 2.4 shows a crash pulse as a function of time against a rigid wall with

32 32 Chapter 2 full overlap optimized for low injury values (HIC and Chest-G, see Section 5.2). This curve is achievable for a large number of current cars (Brantman 1991). Figure 2.4. Achievable optimal crash pulse at 48 km/h against a rigid wall. Figure 2.5 shows the three phases of a collision with impact velocity of 56 km/h in a velocity-deformation graph, calculated with the preceding graph (already optimized crash) but adjusted for the higher velocity. Because higher velocities do not significantly change the time duration (Faerber 1991), the same crash initiation time of 15 ms and an airbag deployment time of 35 ms are assumed. The crash duration is 90 ms with a total deformation length of 78 cm.

33 The Necessity of Improved Crashworthiness Design 33 velocity [km/h] g 9g 24g phase 1 phase 2 phase deformation length [cm] Figure 2.5. Deceleration level during an optimal frontal deformation at 56 km/h. Figure 2.6 shows two additional characteristics on both sides of the graph of Figure 2.5. The upper graph is an example of an offset (40 per cent overlap) crash and the lower graph is an example of a full overlap crash with the same vehicle. The two extreme characteristics are positioned as close as possible around the graph of the optimal (lowest injury values) crash situation. In this case the vehicle has a weighted design, the average deceleration of two extreme crash situations is optimal. In fact, a vehicle must be made stiffer as the optimal stiffness for a full overlap crash, to be sure it has enough stiffness in case of an offset collision. The deceleration levels represented by the upper graph are in the first two phases 50 per cent of the deceleration levels of the lower graph (14 g versus 28 g and 6 g versus 12 g) and in the third phase only 30 per cent (11 g versus 36 g). The reason is that the first two phases are comparable with the first crash duration phase mentioned in Table 2.1. The third phase is comparable with the second half of the crash duration. The deceleration level in each phase must change on the same deformation length. If the available deformation length of the stiff car structure in the lower graph is used, in this example at 64 cm, the deceleration level of the 40 per cent overlap crash should not be 30 per cent of the full overlap deceleration level any longer. It can be 36 g,

34 34 Chapter 2 which is as high as acceptable in the full overlap crash (Brantman 1991). This choice results in an 18 cm longer deformation length in case of a 40 per cent overlap crash. This remaining deformable structure (not used in a full overlap collision) must be very stiff to absorb much energy in a very short length. The total deformation length in this example is 82 cm, which may be acceptable for a 40 per cent overlap crash of middle class cars at 56 km/h. 60 Optimal 40% 100% velocity [km/h] g 21g 28g offset full 6g 9g 12g 11g 36g g 24g deformation length [cm] Figure 2.6. Deceleration level during a frontal deformation in three cases Representative crash tests as a design goal The above-mentioned extreme collision situations represent two different cases: A. A stiff front structure yields high deceleration forces. B. A supple front structure yields a high chance on intrusion into the passenger compartment.

35 The Necessity of Improved Crashworthiness Design 35 The differences can be even larger if case A is a full overlap collision against a rigid wall and case B is a 40 per cent overlap collision against a deformable barrier. In case of a crash against a deformable barrier, it is difficult to line up the stiff parts with the supple parts. Hence, the passenger compartment may be loaded before the stiff parts have absorbed enough energy. These extreme cases A and B form a very challenging design goal because a compromise (a weighted design) must be found. Figure 2.6 can be used as a design guide to define the stiffness of the front structure. In fact, all other less critical collision situations must lie between the extreme curves. Hence, the deceleration level of a certain collision can be found by interpolating (i.e. summarizing the involved vehicle parts) between the mentioned graphs and will be situated even closer to the optimal deceleration level. The only way to achieve the situation that all new cars have an acceptable crash behavior in different but realistic frontal crash situations is to define two compulsory tests that represent the two mentioned extreme cases as is done by the European Union since October 1998, see next section. If the test results for the most extreme collision situations are acceptable, the other less critical situations will naturally show good results because in the two compulsory tests the difference in the available structure stiffnesses is maximal. Case A can be considered as a deceleration test, important for belts and airbags. Case B can be considered as a structural test to ensure that intrusion is avoided and sufficient protection is given against irregular deformation. If the speed for the full overlap test against the rigid wall is set to 56 km/h, the test will be even more realistic. The speed for the other test of 40 per cent overlap against the deformable barrier must be a little bit higher to compensate for the energy absorption in the barrier Overview of actual and expected legal test requirements Prescribed research has contributed to the ideas and proposals for a new additional crash test in the European Union (TÜV Rheinland 1992). In Europe a new additional test requirement called EU Directive 96/79 EC is developed and is effective since October of 1998 for new types and models of vehicles, and October of 2003 for all new vehicles. It is an offset collision, involving only 40 per cent of the frontal structure of the vehicle, into a fixed deformable barrier with 56 km/h. A general summary of the current test requirements in the United States is given in Table 2.3 and for the European Union in Table 2.4. Also the additional more severely requirements of the NHTSA s (National Highway Traffic Safety Administration) New Car Assessment

36 36 Chapter 2 Program (NCAP) consumer information program are mentioned. Consumer groups have with their crash test publications a large influence on the vehicle customers by answering the question which car must be bought for more safety and therefore a large influence on car manufacturers to increase the vehicle safety. The European consumer organizations use the Euro NCAP requirements. Japan and Australia have safety requirements that are similar to requirement FMVSS No. 208 of the United States. This standard is most effective in preventing head, femur and chest injuries and fatalities. However, it does not directly address lower limb and neck injuries and it does not produce the vehicle intrusion observed in many real world crashes. The EU directive 96/79 EC has additional test dummy injury response criteria, particularly for the neck and lower limb. In the EU 96/79 EC offset test the lower extremities are loaded more. NHTSA and also Japan and Australia are currently assessing the additional safety benefits of adopting a supplemental regulation similar to the EU 96/79 EC standard. Table 2.3. Frontal crash test requirement in the United States. Requirement FMVSS No. 208 impact speed 48 km/h (NCAP 56 km/h) impact object (obstacle) fixed rigid barrier vehicle place and direction full frontal perpendicular and (not for NCAP) angles between +/- 30 degrees dummy type and conditions unrestrained and belt restrained (NCAP), 50 th percentile Hybrid III adult male injury criteria HIC 1000 chest deceleration 60 g chest deflection 50 mm femur force N

37 The Necessity of Improved Crashworthiness Design 37 Table 2.4. Frontal crash test requirements in the European Union. Requirement 74/297 EC 96/79 EC impact speed 50 km/h 56 km/h ( Euro NCAP 64 km/h) impact object (obstacle) fixed rigid barrier fixed deformable barrier vehicle place and direction full frontal perpendicular 40 % overlap of the vehicle width directly in line with barrier face dummy type and conditions no dummies belt restrained, 50 th percentile Hybrid III adult male injury criteria or structure criteria steering wheel intrusion horizontal and vertical direction 127 mm HIC 1000 chest deceleration 60 g chest deflection 50 mm femur force N additional criteria on chest (viscous), the neck, the knee, lower leg bending, foot/ankle compression and intrusion of compartment The 50 th percentile (mid-sized) male test dummy represents the mean of an adult male as specified for the total age group. Extending the test requirements to address the effects of occupant size on injuries could be expected. Especially the small 5th percentile adult female dummies may be more at risk by interaction with the airbag because they are in closer proximity than larger occupants (Park 1998). Another important issue is the compatibility of vehicles. There could be adverse effects on vehicle fleet compatibility after structural changes. A vehicle which has a stiffer or more aggressive front structure for his own increased frontal safety could be more dangerous for another car, especially if that other car is involved in a side impact crash. Also the use of the same fixed deformable barrier for light and heavy cars could lead to less compatibility in crashes between small and large cars. The amount of energy absorbed by the barrier is for a light car a larger proportion of the total crash energy as for a heavy car. To achieve a level of performance comparable to a small car, the front structure of the large car must be designed to crush more or to crush at a higher force level to absorb the additional energy. It is possible that a

38 38 Chapter 2 small car becomes softer because a lot of its energy was absorbed by the barrier. This is another reason why a second rigid barrier test is important. The increased crash velocity by Euro-NCAP from 56 km/h to 64 km/h has also a negative influence on the compatibility. This velocity increase yields a 30 per cent higher amount of crash energy. That means that for the same deformation length the force level and thus the stiffness of all cars has to grow with 30 per cent. This effect increases the absolute difference in force levels between light and heavy cars, which deteriorates the compatibility. Otherwise the test velocity must be higher as where collision statistics ask for, because for a comparable vehicle deformation as in a car to car crash the initial kinetic energy must be higher to compensate the absorbed energy in the barrier. Another interesting test for the compatibility problem is a test with a moving deformable barrier. Such a test simulates much better collisions between cars and could improve the fleet compatibility. In this case the smaller vehicle is subjected to a harsher crash environment due to the higher energy absorption and a higher velocity change yielding a stiffer structure. On the other hand the large car would be subjected to a less severe crash environment in terms of velocity change, so a softer front structure gives a temperate crash pulse. There could also be expected in the European Union a requirement on the compatibility with pedestrians. The large difference in mass between a vehicle and a pedestrian requires a soft front structure (bonnet, headlights, bumper) for easy deformation to lower the acceleration forces on the human body. Therefore the vehicle front must be made larger to cover the necessary stiff vehicle components Conclusions For the basic crashworthiness of vehicles two extreme frontal collision situations have been defined to be used as representative tests (and since October 1998 also compulsory tests in the European Union) for the most severe cases within the actual collision spectrum, showing a very different yet realistic crash behavior. If a car shows acceptable results in these two tests, it can be expected that the car has also acceptable crashworthiness in other frontal collision situations. Although there are much more collision configurations (e.g. 30 degrees collisions, compatibility between cars or with pedestrians, other dummy sizes), the mentioned two extreme crash situations have the largest impact on the vehicle structure design and this essential design problem must first be solved before further optimization with additional tests for less demanding crash situations is useful.

39 Numerical Design of Stable Energy Absorbing Longitudinal Members 39 Chapter 3 Numerical Design of Stable Energy Absorbing Longitudinal Members 3.1. Introduction A car has to pass the compulsory crash test as issued by the authorities. However, this regulatory test cannot guarantee that a car will be safe in all likely crash situations. Hence, the total collision spectrum within which the car must be safe must be investigated. For the design of a car structure that has to sustain a frontal collision, multiple aspects must be considered, i.e. collision velocity, crash direction, overlap percentage, and obstacle type. The energy dissipation in a frontal crash normally occurs by deformation of the longitudinal members. These must absorb a large part of the kinetic energy. In the present chapter, the design spectrum for a longitudinal member is further analyzed and a set of design requirements is formulated. With the aid of Finite Element Method (FEM) models of several longitudinal cross-sections, numerical simulations have been executed to evaluate the influence of the design parameters on the crash behavior. Based upon the results of these simulations, an advanced geometric form for a longitudinal member is presented Simulation parameters as design requirements To enable the geometric design of the longitudinal members, it must first be clear what kind of loads can occur in real frontal collisions. These loads are determined by the parameters describing the frontal crash and their predominant values (see Section 2.2). Besides the collision speed, the obstacle type, the impact location and direction, also the vehicle mass has a large influence on the crash behavior. An average car with two occupants and luggage has a mass of approximately 1100 kg. The goal for these simulation studies is to find a single geometric profile optimally suited for all the mentioned parameters rather than a profile that is optimized for a single specific crash situation.

40 40 Chapter 3 To limit the amount of simulations, the following simplifications were chosen in relation to the above mentioned parameters: 1. Two collision speeds, viz. 28 and 56 km/h. 2. Obstacle type: only rigid walls, although a deformable barrier is more realistic, it has a disadvantage in comparing energy absorption s of different geometry s. 3. Two extreme overlap percentages, to be sure a crash load will lead to acceptable energy absorption in extreme crash configurations: a. Full overlap (two longitudinal members and the engine are loaded) b. 40 per cent overlap (only one longitudinal member is loaded) (Witteman 1993). 4. Three impact directions: 0, 15 and 30 degrees. 5. Two vehicle masses: 550 and 1100 kg. The effective mass for a single longitudinal member depends on the specific impact location: In the extreme crash configuration of a full overlap collision, the effective mass changes from half of the entire vehicle mass (at the start of the crash the load is distributed onto two longitudinal members) into less than one third of the entire vehicle mass (the engine can take a considerable part of the load when it hits the barrier (Hobbs 1991)). To assume a constant cross-section of the longitudinal member, the simulations are confined to the first 350 mm from the front end. Under this condition a much simpler comparison between the various concepts is possible. At this specific length, the influence of the engine need not be considered. Hence, the effective mass is equivalent to half of the entire vehicle mass, i.e. 550 kg. In the extreme crash situation of 40 per cent overlap, the effective mass equals the entire vehicle mass, i.e kg (mostly the engine does not hit the barrier with high frontal peak loads in this situation (Justen 1993)). The mentioned values were used as input for the crash simulations: one longitudinal member against a rigid wall with an added mass of respectively 550 and 1100 kg, at a speed of 28 and 56 km/h and with a load direction of 0, 15 and 30 degrees. An optimized geometric profile must have the highest energy absorption per unit of mass for the total range of mentioned crash situations, with acceptable energy absorption for any individual crash situation. This is possible if one can guarantee a stable folding pattern for each parameter combination. Therefore, a good combination of profile thickness and perimeter for a specific shape is necessary. The following basic cross-sectional shapes have been studied: square, rectangle, circle, hexagon and

41 Numerical Design of Stable Energy Absorbing Longitudinal Members 41 octagon. To enable objective comparisons, the mass per unit of length is kept constant. Before the research is described, first a short overview is given of research done by others on crushing columns Research overview of crushing columns In the past there has been done a lot of research on crushing columns. Especially fundamental theoretical research on the mechanics of thin-walled structures is done by Wierzbicki (1983,1989), Abramowicz (1989) and Jones (1983), where relationships for the column width, wall thickness and shape geometry on the energy absorption s for different folding modes are derived. Wang (1992) and Yuan (1992) made analyses only on circular tubes, which could have different folding modes. For rectangular tubes experiments and geometrical folding analyses are done by Kim (1996). The influence of the used material of the column on the energy absorption is also interesting. Especially the strain rate dependency of steel for dynamic collisions (Beermann 1982, Behler 1991, Markiewicz 1996) or the crash behavior of aluminum tubes (Belingardi 1994) or comparison of aluminum and steel (Albertini 1996, Magee 1978, Wheeler 1998) are interesting. Comparison between different cross-section geometry s is mostly based on experiments with only a few different shapes (Groth 1991, Mahmood 1981). For a total overview of different cross-sections it is difficult to compare results of several publications, because the conditions like profile dimensions and material type are different. Also the shape of rectangular profiles is difficult to compare, in most publications additional flanges or stiffeners are used (Wheeler 1998, Kormi 1995) which have a less stable folding behavior as a basic rectangular shape. Giess (1998) describes a numerical simulation with a square profile where the wall thickness is varied over the cross-section yielding an optimized buckling load. Other research is about experiments on the way of welding box sections (or with adhesives), where the distance between the spot welds has influence on the folding behavior (Nishino 1992, Eichhorn 1984, Barbat 1995) or the influence of triggering (Krauss 1994, Yamaguchi 1985). In general most research is based on only an axial load, while more realistic load cases are with an angle of incidence (Crutzen 1996). From this literature overview it is clear that a practical research with usable design rules on all the mentioned basic shapes and in perspective of realistic crashworthiness requirements (variation of mass, velocity, shape and load direction) is not available. Especially for oblique load directions too little results have been published, and as already mentioned, it is difficult to compare research results based on different

42 42 Chapter 3 properties (material, thickness, load conditions, shape) for making an overview with all desired parameters where profiles can be compared under the same circumstances Simulation results The material selected for the five mentioned profiles was FeP03 (Euro), a commonly used steel (for specifications see Table 3.11). Over the length of the longitudinal member, the profile thickness was kept constant at 2.0 mm. This value is realistic and generally gives a stable folding pattern. The dimensions of the profiles were chosen to have the same perimeter resulting in a constant mass per unit of length, see Table 3.1. The undeformed length of each profile is 350 mm. Table 3.1. Dimensions of five profiles. profile shape perimeter [mm] A square 300 = B rectangle 300 = C circle 300 = π 95.5 D hexagon 300 = E octagon 300 = The results of the simulations (Baaten 1994), which concern a collision with a 56 km/h impact speed, a mass of 1100 kg, and a normal angle of incidence are shown in Figure 3.1. The simulation method used in this research is described in Appendix A. Both extremities of the column have an undeformable plate (rigid body), this is more realistic (normally they have a connection with stiffer vehicle components) and better for mutual comparison and to exclude different end effects. The energy absorption is plotted as function of the deformation length (rather than a function of time) because this facilitates the comparison of different structural design concepts. With the deformation length is meant the shortening of the profile. The simulation was terminated at the moment that the load shows a large increase as a result of reaching the maximum available deformation length. Figure 3.2 shows the deformed profiles.

43 Numerical Design of Stable Energy Absorbing Longitudinal Members 43 Figure 3.1. Energy absorption of five different profiles with a load direction of 0 degrees. Based upon these five simulations, it can be concluded that a square and a rectangular profile have significantly lower energy absorption than the other three profiles. The octagonal profile absorbs slightly more energy during the deformation than the circular and the hexagonal profiles, which absorb nearly the same amount of energy. The circular profile yields the longest possible deformation length and, hence, is capable of absorbing slightly more energy in its final deformation. This highest energy absorption for a circular profile is in agreement with the observations by other authors (Beermann 1982, Belingardi 1994, Groth 1991). Note that there were no geometrical imperfections used and the folds have to fit between the undeformable extremities.

44 44 Chapter 3 Figure 3.2. Five profiles with a different cross-section, deformed with a load direction of 0 degrees.

45 Numerical Design of Stable Energy Absorbing Longitudinal Members 45 The following simulations again concern a collision with a 56 km/h impact velocity and a vehicle mass of 1100 kg but here with an angle of incidence of 30 degrees. With an oblique load direction, numerical simulations of the cross-section of the rectangular profile are carried out in two orientations: standing (profile code B1) and lying (profile code B2). It can be expected that in lying position the higher bending resistance of the profile results in a better energy absorption. To avoid simulations with too much unrealistic bending, because a rigid fixation of the profile is assumed, the calculations were terminated after a deformation length of half of the original length. This is illustrated in Figure 3.3. In Figure 3.4 the simulation results are presented, the energy absorption as function of a maximum deformation length of 175 mm. These six deformed profiles are presented in Figure 3.5. Figure 3.3. Simulation with a load direction of 30 degrees until 175 mm.

46 46 Chapter 3 B2 B1 Figure 3.4. Energy absorption of six different profiles with a load direction of 30 degrees. From these simulation results, it can be observed again that the profiles: circular, hexagonal and octagonal perform much better than the square and rectangular profiles. The octagon has the highest energy absorption. Comparison of Figure 3.1 with Figure 3.4 shows that at the common deformation length of 175 mm the absolute amount of absorbed energy with a load direction of 30 degrees averages about 65 per cent of the energy absorption with a load direction of zero degrees. This is also evident from Table 3.2 and Figure 3.6 in which the percentages represent the energy absorption with respect to the energy absorption with a load direction of zero degrees for each profile. Only the rectangle in lying orientation looks less sensitive for variation of the load direction; the energy absorption decreases to 76 per cent with a load direction of 30 degrees.

47 Numerical Design of Stable Energy Absorbing Longitudinal Members 47 SQUARE (A) HEXAGON (D) RECTANGLE STANDING (B1) RECTANGLE LYING (B2) CIRCLE (C) OCTAGON (E) Figure 3.5. Six profiles with a different cross-section, deformed with a load direction of 30 degrees.

48 48 Chapter 3 Table 3.2. Energy absorption of six profiles with three different load directions and equal deformation length. profile shape energy absorption [Nm] 0 degrees 175 mm deformation A square % B1 rectangle % energy absorption [Nm] 15 degrees 175 mm deformation % % energy absorption [Nm] 30 degrees 175 mm deformation % % B2 rectangle see B % % C circle % D hexagon % E octagon % % % % % % % It is noted that if the simulations for all profiles had been executed to include a longer deformation length the difference in energy absorption between normal and oblique would be even larger because no further buckling of the material is possible under this oblique direction. The group of simulations at the same velocity, with the same mass but with a load direction of 15 degrees gives, as expected, an energy absorption value between the values presented in the Figures 3.1 and 3.4, see Table 3.2 and Figure 3.6. In this situation, larger differences between the various profiles of the decrease of absorbed energy are shown with respect to the energy absorption with a direction of zero

49 Numerical Design of Stable Energy Absorbing Longitudinal Members 49 degrees. The energy absorbed by the rectangle in lying orientation decreases only to 93 per cent whereas the octagon decreases to 73 per cent, see Table percentage hexagon octagon circle rectangle lying shape 10 rectangle standing impact angle 0, 15, 30 degrees 30 square Figure 3.6. Percentage energy absorption decrease as function of the angle of incidence of six different profiles, based on Table 3.2. In general, it can be concluded that for most profiles with load direction variations the energy absorption decreases with a similar percentage (viz per cent) if the load direction is increased from zero to 30 degrees. Only the rectangular profile especially in a lying orientation is significantly less sensitive for variations in the load direction (see Table 3.2, 24 per cent). The remarkable difference between the less energy absorbing profiles (the square and the rectangle) and the more energy absorbing profiles (the circle, hexagon and the octagon) is found with each load direction (see Table 3.3 and Figure 3.7 in which the percentages represent the relative energy absorption between the profiles for each load direction).

50 50 Chapter 3 Table 3.3. Energy absorption of six profiles with three different load directions. profile shape energy absorption [Nm] 0 degrees max. deformation A square % B1 rectangle % energy absorption [Nm] 15 degrees 175 mm deformation % % energy absorption [Nm] 30 degrees 175 mm deformation % % B2 rectangle see B % % C circle % D hexagon % E octagon % % % % % % %

51 Numerical Design of Stable Energy Absorbing Longitudinal Members percentage hexagon octagon circle rectangle lying shape 10 rectangle standing impact angle 0, 15, 30 degrees 30 square Figure 3.7. Relative percentage energy absorption between 6 different profiles as function of the angle of incidence, based on Table 3.3. To verify the influence of the mass and the crash velocity, the last group of simulations are carried out with a different mass or at a different velocity with respect to the reference situation of 56 km/h and 1100 kg. A profile with a relatively low and a profile with a relatively high energy absorption are chosen: the square and the octagon.

52 52 Chapter 3 Figure 3.8. Energy absorption of a square profile with two different masses and at two different velocities. From the simulation results presented in Figure 3.8, it is clear that for the square profile reducing the mass from 1100 kg to 550 kg does not give a significantly different value for the energy absorption within the considered deformation length. This is also true for the octagonal profile in Figure 3.9, although at the end of the deformation some difference occurs.

53 Numerical Design of Stable Energy Absorbing Longitudinal Members 53 difference in deformation length Figure 3.9. Energy absorption of an octagonal profile with two different masses and at two different velocities. Reducing the crash velocity from 56 km/h to 28 km/h for the square profile (see Figure 3.8) yields a slightly lower energy absorption, this is due to the fact that the material behavior is strain rate dependent: at higher deformation speeds the profile acts slightly stiffer (Hop 1993, Radlmayr 1993). For the octagon a similar difference in energy absorption is observed (see Figure 3.9). Note that this difference is especially due to a somewhat shorter deformation length at lower velocity. The energy absorption at 28 km/h averages about 91 percent of the energy absorption at 56 km/h for both profiles. The simulations with both profiles result in the same conclusion that reducing the mass does not have much influence on the limited energy absorption and reducing the crash speed yields a comparable decrease of energy absorption in both profiles. It seems to be allowable that only these two profiles as representatives were simulated and that

54 54 Chapter 3 the conclusion may be general for all profiles with a load direction of zero degrees, because there is also no physical reason that the geometry could have a significant influence. Simulations with a load direction of 30 degrees (not presented here) result in the same conclusion (Baaten 1994) Discussion of the simulation results The simulation results show clearly (see Figure 3.7) that the circular, hexagonal and octagonal cross-sections offer a much higher energy absorption for each of the considered load directions than those offered by the square and rectangular profiles. These results were also expected, because the much energy absorbing profiles deform mostly in a symmetric way: all the material in a cross-section has to deform. However, the circular profile seems less stable, after a few axis-symmetric modes the non axissymmetric diamond mode is created. In case of the square and the rectangle an asymmetric deformation mode exist, strong deformation is only at the corners, in a fold only energy is absorbed in the plastic hinges. The differences between the three higher energy absorbing profiles are marginal (see Table 3.3). Hence, from a crashworthiness point of view no dominant choice can be made. They are also about equally sensitive for variations in the load direction. Although the rectangular profile absorbs the least energy, this profile in lying orientation is the least sensitive for variations in the load direction (see Figure 3.6). Variation of mass does not noticeably influence the energy absorption of the profiles (assuming that the considered length is fully deformed). Reducing the crash velocity from 56 km/h to 28 km/h results in an energy absorption decrease of about 9 per cent for the same considered deformation length. To put these conclusions into the practice of designing a car that has to sustain a wide range of frontal collisions, it is advocated to use a rectangular profile in a lying orientation instead of the profiles which are expected to absorb significantly more energy. The reason is that it is more important to give adequate protection for the whole range of load directions rather than to optimize the longitudinal member for minimal mass only. This rectangular profile, which must be constructed heavier to match the energy absorption of the well absorbing profiles with a load direction of zero degrees, will offer a much better energy absorption at oblique load angles in return. For the same amount of energy absorption as the circular profile, the thickness of the rectangular profile must increase with about 2 resulting in about 700 gram added mass (see Relation 3.4).

55 Numerical Design of Stable Energy Absorbing Longitudinal Members 55 It is concluded that with the cross-sectional shapes presented until now it is not possible to meet the various demands with regard to absorption potential to compensate for changes in crash velocity and overlap percentage. One profile can not absorb the different amounts of energy of various crash situations. A possible solution for this challenging design problem is a rigid connection between the front ends of the two longitudinal members. If only one longitudinal member is loaded axially (i.e. in cases of per cent overlap) the other longitudinal member can be loaded with a transverse load as a result of the tensile force within the rigid connection. This load results in extra deformations for instance through bending of the unloaded front side (see for example Figure 3.34). In this case, additional energy absorption is possible. The rigid transverse bar can also load the engine in case of a per cent overlap. If the longitudinal members are very stiff in the transverse direction, the mentioned factor of 2 ½ stiffness difference (see Table 2.2) can be substantially reduced (17 per cent more energy absorption, see Section 4.3). In Section 3.7, a design study is described for a longitudinal cross-section that meets these demands Design of the triggering for a constant stable force level A stable force level of the crushing column over the whole length can be reached by triggering of the beam. By applying specific weaknesses on the proper position at the front end of the beam, a stable regular folding process will start at that position without a much higher peak force level to introduce the first fold. These crash initiators prevent overly high loads, which could cause a bending collapse of the still long undeformed length, or other not programmed deformations of the structure. The folding process starts controlled at the front end and proceeds regularly towards the rear end, giving more stability. In this case, the whole length of the column could be used for folding lobes of the same size, finally leading to the highest energy absorption. In the case of bending, only a few folds could be formed. See Figure 3.10 for the difference in deformation between an equal square crash column of 1160 mm length and one with and the other without triggering for an axial load. The perimeter is 300 mm and the wall thickness is 2.0 mm. For clearness the right picture is zoomed in. Note that this is a special example with a relatively long column, yielding lateral buckling of the not triggered column. In case of a shorter column where no lateral buckling occurs, the final difference in energy absorption is smaller.

56 56 Chapter 3 Column without triggering during two time steps. Column with front triggering. Figure Folding of a not triggered and a triggered longitudinal member. In Figure 3.11, the energy absorption of both longitudinals is compared. Because the column without crash initiators is stiffer, it absorbs more energy during the first deformation part but afterwards the additional energy absorption is strongly reduced by bending, resulting in a lower overall energy absorption. In Figure 3.12, the much lower force level of the triggered beam can be seen at the start of the crash. 4.5 x Energy absorption [J] Deflection [mm] Figure Energy absorption of a longitudinal member with (-----) and without ( ) triggering.

57 Numerical Design of Stable Energy Absorbing Longitudinal Members Force [kn] Deflection [mm] Figure Force level of a longitudinal member with (-----) and without ( ) triggering. In the following three subsections research is described to find the optimal position for the crash initiator at the front end of the longitudinal, which form of initiator is the most efficient, and what is for the found weakness the best dimension Determination of the optimal trigger position The initiator must be positioned in such a way that as much as possible material contributes to the energy absorption. Therefore the position where the first fold will be formed must be determined. This position is a function of perimeter and wall thickness. For this research, 25 square geometry s as mentioned in Table 3.4 are used. The optimal position for a weakness is where the first top of a fold (a plastic hinge) is formed. Although this is determined by the folding wavelength, it also depends on the fact if the front end of the longitudinal has a rigid connection, by example with the cross beam, or if it is free to deform. In the last case, a shorter distance from the front end to the first top of a fold could be expected. In case of a rigid connection the wall near the weld will not be able to deform, so a longer distance is necessary. The folding wavelength is defined as the distance from one fold top to the next fold top on the adjoining perpendicular side. See Figure 3.13 for an illustration.

58 58 Chapter 3 4Hfold=λ 4Hfold=λ Figure Definition of the plastic folding wavelength 4H fold. The half length of the plastic folding wavelength 4H fold can be calculated with the next formula (Abramowicz 1989, Wierzbicki 1983, Wierzbicki 1990) with thickness t and edge dimensions a and b. 2 H fold 3 a + b = t (3.1). The length of the front end of the column until the first fold top is determined by plotting the deformed geometry of the longitudinal cross-section. In Figure 3.14, the measurement is shown. d Figure Longitudinal cross-section of a folded column for determining the distance of the first fold to the front end of the column.

59 Numerical Design of Stable Energy Absorbing Longitudinal Members 59 In Table 3.4, the measured values are present for each geometry and also the percentage of the folding wavelength as calculated later in Table Table 3.4. Distance from the first fold top to the rigid front end & percentage of the folding wavelength of 25 different longitudinal cross-sections. thickness perimeter [mm] perimeter [mm] perimeter [mm] perimeter [mm] perimeter [mm] [mm] 4 25 = = = = = 400 Distance from Distance from Distance from Distance from Distance from first fold top to first fold top to first fold top to first fold top to first fold top to column end column end column end column end column end [mm] & [mm] & [mm] & [mm] & [mm] & Percentage of Percentage of Percentage of Percentage of Percentage of folding wave folding wave folding wave folding wave folding wave length length length length length 1 7.9, 49% 10, 45% 14, 52% 20, 54% 20.2, 45% , 54% 12, 48% 14.5, 48% 21, 53% 25.3, 49% , 53% 15.1, 56% 16.5, 50% 20.3, 46% 27.5, 52% , 53% 15.7, 51% 22, 58% 27.5, 56% 30.9, 52% , 50% 18.8, 54% 23.4, 56% 30, 56% 32.2, 49% For comparable accuracy, larger columns are simulated with larger finite elements. Of course the geometry of the fold is influenced a little bit by the mesh size (see Figure 3.14), which introduce some inaccuracy. In addition, the used wavelengths were rounded off, which could explain the deviations around an average of 50 per cent. Nevertheless, in general the conclusion is that in the case of a rigid front connection the first weakness point must be positioned at half a folding wavelength. The formula for the folding wavelength could also be used for a rectangular crosssection. To see if the conclusion is also valid for rectangular cross-sections, comparable simulations are done (van der Poll 1996). The result is also half of the folding wavelength. To check if a crash initiator placed on this position guarantees a stable folding pattern, simulations have been carried out. Figure 3.15 shows a simulation of a rectangular cross-section 100x50x2 mm with and without triggering.

60 60 Chapter 3 Figure Folding pattern of a rectangular column with (l) and without (r) triggering. For simulations with a square column without a rigid front end, the same crosssections are simulated. From these simulations the conclusion could be made that the average distance from the column front end to the first fold top was 35 per cent of the folding wavelength Determination of the most efficient trigger geometry Two types of crash initiators are researched, weaknesses formed by pressing a stamp into the wall or corner, or weaknesses formed by punching (taking material away). All simulations are done with a square column of 75x75x2 mm with a rigid front end and with a crash speed of 56 km/h, resulting in an initiator distance of ½ 44 = 22 mm. In the upper row of Figure 3.16, four stamp forms are shown and in the lower row four punch forms.

61 Numerical Design of Stable Energy Absorbing Longitudinal Members 61 bead initiator diamond notch spheres plastic fold smaller thickness circular notch circular holes oval hole Figure Eight types of initiators positioned on a square column. To compare the simulation results, all initiators cause a surface of the cross-section area or residual perimeter reduction of 10 per cent. This reduction factor is also used by others (Krauss 1994). This is made clear in Figure 3.17 and A bead initiator 0.9 a a total area A=a.a Figure Decrease of the cross-section area of a bead initiator.

62 62 Chapter 3 perimeter 3.6a circular hole initiator a total perimeter 4a Figure Decrease of the perimeter at places where material is taken away. In Figure 3.19, the force levels during simulation of the stamp forms are viewed. The column without an initiator has the highest initial peak load. With initiator the peak loads are clearly reduced. The bead initiator has the lowest peak load. The distance between two force peaks is an indication for 72.5 per cent of the folding wavelength (the remaining length of folds pressed together is 27.5 per cent of the original undeformed length). The repeatability of about the same distance in the following force peaks is an indication for a stable regular folding pattern. It is clear that the not triggered column starts with an irregular unstable fold.

63 Numerical Design of Stable Energy Absorbing Longitudinal Members force [kn] deflection [mm] no initiators Figure Development of force for square columns with initiators formed by pressing a stamp. In Figure 3.20, the force levels of the punch forms are viewed. Again the peak force of the triggered columns is lower but the force decrease is less compared with the stamp forms. Also all triggered columns show a less regular and stable force curve, the folding wavelength is difficult to detect. Note the strong force level decrease after the first peak force of the profile with punched circular notches, weakening the stiff corners too much has a large influence on the initial stability.

64 64 Chapter force [kn] deflection [mm] no initiators Figure Development of force for square columns with initiators formed by punching. In Table 3.5, the peak load values and the energy absorption s of the columns after 175 mm deformation are mentioned. Because only a limited length of the column is deformed, the energy absorption of the initiators with less or no force peak reduction have more energy absorption due to the higher average force level. If the initiator is more effective, the local weakening of the cross-section is the reason that the force level and therefore the energy absorption decrease. Of course stability and a regular force level are more important than the energy absorption, because higher energy absorption can also be reached by increasing the wall thickness by a small amount.

65 Numerical Design of Stable Energy Absorbing Longitudinal Members 65 Table 3.5. The first peak load level and the energy absorption of columns after 175 mm deformation with different crash initiators. Initiator form First peak load [kn] Energy absorption [J] no initiator bead diamond notch spheres plastic fold smaller thickness circular notch circular holes oval hole To make the final choice for the most suitable trigger form, the visualization of the deformed columns is also important. From the force curves it is not clear where the folds are formed and how regular it looks. In Figure 3.21 the deformation of eight different triggered columns are presented, the upper row are the pressed forms and the lower row are the punched forms.

66 66 Chapter 3 bead diamond notch spheres plastic fold smaller thickness circular notch circular holes oval hole Figure Deformed geometry s of eight different trigger forms. From these pictures, it is clear that especially the notches on the corner and the both forms with holes are not stable. The relative stiff corners of a square column must not be weakened, because this weakening could cause a small rotation or translation of the whole cross-section. If a deviating fold is formed, it disturbs all next folds with a much higher risk of a bending collapse due to the decreasing moment of inertia. Based on the lowest first peak level, the most regular force curve with expected wavelength and the very stable folding pattern as visible, the bead initiator must be preferred as most suitable trigger form Determination of the best dimension of the bead initiator The objective of triggering is to weaken the cross-section in such a way that the folding pattern develops from the front end to the rear. Weakening too little gives probably not the desired folding pattern and the initial peak load is not reduced enough. The still high load could use the weakening for an irregular fold. Too much weakening is useless because the energy absorption decreases and the stiffness in

67 Numerical Design of Stable Energy Absorbing Longitudinal Members 67 that cross-section could be too low for enough bending resistance. For this research, several square perimeters are used, but only the results are shown of the 25x25x2 mm square because it gives the clearest difference between the used cross-section area reductions of 0, 5, 10 and 15 per cent. In Figure 3.22, the force levels are displayed and in Table 3.6, also the energy absorption s at 175 mm deformation. It is clearly seen that a larger cross-section area reduction gives a lower first peak force. Because of instability at 0 and 5 per cent reduction, the energy absorptions are lower than in the case of a 10 per cent reduction. At 15 per cent reduction, the folding process is very regular but the energy absorption is also lower because the peak loads are lower force [kn] deformation length [mm] no initiator % bead 10 % bead % bead Figure Development of force for square columns with different sizes of bead initiators.

68 68 Chapter 3 Table 3.6. First peak load and energy absorption of columns with different bead size. Dimension of bead initiator [percentage of crosssection area reduction] First peak load [kn] Energy absorption at 175 mm deformation [J] no initiator % bead % bead % bead In Figure 3.23, the deformed columns are viewed to see the instability of the folding process of columns without or with only 5 per cent triggering. To see where the folding process goes wrong the plot is given in an early stadium of the deformation process. The deformation of the 10 and 15 per cent bead is very regular. no initiator 5 % bead 10 % bead 15 % bead Figure Deformed columns (25x25x2) with bead initiators of different sizes. The general conclusion also for other perimeters is that 5 per cent bead initiators do not always show a stable folding process or start at the desired position. The simulation results for 10 and 15 per cent bead initiators are not so different. They are

69 Numerical Design of Stable Energy Absorbing Longitudinal Members 69 regular but the energy absorption for 15 per cent is somewhat lower. Since 10 per cent reduction gives stable enough results and further weakening of the cross-section is not necessary, a 10 per cent bead initiator is preferred Design of a longitudinal cross-section In designing a longitudinal member for optimal energy absorption, it is important to obtain a maximum value for the product of the mean crash load and attainable deformation length. To guarantee that the required average force level is maintained over the entire available deformable length of the column, a steady progressive collapse pattern must be maintained in the column (Mahmood 1981, Thornton 1983). It is very important that the folding pattern rapidly converges to a stable, repeatable mode. If the progressive folding pattern is not sufficiently stable, the folding process can easily be disturbed and the structure will fail under a premature bending collapse. In this case, no further energy absorption by progressive folding is possible. To reach a stable progressive collapse pattern for many progressive lobes, an asymmetric folding pattern is preferred. Wierzbicki (1989) has analyzed that an asymmetric folding mode exists with internal angles of the cross-section < 120. This is also observed in our numerical simulations in which hexagonal, octagonal and circular cross-sections (internal angle of the cross-section 120 ) mostly fold in a less regular symmetric mode, and the square and rectangular cross-sections fold in a stable asymmetric mode (see Figure 3.2), despite having a lower energy absorption at the same mass. As seen in our research (Witteman 1994) a square cross-section under axial load can absorb more energy at the same mass than a rectangular cross-section (see Table 3.3). Because stability is more important than energy efficiency, a square cross-section will be used as baseline element for further improvements. The rectangular cross-section, which performs better under oblique load directions, will now not be chosen for the reason that compactness of the profile is desired. For the required bending stiffness another solution will be proposed in Section 3.8. To prevent a bending collapse of the longitudinal member in case of an oblique impact angle, its stiffness has to be increased to yield a higher bending resistance. Higher bending resistance, causing non-axial forces to be transformed into axial forces, can be reached by larger width and increased thickness of the cross-section. For a square cross-section the bending resistance W in undeformed state can be calculated from (PBNA 1993)

70 70 Chapter ( ) with H is the width outside and h is the width inside. W = H h 6H [mm 3 ], (3.2) To see in a well-ordered way the influence of various realistic width and wall thickness dimensions on other important crash behavior aspects, twenty-five different square cross-sections have been analyzed with the aid of numerical simulations (Witteman 1995). For a fast calculation of these simulations of basic profiles with only an axial load, use is made of the program CRASH CAD. It is based on super folding elements, using the theory of Abramowicz and Wierzbicki. The results are shown in Tables The width and the thickness are varied within realistic values attainable within passenger cars. The material used for the longitudinal member is again FeP03 and the length of the column is 5 times the folding wavelength, but this has no influence on the researched aspects. The energy absorption for a specific longitudinal member can be found by multiplying the mean force with the total deformable length of the longitudinal member. The values for the mean force and the energy absorption are calculated for a quasi-static axial load. For a dynamic crash situation of 56 km/h, these values must be multiplied with a mean factor of average 1.5, due to the strain rate dependency of the simulated material. This factor is calculated with the constitutive relation of Cowper and Symonds, describing the influence of the strain rate on the increase of the force level in dynamic impacts, with strain rate parameters D = 1300 and p = 5. This relation is often used in automotive crash simulations (Behler 1991, Du Bois 1987, Richter 1993). Formula 3.3 (Jones 1983) must be used to calculate the difference between the dynamic σ ydyn and the static load σ ystat with ε is the strain rate. It has no noticeable influence on the number and geometry of the folds. σ ydyn / σ ystat = 1 + (ε / D) 1/p (3.3).

71 Numerical Design of Stable Energy Absorbing Longitudinal Members 71 Table 3.7. Quasi-static mean force of 25 different square cross-sections during deformation of longitudinals. thickness [mm] perimeter [mm] 4 25 = 100 Mean Force [kn] perimeter [mm] = 150 Mean Force [kn] Perimeter [mm] 4 50 = 200 Mean Force [kn] perimeter [mm] 4 75 = 300 Mean Force [kn] perimeter [mm] = 400 Mean Force From these calculations the following conclusions can be drawn: 1. The wall thickness has a large influence on the energy absorption. Increasing the thickness from 1 to 4 mm with unchanged perimeter results in a more than tenfold increase in energy absorption (mean force multiplied with deformation length), see Table 3.7. Increasing the width with unchanged wall thickness results in a less than proportional increase of energy absorption. The following exponential relation between energy absorption E, profile thickness t and column width H is observed, which is based on the theory of Wierzbicki ( ) : [kn] E ~ t H (3.4). Adding the wish for a minimal mass design, a cross-section having a small width and a large wall thickness is the most efficient choice. This can also be seen in Table 3.8 where the cross-section with the smallest perimeter and the largest wall thickness has the highest mass specific energy absorption. This conclusion is confirmed by the theory, if relation 3.4 is divided by the cross-section area (~ t H ), the specific (mass or volume independent) absorbed energy relation becomes: E s ~ t H (3.5). E s

72 72 Chapter 3 Table 3.8. Mass specific energy absorption of 25 different longitudinal cross-sections after quasistatic deformation. thickness perimeter [mm] perimeter [mm] perimeter [mm] perimeter [mm] perimeter [mm] [mm] 4 25 = = = = = 400 Energy Energy Energy Energy Energy absorption/ absorption/ absorption/ absorption/ absorption/ mass mass mass mass mass [kj/kg] [kj/kg] [kj/kg] [kj/kg] [kj/kg] It is readily seen that the energy efficiency is the same for all cross-sections that have a common perimeter to wall thickness ratio, which can also be seen from Relation 3.5. For a constant ratio of thickness and width, the specific energy absorption remains constant due to the similar exponents but with opposite sign of the thickness and the width. In Table 3.8, this is a.o. the case for the five crosssections on the diagonal. The mass specific energy absorption is 16 kj/kg for all five cross-sections. 2. The bending resistance (calculated with Formula 3.2 where the perimeter is four times the middle of the width inside and outside) of a longitudinal varies nearly linearly with the wall thickness and nearly quadratically with the profile width, as can be seen in Table 3.9. The most efficient choice for a high bending resistance is therefore, a cross-section with a large thickness and profile width. However, there are limitations to increase the wall thickness, a thick wall has a relatively high energy absorption in axial direction (see Table 3.7), causing the structure to be too stiff in a full overlap frontal crash. Note that a dynamic mean force of 100 kn yields a vehicle deceleration only caused by the two longitudinals of already 18.5 g. Increased width in conjunction with reduced thickness may result in instability. A larger width gives longer lobes and a higher force amplitude.

73 Numerical Design of Stable Energy Absorbing Longitudinal Members 73 Table 3.9. Bending resistance of 25 different longitudinal cross-sections. thickness perimeter [mm] perimeter [mm] perimeter [mm] perimeter [mm] perimeter [mm] [mm] 4 25 = = = = = 400 Bending Bending Bending Bending Bending resistance resistance resistance resistance resistance [mm 3 ] [mm 3 ] [mm 3 ] [mm 3 ] [mm 3 ] The total deformable length of a longitudinal member divided by the folding wavelength (see Equation 3.1) gives the number of folds after full deformation. In Table 3.10 can also be seen that increasing the thickness or the width results in a less than proportional increase of the folding wave length. Cross-sections with small width or thickness give more folds than cross-sections with large width or thickness at the same crush length. Table Folding wavelength of 25 different longitudinal cross-sections after deformation. thickness perimeter [mm] perimeter [mm] perimeter [mm] perimeter [mm] perimeter [mm] [mm] 4 25 = = = = = 400 Folding wave Folding wave Folding wave Folding wave Folding wave length [mm] length [mm] length [mm] length [mm] length [mm] Cross-sections with a common value of the perimeter to wall thickness ratio also have a common value for the width to the folding wavelength ratio and a common

74 74 Chapter 3 modulation depth (relative fluctuation around the mean value) of the force level (not presented in the table). For the stability it is important that the bending resistance during deformation is kept high enough to prevent a premature bending collapse. A small thickness in conjunction with a large width yields a folding wavelength that is a smaller percentage of the width. This implies a relative smaller disturbance of the bending resistance due to a smaller decrease of the cross-section width during deformation, because folds moves relatively less inside. 4. The calculation results also show that profiles with large wall thickness and small profile width give a low relative fluctuation of the crushing force (i.e. a more stable folding process), while profiles with small wall thickness and large profile width give a large relative fluctuation of the crushing force (i.e. a less stable folding process). To summarize: To determine the optimal wall thickness in relation with the optimal perimeter of the square cross-section, the following crash behavior aspects have to be considered: 1. The highest mass specific energy absorption is reached with a large wall thickness in conjunction with a small perimeter value (bottom left cell in Table 3.8). 2. The highest bending resistance (necessary to prevent a bending collapse in case of an off-axis load) is reached with a large wall thickness in conjunction with a large perimeter value (bottom right cell in Table 3.9). 3. The bending resistance decreases less in a fold if the folding wavelength is a smaller percentage of the profile width. The lowest relative disturbance of the bending resistance is reached for a small wall thickness in conjunction with a large perimeter value, due to a smaller decrease of the cross-section width during deformation (top right cell in Table 3.10). 4. The most stable force level is reached with a large wall thickness in conjunction with a small perimeter value (bottom left cell in a table). From these results it is clear that it is impossible to select a profile that simultaneously meets the following four design criteria. See also Figure 3.24.

75 Numerical Design of Stable Energy Absorbing Longitudinal Members Lowest mass (highest energy absorption). 2. Highest bending resistance. 3. Lowest relative decrease (disturbance) of the bending resistance in a fold. 4. Most stable force level. Increasing perimeter Increasing wall thickness Figure Design paradox, four design requirements resulting in different perimeter / wall thickness choices. The choice between the various cross-sections is also restricted by the desired energy absorption, i.e. a choice can be made between cross-sections with similar mean force values. For example a cross-section of or a cross-section of 100-2, see Figure However, it is possible to combine some small cross-sections. For example, four separate profiles have the same energy absorption as a single profile 200-2, see Figure In this case the smaller profiles could be optimized without decreasing the bending resistance, because if the four profiles have at the extremities a stiff connection to each other the total structure could give more bending resistance and stability as a single profile with the same total mass. For optimal bending resistance as much as possible material must be positioned far from the central line. Disadvantages are of course the increased outside dimensions which have to fit inside the vehicle front and increased costs.

76 76 Chapter 3 100x100x1.5 25x25x2 Figure Example of two profile cross-sections yielding nearly the same energy absorption. 50x50 x2 25x25x1 Figure Example of a profile cross-section with the same energy absorption as four smaller profiles together. Other possibilities to influence the bending resistance or the outside dimensions are changing the orientation of the four profiles or an increased outside wall thickness in combination with a decreased inside wall thickness for the same total mass, see Figure However, in the last case unstable folding could be expected and it is more expensive to manufacture. 25x25x1 25x25x1 25x25x1.5/0.5 A B C Figure Example of four cross-sections (A) in different orientations (B) and other wall thickness distribution (C) yielding a different bending resistance but with the same total mass.

77 Numerical Design of Stable Energy Absorbing Longitudinal Members 77 Changing with the same mass the square cross-section into a lying rectangular crosssection increases the bending resistance in the horizontal plane while it decreases in the vertical plane. This is more extreme in case C of Figure 3.28, where also the thickness is increased of the most outside vertical walls while reducing the wall thickness of the horizontal walls. Dimension changes from A to B or C must be not too large to prevent folding instability and a decreased Euler buckling load due to a too much decreased bending resistance in the vertical plane. 50x50 x2 60x40x2 60x40x 1.33/3 A B C Figure Example of a cross-section (A) which has an increased bending resistance (in horizontally plane) in case of a rectangular cross-section (B) or more increased with changed wall thickness distribution (C). Although the examples in Figures have potential in finding a more optimal solution, large improvements could not be expected due to the contradictory design requirements. For the crash problem with high asymmetric loads due to the offset crash or the oblique load directions, other design concepts must be found which give a large improvement in bending resistance in combination with also optimal folding behavior. Also, the addition of flanges or ribs to the basic square cross-section for more bending resistance should be avoided, as each attachment to the profiles can disturb the normal stable folding pattern (van Oirschot 1995). These disturbances can result in unstable folding patterns, their irregular perimeters introduce weaknesses, which in turn result in premature bending collapses. For optimal folding behavior, the column should be ideally positioned without any fixed attachment to disturb the natural folding process. The contradictory design requirements yield an unsatisfactory result. The demands cannot be met by a single or assembled cross-section configuration. To solve this design problem with its contradictory requirements, an approach is needed in which the design for the frontal car structure is decomposed into separate components each fulfilling a separate function. The combination of these components yields an integral vehicle structure which meets the requirement that in each crash situation (off-axis, offset and full overlap) nearly the same energy is absorbed and a similar deceleration level is obtained. The next section presents a design solution based on

78 78 Chapter 3 this approach. It consists of a longitudinal with conventional axial stiffness but offering a much higher bending resistance A new design concept for functionally decomposed longitudinal members The new concept is based on the design philosophy that an optimal longitudinal member must be functionally decomposed into two separate systems: the first, called the crush component, guarantees the desired stable and efficient energy absorption. The other, called the support component, guarantees the desired stiffness in the transverse direction, see Figures 3.29 and This latter component is necessary to allow enough energy absorption during an off-axis collision and to give enough support with a sliding wall to protect the crush component against a bending collapse. Figure D View of the longitudinal member with the crush component inside the support component.

79 Numerical Design of Stable Energy Absorbing Longitudinal Members 79 Figure Interior view of the longitudinal member. A square profile is chosen for the crush component with a width of 70 mm outside and a thickness of 2.0 mm. The width dimensions of the crush component are limited, as it has to fit within the available interior dimensions of the support component, depending on the available space between the engine and wheel envelope. The total length of 1160 mm of the crush component is based on a large mid class car. With a theoretical effective deformation (deflection) of about 72.5 per cent of the original length (27.5 per cent remains as packaged folds) (Wierzbicki 1990), the deformation that can be obtained by the crush component measures 840 mm. The support component consists of four very stiff square tubes that fit into each other and may slide over each other, like a telescope. Flanges prevent the telescope from falling apart in forward direction. Two support squared rings are necessary to prevent a bending collapse of the crush component in the wider rear tubes of the telescope. The support component has the same total length of 1160 mm and can be telescoped to 360 mm. See Figure 3.31 for details. The resulting deformation length of 800 mm is an acceptable value for a middle class car (Witteman 1993) in the case of a per cent overlap crash. The space between the edges of the crush component and the inside of the support component is only 0.5 mm. At both ends of the longitudinal member, the two functional components must be joined with a rigid plate.

80 80 Chapter 3 tube1 tube 2 tube 3 tube 4 Figure Drawing of the longitudinal member. An angle of 45 is chosen for the crush component along the longitudinal axis with respect to the orientation of the enveloping support component. This unusual angular orientation has several advantages: 1. The crush component is supported at its edges by the enveloping square. At the edges, no material deforms to the outside, which implies that at this position contact with the support component does not disturb the folding process. The narrowly fitting position of the crush component in the enveloping support component gives a continuous sliding force acting as a support against bending. Note that during the deformation process the first tube of the support component with the smallest inner dimensions slides together with the folding front to the rear. After full deformation, all the folds are packaged in the first support tube. Figures 3.32 and 3.33 show the lobes of the crush component inside the support component after deformation. 2. In reverse, the crush component supports also the support component against a bending collapse in case of a transverse load. There is not much space to fold material from the support component inwards. 3. The space needed for undisturbed folding is always guaranteed. The width increase of an asymmetric fold is nearly half of the undeformed width (based on several experiments in our laboratory and also found by Beermann 1982). This extra needed space is available due to the rotated orientation of the crush component with respect to the enveloping support square (see Figure 3.32). 4. Other vehicle components can be built near the outside of the support component, since no deformed material will bulge out.

81 Numerical Design of Stable Energy Absorbing Longitudinal Members A decrease of the bending resistance, occurring during deformation in the asymmetric folding pattern of the crush component, will not be fully oriented in the critical horizontal or vertical directions where additional moments are possible, due to offset loads (working in horizontal plane) or due to a vehicle center of gravity above the longitudinal horizontal axis (working in vertical plane). Figure Front view of the crush component inside the boundaries of the support component after deformation. Figure Side view of the crush component inside the boundaries of the support component after deformation.

82 82 Chapter 3 The result is a longitudinal member having a conventional stiffness for stable energy absorption during a full overlap crash. In addition it has an extremely high bending resistance in order to absorb energy during an offset or off-axis collision. Combining two of such longitudinals with a rigid connection beam (a cross member) at the front ends results in a frontal structure. It is supposed to be fixed to the stiff firewall at the rear ends. A numerical simulation (Slaats 1996a) with an axial 30 per cent offset load of 1100 kg and 56 km/h showed that the loaded longitudinal deforms very regularly with a maximal energy absorption and the unloaded longitudinal deforms by bending, due to the tensile force in the cross member. To provide a realistic simulation, a rigid engine block was modeled between the two longitudinals. See Figure 3.34 for the deformed state after 800 mm deformation during an offset load. engine 30% offset Figure Thirty per cent offset crash of the new design concept after 800 mm axial shortening. Because of the high transverse stiffness of the support component of the bending longitudinal, it absorbs a considerable amount of extra energy in comparison to a regular longitudinal. Nevertheless, it is better to reach the same amount of energy absorption in the case of an offset crash as in a full overlap crash to prevent passenger compartment deformation. Although this can be reached by further increasing the wall thickness of the support component, it is impossible to maintain an acceptable mass for the entire structure. Energy absorption by bending is very inefficient since mostly only one fold is formed instead of a range filling the total length. A considerable amount of energy absorption is only possible with heavy

83 Numerical Design of Stable Energy Absorbing Longitudinal Members 83 structures (one bending collapse comparable with several regular folds). Also an additional ring around the first tube and joined with the wheel house structure could improve the bending resistance. Problem is however, that in the case of a too stiff structure the axially loaded longitudinal is also forced to bend by the cross member, because if only one longitudinal deforms axially the distance between the front ends of both longitudinals has to increase which is not possible with a strong cross member. The only way to reach the same energy absorption and deformation length for an offset crash compared to a full overlap crash is to force the unloaded longitudinal to crumple as well, by means of a stable axial folding process. In Chapter 4 a solution for the mentioned design problem will be presented to force the unloaded longitudinal of an offset crash to crumple as well Manufacturing and axial crushing of the advanced longitudinal member Although the crash simulation software (PAM-CRASH ) is in use at many car manufacturers and shows good similarity between simulations and practical tests (Richter 1993, Ducrocq 1997, Arimoto 1998), it is important to see if the folding process of a column in a test has the same folding pattern as predicted by the numerical simulation. Especially for the narrow dimensions where the folds of the crush component have to fit inside the support component and around a rod (positioned in the center of the crush component), which is necessary for the solution of the design problem as described in Chapter 4. For this reason a telescope structure with crushing component and a rod inside has been manufactured and partially crushed on a hydraulic press in our own laboratory. At that moment our new dynamic crash test facility for a crash mass of maximal 500 kg and a velocity of up to 64 km/h was not ready. The experiment was also a good test to see if and how the structure could be manufactured. For manufacturing of the telescope some design changes were necessary (which have no influence on the test results). For the material is chosen Fe360 instead of Aluminum AA6061-T6 as used in the simulations. Reasons are expected problems with welding of two U-forms together, the straightness of the corners after welding and the costs. Default columns in the desired dimensions (not available in aluminum) have imprecise dimensions; tolerances of wall thickness and perpendicularity are too large to ensure good sliding. Therefore, the columns must be assembled with two pressed U-parts. Before welding, the ends are milled for good parallelisms. Of course in mass production, extrusion of aluminum profiles is possible.

84 84 Chapter 3 The crushing component is also of Fe360 instead of FeP03, which was easier available in square profiles (70x70x2). The measured thickness was 1.9 mm. In Figure 3.35 two views of the manufactured telescope with the crushing component inside can be seen. The crushing component is moved a little outside to show the bead triggering. Detailed manufacturing drawings are reported by Morlog (1998). Figure Manufactured telescope with crushing component inside. To test if the folds are regularly formed at the trigger side inside the first narrowest telescope tube, with a massive square rod inside the crushing component, an axial quasi-static load test is done. For this test, 190 mm of the front telescope is used with a crushing component of 450 mm length. Therefore, the maximum deformation length will be 260 mm. See Figure 3.36.

85 Numerical Design of Stable Energy Absorbing Longitudinal Members 85 Figure Used component for the axial test. In Figure 3.37, the hydraulic press with the clamped structure can be seen. The maximum hydraulic load is 250 kn and the used speed is 250 mm/min. Figure Hydraulic press with clamped longitudinal structure. In Figure 3.38, two top views are given after deformation. It can be seen that the folds are packaged between the inside rod and the outside supporting column. The folds are started at the trigger side inside the supporting column.

86 86 Chapter 3 Figure Top views of the deformed crushing component with inside the rod and outside the supporting component. In Figure 3.39, a side view is given of the deformed crushing component. To see the folds, the telescope is removed. To compare the number and the shape of the folds, a simulation is added. Using Equation 3.1 the plastic folding wavelength can be calculated, yielding with a = b = 70 mm and t = 1.9 mm to 41.6 mm. After full deformation, theoretically 72.5 per cent of the original length is shortened, so the deformation of 260 mm uses 1 /.725 x 260 mm = 359 mm length. From this length 359 mm / 41.6 mm = 8.6 folds could be formed. This is exactly the number of folds that can be seen in Figure The ninth fold is already half formed.

87 Numerical Design of Stable Energy Absorbing Longitudinal Members 87 Figure Side view of the deformed crushing component after the experiment (left) and after the simulation (right). In Figure 3.40 the force level against the deformation length of the experiment is given. It is clearly seen that the nine folds of the crushing component in Figure 3.39 correspond with the nine force peak levels in the curve of Figure The very regular force peaks repeat after the first peak each 30.1 mm deformation length, which means a folding wavelength of 30.1 /.725= 41.5 mm, this is much like the calculated 41.6 mm. The average force level after the start is about 51 kn, fluctuating between 27 and 75 kn. Another effect which can be seen in Figure 3.40 is that after the start of the deformation, a high force peak is followed by a little lower peak, which is followed again by a high peak and so on. This is explained by the fact that a folding square column generates a free inner space, which is not exactly a square but more a rectangle. Repeating folds move a little outside, and inside on the adjoining side around the corner. This can also be seen in Figure 3.38, where the fold in the left under quadrant and the fold in the right upper quadrant come more to the outside (expand laterally) and have a small space with the inner rod (most clear in the right picture), while the fold in the right under quadrant and the fold in the left upper

88 88 Chapter 3 quadrant hit the inner rod and come less to the outside. The contact force between the inner rod and half of the formed folds causes a small difference in the force peaks of repeating formed folds (see also Section 4.3) force deflection curve of quasi static experiment Force [kn] Deflection [mm] Figure Force level of the quasi-static experiment after 260 mm deformation. To compare the simulation results with the experiment, it is important that the same material properties of Fe360 as used in the experiment are also used in the material model of the simulation. Therefore the properties are investigated with a separate tensile test on a test piece of the same column, see Table For completeness also the properties of FeP03 as used in all other simulations are presented.

89 Numerical Design of Stable Energy Absorbing Longitudinal Members 89 Table Material properties of FeP03 and Fe360. Property FeP03 Fe360 Young Modulus E = 210 kn/mm 2 E = 210 kn/mm 2 Yield stress R p0.2 = 185 N/mm 2 R p0.2 = 357 N/mm 2 Ultimate tensile strength R m = 325 N/mm 2 R m = 412 N/mm 2 Strengthening coefficient N = N = Elongation at break A 5 = 47.2 % A 5 = 29 % Equal strain A g = 30.3 % A g = 7.2 % In Figure 3.41 the force deflection curve of the simulation with a crush speed of 40 km/h is shown until 250 mm deflection. To compare the dynamic simulation results with the results of the quasi-static experiment, a correction on the force level of the experiment is done, due to the strain rate dependency of the simulated material. With Formula 3.3 the correction factor σ dyn σ stat is calculated, where the strain rate for the used geometry is determined with the simulation program CRASH-CAD. For the quasi-static experiment the velocity was 250 mm/min, yielding already a correction factor of For a crush speed of 40 km/h the correction factor is Therefore the force level of the experiment is increased with a factor / = 1.336, showed with the dashed line in Figure The influence of the triggering is difficult to compare since in the real crushing component probably some material weakening occurs by pressing the stamp form in the material, yielding a lower initial force peak. Also the bead triggering was not modeled with the same geometry as used in the real column caused by the mesh distribution. In the simulation this first peak is higher, yielding already the initiation of the second peak during the creation of first peak. Therefore the second peak in the simulation is lower. The following force levels fit very well, the mean force level is almost equally.

90 90 Chapter force deflection curve of compensated experiment and simulation 200 Force [kn] simulation experiment Deflection [mm] Figure Force level of the compensated experiment (dashed line) and the simulation (solid line) at 56 km/h. The force level is very dependent on the properties of the materials employed. Besides the yield stress that determines the first peak force, also the strengthening coefficient and the ultimate tensile strength are important (Magee 1978, Thornton 1983). It is difficult to determine the exact material properties of a tested object, because making a tensile strip already influences the properties by thermal heating. If a test piece could not be made of the same material sheet, also material deviations caused by manufacturing or orientation (Richter 1993) could be expected. Besides this, also measurement or statistical faults are possible. For verification of the crash behavior of the longitudinal member, the exact absolute value of the predicted force level by the simulation is not so important, this level can easily be changed in practice by using another material or another thickness. Optimizing a design with the aid of numerical simulations is mostly based on comparing mutually design changes. With the same material, also the same deviations are compared.

91 Numerical Design of Stable Energy Absorbing Longitudinal Members Conclusions Numerical simulation studies have been performed analyzing the effects of wall thickness and profile width variations on the crash performance of longitudinal members. These show that it is impossible to design a cross-sectional profile that provides an efficient and stable folding process with an acceptable axial stiffness and at the same time satisfies the demand for sufficient transverse stiffness. It is essential that both requirements are met simultaneously, since in actual car to car crashes about 60 per cent of the crash energy cannot be absorbed by the frontal structure if that car was designed to only meet the compulsory full overlap test. A new structural concept is proposed for the vehicle front, made up of two interconnected longitudinal member systems, each consisting of an interior crush component that resides inside a telescoping outer support component. This concept is capable of providing regular axial stiffness values in conjunction with very high transverse stiffnesses. The latter is necessary to also absorb a considerable amount of energy during an offset or off-axis crash situation. However, to obtain the same level of energy absorption during an offset crash as during a full overlap crash, extra measures are necessary to better involve the unloaded longitudinal in the energy absorption. To protect the passenger compartment, no extra deformation length must be necessary in case of an offset crash.

92 92 Chapter 3

93 Design of a Frontal Safety Structure Suited for Different Crash Situations 93 Chapter 4 Design of a Frontal Safety Structure Suited for Different Crash Situations 4.1. Introduction As explained in the preceding chapter, increased protection for the entire collision spectrum can be obtained by a frontal structure consisting of two special longitudinal members, which combine a higher bending resistance without increased axial stiffness. However, to reach the same energy absorption and deformation length for an offset crash compared to a full overlap crash, the unloaded longitudinal must be forced to crumple as well, by means of a stable axial folding process. This is not easy with a very stiff transverse structure in the front. Bending moments will be introduced that disturb the folding process of the longitudinals. If the front cross beam is stiff enough, it will force the beam connection with the longitudinal front ends to bend in the case of an offset load. With additional beams from the middle of the front cross member in oblique direction to the middle of the longitudinals, bending of the connection with the longitudinal front ends could be prevented. However, these oblique beams need additional free space and could create new bending moments in the middle of the longitudinals. In addition, the engine between the longitudinals reduces extremely the available deflection length of the longitudinals after a wide rigid transverse structure in front of the engine hits the engine. Alternatives are to bring the structure outside the front of the car by means of a hydraulically extended bumper system (Jawad 1996). With radar detection a collision is predicted and the bumper moves with hydraulic cylinders rapidly forward before the crash to enlarge the deformation length. Problems are the space requirements and the weight of the equipment, the provision of hydraulic power and facilitating high flow-fast response control valves. But due to the much longer available deformation length, also in the case of an offset collision enough energy could be absorbed. Another hydraulically controlled frontal car structure is presented in Chapter 5 (Section 5.7) as a future possibility. Also in this solution the weight and the necessary space could be a problem. Another solution to reduce the crash severity and to make the vehicle less sensitive for the collision situation is an extended airbag bumper system (Clark 1994). With radar detection a collision is predicted and a large airbag deploys in front of the bumper. However, what happens if a pedestrian is detected too late or if the predicted collision could be avoided.

94 94 Chapter 4 A more practical and reliable design solution is to support both longitudinal members by a cable connection system for symmetric force distribution. If only one of the longitudinal members is loaded during a partial overlap crash, the cable connection system will force the other longitudinal member to crumple as well, see Figure 4.1. This results in a similar programmed energy absorption as in the case of a full overlap collision. This solution needs not a heavy and expensive hydraulic system or an insecure radar detection system and it needs hardly space in front of the passenger compartment. With this novel concept it is possible to design a frontal car structure with the same stiffness for all overlap percentages and impact angles, resulting in an optimal crash pulse. The influence of various crash situations on the amount of energy absorbed by this system will be demonstrated by means of simulations and analyses. Rod Figure 4.1. Principle sketch of a cable-supported longitudinal structure The cable connection system for a symmetric force distribution In Figure 4.1 a schematic sketch of a cable-supported frontal car structure is given. The system consists of two rods, two cables and four cable guides. The stiff rods are placed within the longitudinal members. At the front of the vehicle, they are interconnected with the cross member. The rods are longer than the longitudinals and extend beyond the vehicle's firewall. A cable is connected to the end of each rod. This cable is guided to the front end of the other longitudinal via two cable guides,

95 Design of a Frontal Safety Structure Suited for Different Crash Situations 95 where it is connected to the cross member. The working principle is simple: if one longitudinal is loaded and starts deforming, the corresponding rod moves back to the rear and pulls the cable, which leads the crushing force via the cable guides directly to the other, unloaded, longitudinal. This force transmission occurs without loss of energy. Note that if both longitudinals are loaded (full overlap crash), the cable construction has no influence on the crash behavior. Although this cable concept can be built into most cars with conventional frontal structures, the crush/support component concept described is very suitable for combination with the cable construction. The rods need to have sufficient space to move backwards. Because intrusion of the passenger compartment is not desirable, the rods must move under the vehicle. This means that the longitudinal members need to be positioned under a slight angle (higher on the bumper side, lower on the firewall side), due to the prescribed compulsory height at which the forces must be led into the structure. The crush/support component concept offers two important advantages to combine it with the cable system. The first advantage is that the crush/support component concept is well suited to be positioned under an angle. Its high bending resistance guarantees that the structure will not collapse in a premature bending collapse, unlike most conventional longitudinal members. The second advantage is that the crush/support component concept guarantees enough free inner space in the crush component by stable folding of the crush component under all circumstances. Most conventional longitudinals have all kinds of connections with other components under the bonnet, which can easily disturb the folding process. A stable folding process is necessary, because the rod is placed within the crushing longitudinal and should always be free to move backwards. Unstable folding would prevent the rod from sliding within the narrow space of the crush component. This would cause the cable system to stop working correctly. To avoid any transverse forces on the sliding rod, the cable is guided through the center of the rod. This is possible if the two rods are formed like a U-profile and the cable guides fit into these U-rods. See Figures 4.2 and 4.3 for more details.

96 96 Chapter 4 A A Figure 4.2. Top view of the cable-supported longitudinal structure. Rod Bar Disk Cable Figure 4.3. Cross-section A-A of the cable around the cable guide disk inside the rod. A steel 8-string wire-rope cable with a diameter of 20 mm has a fracture force of 279 kn (Staalkabel B.V. 1995). That means that a force up to 167 kn does not deform the cable plastically. Simulations (Slaats 1996a) showed that in a real offset crash the peak load is below this value.

97 Design of a Frontal Safety Structure Suited for Different Crash Situations 97 The free space inside the stable asymmetrically deformed square crush component is decreased to about half the original width. This was confirmed by simulations and experiments in our laboratory and done by others (Beermann 1982). This 50 per cent decrease means that the inner dimension of the square crush component after fold forming will be nearly 33 mm. To prevent each disturbance of the regular folding process, and to guarantee enough sliding space, an outside width of 32 mm of the square sliding rod is chosen. The cable guide disk with a diameter of 170 mm (minimum size for cable 20 mm, Staalkabel B.V. 1995) has a height of 20 mm, the same as the cable diameter. This is important because the rod must not be weakened more than necessary. The cable guide disk suspensions must be mounted on a stiff floor structure, since the transverse forces on the disk shafts are 2 times the tensile forces in the cable which is guided 90 degrees around the disk. A stiff cross member frame under the firewall and joined between the ends of the supporting tubes could give the desired stiffness. The buckling load of the rod was calculated to be 171 kn. This is more than the expected peak load during a crash. For the buckling load, the free length of the rod is the same as the longitudinal length. Behind the firewall extra leading support for the sliding rod is necessary to ensure that movement only occurs in the axial crush direction. During the crash, the free buckling length of the rod inside the crush component decreases by shortening of the crush component and by additional support from the folds formed. In Figures 4.4 and 4.5, the assembly of the new front structure concept with the cable connection system can be seen. In addition, extra cable guide rings preventing the cable from slipping off the disk and a pin mounted to the firewall at the crossing point of the cables are shown. Note that the center lines of the cable, rod and longitudinal fall together yielding axial forces only. The position of this cable-supported structure built inside a car is shown in Figure 4.6. Note that the two rods can move freely to the rear under the car floor during a crash, due to the position under a slight angle. To prevent buckling of the growing free length of the rod behind the cable guide disks under the car, a guiding pipe like an exhaust pipe with a diameter of 50 mm and a length of 800 mm could be used. Of course the rod has to bent a little back horizontally, but this could also absorb energy.

98 98 Chapter 4 cable guide rings pin on firewall Figure D view of the cable-supported longitudinal structure. Figure D open view of the cable-supported longitudinal structure.

99 Design of a Frontal Safety Structure Suited for Different Crash Situations 99 A-pillar front rod cable guide disks guiding pipe Figure 4.6. Side drawing of the cable-supported longitudinal structure positioned in a car Numerical simulation of the cable-supported longitudinal structure To evaluate whether the designed cable-supported structure theoretically works, and whether the absorbed energy is about two times the energy absorption of one longitudinal during an offset crash, a numerical model of the cable-supported longitudinal structure was built with the finite element program PAM-CRASH. The model consists of 60,000 elements and has 14 sliding surfaces to describe the folding and sliding process accurately. To avoid unrealistic simulation results, and to focus on the longitudinals only, the calculation was stopped at the moment the rigid engine block became involved. In this model, a shortening of 510 mm is used as the deformation position to compare the amounts of absorbed energy between various crash situations. Figures 4.7 and 4.8 show the deformed state of a 30 per cent offset crash with a moving rigid barrier of 1100 kg and an initial velocity of 56 km/h (Witteman 1996).

100 100 Chapter 4 30% offset Figure D view of 30 per cent offset crash of the cable-supported longitudinal structure after 510 mm shortening. 30% offset Figure 4.8. Top/inside view of 30 per cent offset crash of the cable-supported longitudinal structure after 510 mm deformation length.

101 Design of a Frontal Safety Structure Suited for Different Crash Situations 101 It is clear that both longitudinals are deformed identically despite the asymmetric external offset load. The crushing force in the loaded longitudinal is transmitted to the unloaded longitudinal. In Figure 4.8, the telescoping tubes are left away to expose the regular folding pattern in both crush components. The energy absorption of this 30 per cent offset crash is 58.5 kj. This is plotted in Figure 4.9 with the solid line. This energy absorption is compared to the energy absorption of the crush/support component concept without the cable system during a 30 per cent offset (dashed line, 35.5 kj) and a 100 per cent overlap (dash/dot line, 61 kj). The last one is two times the energy absorption of one crush/support component longitudinal without internal sliding rod (dotted line, 30.5 kj). Energy Energy absorption [J] [J] 7 x NDC 100% - Cable 30% -- NDC 30%.. Single 100 % overlap 30 % offset with cable 30 % offset without cable single longitudinal deformation Shortening [mm] Deformation length [mm] Figure 4.9. Comparison of the energy absorption between four situations. The high bending resistance of the crush/support component structure gives during an offset crash an increment in energy absorption of nearly 17 per cent with respect to a single longitudinal (35.5 kj vs kj), resulting in 58 per cent of the full overlap crash energy (61 kj). However, the addition of the cable system increases the energy absorption by nearly 92 per cent (58.5 kj vs kj), resulting in 96 per cent of the full overlap crash energy. These percentages are for the same axial

102 102 Chapter 4 deformation length. The difference in energy absorption between a full overlap (without cables) and the 30 per cent offset with cables is 4 per cent (61 kj vs kj). This is caused by the fact that a longitudinal with an internal sliding rod absorbs about 4 per cent less energy than the same longitudinal without the internal rod as first used in the crush/support component concept. The reason for this is that a square tube does not naturally deform with an exactly square free inner space. The space is a little likely to be rectangular. The inside formed folds that hit the rod, can use the rod as a support in forming the following outside folds with a little lower force level. Of course, the energy absorption with the same cable-supported structure in a full overlap or in an offset crash is comparable Building the longitudinal structure in a numeric frontal car model Since the simulations have shown that the principle of the cable support works without disturbing the regular folding process of the longitudinals, a complete frontal car model is useful for evaluating more realistic crash situations. Just in case of an offset or an oblique crash situation, it is important that the model can move more like the reality. In a realistic test procedure, a vehicle moves freely with a velocity against a rigid stationary obstacle. Due to asymmetric forces during offset or oblique collisions, the back of the vehicle could turn a little, which implies an extra bending moment on the longitudinals. In addition, the mass inertia of the structure has an influence on the crash behavior. Especially the rods inside the longitudinals are stopped abruptly. The longitudinals are responsible for the largest part of the energy absorption. For the necessary repeating folding process it is important that there is a good load introduction in axial direction. In case of a 30 degrees collision against a rigid barrier, the stiff edge of the first support tube at the front of the longitudinal will generate a bending moment, which can cause a bending collapse. To avoid this rotation, the front of the first tube is changed into a tip. This tip can also provide the connection between the first telescope tube and the bumper. Three geometry s of bumper stays are evaluated, a tip angle of 15, 30 and 60 degrees. In case of a 60 tip, the tip is flattened to reduce the length (see Figure 4.10).

103 Design of a Frontal Safety Structure Suited for Different Crash Situations 103 tube 1 tip Figure Bumper stays with different tip angles. The tips deform in such a way that the moment applied on the first tube will be reduced. The rod inside the crush component, which has a rigid connection with the stiff support component, has in case of a deformable tip a less abrupt deceleration during the impact. This is important to prevent buckling of the rod under its own inertia. To test the best alternative, a simulation with the first two telescoping tubes against a 30 degrees barrier during 10 ms has been carried out. In Figure 4.11, the energy absorptions of the longitudinal only with different bumper stays and in Figure 4.12, the energy absorptions of the longitudinal including the different bumper stays are compared. In all cases, the energy absorption increases by using a tip. Because the 60 degrees bumper stay deforms easier during a longer length, the energy absorption is at the start a little lower. More important is the final bending distance, measured at the displacement of the second support tube in transverse direction. This is shown in Figure 4.13, where the maximum transverse displacement is limited to 40 mm since a larger transverse displacement is not realistic in a car with limited free space around the longitudinals.

104 104 Chapter Longitudinal internal energy: 30 degrees impact 15 degr. tip 30 degr. tip 60 degr. tip No tip 8000 Energy [J] Deflection [mm] Figure Energy absorption of longitudinal only at different tip profiles Longitudinal + tip internal energy: 30 degrees impact 15 degr. tip 30 degr. tip 60 degr. tip No tip 8000 Energy [J] Deflection [mm] Figure Energy absorption of longitudinal and tip with different tip profiles.

105 Design of a Frontal Safety Structure Suited for Different Crash Situations Displacement 2nd profile: 30 degrees impact Displacement [mm] degr. tip 30 degr. tip 60 degr. tip No tip Time [ms] Figure Tube transverse displacement if using different tip geometry s. It is clear that a 30 or 60 degrees tip angle of the bumper stay gives hardly a transverse displacement. This implies that no serious distortion occurs in the continued folding process of the crush component. The favorable results of the 60 degrees geometry in combination with a better bumper connection possibility leads to the choice of the 60 bumper stay. The frontal model should contain the most important structural components that influence the method of energy absorption. The aim is to avoid deformation of the passenger compartment. All crash energy must be fully absorbed by the front structure. This implies that the frontal model boundary starts at the deformable firewall, which is connected to an undeformable cage. Rigid nodes on the wall borders, which also contain a part of the mass distribution of the rear of the vehicle that has not been modeled, can simulate this cage (De Santis 1996). In front of the cage the following deformable and rigid components are necessary (see Table 4.1):

106 106 Chapter 4 Table 4.1. Important frontal vehicle components. component deformable body rigid body fire wall longitudinals bumper bumper stay lateral wing lateral wing reinforcement front panel engine mountings engine and gearbox front wheel suspension front wheel rim front wheel tire For the simulation of a crash, also the surroundings like the road surface and an obstacle should be modeled as rigid body. Although the rigid bodies in the front model do not absorb energy by deformation, their masses represent energy contributions, which correspond to their initial velocities. The rigid volumes influence the available space and sequence for motion of the deformable components. Their own motion could deform the surrounding deformable structures. See Figures 4.14 and 4.15 for the complete model.

107 Design of a Frontal Safety Structure Suited for Different Crash Situations 107 Figure D views of the complete frontal vehicle model.

108 108 Chapter 4 bumper stays rods engine fire wall front panel bumper longitudinals Figure Top view of the complete frontal vehicle model. The dimensions are based on a popular compact class car. To fit the already designed longitudinal structure, the overlap of the four support tubes is increased from 40 mm to 80 mm. This yields a higher bending resistance and the telescoping tubes also slide better into each other. The total length is reduced to 980 mm. To achieve a maximum deformation length, while taking into account the support rings, the length of the telescoping tubes must be: 280 mm, 300 mm, 320 mm and 320 mm successively from bumper to fire wall side. For the material of the tubes is chosen Aluminum (AA6061-T6), which gives a total weight for the telescope and rings of 6.5 kg. In Figure 4.16 the new dimensions and the connected bumper stay are shown. In this case, the available deformation length is 660 mm, 67.3 per cent of the original length. This must be enough for a compact class car (acceptable decelerations achievable). Note that due to the presence of the rigid engine, in most collision situations the residual length of the longitudinal could not be less than the engine

109 Design of a Frontal Safety Structure Suited for Different Crash Situations 109 length. See Figures 4.17 and 4.18 for details of the longitudinal positioning with an angle of 10 degrees with the horizontal plane Figure New longitudinal dimensions to fit in a compact class car in mm. Figure Cross-section of the frontal vehicle model.

110 110 Chapter α=10 α Figure Dimensions of the compact class longitudinal in mm. The design of the engine mountings and their locations has an important influence on the crash behavior of the structure it is connected to. Because the longitudinal members are a relatively stiff body component and therefore ideally for connecting heavy components, they are often used to carry the engine. However, if the rigid engine block is mounted with two points to the longitudinal, the part between the mountings is bridged and can not deform as programmed. In addition, in case of one mounting point, the fixed connection of the engine with the drive line will bridge the longitudinal as well with a rigid link. This is more critical with the mounting point more in front of the car. Otherwise, if the mounting points are positioned opposite to each other on both longitudinals, it gives a rigid support against bending in case of asymmetric loads. To connect the engine with the outside of the support component, it must be fit at the front of the fourth support tube at the firewall. Because the engine geometry requires a second mounting point on the same longitudinal, it can be fitted on the front of the third tube. Because the first (rear) mounting point is a deformable beam which collapses during a crash, the third tube can slide inside the fixed fourth tube while deforming the mounting. In this way, the first two tubes can slide into the third tube during the first part of the crash where the engine is not directly involved. During the second part of the crash the engine moves backward together with the movement of the third tube into the fourth tube, yielding a normal deformation length. Figure 4.19

111 Design of a Frontal Safety Structure Suited for Different Crash Situations 111 shows the principle working of the engine mounting in four simulation steps. After the first time step the engine is moved backwards by the rigid barrier. Note it is a top view with the longitudinal in a vertically rotated position with finally an angle of more as 10 degrees. tube 4 tube 3 mountings engine step 1 step 2 step 3 step 4 Figure Deformation of engine mountings. To investigate the vehicle s deformation and its influence on the longitudinals with support and cable system, different frontal crashes are simulated in the next sections Numerical simulation of a full overlap crash The following numerical simulations are performed with the complete frontal vehicle model, with the cable supported crush/support component concept built in a compact class vehicle front structure. The total mass is 1053 kg, the crash velocity of the car is 56 km/h and an infinite friction between the vehicle model and the rigid barrier is assumed.

112 112 Chapter 4 In the full overlap crash against a rigid wall, the cable system has no function. Both longitudinals are loaded directly with an axial force direction. However, the cables and rods might not disturb the folding process. Both rods should slide backwards without pulling the cables. In Figure 4.20 six simulation steps are shown (Witteman 1998). In this top view the front panel is only shown for t=0 ms to have a better view on the deformation of both longitudinals and the engine. t=0 [ms] t=10 [ms]

113 Design of a Frontal Safety Structure Suited for Different Crash Situations 113 t=20 [ms] t=30 [ms] t=40 [ms] t=60 [ms] Figure Top view of a full overlap crash (56 km/h) in six time steps.

114 114 Chapter 4 In Figure 4.21, a side view of the vehicle deformation is shown in four time steps. It can be seen that the position angle of the longitudinal increases during the crash starting from 10 degrees. In both figures it is also clear that the rods can freely move backwards, the cables are not tightened and both longitudinals show a nearly identical deformation behavior. In Figure 4.22 the regular folding patterns of the crush components inside the telescopes are viewed. To ensure that the folding process starts at the front side without a too high peak force a little fold is modeled. In the undeformed state, this triggering for the folding process can be seen. At t=40 ms a little extra deformation can be found due to the collision of the engine with the left support component. t=0 [ms] t=20 [ms] t=40 [ms] t=60 [ms] Figure Side view of a full overlap crash in four time steps.

115 Design of a Frontal Safety Structure Suited for Different Crash Situations 115 LH triggering RH t=0 [ms] LH RH t=20 [ms] LH RH t=40 [ms] LH RH t=60 [ms] Figure Inside view, folding process of both crush components in four time steps.

116 116 Chapter 4 The calculations have been carried out until 60 ms and at that moment, the velocity is reduced to 0 km/h. This is shown in Figure The division of the weight of the engine mainly causes the little difference between the movement of the left and right A-pillar (see Figure 4.6). The more accurate shell mesh around the undeformable rigid body of the left engine side, which was necessary on several engine planes to prevent numerical penetration of other vehicle components, results in more nodes with additional masses. The same effect can be seen in Figure 4.24 in which the deceleration level is plotted. During the first part of the crash, the deceleration of the vehicle model is about 20 g. After t=30 ms the engine hits the rigid barrier causing the vehicle to decelerate up to about 35 g Full overlap, 56 [km/h] Vehicle centre Left A pillar Right A pillar 40 Velocity [km/h] Time [ms] Figure Velocities of a full overlap crash with 56 km/h.

117 Design of a Frontal Safety Structure Suited for Different Crash Situations Full overlap, 56 [km/h] (Sae 180_5) Vehicle centre Left A pillar Right A pillar 40 Deceleration [g] engine hits rigid barrier Time [ms] Figure Decelerations of a full overlap crash with 56 km/h. To see the influence of the crash speed on the resulting decelerations of the vehicle model, the same simulation is done with 28 km/h. In this case, the engine will not hit the barrier. In Figure 4.25 two interesting time steps in top view are shown. In Figure 4.26 and Figure 4.27 the velocities and decelerations are plotted. The crash time is in both crashes (i.e. at 56 km/h and 26 km/h) about 60 ms. The deceleration of the 28 km/h crash is lower than the deceleration of the 56 km/h crash, it fluctuates between 10 g and 20 g until it further drops after 45 ms.

118 118 Chapter 4 t=20 [ms] t=60 [ms] Figure Top view of a full overlap crash (28 km/h) in two time steps Full overlap, 28 [km/h] Vehicle centre Left A pillar Right A pillar 40 Velocity [km/h] Time [ms] Figure Velocities of a full overlap crash with 28 km/h.

119 Design of a Frontal Safety Structure Suited for Different Crash Situations Full overlap, 28 [km/h] (Sae 180_5) Vehicle centre Left A pillar Right A pillar 40 Deceleration [g] Time [ms] Figure Decelerations of a full overlap crash with 28 km/h. In Figure 4.28, the rigid barrier force is plotted of both full overlap crashes (i.e. at 56 km/h and 26 km/h). The familiar first force peak to start the folding process can be seen in both crashes. In addition, the peak of the 56 km/h crash at 30 ms when the engine hits the barrier can be easily recognized.

120 120 Chapter Full overlap, rigid barrier force (Sae 180_5) 56 [km/h] 28 [km/h] engine hits rigid barrier 500 Force [kn] first fold Time [ms] Figure Rigid barrier forces of a full overlap crash at 56 and 28 km/h Numerical simulation of a 40 per cent offset crash The vehicle model is impacted with an initial velocity of 56 km/h against a rigid barrier with infinite friction. Although using a deformable barrier will result in more realistic deformations for a car to car collision, a rigid barrier is used in order to keep the computing time down. With the rigid barrier, it is also possible to evaluate the working principle of the crush/support component design with cable system. For a well functioning of the cable system, it is important that there is a rigid side support of the sliding rods behind the longitudinals near the cable disks. This has an important influence on the Euler buckling load of the cable rod. If the rod collapses, it could not tighten the cable of the unloaded longitudinal. In the compact class design, the collapse load is increased from 171 kn to 239 kn, due to the shorter length of the rod inside the crush component. Without a slide contact the rod has a hinge at the end of the longitudinal, the maximum buckling load will be only half due to the changed Euler buckling mode. In the physical design the cable guide disks and a stiff rod leading profile connected with the vehicle floor could perform this function. In the

121 Design of a Frontal Safety Structure Suited for Different Crash Situations 121 numeric model, the cable guide disks are substituted by slip rings. Therefore, the necessary side support is modeled with additional rigid planes (see Figure 4.29). To ensure the forces on the cable rod should be kept lower than the calculated buckling load, it is important that the load which stops the rod from moving is not combined with the load necessary to start the folding process of the unloaded longitudinal. Simulations have shown (van Leeuwen 1997) that the impact from a single rod against the rigid barrier generates a force with a maximum of 150 kn. The initial load to start the folding process can be estimated from the results of the full overlap crash simulation. In Figure 4.28, the maximum rigid barrier force at crash initiation is about 360 kn. This peak value is the result of the familiar peak force of forming the first fold in both longitudinals. This means that in order to deform the vehicle, the necessary peak load at one side will maximally be about 180 kn. Both values (rod impact of 150 kn and the first folding load of the crush component of 180 kn) are safe below the buckling load of 239 kn. To separate these loads in time, the cable length between the longitudinals is elongated by 30 mm. After the loaded longitudinal rod is stopped, and the corresponding crush component is 30 mm shortened, the cable is tightened and starts deforming the not involved side. With these model adjustments numerical simulations are performed, see Figure 4.29 for a top view of the deformation in six time steps. Again the front panel is only shown at t=0 ms. t=0 [ms] t=10 [ms]

122 122 Chapter 4 t=20 [ms] t=30 [ms] numerical rigid planes t=40 [ms] t=60 [ms] Figure Top view of a 40 per cent offset crash in six time steps.

123 Design of a Frontal Safety Structure Suited for Different Crash Situations 123 From these figures, it can be concluded that the cable system is working. The unloaded vehicle side is also deformed, and both rods slide backwards. The unloaded longitudinal collapses in a folding mode and during the first stages (until 20 ms) of the impact both longitudinal members show a stable folding process. See Figure 4.30 in which the folding process of the crush components is visualized. It can be seen that after the first fold is formed in the loaded longitudinal, the unloaded longitudinal starts forming folds. However, after 20 ms the folding process in the loaded longitudinal is disturbed and it will collapse in a bending mode. Although the rod inside the longitudinal starts to bend after 20 ms, it is still pulling the cable resulting in continuing of the folding process in the unloaded longitudinal until about 40 ms.

124 124 Chapter 4 LH RH t=20 [ms] LH RH t=40 [ms] Figure Inside view, folding process of both crush components during an offset crash in two time steps. One important reason for this undesirable distortion of the loaded longitudinal is a limitation of the numeric model. Normally, a complete vehicle on four wheels with a normal mass distribution (e.g. and luggage in the back) crashes against the barrier. In this simulation, the masses of the sections not modeled are distributed over the front section model components such as the firewall, the wheels, wing and reinforcement, and the longitudinals. However, this leads to an unrealistic rotation of the heavy vehicle front around the rigid barrier. Due to this rotation the loaded longitudinal, which has a stiff connection with the bumper which fits to the barrier, has to bent because the backside is not fixed and moves sideways. A real vehicle has a

125 Design of a Frontal Safety Structure Suited for Different Crash Situations 125 much higher mass inertia due to the longer distance from the impact point to the center of gravity. This means it takes a longer time with a higher force level to rotate the vehicle. This can be proven with photos of real crashes, in which the vehicle does not rotate during the crash, despite the cars mostly lift up the back wheels. See Figure 4.31 in which offset crashes with 55 km/h against a rigid barrier are shown with comparable compact class cars. The alignment until the end of the crash with the floor squares is striking.

126 126 Chapter 4 Nissan Sunny VW Golf

127 Design of a Frontal Safety Structure Suited for Different Crash Situations 127 Mitsubishi Colt Ford Escort Figure Examples of offset crashes with no vehicle rotation (Auto motor und sport spezial 1992, photos H.P. Seufert).

128 128 Chapter 4 A possibility to model this mass deviation more realistically in future is by means of a centralized mass behind the firewall connected with beams. A more usable solution will be a rigid passenger cage (rigid bar elements) with a correct mass distribution in which a finite element dummy can be placed for injury research (Landheer 1996,1997). In the case of a rigid cage, most calculation time is spent to the vehicle components that has to deform. The extreme rotation of the vehicle also appears in the velocity curve of the vehicle front in which a large difference arises between the velocity of the left and the right A- pillar. If the left side is stopped after 43 ms the unloaded right side still moves with 44 km/h. See Figure % offset, 56 [km/h] Velocity [km/h] Vehicle centre Left A pillar Right A pillar Time [ms] Figure Velocities of an offset crash. In Figure 4.33, the rigid barrier force is plotted. It is clear that the maximum load until the engine hits the barrier (after t=30 ms, see also Figure 4.29) fluctuates below a safe level of nearly 200 kn. After 23 ms the force increases due to deformation of the front by the engine. This load curve has much similarity with the rigid barrier force of the 56 km/h full overlap crash in Figure 4.28, except the higher starting peak due to two starting longitudinals simultaneously. This offset load level of nearly 200 kn is of similar magnitude as the load level in case of a full overlap crash. However, it is not

129 Design of a Frontal Safety Structure Suited for Different Crash Situations 129 symmetrically distributed on the vehicle front but asymmetric at one side. Thus, another reason for the extreme vehicle rotation is the fact that the offset load due to the cable system is much higher yielding a higher rotation force compared with a familiar vehicle front. Also the fact that for this particular lighter compact class car the longitudinal is a little too stiff in the axial direction, makes a rotation easier. If a longitudinal is too stiff, the vehicle is stopped with a higher deceleration than necessary and before the engine hits the firewall, as is here the case. The vehicle now has design potential for increased mass or higher crash velocities % offset, 56 [km/h] (Sae 180_5) Rigid barrier force Force [kn] engine hits front Time [ms] Figure Rigid barrier force of an offset crash. The decelerations of the offset crash for the left and right A-pillar are shown in Figure The same large differences as found in the velocity curves are found here. In the first 43 ms, in which the left side decelerates to a halt with deceleration peaks of up to 60 g, the maximum deceleration on the right side is only about 20 g.

130 130 Chapter % offset, 56 [km/h] (Sae 180_5) Vehicle centre Left A pillar Right A pillar 60 Deceleration [g] Time [ms] Figure Decelerations of an offset crash Numerical simulation of a 30 degrees crash The same model adjustments as carried out for the offset crash simulation are necessary for a 30 degrees crash simulation. In Figure 4.35, the top views of the deformation are shown in six time steps. Again the cable system is working until t=40 ms. Note that the first 10 ms only the bumper deforms, after that time the longitudinal starts to deform. However, after 30 ms the loaded longitudinal starts to collapse, and after t=40 ms the rod which pulls the cable is not moving backwards. This leads to the end of the folding process in the unloaded longitudinal. See also Figure 4.36 in which the folding process of the crush components is visualized. At t=40 ms the engine is crushed against the left longitudinal. More folds are formed after the unloaded longitudinal encounters the rigid barrier after 60 ms. As can be seen in Figure 4.35 at t=75 ms, the unloaded longitudinal has not bent and is still able to form folds.

131 Design of a Frontal Safety Structure Suited for Different Crash Situations 131 t=0 [ms] t=10 [ms] t=20 [ms] t=30 [ms]

132 132 Chapter 4 t=40 [ms] t=75 [ms] Figure Top view of a 30 degrees crash in six time steps.

133 Design of a Frontal Safety Structure Suited for Different Crash Situations 133 LH RH t=20 [ms] LH RH t=40 [ms] Figure Inside view, folding process of both crush components during a 30 degrees crash in two time steps. The velocity curve of the 30 degrees crash is plotted in Figure The large difference in velocity between the left and right A-pillar again indicates the vehicle rotation. At t=75 ms the right side of the vehicle has still a velocity of 25 km/h, which can be further reduced by the right longitudinal, while the left side has absorbed enough energy to stop at t=49 ms.

134 134 Chapter degrees impact, 56 [km/h] Velocity [km/h] Vehicle centre Left A pillar Right A pillar Time [ms] Figure Velocities of a 30 degrees crash degrees impact, 56 [km/h] (Sae 180_5) Rigid barrier force 600 Force [kn] bumper collapse longitudinal deformation engine hits rigid wall engine hits longitudinal right longitudinal makes more folds Time [ms] Figure Rigid barrier force of a 30 degrees crash.

135 Design of a Frontal Safety Structure Suited for Different Crash Situations 135 The rigid barrier force is plotted in Figure In the first 10 ms, it is very low because only the bumper collapses. After 33 ms the engine is crushed increasingly more against the bending longitudinal in front of the engine, this generates a force peak to decelerate the engine. From t=57 ms to the end of the crash the right longitudinal is making more folds, as can be seen by the fluctuating force level. This fluctuating level is lower as in the first stage in which both longitudinals form folds. It is obvious that the rotating of the vehicle resulting in the velocity difference on both sides also has its influence on the decelerations. The same difference between both sides is found. At the beginning of the crash of the loaded longitudinal, the unloaded side is accelerated just a little (see the negative deceleration in Figure 4.39) degrees impact, 56 [km/h] (Sae 180_5) Vehicle centre Left A pillar Right A pillar 60 Deceleration [g] Time [ms] Figure Decelerations of a 30 degrees crash A non-axial component test as verification is not realistic An experimental verification of the longitudinal member as vehicle component as done in Section 3.9 for an axial load is not realistic for a 30 degrees impact. Looking in more detail to the differences between a 30 degrees full-scale barrier test and a real oblique crash between two cars explains this statement. For a full-scale 30

136 136 Chapter 4 degrees crash direction it is important to make a difference between a rigid wall with or without anti-slide. Without anti-slide, the vehicle changes from direction (glanceoff) and the forces on the longitudinal are much lower (Justen 1993, Seiffert 1992). With anti-slide, as normally used in full-scale experiments, see also the vertical strips on the rigid wall in Figure 4.31, the position of the contact point between the front of the longitudinal and wall remains constant. Of course the less stiff corner of the vehicle deforms (see also the simulation results in Figure 4.35 where infinite friction with the wall is supposed). In the contact point between wall and longitudinal front, the transverse force component from the wall (half of the axial force) is compensated by the infinite friction force of the wall or strip. If the contact point can not move sideways, it has no transverse influence on the structure. So there is in fact only an axial force. Note that in a real frontal car to car crash where the opposite car has an oblique direction, the contact point moves sideways (law of conservation of momentum) resulting in an additional transverse force on the already axially loaded structure. For this reason, an oblique crash test is only useful in a realistic full-scale test between two freely movable cars. This goes beyond the possibilities of the research scope. Although it is not expected such oblique collisions become soon a compulsory crash test (proposals are substituted by the compulsory axial offset test for October 1998), the function of additional bending resistance of the telescope must not be underestimated. In axial loads, there is often a chance for a bending collapse of the crushing component or there are stiff components like bumper or engine pulling or pushing the longitudinal. The telescope gives with the additional rings continuous side support. Another oblique load direction on the telescope could be the positioning angle of 10 degrees from bumper height to passenger floor in a traditional built vehicle. The backside part of the telescope must be very stiff mounted to the firewall and floor section. Additional supporting beams may be connected to the outside of this largest supporting tube. During a crash, the telescope length reduces and the bending resistance increases. The bumper has to move down or the passenger compartment has to move up about 115 mm. This last option is conceivable because vehicles mostly lift up their backside, see also Figure These effects dependent on frontal friction and inertia forces can only be tested in a full-scale crash.

137 Design of a Frontal Safety Structure Suited for Different Crash Situations Conclusions A structure consisting of two stiff sliding rods and two cables connecting the rear of one rod inside one longitudinal to the front of the other longitudinal is added to the crush/support component concept to transmit the crushing force from the loaded to the unloaded longitudinal. Numerical simulations with a complete frontal vehicle model showed that both longitudinals have a stable folding pattern during the first half of an offset or a 30 degrees crash. In the second half of the crash, the vehicle rotates too much which creates a bending in the loaded longitudinal. Model modifications like a better mass distribution (yielding correct inertia properties) and a less stiff crush component could prevent this problem in future. In addition, some numeric problems between the cable and the gliding rod resulting in unrealistic deformation of the rod and the crush component should be solved. The rod collapses after 40 ms. This can also be seen in Figure 4.21 at t=60 ms where the rod has no straight shape in side view. The reason for this contact problem is that the cable has been modeled in a too simple fashion for minimizing the computation time. As a result, the total energy of the integral system increases after 40 ms, while it has to be constant by principle. The kinetic energy of the model decreases as expected after 40 ms, which means that most simulation results are usable. The increasing total energy involve also the internal energy which means that the graphs of the absorbed energy in Figure 4.40 are not valid for times after 40 ms. Note the abrupt increase at that point. One of the objectives of the cable system is to obtain the same energy absorption for all impact configurations, to be achieved by a stable folding process of both loaded and unloaded longitudinal. In Figure 4.40 the energy absorption of both longitudinals are presented for a full overlap, a 40 per cent offset and a 30 degrees impact all with 56 km/h. As a reference, also the energy absorption of the longitudinal in case of a 40 per cent offset without the cable system is shown. As the longitudinal energy absorption is not accurate for times later than 40 ms, in Figure 4.41 the energy absorption s are presented until 40 ms. In Figure 4.40, the impression could be made that there is less energy absorbed by the longitudinal deformation in the 30 degrees crash than in the other collision situations. However, in this case the longitudinals are deformed later in time, since first the bumper has to deform before the longitudinals hit the rigid barrier. This happens only after 11 ms after impact. For a better comparison in Figure 4.41, this delay of longitudinal deformation is taken into account by shifting the time axis of the 30 degrees impact.

138 138 Chapter 4 10 x Longitudinals energy absorption, 56 [km/h] 104 Full overlap 9 40% offset 30 degress e impact 8 40% offset (no cable) 7 Energy [J] no valid evaluation Time [ms] Figure Total energy absorption of the longitudinals in different crashes. 10 x Longitudinals energy absorption, 56 [km/h] 104 Full overlap 9 40% offset 30 degress e impact 8 40% offset (no cable) 7 Energy [J] Time [ms] Figure Energy absorption of the longitudinals in different crashes during 40 ms.

139 Design of a Frontal Safety Structure Suited for Different Crash Situations 139 The conclusion can be drawn that using an advanced longitudinal design with cable system increases the energy absorption considerably in case of an offset and an oblique impact. However, one should note that the energy absorption is still less than the energy absorbed in a full overlap crash. Several reasons (unrealistic mass distribution, too stiff longitudinal, higher rotation moment caused by higher offset load) are mentioned to explain this difference. Another reason is of course the fact that the not involved longitudinal is not loaded by the cable at the crash initiation, but after the loaded longitudinal has deformed 30 mm (due to the desired cable margin of 30 mm) to prevent a peak load. It can be seen that the energy absorptions of the offset and oblique crash initially stay below the energy absorption of the full overlap and after about 5 ms the difference remains relatively constant for a longer time. In Figure 4.42 the deceleration levels of the full overlap, 40 per cent offset and the 30 degrees collision are shown together in one picture as a function of the deformation length instead of the elapse time. Again for a better comparison with the 30 degrees collision, for this crash situation a time correction of 11 ms resulting in 171 mm displacement is taken into account. The deceleration levels are accurate and plotted until 485 mm deformation length (about 40 ms), after that time numerical instability occurs as already mentioned % overlap 40% overlap 30 degrees Decelerations at vehicle centre, 56 [km/h] Deceleration [g] Deformation length [mm] Figure Comparison of the deceleration levels in three different crash situations.

140 140 Chapter 4 Although the shape of the deceleration curves is sometimes erratic, on several deformation length values the deceleration level is similar. Figure 4.43 shows smoother curves where the velocities are plotted against the deformation length (until 485 mm). The decrease in velocities appears to be quite similar in spite of the completely different collision situations Velocities at vehicle centre, 56 [km/h] 100% overlap 40% overlap 30 degrees 50 Velocity [km/h] Deformation length [mm] Figure Comparison of the velocities in three different crash situations. In the prescribed research, an advanced new longitudinal front structure is designed to protect the occupants for different collision situations. With the aid of numerical simulations the concept ideas are analyzed and further optimized. The results show that the designed structure could function as expected. Mentioned problems have mainly numerical reasons. Although the used numerical models of the complete vehicle front were not validated with full-scale crashes, as done regularly by car manufacturers with their own crash test data, some checks are possible to make the results more plausible for the first 40 ms. The absorbed energies of the important vehicle parts like the longitudinals, the frontal region, wing and firewall are analyzed and show realistic values. For example, the absorbed energy in the longitudinals is at 40 ms (56 km/h crash velocity) for the full overlap crash 55 per cent of the total absorbed energy and for the 40 per cent overlap and the 30 degrees crash both 48

141 Design of a Frontal Safety Structure Suited for Different Crash Situations 141 per cent of the total absorbed energy. These are realistic values as mentioned in Chapter 2. Further it must be noted that it is not the intention of this research to exactly predict the absolute value of the absorbed energy or deceleration levels of the vehicle structure for several collision situations. The numerical simulations are used as tool to optimize the design and to analyze if the designed cable-supported longitudinal structure could work and if the crash results of the different collision situations are mostly similar. Possible model deviations can have the same influence in all analyzed crash situations and have less influence on the relative difference between the analyzed situations. To verify that the designed structure really generates the desired deceleration levels in different collision situations, full-scale vehicle crashes are necessary. In this case, the new longitudinal structure should be built inside a number of vehicles and afterwards several realistic crash tests must be done. Of course, this is very expensive and time consuming. Secondly, a problem is that the result is dependent on the used car body and engine positioning. For optimal results, the designed system must be customized for the applied vehicle. If in future industrial parties are interested to build this structure in a new concept car, it is possible to test the system with the manufacturing and crash facilities of that manufacturer. Further optimization of the numeric model has not been pursued. The reasons are that this is rather time consuming and will cost a lot of computer time. For example, the 30 degrees crash simulation on a Cray C90 Supercomputer had a cpu time of 80 hours and the 40 per cent offset crash simulation on a HP 715/80 workstation had a cpu time of more as 300 hours. In practice the final elapse times are twice as long, depending on the availability of cpu time in a multi user/task environment. Moreover, the simulation results will certainly show deviations with a real crash experiment. Reason is that many unknown factors like the final engine geometry and possible position and other not modeled components have an influence on the crash behavior of the longitudinals. Moreover, adjustments in the structure are necessary to reach a desired deceleration level, which also depends on the final weight of the car. Main goal for these simulations was to show the designed system could work. It further showed, despite all limitations, that the difference in energy absorption between the most important crashes can be considerably reduced. This is a very important result, because with this advanced design the same deceleration level of the car could be reached for every crash overlap percentage. Now it is possible to design a frontal car structure with one optimal stiffness that hardly varies for different crash situations. Hence, one optimal occupant deceleration level yielding the lowest injury levels can be obtained for the entire collision spectrum. In the next chapter, the optimal deceleration level is obtained and design solutions are presented for realization.

142 142 Chapter 4

143 A Structural Solution to Realize the Desired Deceleration Pulse 143 Chapter 5 A Structural Solution to Realize the Desired Deceleration Pulse 5.1. Introduction The aim of this research is to minimize the injury level of the occupant in several frontal collisions. Therefore, it must be clear which parameters influence the injury level. If an undeformable passenger compartment and no intrusion of vehicle parts like steering wheel, dashboard and pedals are assumed, the injury level is only influenced by means of g forces of the deceleration pulse generated by the vehicle front. To be sure that the injury level is the lowest possible, a numerical model is necessary to calculate the expected injury level by variation of the deceleration pulse. If the optimal deceleration pulse for a specific crash velocity is known, the structure must be designed to generate such a desired crash behavior. With an ideal not deforming passenger compartment, it is acceptable to use an uncoupled model of the dummy and the frontal deforming structure. A common method is, to predefine a deceleration pulse as input on the passenger cage. A full frontal coupled model has a longer calculation time, also because the dummy movement has a longer crash duration time while the frontal structure is already fully deformed. The usual real time interactions between the occupant and the vehicle structure during a crash (Khalil 1995), which influence the vehicle deceleration a little (Seiffert 1992) by means of the restrained dummy mass, can be compensated in the input pulse. Of course for exactly determining the deceleration pulse of a vehicle structure (not for determining the pulse that causes the lowest injury level) the dummy masses must be added to the vehicle model with restraint characteristics. Note that in Chapter 4 it was acceptable to simulate the structure deformation without dummy, because especially the differences between vehicle decelerations for different collision situations have been studied, which are not really influenced by the dummy. Of course in case of a side impact an uncoupled method is not allowed, the dummy mass and its close position to the door have a not negligible influence on the deformation behavior of the relatively low mass of the side structure (Landheer 1996). With the aid of an interior model, variations of the deceleration pulse can be compared on basis of a calculated injury level and an optimal pulse can be obtained

144 144 Chapter 5 for several crash velocities. Structural design specifications are presented to realize such an optimal pulse and conceptual design ideas will be proposed which fulfil these desired deceleration levels for different crash velocities Example of a method for optimizing the deceleration pulse To compare the injury severity for different vehicle collisions, some sort of index or formula is needed. The regulations for vehicle crashes only prescribe a maximum value not to be exceeded for several different injury criteria. Because the proposed vehicle model has no intrusion, only the injury criteria as mentioned in Table 5.1 with their limiting values (Levine 1994, Mertz 1993, Seiffert 1997) are useful. An overall severity index can be a specific weighted combination of these injury criteria, and which takes also into account the relative importance of individual changes of these injury criteria (Bakker 1997). The relative importance to an overall severity index can be expressed by a weight factor (Viano 1990). For an extended description of an overall severity index see a separate research of Witteman (1999). Table 5.1. Relevant injury criteria with their by legislation limited values. Injury Criterion HIC CHEST-G CHEST-D FEMUR-F NECK-M Limit value g 50 mm N 189 Nm Weight factor The Head Injury Criterion (HIC) is calculated on a specific time interval around a deceleration peak of the dummy head to reach a maximum value as shown in next formula, where t 1 and t 2 are the start and end time of the considered deceleration interval in seconds and a(t) is the head deceleration in g as function of time: 25. t 2 1 HIC = max at dt t t t t () t ( ) ( 2 1 ) (5.1). CHEST-G is the peak deceleration in g of the dummy chest. CHEST-D is the peak compression of the dummy chest, mostly a result of the belt force.

145 A Structural Solution to Realize the Desired Deceleration Pulse 145 FEMUR-F is the maximal longitudinal force in the upper leg caused by the dashboard. NECK-M is the flexion bending moment of the dummy neck by forward head rotation. To simulate the movement behavior of an occupant and to measure the forces working on the body, use can be made of a modern deformable frontal finite element dummy Hybrid III (ESI 1996). This dummy consists of rigid body elements and a full deformable thorax and pelvis and is developed by the safety group of ESI SA (Rungis, France) in collaboration with the Biomechanics Department of Wayne State University (Detroit, USA). This numerical dummy simulates the crash dummy of a full-scale frontal crash. In literature a good correlation is reported between computed accelerations of the basic rigid body dummy and measured accelerations in a sled test (Ni 1991). Also the new dummy with deformable chest and pelvis, shows good correlation s with low and high speed pendulum impact tests and with a sled test (ESI 1996, Schlosser 1995). In Figure 5.1 the full scale 50 per cent test dummy Hybrid III is shown together with its finite element equivalent. A B Figure per cent Hybrid III dummy (A) and the numerical model (B).

146 146 Chapter 5 The dummy model must be positioned inside a realistic vehicle interior model (Bosch 1993, Relou 1995, Seiffert 1989, Wijntuin 1995). To this aim, a seat, a dashboard with steering wheel, a floor plane and a restraint system must be modeled, see Figure 5.2. The restraint system consists of a belt with automatic lock and a retractor, and a folded (as far as necessary) airbag (Hoffman 1989). The folding FEM airbag has a good interaction with a dummy and shows a good agreement with experimental data (Hoffmann 1990, Lasry 1991). The seat has a so-called anti submarining plate, which prevents forward moving of the dummy pelvis under the lap belt. Figure 5.2. Dummy positioning within the interior with restraint systems. Optimization for a single collision velocity, a single collision direction, and a single overlap percentage seams not realistic. But as mentioned in the conclusions of Chapter 4, with the designed longitudinal structure it is possible to generate one, not yet optimized, deceleration pulse that is quite similar for different overlap percentages and crash directions. In Chapter 2 it was found that the speed of 56 km/h against a rigid barrier is a realistic test speed because it gives a balance between acceptable injuries at lower speeds and minor fatalities at higher speeds. Research has been carried out (Witteman 1999) to find a deceleration pulse (the resulting deceleration-time signature on the vehicle passenger compartment generated when a collision occurs) with a minimal injury risk, based on the lowest value of the overall severity index, at a 56 km/h crash during 90 ms. This pulse determines the occupant loading and hence the injury risk for a passenger in a vehicle involved in an accident. In this research the reverse approach is used by answering the question which crash pulse gives the lowest injury levels with an already optimized restraint system, instead of finding the optimized restraint system

147 A Structural Solution to Realize the Desired Deceleration Pulse 147 for a given crash pulse. In Figure 5.3 the optimal pulse for 56 km/h is given with the corresponding velocity and deformation length curve against time. The pulse has a quite similar shape as the pulse used in Chapter 2 (Figure 2.5). Deceleration Deformation Velocity deceleration [g], deformation [cm], velocity [km/h] time [ms] Figure 5.3. Optimal deceleration pulse and the velocity and deformation curve. In Figure 5.4 the specific injury time plots of this new optimal pulse are given as an example. The injury values are plotted with a SAE180_5 filter with the times in ms on the x-axis and on the y-axis for the HIC the deceleration in mm/ms 2, for the CHEST- G also the deceleration in mm/ms 2, for the CHEST-D the negative elongation (deflection) in mm of 7 bar elements perpendicular to the chest (where the CHEST-D is calculated as the average of this 7 distances), for the FEMUR-F the force in kn, and for the NECK-M the flexion moment in knmm. For this optimal pulse the calculated injury values are given as indication in Table 5.2. Table 5.2. Simulation results for the injury types of an optimized pulse at 56 km/h. Injury Criterion HIC CHEST-G CHEST-D FEMUR-F NECK-M Simulation value g 21 mm 5066 N 29 Nm

148 148 Chapter 5 P u ls e - V mm/ms M a g n i tu d e ( S a e _ 5 ( A c c e le r a ti o n [ ]) H IC ) ) H IC = T im e [m s ] P u ls e - V HIC mm/ms M a g n i tu d e ( S a e _ 5 ( A c c e le r a ti o n [ ]) M S ) ) C H E S T G = T i m e [m s ] P u ls e - V B a r A x i a l E lo n g a ti o n [ t/m ] CHEST-G mm C H E S T D = T i m e [m s ] P u ls e - V CHEST-D 7 S a e _ 5 ( S P R IN G L o c a l F o r c e - M a g n i tu d e [ ] = -. S a e _ 5 ( S P R IN G L o c a l F o r c e - M a g n i tu d e [ ] = - F E M U R F = kn FEMUR-F Figure 5.4. Injury values of an optimal pulse for 56 km/h. NECK-M

149 A Structural Solution to Realize the Desired Deceleration Pulse Structural design specifications for different crash velocities Since more than 90 per cent of all frontal collisions occurs at a velocity lower than the prescribed crash velocity of 56 km/h, see Figure 2.1, also an optimal pulse for lower collision velocities is necessary to minimize the occupant injury level at that lower velocity. Although only 2 to 10 per cent (dependent of the overlap percentage) of all crashes takes place at a velocity higher than 56 km/h, also a pulse optimized at such a velocity is interesting because of the higher energy level yielding larger vehicle deformations and higher injury levels. If the initial crash velocity is decreased to 32 km/h, this results in a decrease of crash energy of 67 per cent with respect to a crash speed of 56 km/h, so the vehicle might just be too stiff. An increase of the initial crash velocity to 64 km/h results in an increase of energy of 31 per cent, so the structure might be too supple. An optimal pulse for a speed of 64 km/h (total deformation length of 762 mm) is plotted in Figure 5.5 together with the already mentioned optimal pulse for 56 km/h (total deformation length of 724 mm) and with an optimal pulse for a collision with 32 km/h (total deformation length of 448 mm). These additional pulses are obtained with the same numerical research. Before designing structural solutions to realize the desired deceleration pulses (see next section), first the specifications for this design will be mentioned. For these specifications the optimal curves that were obtained for three different crash velocities as shown in Figure 5.5 will be used. In this figure the velocity decrease is plotted against deformation length instead of time, because it is more interesting to know on which length position in the car structural measures are necessary to realize a change in stiffness corresponding with the desired change in deceleration level. In Table 5.3 the time duration and deformation length of each deceleration interval are presented. As can be seen the difference in deformation length in the first interval of the 56 km/h and the 64 km/h collision is small. So for simplification the lengths of 170 mm and 188 mm could be joined together on 179 mm. At the end of the second interval the deformation length is already identical for both velocities, vz. 586 mm. These interval borders are visualized in Figure 5.5 as two vertical lines. For the 32 km/h collision there is no difference in deceleration between the first two intervals.

150 150 Chapter 5 velocity [km/h] km/h 64 km/h 32 km/h 45g 32g 9g 9g 9g 23g 23g 9g deformation length [cm] Figure 5.5. Three optimal decelerations curves in three phases (Witteman 1999). The optimal pulse obtained for higher velocities has a higher deceleration level in the first interval and the levels of the middle and the third interval remain unchanged in comparison with an optimized pulse for 56 km/h. The obtained optimal pulse for a collision with 32 km/h has a constant deceleration level of 9 g, the same level as the higher velocity pulses have during their middle interval. From these observations it can be concluded with the considered numerical model, that for minimal injury for crash velocities starting at 32 km/h the vehicle structure needs a constant stiffness to decelerate 9 g during the first 586 mm. For higher velocities as 32 km/h the stiffness of the first 179 mm must be directly increased to decelerate up to 45 g for the highest velocity of 64 km/h. After 586 mm deformation has been reached the stiffness must be increased to decelerate to 23 g for all relevant velocities.

151 A Structural Solution to Realize the Desired Deceleration Pulse 151 Table 5.3 Deceleration parameters of 3 crash velocities (Witteman 1999). Crash velocity 32 km/h 56 km/h 64 km/h Phase 1 Deceleration 9 g 32 g 45 g Deformation length 170 mm 188 mm Time duration 12.5 ms 12.5 ms Phase 2 Deceleration 9 g 9 g 9 g Deformation length total 448 mm 416 mm, total 586 mm 398 mm, total 586 mm Time duration ms 42.5 ms, total 55 ms 37.5 ms, total 50 ms Phase 3 Deceleration 23 g 23 g Deformation length 138 mm, total 724 mm 176 mm, total 762 mm Time duration 35 ms, total 90 ms 39.4 ms, total 89.4 ms 5.4. The necessity of an adaptive structure In the following sections, conceptual design ideas will be presented which can fulfil the specifications of different deceleration levels for an optimal deceleration pulse as given in Table 5.3. Although the obtained pulses must be adapted a little to compensate for the dummy mass(es) during a crash test, this will not be done in this conceptual research because the separate dummy influence is not simulated with the used uncoupled model. To realize the lowest deceleration level of 9 g in the second phase of the optimal deceleration curves, which is independent of the considered crash velocities, the average constant crumple force of the longitudinals can be used. However, to realize a higher deceleration in the first phase with a stiffer structure in the front part is more difficult, because normally deformation starts in the weakest part of the loaded structure. In addition, the desire to obtain different deceleration levels for different collision speeds means that an adaptive structure is needed to adapt the stiffness at the beginning of the crash. Note that in case of adaptable structure stiffness occupant mass corrections are easier to realize by determining the additional vehicle load, and based on this value the structure stiffness must be increased. For the last phase, it is easy to use a stiffer cross-section of the longitudinal to increase the deceleration to a velocity independent level. However, its more likely that after 586 mm deformation the engine and other components will be

152 152 Chapter 5 impacted against the firewall which generates already a much higher deceleration. It is more important to prevent an early increase in the crash force by the engine. In the next sections concept ideas are presented as a solution for the mentioned problems. Note that the ideas and new alternatives are given to complete the proposed design solution for the whole frontal vehicle crashworthiness problem, but simulation or expensive validation of these ideas goes beyond the possibilities of the research scope. However, the ideas have enough potential for further research in future Energy absorption by friction A practical method to absorb kinetic energy is by means of friction. Changing the pressure force on a friction block regulates the energy absorption. The well functioning idea of hydraulic vehicle brakes can be used on the backwards moving cable rod. The application of friction blocks around the two square rigid rods can generate the desired additional deceleration forces. In case of a 64 km/h collision the additional deceleration, next to the 9 g generated by the crushing longitudinals, must be 36 g. For this deceleration an axial friction force of 388 kn is needed in the case of a 1100 kg vehicle. The choice for friction material must be tungsten, because a high temperature could be expected and melting of the material must be avoided for a necessary high friction coefficient. As calculated in Appendix B, the temperature at the contact surface of the friction block rises 2328 K and on the rod the temperature rises 1698 K, but the temperatures drop very fast inside the material. Therefore both surfaces must contain a coating or plate of tungsten. In this case the expected friction coefficient is at least about 0.45 (normally for different dry metals without lubricator) but in case of two equal metals the friction coefficient could rise to 1.0 or more (Landheer 1993). This means that for a safe value of 0.45, a total normal force of at most 862 kn is needed. For this high force level, a hydraulic system is the right choice. If an available hydraulic pressure of maximal 1350 bar is supposed (the same pressure as in common rail diesel engines), the necessary piston area behind the friction surface must be 6385 mm 2. For this surface, 10 pistons with a diameter of 30 mm (total 7069 mm 2 ) are sufficient to compensate also some pressure loss in the pipe between a radial piston pump and the pistons. The pistons could be positioned in two rows of five pistons with 20 mm space between, opposite to each other and connected with a frame on the cable disk axles under the vehicle floor. Of course a strong connection with the car structure is required. See Figure 5.6 for more details.

153 A Structural Solution to Realize the Desired Deceleration Pulse 153 top view rod disk A A frame rod pistons cables pistons 10x disk cable oil disk frame tungsten plates firewall longitudinal tube 4 cross-section A-A Figure 5.6. Sketch of the friction pistons against the cable rod. In the case of an offset or an oblique collision where only one longitudinal is directly loaded, it is not allowed to use the additional friction force on the other, not directly loaded, longitudinal. The fracture force of the cable (279 kn) is only sufficient to resist the peak load of the folding process (180 kn). For this reason, two sensors are required inside the bumper, in front of the longitudinal, which detect a contact with an object by means of a pressure force or with radar detection. If only one signal is detected (offset collision), only on the cable rod in the longitudinal at that side the maximal needed additional friction force must be generated. In the case of two signals (full overlap collision), both cable rods must be loaded with half of the total needed additional force. To determine the necessary additional force, the velocity information of the vehicle must be used. Since most modern cars use ABS which continuously detect the speed of each wheel, the current speed (or before the last 100 ms from memory to prevent crash influence) of the car is always well known. With this information the pressure generated by the radial piston pump could be controlled. In case of

154 154 Chapter 5 velocities below 32 km/h it could be zero, in case of velocities up to 64 km/h it must be increased to for example 1350 bar. A radial piston pump with zero regulation (no power loss) can be equipped with electronic pressure control. Another possibility is to keep the highest pressure always available and control the magnetic valves on each piston (comparable with common rail diesel engines). Probably this is for faster adjustments in the very short time preferable. In Table 5.4, the required number of opened valves is mentioned. In case of a symmetric collision the number is valid for each longitudinal, in case of an asymmetric collision the number is only valid for the directly loaded longitudinal, the valves of the other longitudinal must be closed. Of course for other collision speeds between 32 km/h and 64 km/h a number between the mentioned numbers could be chosen. Table 5.4. Number of opened valves to reach enough pressure for additional friction force. Crash velocity 32 km/h 56 km/h 64 km/h Symmetric collision 0 valves 3 valves 5 valves Asymmetric collision 0 valves 6 valves 10 valves The additional axial force of maximal 388 kn on the cable rod is too much to prevent buckling of the rod. The Euler buckling load is defined as (PBNA 1993): F b F 2 2 ( c E I ) = π (5.2). b l b where c denotes the buckling mode, which has a value of 4 if one side of the rod is fixed and the other side can only make a longitudinal translation, E is the Young s modulus, I is the smallest principle moment of inertia and lb is the buckling length. With c = 4, E = N/m 2, I = m 4 and l b = 0.98 m, F b = 239 kn. The only way to increase this buckling load with the same geometry is to divide the free buckling length into halves. If an additional sliding support in the middle of the cable rod is used, the buckling force increases to kn = 956 kn. This is enough for the crumple load and the maximal additional friction force. This sliding support could be a square plate around the rod connected with breakable spot-welds in the middle of the rod. Because this plate moves together with the rod to the firewall, it does not disturb the folding process of the crushing component. After 490 mm deformation, this plate stops against the stiff firewall, while the rod breaks the connection and moves further rearwards.

155 A Structural Solution to Realize the Desired Deceleration Pulse 155 After 179 mm deformation, the additional friction force must be removed. This can be done by moving rapidly the zero regulation rod, which controls the eccentricity of the radial piston pump axle, so the oil flows in the opposite direction back and lifts the pistons from the rod. The movement of the zero regulation rod can be done electronically by the pressure control module or it can be done mechanically by a mechanism connected with the zero regulation rod and activated at the right moment by the crossing cable rod. Another possibility is the use of a large electronic valve in the common pressure pipe, which releases the pressure rapidly. The third phase of the deceleration curve starts always at 586 mm deformation length. From this point, the deceleration must be increased from 9 to 23 g as long as the crash lasts. If the engine is involved before this point, which is plausible in smaller cars, a solution could be flexible engine mounting points. As already mentioned in Section 4.4, the engine mounting rods must be easily deformable to ensure rearward sliding of the supporting components. In addition, the connection of the engine with the drive line must be movable to prevent high translation forces on the engine too early. All the aggregates must be positioned in such a way that only after 586 mm deformation has been reached, high contact forces start to press the aggregates together. In addition, the front wheels have to deform the wheel bay and the sill. Finally, the engine hits the stiff firewall, which could deform at high collision speeds. The final deformation forces are very dependent on the positioning and the dimensions of the aggregates and the free space under the bonnet. If the necessary force level is not reached, assistance of the friction force as used in the first phase of the collision is always possible. Signals must be send to the pressure regulation module and the valves to control the correct friction force Future possibilities An optimal regulation for the whole deformation length is of course with a computer controlled system, which measures continuously the actual deceleration level and adjusts at the same time the pressure to reach the programmed optimal deceleration pulse. Maybe when very fast sensors, high-pressure valves and control modules are available this is a realizable idea. In this way, it is also possible to compensate for the stiffness, velocity or weight of the colliding obstacle. This would be an ideal solution for the compatibility problem between small and large vehicles. If this system is fast enough and very reliable, it is possible to think about a structure which has only two very stiff beams, which can fully slide backwards without deformation. A heavy computer controlled break system regulates the desired deceleration. In this case, a

156 156 Chapter 5 cable system to direct asymmetric forces to a symmetric force distribution is not necessary, because sensors already send signals to increase the friction of one loaded beam to reach the same energy absorption. In addition, the telescope structure is not longer necessary, because the new beams have not to crumple to absorb energy so they can be made very stiff with a high bending resistance. Of course the control system with the breaks must be reliable in all crash situations because there is no alternative to moderate the energy absorption, which means that large force level differences must be taken care of. Only problem could be the space behind the firewall or under the vehicle floor. Vehicles with structural space under the passenger compartment have very good possibilities for safety increasing features, also for side impact crashworthiness. A very nice vehicle concept for this application is the new Mercedes-Benz A-class vehicle. Because of the double floor with a higher placed passenger compartment, the longitudinals stay fully horizontally in a stiff ladder chassis. In the floor structure there is enough space for rearward sliding beams and for the positioning of the energy absorbing brake system. Furthermore, the engine does not shorten the available deformation length or penetrate the firewall since it moves to the road surface. Maybe with the popularity of space wagons or mini multi purpose vehicles nowadays, this is an interesting design aspect. Occupants are not longer sitting in the extension of the crumple zone but above, especially at side impact crashes. In the case of structural space behind the firewall, the hydraulic piston solution presented in next section is another possibility Design of a hydraulically controlled frontal car structure To load the missed longitudinal member during an asymmetric collision, it is possible to use a hydraulic system instead of the proposed cable system. In Figure 5.7, a principle sketch of the system is shown with besides the longitudinals two cylinders with pistons.

157 A Structural Solution to Realize the Desired Deceleration Pulse 157 Figure 5.7. A hydraulically controlled frontal car structure. The cylinder rods are fixed to the cross member, just like the front ends of the longitudinals. If one of the longitudinals is loaded during an offset crash, it starts to deform and because of the connection to the cylinder, the rod slides into the cylinder. The oil inside the cylinder is pressed via a tube or pipe to the rod side of the cylinder of the unloaded longitudinal. Under the influence of this oil pressure, the piston of this cylinder is also pushed backwards. Because this piston is connected to the unloaded longitudinal member, it is forced to collapse in an axial folding mode. The pressure that arises in the cylinder of the unloaded longitudinal is led back to the rod side of the cylinder of the loaded longitudinal, where it helps to further move the piston inside the cylinder. Hence, the hydraulic cylinders form a closed-loop system. Note that in the case of a full overlap collision where both longitudinals are loaded, the system is in equilibrium and does not influence the crash behavior. One problem is however, that the oil volume in the cylinder does not fit in the other cylinder at the rod side, because of the volume of the rod itself. Because the rods move inwards, the total available volume decreases. Solution is a piston with at each side a rod, where the second rod has not a force function but causes identical volumes exchanges. For this solution there must be space behind the firewall where the additional rods can move backwards. Advantage is the same area at each piston side, which gives a 1:1 force transmission. A second problem is the available deformation length, because a cylinder with piston can be shortened less as half of the original length. For this reason it is also

158 158 Chapter 5 necessary that there is much horizontal space under the passenger floor, because then the cylinders could be mounted at the rear of the firewall. For the connection pipes, enough space is also important because they must have a large diameter and a short length to minimize the pressure loss at high stream velocities. With a cylinder diameter of 90 mm, the pressure at a crash load of 150 kn is 236 bar. At an initial flow rate of 15 m/s (56 km/h crash), the pressure loss in the connection pipe with a diameter of 30 mm and with oil ISO VG2 is 12 bar/m and 1 bar/m for a pipe diameter of 60 mm (Slaats 1996b). Although the guaranteed maximal velocity for the cylinder s sealing is much lower, the high velocity works a very short time and the system has to work only one movement. The final structure can be built together, the rod of the cylinder can be positioned inside the crushing component as is done with the cable rod, and gives additional bending resistance. The stiff cylinder behind the longitudinal can be used as support structure for the axial crushing forces. This hydraulic supported structure generates a constant deceleration force, independent of the overlap percentage. However, to reach the optimal crash pulse, control of the oil flow is necessary. In this case, a valve with a controllable flow restriction (Janssen 1994) or several valves must be used in the outlet of the backside of the cylinders. Reducing the outlet area increases the pressure and therefore the stiffness of the system. After the first deceleration interval, the valve can be fully opened and for the third interval, if necessary the total outlet area can be reduced again Conclusions A method has been described how a deceleration pulse can be optimized. As an example three pulses are mentioned for three different velocities to use as specification for conceptual design ideas. To fulfil the requirements for different velocities an intelligent structure is desirable. With the use of an additional friction force on the rearwards moving cable rods, it must be possible with a hydraulic system to control the deceleration pulse to the optimal level dependent on the crash velocity. In case of a multi purpose vehicle concept (component space under the passenger floor) a new hydraulic brake or flow system for controlled energy absorption is a promising idea. This intelligent structure with adaptable stiffness is also a solution for the compatibility problem between different vehicles or for compensating the additional occupant and luggage masses.

159 Conclusions and Recommendations 159 Chapter 6 Conclusions and Recommendations 6.1. Overview of the research To assess vehicle crashworthiness, two crash tests (full and partial overlap) representing extreme cases within the frontal collision spectrum have been proposed. If a car shows acceptable results in these two tests, it will be plausible that the car has an optimal crashworthiness in other frontal collision situations. The influence of the overlap percentage on the energy absorption in the frontal structure has been analyzed. To reach acceptable deceleration levels in a full as well as in a partial overlap crash situation, the stiffness of the structure must be chosen in such a way that both different deceleration levels match the optimal characteristic as closely as possible. Intrusion of the passenger compartment must always be avoided, because this gives more injuries than a high deceleration level, since the latter can be compensated by an optimized restraint system. To design an optimal longitudinal member that absorbs sufficient energy in different collision situations, numerical simulations have been executed to find optimal crosssection geometry. However, with a simple crushing column, no solution exists for the various demands with regard to absorption potential to compensate for changes in crash velocity and overlap percentage. A single cross-section configuration can not fulfil all design requirements with respect to the highest mass specific energy absorption, the highest bending resistance, the lowest relative disturbance of the bending resistance in a fold and the most stable force level. To solve these contradictory design requirements, a new longitudinal member has been designed that consists of two functional components. One inside crushing component that guarantees the desired stable and efficient energy absorption and an outside supporting component that guarantees the desired stiffness in the transverse direction. With a rigid connection between the front ends of both longitudinals, the other longitudinal, which is not loaded axially during an offset crash, absorbs additional energy by bending. However, to reach similar energy absorption in case of a full overlap as in case of an offset crash, both longitudinals have to crumple axially. Therefore, the compound longitudinals are supported by a cable connection system for a symmetric force distribution. If only one longitudinal is loaded during a partial overlap crash, the cable forces the other longitudinal to crumple as well. The

160 160 Chapter 6 influence of different collision situations on the energy absorption of this new system has been demonstrated by numerical simulations. A frontal vehicle model has been numerically crashed in a full overlap, a 40 per cent offset and a 30 degrees collision to evaluate the working principle of this new design with cable system. Due to a too stiff crushing component and an incorrect mass distribution of the not modeled vehicle parts, the vehicle rotates too much around the barrier. In addition, some numerical problems cause incorrect energy absorption in the second part of the crash. However, it has been demonstrated clearly that in the first part of the crash the system works. The energy absorption for an offset crash situation has been nearly doubled with the aid of the cable system. The deceleration levels at the three different crash situations (full overlap, offset and 30 degrees) are quite similar. With this result has been demonstrated that the vehicle structure has one stiffness that hardly varies for different crash situations. To obtain the lowest injury levels for the occupant over the entire collision spectrum, the more interesting case of a reverse question is answered: which crash pulse gives the lowest injury levels with an already optimized restraint system. Therefore, a method is described in which a numeric model of an interior and a FEM dummy has been used to find the levels of the injury criteria. To compare the results of different crash pulses, an overall severity index has been used. At a crash speed of 56 km/h, an optimal pulse has been found after several considered pulse variations in a separate research (Witteman 1999). This pulse has three deceleration intervals and deviates much from a traditional pulse, but gives as it seems much lower injuries. For a crash speed of 64 km/h, the deceleration level has been increased only in the first interval. For low velocities, the level of the first interval must be the same as the lower second interval of the higher velocity characteristics. The optimal deceleration pulses found for different velocities have to be generated by stiffness adaptations of the designed structure. Therefore, intelligent design solutions have been presented, which are based on controlling energy absorption. The collision situation is detected with sensors and signals control the hydraulic pressure for additional friction on the backwards moving cable rods. After 18 cm deformation, the pressure is removed for the lower deceleration interval. At the end of the crash in the last (third) deceleration phase, the pressure must be increased again if the deformation forces of the surrounding structure are not high enough to reach the desired deceleration level. To maintain a constant stable force level in the crushing components, a triggering with a bead initiator placed on a half folding wavelength from the rigid longitudinal front is necessary. Another solution to control the energy absorption is a complete hydraulic system, which substitutes the cable system by two

161 Conclusions and Recommendations 161 coupled cylinders. During an offset crash, the backwards moving piston connected with the loaded longitudinal forces the unloaded longitudinal to crumple as well by means of hydraulic equilibrium. A controllable flow restriction regulates the flow friction, yielding additional stiffness. Finally, it has been demonstrated that the new compound longitudinal member could be manufactured and that the crushing component has a stable folding pattern that fits between the rod and the surrounding supporting component. The test results agree with the numerical simulation. A non-axial component test is not realistic, only a full-scale test with two freely movable cars is sensible Conclusions Several conclusions with respect to the improved vehicle crashworthiness design by control of the energy absorption for different collision situations can be drawn. In general it can be concluded that the new designed concept structure satisfies the research objectives, because optimized deceleration pulses can be obtained for different crash velocities and overlap percentages. First some general conclusions are given and then more specific conclusions are drawn. General conclusions: At least two frontal crash tests are necessary to assess vehicle crashworthiness: like a full overlap crash against a rigid wall with 56 km/h and a 40 per cent offset crash against a deformable barrier with a little higher speed to compensate the energy absorption of the barrier. The stiffness of a frontal vehicle structure must generate a deceleration level that is acceptable in both extreme different collision situations. However, the structure must be stiff enough to avoid intrusion of the passenger compartment. For a crushing column, a rectangular cross-section in a lying position gives stable energy absorption and is less sensitive for oblique crash directions than the more energy absorbing geometry s of the hexagonal, the octagonal or the circular profiles. Increasing the wall thickness of a profile increases the mass specific energy absorption much more than increasing the profile perimeter.

162 162 Chapter 6 Optimal triggering of a crushing column for a stable folding process (which starts at the front end), can be done with a bead geometry, which gives a 10 per cent cross-section area reduction of the profile. The initiator must be positioned a half folding wavelength from the rigid front end or 35 per cent of the folding wavelength from a free front end. Multi purpose vehicles with a higher passenger floor have very good possibilities for safety increasing features. The space under the floor can be used for intelligent energy absorbing systems, like controllable hydraulic pressure for additional friction on backwards moving beams or a flow restriction on the cylinder outlet of a backwards moving piston. It is possible to design a vehicle with an intelligent structure, which adapts the stiffness of the structure for an optimal deceleration characteristic at each different crash velocity. Specific conclusions: To increase the bending resistance of a longitudinal cross-section without increasing the axial stiffness, a separate telescope structure is a promising concept giving additional side support to prevent bending or a premature bending collapse of the inside positioned crushing column. To obtain about the same amounts of energy absorption in a full and in a partial overlap crash, both longitudinals have to collapse axially in a folding mode. In case of only bending of the not directly involved longitudinal during an offset crash, 58 per cent of the total energy has been absorbed. An additional cable connection system can force the other longitudinal to crumple as well. It has been demonstrated with numerical simulations that with the designed longitudinal structure and additional cable system, it is possible to design a vehicle that has a similar deceleration characteristic or energy absorption in very different frontal collision situations. However, in the used vehicle model the longitudinals are too stiff yielding a too high deceleration level. With a numerical example optimal crash pulses are mentioned for several crash velocities. For a crash speed of 56 km/h during the first 18 cm deformation length, a high (32 g) deceleration level is desired, then an interval with a low (9 g) deceleration level is desired (dummy is colliding with belt and airbag), and at the

163 Conclusions and Recommendations 163 end (dummy is restrained by belt and airbag) the deceleration level can be high (23 g) again. For a crash speed of 64 km/h, the deceleration level can be increased (45 g) during the first 18 cm deformation length. For 32 km/h, the deceleration level of the first interval can be 9 g, the same level as the lower second interval of the higher velocity characteristics. Experimental verification was only possible with a quasi-static component test. For a realistic verification, the new structure must be built in several vehicles and full-scale tests must be done in different crash situations Recommendations The following recommendations can be given. Increasing the total energy of the model must be avoided. The increase of energy is caused by too simple modeling of the cable with beam elements to save computer time. This gives contact problems with the shell meshed cable rod yielding numerical instability. It is recommended to use shell elements with another sliding interface. For the simulated compact class vehicle the longitudinals must be less stiff. Decreasing the wall thickness of the crushing components can solve this problem. The mass distribution must be more realistic to prevent vehicle rotation around the barrier. A possible solution is presented in Figure 6.1. For a more realistic material behavior in a numerical simulation, the material properties must be determined before the simulation is started by means of a tensile test of a material strip of the same sheet as will be used for an experiment. Material properties found in literature differ too much from the real properties. Using a deformable barrier instead of a rigid wall in case of an offset crash provides a more realistic test for simulation of a car to car crash. In this case, a larger stiff front surface is necessary to prevent a deep perforation of the barrier. However, the computation time will increase largely due to necessary small elements for a good contact definition between barrier and vehicle front. To evaluate injuries caused by intrusion of the passenger compartment or for dummy out of position problems due to rotation of the vehicle at the end of an

164 164 Chapter 6 asymmetric test, a simultaneous numerical simulation of the structure deformation and the dummy behavior is favorable. In this case also the influence of the dummy mass can be determined on the deceleration pulse. An example is showed in Figure 6.1, a simulation with the designed structure without cable system at state t=70 ms. Note the much smaller rotation of the vehicle at this time step, due to the appended passenger compartment with dummy yielding a better mass distribution. Figure 6.1. Offset crash with dummy and airbag at simulation state t=70 ms. The optimization can also be done for different dummies (e.g. a 5 percentile female and a 95 percentile male) and seat positions, but for these kind of variables an intelligent airbag and safety belt system that adjusts itself to the occupants is more favorable. For the occupant simulation it is better to use a human like dummy (which are under development see Figure 6.2) to simulate a more realistic behavior. However, until now legislation is based on measurements at full-scale dummies in crash tests. As soon as reliable numerical human like dummies are available, car manufacturers should use these dummies to design their vehicles with more reliable injury prediction resulting in really safer cars.

165 Conclusions and Recommendations 165 Figure 6.2. Human body model (ESI). The total front structure design is only verified with numerical crash simulations. To verify that the designed structure really generates the desired deceleration levels in different collision situations, full-scale vehicle crashes are necessary. In this case, the new longitudinal structure should be built inside a number of vehicles and afterwards several realistic crash tests must be done. Although the proposed vehicle structure shows good results concerning crashworthiness, it must be verified in the perspective of car body stiffness and car body dynamic response (vibration behavior). Also the comfort aspects influenced by changed engine mounting positions and probably changed wheel suspension must be verified. The mean focus point of this research was optimal crashworthiness. Although the increased vehicle safety could save a lot of economic and medical costs after an accident (besides savings of emotional distress), it must be balanced with higher structure costs and weight. To reduce the additional weight (which consumes more fuel) the new design must be more integrated in the traditional vehicle structure so that with less mass more stiffness requirements are fulfilled. The presented idea of energy absorption by friction between surfaces covered with tungsten must be verified with experiments. Especially the friction coefficient at high contact velocities must be find to determine the correct pressure force.

166 166 Chapter 6

167 Numerical Simulation Method with PAM-CRASH 167 Appendix A Numerical Simulation Method with PAM-CRASH PAM-CRASH is a trade-oriented industrial software dedicated to crashworthiness design in the transport industry, through the efficient numerical simulation of mechanical structural testing (i.e., the regulatory crash tests). PAM-CRASH has been extensively validated by industry worldwide through detailed simulation of full-scale experiments in the automotive, railway, and aerospace industries. By drastically reducing the number of physical prototypes for testing prior to final design completion, PAM-CRASH predictive test simulations translate into better vehicle designs and shorter, less expensive design cycles. PAM-CRASH is based on the most advanced numerical techniques using the Explicit Finite Element Method. It allows the model the necessary details of complex industrial structures, under arbitrary static and dynamic loading conditions such as crash tests, or other extreme conditions. The element library includes solids, shells, beams, bars, and joints (flexible, cardan or spherical). Specific models exist for rivets, spot welds and gaps. Free numbering and keyword input format allows to easily merge models and to create a model database (i.e., car components, dummies, etc.). PAM-CRASH offers many significant improvements, such as foam models and specific and new contact options (including a body-to-plane contact, etc.). To understand the method, a finite element overview is given (ESI 1994): The integration volume is discretized into elements interconnected at nodal points. The equations of motion are solved according to the weak formulation using weighing (interpolation) functions. The displacements are the unknowns and approximated using interpolation functions; continuity is presumed at the nodes. Stresses are obtained from the strains (or their increments) using the constitutive equations of the material. Adequate boundary conditions are required; for dynamic problems initial conditions are also required, but free boundaries are allowed (rigid body motion). As usual for structural mechanics a Lagrangian formulation will be adopted i.e. the coordinates follow the material motion. The explicit solution method employed by PAM-CRASH does not require simultaneous solution of the system of equations. Contrary to implicit/static Finite

168 168 Appendix A Element solvers, no stiffness matrix is assembled and the mass matrix is diagonal. The used explicit integration scheme is given below. A vibrating free-free spring-mass system consists of two masses, m, a spring with constant, k, and an external load, F(t). Consider the second order ordinary differential equation which expresses dynamic equilibrium of this system, m (d 2 x/dt 2 ) + k x = F(t) (1 DOF) F(t) m k m F(t) and a discretized time-axis t n t n-½ t n+½ t n-1 t n t n+1 time t t n+½ Let the known quantities be x n and (dx/dt) n-½. The wanted quantities are x n+1 and (dx/dt) n+½. Dynamic equilibrium at time t n is expressed as m (d 2 x/dt 2 ) n = F n - k x n. Since all terms on the right hand side are known, one can solve for (d 2 x/dt 2 ) n and apply central differences (d 2 x/dt 2 ) n = (F n - k x n ) / m (dx/dt) n+½ = (dx/dt) n-½ + t n (d 2 x/dt 2 ) n x n+1 = x n + t n+½ (dx/dt) n+½. This scheme works as long as the condition on the solution time step, t n < (2m/k) ½, holds, where m is the nodal mass and k is the nodal stiffness or the force needed to obtain a unit displacement at that node in the considered DOF, when all the surrounding nodes are fixed in the context of the structure with more than one DOF. The scheme is therefore said to be conditionally stable. Care has to be taken in the choice of t, i.e.

169 Numerical Simulation Method with PAM-CRASH 169 t n+½ = t n+1 - t n t n = t n+½ - t n-½. After creation of a CAD model and a FEM mesh distribution with the aid of Unigraphics, the data is converted into Generis, a preprocessor of PAMCRASH. In this phase all physical data is added and read by the solver. The following principal operations as input subroutine are read (ESI 1994): control cards material data nodal point data element data boundary condition data nodal and element printout data load curve data concentrated load data velocity boundary condition data rigid wall data initial velocity data sliding interface data section force data airbag data In the solution phase the program performs a loop over the time steps in which the dynamic solution is advanced according to the central difference explicit solution algorithm. Within each time increment the program mainly performs the operations described below. set the time step set global nodal force array to zero add concentrated nodal loads to global force vector treat finite elements update applied pressure initialize energies per material loop over the element groups/elements: - find element nodes global coordinates - define local element coordinate system - transform coordinate and translational velocities from global to local

170 170 Appendix A - calculate shape functions at element center and elemental surface area - calculate new stable time steps per element and apply time step control - eliminate elements with too small time steps - internal force calculation per element: - compute strain increments - update element thickness - loop over thickness integration points: - calculate elastic stress - check for plasticity - calculate internal force contribution - calculate internal energy increment - calculate and add hourglass damping forces - transform forces to global coordinate system - assemble forces calculate rigid body forces treatment of sliding interface contact calculate global translational and nodal rotational accelerations introduce boundary conditions calculate rigid body accelerations introduce nodal constraints treatment of rigid walls calculate velocities and displacements calculate nodal principal inertia coordinate systems compute new airbag volume, pressure, temperature, gas flow. With the following model description of a crushing longitudinal member, some modeling conditions are reviewed. The obstacle with which the longitudinal member collides, is a flat, undeformable wall with infinite dimensions and infinite mass (rigid wall). The friction which occurs perpendicular to the member is taken into account by means of a friction coefficient between the longitudinal and the wall. Furthermore, for the wall and for the member there are a number of displacement boundary conditions, which are mentioned in Figure A.1. At the front of the vehicle, where the longitudinal is connected to the crossbeam, no boundary conditions apply to the longitudinal. For the wall, all degrees of freedom are suppressed, so the wall does not move. The added mass M crash results in a load on the longitudinal and covers the masses of the vehicle, occupants and luggage.

171 Numerical Simulation Method with PAM-CRASH 171 Rigid wall: M Ux=Uy=Uz=Θx=Θy=Θz=1 Rigid body 2 Ux=Uy=Uz=Θx=Θy=Θz=0 µ=0.3 Mcrash v0 Rigid body 1 Ux=Uz=Θx=Θy=Θz=1 Uy=0 0=free direction or angle 1=fixed direction or angle z Uz Θz Θx Uy y Θy Ux x Figure A.1. Model of a longitudinal member and a rigid wall as used in Section 3.4. In Generis, tools are available to adjust the above mentioned boundary conditions and other properties. The entities listed below can be used for modeling boundary conditions and properties of the longitudinal and the rigid wall. Rigid body s are used for modeling a rigid attachment like the connection between the firewall and the longitudinal or the crossbeam with the longitudinal. The speed of the vehicle before the start of the crash is given by the initial velocity. The entity rigid wall is available for modeling an undeformable barrier. For the rigid wall, different properties can be adjusted, like mass, friction with the longitudinal member, velocity of the wall and angle of the wall with the axis of the longitudinal. For the entity rigid wall, slave nodes have to be defined. These slave nodes are nodes of the FEM model which cannot move through the wall. Normally, these will be all nodes of the whole model. The entity added mass is used for specifying the value of the added mass. Furthermore, the place where this added mass is connected to the tube can

172 172 Appendix A be specified by selecting a node of the FEM model. The nodal time history is provided for analyzing the displacement of the model during the crash. The displacement information of different nodes can be saved when the entity nodal time history is used. When two nodes are chosen on the axis of the longitudinal, one at the front end and one at the rear end, the real deflection can be reconstructed. Nodal constraints are used for the definition of boundary conditions for nodes which are found on a plane of symmetry. A contact algorithm is used to avoid certain parts of the longitudinal from moving through one another, which is physically impossible. This so-called sliding interface, checks whether two parts of the structure have approached each other to a minimum distance t min, during for example the folding process. When this distance t min is reached, a reaction force forces the displacement not to be smaller than this minimum distance. The sliding interface can be defined as a self-contact algorithm (between two parts with the same material properties) and as a contact algorithm between different material types. The calculation time highly depends on the application of sliding interface definitions.

173 Calculation of the Temperature Increase after Energy Absorption by Friction 173 Appendix B Calculation of the Temperature Increase after Energy Absorption by Friction To decelerate a vehicle mass of 1100 kg with 36 g during 179 mm length (as specified in Section 5.3), the energy that must be absorbed is 1100 x 36 x 9.81 x = J. As already calculated in Table 5.3, the time duration is 12.5 ms. The heat flux at the contact surface becomes J / 12.5 ms = 5.56 x 10 6 W. The friction surface has a total area of 7069 mm 2 (10 pistons x 30 mm). Per unit area this friction power becomes 5.56 x 10 6 W / 7069 mm 2 = 7.87 x 10 8 W/m 2. The question is what temperatures will be found for the materials. This can be answered by calculating the temperature distribution, which depends on the time and on the distance from inside the material to the contact surface between the two materials. The heat equation will be derived. Consider the heat conductivity in either the steel rod and the friction piston. In order to calculate the maximum temperature increase by the dissipation of the released energy, the heat transfer in the direction perpendicular to the friction area has been treated. The following scheme is used as depicted in Figure B.1. Φ(x) Φ( x + dx) x x+dx x Figure B.1. Heat transfer perpendicular to the friction area.

174 174 Appendix B The heat production occurs at the plane x = 0. The heat flux (power per unit area) through a plane at the distance x is given by U (1) Φ = K x where K is the thermal conductivity and U (x) the temperature rise above room temperature. Next the volume bounded by the planes at x and x + dx with cross-section of unit area is considered. The heat transferred into this volume becomes U Φ( x) Φ( x + dx) = K x U + K x 2 U + 2 x 2 U dx +... = K 2 x (2) If the mass density of the material is ρ and its specific heat c, then the heat capacity of the volume element (with cross-section of unit area) becomes ρ cdx. The temperature increase U t is then given by the partial differential equation U t 2 U = k 2 x (3) with k = K ρc. For calculating the temperate U ( x, t), a constant heat flux Φ 0 at x = 0 and U ( x,0) = 0 is considered. Primarily the maximum value of U is of interest, which occurs at x = 0 just at the end of the heat dissipation (after t = 12.5 ms). It will be found that the absorbed heat does not penetrate far into the material within 12.5 ms, so that is considered U ( x L, t) is still zero. The thickness of the material is xl. The following boundary conditions are used: Φ 0 = KU x ( 0, t) and lim U ( x, t) = 0 (4) x Equation 3 can be solved by Laplace transforms where u ( x, p) = L U e U ( x, t) dt and { ( x, t) } = Ψ ( ) = L { Φ = Φ p (6) p } ( ) pt p 0 (5)

175 Calculation of the Temperature Increase after Energy Absorption by Friction 175 Equation 3 transforms into pu( x, p) U ( x,0) = ku ( x, p) (7) xx Since U ( x,0) = 0 Equation 7 becomes pu( x, p) kuxx ( x, p) = 0 (8) with the solutions u x p k x p k ( x, p) = C1 ( p) e + C2e (9) The coefficients C1 and C2 are obtained by substituting the Laplace transforms of the boundary conditions as given by Equation 4 Ku x ( 0, p) = Ψ( p) and lim u( x, p) = 0 (10) Then is obtained x = 0 k Ψ( p) C 1 and C2 = (11) K p Substituting these values into Equation 9 gives u( x, p) k K Φ e p p = 0 x p k (12) Taking the inverse transform of Equation 12, the following temperature distribution is finally obtained Φ U ( x, t) = K 0 2 kt e π 2 x 4kt x x. erfc 2 kt (13) Since the maximum temperature, that occurs at x = 0 is of interest, therefore it will be continued with 2Φ0 kt 2Φ 0 U (0, t) = = K π Kρc t π (14)

176 176 Appendix B It is seen that the surface temperature is inversely proportional to the square root of the product of heat conductivity, specific heat and mass density. For this reason a friction material must be chosen with a high product value of this three properties. The heat is produced by friction along the contact surface between the friction piston and the steel rod. Since a very high surface temperature for the friction piston as well as locally on the steel rod can be expected, a temperature calculation will be made with the material tungsten as a thin plate mounted on the steel friction piston and on the two friction surfaces of the steel square rod. Tungsten has a high value for mass density and thermal conductivity. The melting point of steel is too low. In Table B.1 physical values of steel and tungsten are given. Table B.1. Material Density ρ 10 3 [Kgm -3 ] Thermal conductivity K [Wm -1 K -1 ] Specific heat c [JKg -1 K -1 ] Melting point C Steel FeP Tungsten The total heat flux of 7.87 x 10 8 W/m 2 is split up into two equal amounts, since the temperature at the contact surface of both tungsten plates must be initially the same. In Equation 15 the subscript t refers to tungsten. Φ t = 3.93 x 10 8 W/m 2 (15) The area of the friction piston ( 30 mm) has an equivalent square area of 26.6 x 26.6 mm. Between the five pistons in one row there is each time 20 mm space, in case of a fictive square piston area this space increases to 23.4 mm. This means that during the moving time of 12.5 ms the rod experiences the heat flow approximately during 26.6 mm / (26.6 mm mm) x 12.5 ms = 0.53 x 12.5 ms = 6.65 ms. Using Equation 14, the tungsten temperature rise on the rod can be calculated after the heating period of 6.65 ms, where subscript r in Equation 16 refers to the rod U r = = K (16) π 1 2 However, the tungsten plate on the piston will experience the heat continuously up till the end of 12.5 ms. This yields that the temperature rise will not be higher as (subscript p refers to the piston) U p = 2328 K.

177 Calculation of the Temperature Increase after Energy Absorption by Friction 177 The calculated temperature increases of tungsten lie far below the melting point of 3409 degrees, but are higher as the melting point of steel. Therefore the temperature decrease as function of the thickness of the tungsten plates must be calculated. Equation 13 can be written as U ( x, t) = U (0, t) e 2 x 4kt x x π.. erfc 2 kt 2 kt (17) In Figure B.2 Equation 17 is plotted as function of x 2 kt. In this way the temperature distribution within a material as function of time can be found. 1 plot of Equation U(x,t) / U(0,t) x 2 kt Figure B.2. Temperature distribution within a material as function of x 2 kt. With k = K ρc, for tungsten is found k t = =

178 178 Appendix B (18) Equation 17 is plotted in Figure B.3 for the piston with t =12. 5 ms and for the rod with t = 6.65 ms temperature as function of penetration depth in piston and rod 2000 temperature rise [K] tungsten thickness [mm] Figure B.3. Temperature rise as function of the penetration depth in tungsten. In Figure B.3 the temperature distribution within the tungsten as function of the distance in mm from the friction surface is plotted. Both curves start at the calculated temperature rises of respectively 2328 K for the piston surface and 1698 K for the rod surface. It is seen that for the piston for x = 0.5 mm the temperature rise drops to 1338 K, and for the rod the temperature rise drops to 775 K. Therefore tungsten plates of already 0.5 mm thickness will comply with the technical condition of shielding the steel behind the tungsten from melting. In case of the rod the temperature rise is already at a thickness of 0.1 mm below the 1500 K, may be a tungsten coating is possible. At a distance of 2.5 mm or more from the contact surface the temperature rise is negligible.

179 References 179 References Abramowicz, W., Wierzbicki, T., Axial Crushing of Multicorner Sheet Metal Columns, in: Journal of Applied Mechanics, No.1, Vol. 56, pp , Albertini, C., Cadoni, E., Labibes, K., Precision Measurements of Crashworthiness of Automotive Longitudinals in Aluminium and Steel, Proceedings of the 29 th ISATA Conference, Paper 96SAF046, pp , Florence, Italy, Arimoto, H., Yasuki, T., Kawamura, K., Kondou, M., A Study on Energy-Absorbing Mechanism of Plastic Ribs, Euro-PAM 98 Seminar, Tours, France, Auto Motor und Sport, spezial, Überleben in der Golf-Klasse, Baaten, W., Numerieke simulatie van het crashgedrag van langsliggerprofielen onder verschillende botsomstandigheden (Numerical Simulation of the crashworthiness of longitudinal members for different collision situations), in Dutch, Master s thesis, Internal report WOC/VT/R/94.43, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, Bakker, S.O., Optimising Deceleration Pulses - Lowering injury levels for different collision velocities, Master s thesis, Internal report WOC/VT/R/97.04, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, Barbat, S.D., Prasad, P.P., Numerical Simulations and Predictions of Spot-Weld/Weld- Bond Separation in Aluminum Closed-Hat Sections in Axial Collapse, AMD-Vol. 210/BED-Vol. 30, Crashworthiness and Occupant Protection in Transportation Systems, ASME, pp , Beermann, H.J., Staisch, A., Aufpralluntersuchungen mit vereinfachenden Strukturmodellen, IV. IfF-Tagung, Braunschweig, Behler, F.J., Höck, H.G., Poth, A., Schrepfer, W., Dynamisch ermittelte Werkstoffkennwerte und numerische Simulation des Crash-Verhaltens von Kraftfahrzeugen, VDI Berichte Nr. 852, pp , Belingardi, G., Avalle, M., Chiara, A., Crushing Behaviour of Aluminium Circular Tubes, 4th International Conference ATA-MAT, Torino, Italy, 1994.

180 180 References Bois, P. du, Chedmail, J.F., Automotive Crashworthiness Performance on a Supercomputer, SAE paper no , Bosch, Automotive Handbook 3 rd edition, pp. 649, Brantman, R., Achievable Optimum Crash Pulses for Compartment Sensing and Airbag Performance, Thirteenth International Technical Conference on Experimental Safety Vehicles (ESV), Paper S9-O-22, pp , Paris, France, Clark, C.C., The Crash-Anticipating Extended Airbag Bumper System, Proceedings of the Fourteenth International Technical Conference on the Enhanced Safety of Vehicles (ESV), Paper 94-S8-W-22, pp , Munich, Germany, Consumentengids (Consumer guide), in Dutch, pp , February Consumentengids (Consumer guide), in Dutch, pp , September Crutzen, Y.R., Inzaghi, A., Mogilevsky, M., Albertini, C., Computer Modelling of the Energy Absorption Process in Box-Type Structures Under Oblique Impact, Proceedings of the 29 th ISATA Conference, Paper 96SAF016, pp , Florence, Italy, Ducrocq, P., Markiewicz, E., Drazetic, P., Guyon, P., An Inverse Approach to Determine the Constitutive Model Parameters Using Drop Tests and Numerical Tools, Proceedings PAM 97 Seminar, Prague, Czech Republic, Eichhorn, F., Schmitz, B.H., Vergleichende Prüfung standardisierter punktgeschweisster Hohlprofile aus Stahlblech mit und ohne zusätzliche Verklebung der Fügespalte, Fachbeitrage aus Schweissen und Schneiden 36, Heft 3, ESI Group Software Product company, PAM-CRASH Version 1994, User s Training Manual, PAM System International, ESI Group Software Product company, Occupant Database, PAM-SAFE Version 1996, PAM System International, Faerber, E., Influence of Rigid Wall Impact Speed on Dummy and Vehicle Loadings, Thirteenth International Technical Conference on Experimental Safety Vehicles (ESV), Paper S9-O-01, pp , Paris, France, 1991.

181 References 181 Giess, M., Tomas, J., Improving Safety Performance in Frontal Collisions by Changing the Shape of Structural Components, Proceedings of the Sixteenth International Technical Conference on the Enhanced Safety of Vehicles (ESV), Paper 98-S1-O-07, pp , Windsor, Canada, Groth, H.L., Johansson, R.E., Lundh, H., Energy Absorption in Stainless Steel Structural Members for Automotive Applications, Proceedings of the 24 th ISATA Conference, Paper , pp , Florence, Italy, Hackney, J.R., Comparison of 0 Degree and Oblique Tests, Proceedings of the Tenth International Technical Conference on Experimental Safety Vehicles (ESV), pp. 436, Oxford, England, Her Majesty s Stationery Office, Road Accidents Great Britain 1990, Hobbs, C. A., The Need for Improved Structural Integrity in Frontal Car Impacts, Proceedings of the Thirteenth International Technical Conference on Experimental Safety Vehicles (ESV), Paper S9-O-12, pp , Paris, France, Hobbs, C. A., The Need for a Deformable Impact Surface for Frontal Impact Testing, Colloquium Comparative Crash Tests within the EC in Brussel on December 1992, pp , TÜV Rheinland, ISBN , Köln, Hoffman, R., Pickett, A.K., Ulrich, D., Haug, E., Lasry, D., Clink lie, J., A Finite Element Approach to Occupant Simulation: The PAM-CRASH Airbag Model, SAE paper no , Hoffmann, R., Ulrich, D., Protard, J.-B., Wester, H., Jaehn, N., Scharnhorst, T., Finite Element Analysis of Occupant Restraint System Interaction with PAM-CRASH, 34th Stapp Car Crash Conference, Orlando, Florida, Hop, R.M., Meyer, L.W., Staskewitsch, E., Bettenhausen, K., Verformungsvermögen von Tiefziehblechen bei quasistatischer und dynamischer Belastung, Bad Nauheim, Werkstoffprühffung, Janssen, R.A.M., Alternatieven voor energiedissipatie bij frontale botsingen (Alternatives for Energy absorption in Frontal Collisions), in Dutch, Master s thesis, Internal report WOC/VT/R/94.48, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, 1994.

182 182 References Jawad, S.A., Frontal Collision Mitigation Using Intelligent Extending Bumper, Proceedings of the Fifteenth International Technical Conference on the Enhanced Safety of Vehicles (ESV), Paper 96-S3-W-21, pp , Melbourne, Australia, Jones, N., Wierzbicki, T., Structural Crashworthiness, pp , , 359, Butterworths, ISBN , London, Justen, R., Scheunert, D., Herrmann, R., Schwede, W., and Zeidler, F., Die Bedeutung des Überdeckungs grades für die Realitätsnähe frontaler Offset-Crashs, ATZ 3, pp , Khalil, T.B., Sheh, M.Y., Chalons, P., Dubois, P.A., Integrated Vehicle-Dummy-Airbag Model for Frontal Crash Simulation by FE Analysis, AMD-Vol. 210/BED-Vol. 30, Crashworthiness and Occupant Protection in Transportation Systems, ASME, pp , Kim, C.W., Han, B.K., Won, C.J., Lim, C.H., A study on crushing behaviour and energy absorption capacity of a thin-walled rectangular tube under axial compression, IJCrash 1996, Vol1, No. 3, Woodhead Publishing Ltd, pp , Kormi, K., Webb, D.C., Oguibe, C.N., Al-Hassani, S.T.S., The Dynamic and Static Axial Crushing of Axially Stiffened Square Tubes a Comparison between Numerical and Experimental Results, AMD-Vol. 210/BED-Vol. 30, Crashworthiness and Occupant Protection in Transportation Systems, ASME, pp , Krauss, C.A., Laananen, D.H., A parametric study of crush initiators for a thin-walled tube, pp , Int. J. of Vehicle Design, Vol. 15, Nos 3/4/5, Landheer, D., Gee, A.W.J. de, Dynamische Contactverschijnselen (Dynamic Contact Phenomena), in Dutch, Lecture notes college 4B500, Eindhoven University of Technology, Eindhoven, The Netherlands, Landheer jr., D., Witteman, W.J., Kriens, R.F.C., Crashworthiness in the Concept Stage: A Parametric Body Design Method, Proceedings of the 26th FISITA Congress, Paper B16.54, 16p, Prague, Czech Republic, Landheer, D., Rapid Design of Crash Properties for Safe Automobiles: A Conceptual Approach, Ph.D. thesis, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, 1997.

183 References 183 Lasry, D., Hoffmann, R., Protard, J.-B., Numerical Simulation of Fully Folded Airbags and Their Interaction with Occupants with Pam-Safe, SAE Paper no , Leeuwen, O.L. van, Numerical Crashworthiness Design: Advanced Longitudinal Members in Frontal Vehicle Model, Master s thesis, Internal report WOC/VT/R/97.47, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, Levine, R.S., Head and Neck Injury, Chapter 5, USA, SAE P-276, ISBN , Magee, C.L., Thornton, P.H., Design Considerations in Energy Absorption by Structural Collapse, SAE Paper no , Mahmood, H.F., Paluszny, A., Design of Thin Walled Columns for Crash Energy Management - Their Strength and Mode of Collapse, SAE Paper no , Markiewicz, E., Ducrocq, P., Drazetic, P., Ravalard, Y., Calculation of the dynamic axial crushing response of complex prismatic sections, IJCrash 1996, Vol1, No. 2, Woodhead Publishing Ltd, pp , Mertz, H., Patrick, L., Strength and response of the human neck, pp , From: Biomechanics of impact injury and injury tolerances of the head-neck complex, USA, SAE PT-43, ISBN , Morlog, M.J.C., Simulatie en beproeving van een crashkoker (Simulation and testing of a crush column), in Dutch, Practical assignment, Internal report W/VT/R/98.57, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, Ni, X., Lasry, D., Haug, E., Hoffmann, R., Advances in Problem-Adaptive Occupant Modelling with PAM-SAFE, Proceedings of the Thirteenth International Technical Conference on Experimental Safety Vehicles (ESV), Paper 91-S9-O-20, Paris, France, Nishino, M., Yoneno, M., Kishimoto, Y., Murata, A., Axial Impact Testing of Weldbonded Box Sections, JSAE Review Vol.13, No.4, 1992.

184 184 References Oirschot, D. van, Numerieke modelvorming en optimalisatie van verbindingen tussen plaatdelen (Numerical Modeling and optimization of joints between sheet metal panels), in Dutch, Master s thesis, Internal report WOC/VT/R/95.46, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, Park, B., Morgan, R.M., Hackney, J.R., Lee, J., Stucki, S.L., Frontal Offset Crash Test Study Using 50 th Percentile Male and 5 th Percentile Female Dummies, Proceedings of the Sixteenth International Technical Conference on the Enhanced Safety of Vehicles (ESV), Paper 98-S1-O-01, pp , Windsor, Canada, PBNA, Koninklijke, Poly Technisch zakboekje (Technical Pocketbook), in Dutch, ISBN , Arnhem, The Netherlands, Poll, C.A.H. van der, Structural Concepts for Controlling the Folding Process and Improving the Energy Absorption of Vehicle Longitudinal Members, Master s thesis, Internal report WOC/VT/R/96.57, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, Radlmayr, K.M., Ponschab, H., Stiaszny, P., Till, E., Comparative Behaviour of Safety Structures from Soft and Higher-tensile Qualities as well as Aluminium Alloys in Crashes, Proceedings of the 26 th ISATA Conference, Paper 93SF061, pp , Aachen, Germany, Ragland, C., Overlap Car-to-Car Tests Compared to Car-to-Half Barrier and Car-to-Full Barrier Tests, Proceedings of the Thirteenth International Technical Conference on Experimental Safety Vehicles (ESV), Paper S9-W-31, pp , Paris, France, Relou, J.J.M.G., Integrated Crash Simulation of a Frontal Crash Structure and a Dummy, Master s thesis, Internal report WOC/VT/R/95.66, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, Richter, V., Holzner, M., Numerical Analysis of Folding Collapse of Hexagonal Profiles, PAM '93 Third European Workshop on Advanced Finite Element Simulation Techniques, pp , Schlangenbad, Germany, Santis, F. De, The modelling of the vehicle frontal structure and the numerical simulation of the crash-test, Master s thesis, Politecnico di Torino, Facoltá di Ingegneria, Corso di specializzazione in Ingegneria Meccanica, Italy, Torino, 1996.

185 References 185 Schlosser, J., Ullrich, P., Industrial Applications of PAM-SAFE in Occupant Simulation at Audi, PAM 95 Fifth European Workshop on Advanced Finite Element Simulation Techniques, Bad Soden, Germany, Seiffert, U., Scharnhorst T., Die Bedeutung von Berechnungen und Simulationen für den Automobilbau - Teil 1, Automobiltechnische Zeitschrift 91, pp , Seiffert, U., Fahrzeugsicherheit, VDI-Verlag, ISBN , Düsseldorf, Seiffert, U., Möglichkeiten und Grenzen der neuen Frontal- und Seitenaufprall- Gesetzgebung, ATZ 9, pp , Slaats, P.M., Numerical Modelling and Crashworthiness Design of Vehicle Longitudinal Members, Master s thesis, Internal report WOC/VT/R/96.04, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, 1996a. Slaats, P.M., Design of a Frontal Car Structure for Offset Collisions, Interim Assignment, Internal report WOC/VT/R/96.61, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, 1996b. Staalkabel B.V., Technisch Vademecum: Staalkabel-, Hijs- en Heftechniek. Handelsdivisie van Nationaal Grondbezit N.V., 3 e editie, Den Haag, Stucki, S.L., Hollowell, W.T., Fessahaie, O., Determination of Frontal Offset Test Conditions Based on Crash Data, Proceedings of the Sixteenth International Technical Conference on the Enhanced Safety of Vehicles (ESV), Paper 98-S1-O-02, pp , Windsor, Canada, SWOV, Schrift nr.75, in Dutch, source: AVV-BG/CBS, June Thornton, P.H., Mahmood, H.F., Magee, C.L., Energy absorption by structural collapse, In: Jones, N., Wierzbicki, T., Structural Crashworthiness, Butterworths, London, TÜV Rheinland, Comparative Crash Tests within the EC, TÜV-Kolloquium in Brussel on December 1992, ISBN , Köln,1993.

186 186 References Viano, D.C., Arepally, S., Assessing the Safety Performance of Occupant Restraint Systems, SAE paper no , Wang, X.G., Bloch, J.A., Cesari, D., Sidoroff, F., On the Axisymmetric Crushing of Circular Tubes, Proceedings of the Eighth International Conference on Vehicle Structural Mechanics and CAE, Paper , pp , Wheeler, M.J., Crashworthiness of Aluminum Structured Vehicles, Proceedings of the Sixteenth International Technical Conference on the Enhanced Safety of Vehicles (ESV), Paper 98-S1-W-20, pp , Windsor, Canada, Wierzbicki, T., Abramowicz, W., On the Crushing Mechanics of Thin-Walled Structures, in: Journal of Applied Mechanics, Vol. 50, pp , Wierzbicki, T., Abramowicz, W., The Mechanics of Deep Plastic Collapse of Thin-Walled Structures, in: Structural Failure, John Wiley, Ed., T. Wierzbicki and N. Jones, Wierzbicki, T., Plastic Folding Wave and Effective Crush Distance, Center for Transportation Studies, M.I.T. Cambridge, MA, 02139, Wijntuin, A.L.M., Evaluation of vehicle interior dimensions and numerical modeling of interior foam padding, Practical assignment, Internal report WOC/VT/R/95.56, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, Witteman, W.J., Insufficiency of a Single Frontal Impact test for Vehicle Crashworthiness Assessment, Proceedings of the 26 th ISATA Conference, Paper 93SF069, pp , Aachen, Germany, Witteman, W.J., Kriens, R.F.C., Requirements for Optimized Crashworthiness Design of the Longitudinal Members, Proceedings of the Fourteenth International Technical Conference on the Enhanced Safety of Vehicles (ESV), Paper 94-S8-W-24, pp , Munich, Germany, Witteman, W.J., Kriens, R.F.C., Crashworthiness Design of Longitudinal Members for Real Crash Situations, Proceedings of the 5 th International EAEC Congress, Conference B Chassis-Body Engineering, Paper SIA 9506B13, 10p, Strasbourg, France, 1995.

187 References 187 Witteman, W.J., Kriens, R.F.C., A Cable-supported Frontal Car Structure for Offset Crash Situations, Proceedings of the Fifteenth International Technical Conference on the Enhanced Safety of Vehicles (ESV), Paper 96-S3-W-24, pp , Melbourne, Australia, Witteman, W.J., Kriens, R.F.C., Modeling of an Innovative Frontal Car Structure: Similar Deceleration Curves at Full Overlap, 40 per cent Offset and 30 Degrees Collisions, Proceedings of the Sixteenth International Technical Conference on the Enhanced Safety of Vehicles (ESV), Paper 98-S1-O-04, pp , Windsor, Canada, Witteman, W.J., Optimizing the Deceleration Pulse, Internal report, W/VT/R/99.09, Eindhoven University of Technology, Laboratory for Automotive Engineering, Eindhoven, The Netherlands, Witteman, W.J., Kriens, R.F.C., Numerical Optimization of Crash Pulses, will be published in Proceedings PAM 99, Ninth European Seminar on Advanced Finite Element Simulation Techniques, Darmstadt, Germany, Yamaguchi, S., Kato, H., Okazaki, T., Efficient Energy Absorption of Automobile Side Rails, Proceedings of the Tenth International Technical Conference on Experimental Safety Vehicles (ESV), pp , Oxford, England, Yuan, Y., Viegelahn, G.L., A New Model for Axisymmetric Collapse of Circular Tubes, Proceedings of the Eighth International Conference on Vehicle Structural Mechanics and CAE, Paper , pp , 1992.

188 188 References

189 Acknowledgements 189 Acknowledgements I want to thank all the persons who have contributed to the work presented in this thesis. Especially the graduate students who did a lot of numerical simulations during their master degree: Wim Baaten, Paul Slaats, Stijn van der Poll, Jack Relou, Francesco De Santis, Olaf van Leeuwen and Sven Bakker. Together we have built a lot of knowledge about crashworthiness simulations. For the practical experiments, I want to thank especially trainee Marcel Morlog and colleague Jan de Vries. My promoters prof.dr.ir. Ruud Kriens and prof.dr.ing. Giovanni Belingardi (Politecnico di Torino), and the commission members prof.dr.ir. Jac Wismans and prof.dr.ir. Piet Schellekens are acknowledged for their critical comments on my thesis, and especially Ruud Kriens who has shown confidence in me by letting me develop this research area. Furthermore, I want to thank Hewlett Packard and the software suppliers for their support and to make it possible to use the professional computer programs like PAM- CRASH and PAM-SAFE (ESI/PSI), UNIGRAPHICS (UG Solutions) and CRASH CAD (Impact Design). This work was partly sponsored by the National Computing Facilities Foundation (NCF) for the use of supercomputer facilities, with financial support from the Netherlands Organization for Scientific Research (NWO).

190 190

191 Curriculum vitae 191 Curriculum vitae Willem Witteman was born on November 28, 1961, in the Netherlands. After his graduation in 1980 from Lyceum de Grundel at Hengelo (Ov) he studied Mechanical Engineering at the University of Twente. His master s thesis within the group Design and Construction was on CAD/CAM techniques. It was carried out at Philips Lighting in Eindhoven. After his graduation, he joined Philips Machine Factories in 1987 as a CAD/CAM specialist. In 1991, he was appointed as scientific staff member in the Laboratory for Automotive Engineering of Eindhoven University of Technology. His specialism is vehicle structures and advanced materials, with emphasis on crashworthiness design. In this position he guides students on crashworthiness topics at the university and within the automotive industries. He was also involved as coach in a second Ph.D. research based on the subject of early design for side impact crashworthiness. He made several contributions to international vehicle safety conferences, which form the basis for this Ph.D. thesis on frontal vehicle design for improved crashworthiness.