5 - Crash Analysis and Dynamical Behaviour of Light Road and Rail Vehicles PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 28

Vehicle System Dynamics

Vol. 43, No. 6–7, June–July 2005, 385–411

Crash analysis and dynamical behaviour of light


road and rail vehicles
JORGE AMBRÓSIO*
Instituto de Engenharia Mecânica, Instituto Superior Técnico, Av. Rovisco Pais,
1049-001 Lisboa, Portugal

The main goal of crashworthiness is to ensure that vehicles are safer for occupants, cargo and other
road or rail users. The crash analysis of vehicles involves structural impact and occupant biomechan-
ics. The traditional approaches to crashworthiness not only do not take into account the full vehicle
dynamics, but also uncouple the structural impact and the occupant biomechanics in the crash study.
The most common strategy is to obtain an acceleration pulse from a vehicle structural impact analysis
or experimental test, very often without taking into account the effect of suspensions in its dynamics,
and afterwards feed this pulse into a rigid occupant compartment that contains models of passengers.
Multibody dynamics is the most common methodology to build and analyse vehicle models for occu-
pant biomechanics, vehicle dynamics and, with ever increasing popularity, structural crash analysis.
In this work, the aspects of multibody modelling relevant to road and rail vehicles and to occupant
biomechanical modelling are revised. Afterwards, it is shown how multibody models of vehicles and
occupants are used in crash analysis. The more traditional aspects of vehicle dynamics are then intro-
duced in the vehicle models in order to appraise their importance in the treatment of certain types of
impact scenarios for which the crash outcome is sensitive to the relative orientation and alignment
between vehicles. Through applications to the crashworthiness of road and of rail vehicles, selected
problems are discussed and the need for coupled models of vehicle structures, suspension subsystems
and occupants is emphasized.

Keywords: Energy absorption; Occupant biomechanics; Vehicle dynamics; Multibody dynamics;


Impact

AMS Subject Classification: 70E55; 92C10; 74M20

1. Introduction

The complete design of road and rail vehicles includes active and passive safety and the comfort
of occupants. The active safety of vehicles includes the manoeuvrability and stability and the
study of all systems that address their improvement such as suspensions, active control and
electromechanical subsystems for driving support, among others. Passive safety of vehicles
addresses the protection of occupants and cargo from the moment that an accident starts
until the vehicle stops. Vehicle system dynamics addresses mostly the active safety aspects
of the vehicle design, whereas vehicle crashworthiness deals with the structural impact, the
occupant biomechanics and all subsystems aimed to provide a better protection during a crash.

*Corresponding author. Email: jorge@dem.ist.utl.pt

Vehicle System Dynamics


ISSN 0042-3114 print/ISSN 1744-5159 online © 2005 Taylor & Francis Group Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/00423110500151788
386 J. Ambrósio

Though the most common numerical tools used to address both areas of vehicle dynamics are
based on multibody methodologies, traditionally these two aspects of vehicle design have
been addressed separately or with little interaction. It is an objective of this work to show
that not only the two aspects of vehicle dynamics can be addressed simultaneously, but also
that in many studies that address the passive safety of vehicles it is fundamental to include
detailed models of the vehicle active safety subsystems. Moreover, the common trend in actual
vehicles is to have active driving support and intelligent safety systems, both requiring the use
of sensors to provide the data required for the decision making processes. Also from this point
of view, a common approach to active and passive safety of road and rail vehicles is of major
interest.
Until the early 1970s, crash studies relied almost exclusively on experimental testing which
focused mostly in full scale testing and in the development of relevant testing scenarios [1]. In
a review work, Tidbury [2] stressed the facts that not only the costs associated with experimen-
tal testing are very high, but also the results cannot be generalized to other impact situations
neither can such methods be used during the design of new vehicles. Therefore, the need
for accurate numerical procedures became apparent at this time. The earlier numerical meth-
ods used for vehicle crashworthiness were based on the use of lumped masses and nonlinear
springs. The models built with these methods, known as lumped parameter models, use lumped
masses to represent parts of the vehicle, such as the engine block or the passenger compart-
ment, considered rigid during the analysis, and springs to represent the structural elements
responsible for the deformation energy management [3]. The lumped parameter models are
mostly one- or two-dimensional and they include at the most very simplified representations
of the suspension elements of the vehicles.
The first numerical model for the simulation of the vehicle occupant biomechanics proposed
by McHenry in 1963 [4] was a two-dimensional model that included four articulated rigid
bodies and the occupant restraint system. Later, Lobdell [5] proposed a lumped parameter
approach with similar numerical characteristics to that used in vehicle structural impact for the
simulation of human thorax loading. The developments of lumped parameter biomechanical
models continued through the years of 1970 and 1980 in parallel with the development of
models for the structural impact of vehicles but without the synergies associated to the use of
a common numerical framework.
The generalization of the vehicle structural and occupant biomechanical models to become
fully three-dimensional required another type of numerical methodologies. The multibody
dynamics methods provided the framework in which more advanced developments could be
undertaken. The plastic hinge approach, proposed by Nikravesh et al. [6] to represent the
structural dynamics of road vehicles in crash situations, uses a general multibody represen-
tation where the full vehicle is described and the structural systems are divided into several
rigid bodies connected by kinematic joints to which nonlinear springs, representing the struc-
tural deformation, are associated. This approach has been further developed by several authors
[7, 8] and applied to the crashworthiness of road and rail vehicles proving to be a versatile
and efficient methodology to deal with structural crashworthiness, especially in the initial
phases of the vehicle design, when extensive re-analysis is required. Still with little or no con-
tact to the developments in structural crashworthiness, the multibody methodologies began
to be extensively applied in the development of biomechanical models of vehicle occupants.
Starting in the initial work of McHenry [4] and going through the developments of Robbins
et al. [9] and Fleck et al. [10], the use of multibody dynamics methodologies for occupant
biomechanics reached a level of maturity that allowed it to become the most popular numerical
approach used today in impact biomechanics. The acceptance of the multibody-based models
by the vehicle industry led to the successful development of commercial occupant simulation
computer codes, such as SOMLA [11] and MADYMO [12].
Crash analysis and dynamical behaviour of vehicles 387

The continuous increase in computer power and the development of the finite element
method ensured the feasibility of finite element-based models for the study of the structural
crashworthiness of vehicles. One of the first successful attempts to apply nonlinear finite
elements to the prediction of a vehicle crush was reported by Thompson [13]. Belytchko and
Hsieh [14] proposed the finite element developments that defined the framework of their use in
structural crashworthiness. During the 1970s, several nonlinear finite element codes suitable
to crashworthiness applications were released. Among these, the codes DYNA3D by Hallquist
[15] and WHAMS by Belytschko and Kenedy [16] are of special importance as they served
as the backbone of the codes LS-DYNA [17] and of PAM-CRASH [18] and RADIOSS [19],
respectively, which are the most common finite element codes used in crash analysis. The
application possibilities of this method are being expanded in order to allow for the complex
modelling of vehicle and integrated occupants, but still, the representation of the suspension
systems of the vehicles and of systems responsible for their active safety are not yet present
in the finite element models, unless at the cost of a large computational effort. The amount of
information required for large systems is still too high making the application of FEM codes
suitable only for the analysis phase of structural and biomechanical systems. For the early
design phases, the multibody models are still the primary numerical tool used in simulation.
Though the features of the different numerical methodologies based on multibody dynamics
and on finite elements include most of the ingredients required for the crash analysis of
road and rail vehicles, the models more commonly developed neither integrate the structural
crashworthiness with the occupant biomechanics nor take into account the overall dynamics
of the vehicle. Applications that include low and medium vehicle impact speeds or multiple
impacts or long contact periods or even for which the alignment between vehicles is important
for the outcome of the crash are examples of cases for which the vehicle dynamics associated to
the suspensions and to the wheel road–rail contact can play an important role. Traditionally, the
most common numerical techniques for vehicle dynamics modelling are based on multibody
dynamics methodologies, which are the same used for vehicle impact models and for occupant
biomechanics modelling. Therefore, no reason exists for not taking the advantage of their
features to develop more accurate and efficient models of road and rail vehicles.
In this work, it is shown that procedures based on multibody dynamics provide a unified
framework for the development of design and analysis tools of crashworthy structures and of
occupants. The models developed are computationally efficient in today’s computers to allow
for the extensive re-analysis cycles required in the optimal design of vehicles for crashwor-
thiness. Furthermore, in applications for which the structural patterns of deformation are not
known beforehand or for which subcomponent finite element models are available, the multi-
body dynamics methods enable the effective coupling between the lumped parameter models
and the finite element description of the structural deformations. The application of the meth-
ods presented is addressed here through the study of the development of energy absorption and
anti-climber devices for railway vehicles, the development of occupant protection strategies in
side impact of a road vehicle and the rollover of an off-road vehicle with occupants included.

2. Multibody dynamics

There are different coordinates and formalisms that lead to suitable descriptions of multibody
systems, each of them presenting relative advantages and drawbacks. The objective of this
work is not to discuss what is the most efficient multibody methodology that can be applied
for vehicle dynamics crash analysis. In this work, the methods presented are based on the use
of Cartesian coordinates, which lead to a set of differential-algebraic equations that need to
388 J. Ambrósio

be solved. It is assumed that appropriate numerical procedures are used to integrate the type
of equations of motion obtained with the use of Cartesian coordinates. It is also assumed that
the different numerical issues that arise from the use of this type of coordinates, such as the
existence of redundant constraints and the possibility of achieving singular positions, are also
solved. For a more detailed discussion on the numeric aspects of this type of coordinates, the
interested reader is referred to [20–22].

2.1 Multibody equations of motion

A typical multibody model is defined as a collection of rigid or flexible bodies that have their
relative motion constrained by kinematic joints and that are acted upon by external forces. A
generic description of such a model is represented in figure 1.
Let the multibody system be made of nb bodies. The equations of motion for the system of
unconstrained bodies are
Mq̈ = g (1)
where M is the mass matrix, which includes the masses and inertia tensors of the individual
bodies, q̈ the acceleration vector and g the vector with applied forces and gyroscopic terms.
The relative motions between the bodies of the system are constrained by kinematic joints,
which are mathematically described by a set of nc algebraic equations, written as
(q, t) = 0. (2)
The first and second time derivatives of equation (2) constitute the velocity and acceleration
constraint equations, respectively, written as
˙
(q, t) ≡ Dq̇ = v
(3)
¨
(q, q̇, t) ≡ Dq̈ = γ
where D is the Jacobian matrix. For a system of constrained bodies, the effect of the kinematic
joints can be included in equation (1) by adding to their right-hand side the equivalent joint
reaction forces g(c) = −DT λ leading to
Mq̈ = g − DT λ (4)
where λ is a vector with nc unknown Lagrange multipliers. Equation (4) has nb + nc unknowns
that must be solved together with the second time derivative of the constraint equations. The

Figure 1. Generic multibody system.


Crash analysis and dynamical behaviour of vehicles 389

Figure 2. Flowchart representing the forward dynamic analysis of a multibody system.

resulting system of differential-algebraic equations is


    
M DT q̈ g
= (5)
D 0 λ γ

Note that the solution of equation (5) presents numerical difficulties resulting from the
need to ensure that the kinematic constraints are not violated during the integration
process.

2.2 Solution of the equations of motion

The forward dynamic analysis of a multibody system requires that the initial conditions of the
system, i.e. the position vector q0 and the velocity vector q̇0 , are given. With this informa-
tion, equation (5) is assembled and solved for the unknown accelerations, which are in turn
integrated in time together with the velocities. The process, schematically shown in figure 2,
proceeds until the system response is obtained for the analysis period.

3. Vehicle structural impact

The analysis of road or rail vehicles in crash events requires that the deformation of the
structural components during impact is properly represented and that the contact forces
are described. A proper description of the structural deformation means that not only the
mechanisms of deformation and the relative displacements between the different structural
components are captured with accuracy, but also that the deformation energy distribution is
well represented. Methodologies based on lumped deformations, such as the finite segment
and the plastic hinge approaches, or on a continuous description of the deformation, such
as the finite element method, are presented here to describe the structural deformation of
the vehicle. The contact force models that are used in vehicle impact should be consistent
with the vehicle model used. Continuous contact force models based on penalty formula-
tions are generally preferred for rigid body impact, being a representative force model of
this type described here. The tyre contact models used for road vehicles and the wheel–rail
contact models of railway vehicles are also used in some problems of vehicle impact to
describe the ground interaction forces. However, their description falls out of the scope of this
work.
390 J. Ambrósio

Figure 3. Slender component and its finite segment model.

3.1 Finite segment approach to flexible multibody systems

Let the flexible components of the multibody system be made of slender components such as
the connecting rods of high-speed machinery or the structural frame of buses and trucks. In
this case, each slender component can be modelled as a collection of rigid bodies connected
by linear springs, as shown in figure 3. These springs, representing the axial, bending and
torsion properties of the beams, capture the flexibility of the whole component.
Twelve generalized displacements are associated to each finite segment, that is, three trans-
lations and three rotations at each end. When the beams deform, the reference frames attached
to the rigid bodies used for their model rotate and translate with respect to each other. Forces
and moments applied to the rigid bodies can be calculated from the relative end displacements
and rotations assuming that each two adjacent bodies are connected by springs and, eventually,
by dampers. For a rigid body, there are deformation elements attached to each end, as shown
in figure 4. The characteristics of these springs are related with the material and geometric
properties of the system components [23].
Using the principles of structural analysis, the stiffness coefficients for these springs are
calculated. For instance, the straight extensional segment and the straight bending, represented
in figures 4(a) and (b), have their stiffness coefficients, respectively, given by

Ei A i
kie = kje = 2 (6a)
li
E i Ii
kie = kje = 2 (6b)
li

where Ei is the Young modulus, Ai the cross-section area, Ii the cross-section moment of
inertia and li the length of the finite segment.
Note that the finite segment methodology, as described in the original work by Huston
and Wang [23], only applies to structural components made of beams that have linear elastic
deformations. However, this formulation can be extended to cases with nonlinear deformations,
which is relevant to vehicles models undergoing impact scenarios.

3.2 Plastic hinges in multibody nonlinear deformations

In many impact situations, the individual structural members are overloaded, principally in
bending, giving rise to plastic deformations in highly localized regions, called plastic hinges.
These deformations, presented in figure 5, develop where maximum bending moments occur, at
load application points, joints or in locally weak areas. Therefore, for most practical situations,

Figure 4. Finite segments and their combinations: (a) extensional straight; (b) bending straight; (c) tapered bending.
Crash analysis and dynamical behaviour of vehicles 391

Figure 5. Localized deformations on a beam and a plastic hinge.

their location is predicted well in advance. The methodology described herein is known as
conceptual modelling in some industrial areas.
The plastic hinge concept has been developed by using generalized spring elements to rep-
resent the constitutive characteristics of localized plastic deformation of beams and kinematic
joints in order to control the deformation kinematics, as illustrated in figure 6. A more complex
structure, such as the B-pillar of a road vehicle or the end-underframe of a train car shown
in figure 7, can be represented by a collection of rigid bodies connected by the joint-spring
arrangements that describe its deformation and energy absorption characteristics.
For a flexural plastic hinge, the spring stiffness is expressed as a function of the change
of the relative angle between two adjacent bodies connected by the plastic hinge, as shown
in figure 8. For a bending plastic hinge, the revolute joint axis must be perpendicular to the
neutral axis of the beam and to the plastic hinge bending plane simultaneously. The relative
angle between the adjacent bodies measured in the bending plane is

θij = θi − θj − θij0 (7)

where θij0 is the initial relative angle between the adjacent bodies. Note that for the case of
flexible adjacent bodies, the relative angular values also include information on the nodal
rotational displacements.
The characteristics of the spring-damper that describes the properties of the plastic hinges
are obtained by experimental component testing, finite element nonlinear analysis or simplified
analytical methods. For instance, the typical torque–angle constitutive relation, as in figure 8,
is found based on a kinematic folding model for the case of a steel tubular cross section. This

Figure 6. Plastic hinge models for different loads: (a) one-axis bending; (b) two-axis bending; (c) torsion; (d) axial.
392 J. Ambrósio

Figure 7. Plastic hinge models for vehicle substructures: (a) B-pillar of a road vehicle; (b) end-underframe of a rail
vehicle; (c) door of a car.

Figure 8. Plastic hinge bending moment and its constitutive relation.

model is modified accounting for elastic–plastic material properties including strain hardening
and strain rate sensitivity of some materials. A dynamic correction factor is used to account
for the strain rate sensitivity [24]

Pd
= 1 + 0.07 V00.82 . (8)
Ps

Here Pd and Ps are the dynamic and static forces, respectively, and V0 is the relative velocity
between the adjacent bodies. The coefficients appearing in equation (8) are dependent on the
type of cross section and material [24].

3.3 Nonlinear finite elements for multibody systems

The description of the structural deformations during vehicle impact using plastic hinges
requires that the mechanisms of deformation are known beforehand, which is very often the
case in most of the practical applications. However, in cases for which the structural deforma-
tions are generalized, rather than concentrated in specific points, the modelling capabilities
of the finite element method are irreplaceable. The nonlinear finite element formulation is
Crash analysis and dynamical behaviour of vehicles 393

Figure 9. General motion of a flexible body.

summarized here in the framework of flexible multibody dynamics to be compatible with the
standard description of the multibody formulation presented before.
The motion of a flexible body, depicted in figure 9, is characterized by a continuous change
of its shape and by large displacements and rotations, associated to the gross rigid body motion.
Let XYZ denote the inertial reference frame and ξ ηζ a body fixed coordinate frame. Let the
principle of the virtual work be used to express the equilibrium of the flexible body in the
current configuration t + t and an updated Lagrangian formulation be used to obtain the
equations of motion of the flexible body [25]. The finite elements used in the discretization of
the flexible body are assembled, leading to the equations of motion of the flexible body
           
Mrr Mrf Mrf r̈ gr sr 0 0 0 0 0
 
Mφr Mφφ Mφf  ω̇  = gφ  − sφ  − 0 − 0 0   0  (9)
0
Mf r Mf φ Mff ü gf sf f 0 0 KL + KNL u

where r̈ are ω̇ are, respectively, the translational and angular accelerations of the body fixed
reference frame and ü denotes the nodal accelerations measured in body fixed coordinates.
The local coordinate frame ξ ηζ , attached to the flexible body, is used to represent the gross
motion of the body and its deformation. The right-hand side of equation (9) contains the vector
generalized forces applied on the deformable body g and matrices KL and KNL , which are
the linear and nonlinear stiffness matrices, respectively. Vector f denotes the equivalent nodal
forces due to the state of stress.
In order to improve the numerical efficiency of the solution of equation (9), a lumped mass
formulation is used and the nodal accelerations ü , measured with respect to the body fixed
frame, are substituted by the nodal accelerations q̈f relative to the inertial frame. Furthermore,
it is assumed that the flexible body has a rigid part and a flexible part and that the body fixed
coordinate frame is attached to the center of mass of the rigid part, as shown in figure 10. The
flexible and rigid parts are attached by the boundary nodes ψ. The procedure is described in
Ambrósio and Nikravesh [26] leading to the new form of the equations of motion
    
fr + ACδ
T
mI + AM∗ A −AM∗ S 0 r̈
  
0   ω̇  = n − ω̃ J ω − ST Cδ − I Cθ  (10)
  T
 − AM∗ S T J + ST M∗ S

0 0 Mff q̈f gf − f − (KL + KNL )u

where J is the inertia tensor, expressed in body fixed coordinates, fr the vector of the external
forces applied to the body and n the vector with the force transport and external moments.
Vector u denotes the nodal displacement increments from a previous configuration to the
394 J. Ambrósio

Figure 10. Flexible body with a rigid part.

current configuration, measured in body fixed coordinates. In equation (10), M∗ is a diagonal


mass matrix containing the mass of the boundary nodes and

T
A = [A A · · · A]T ; S = −[x̃1 x̃2 · · · x̃n ]T ; I = [I I · · · I]T

where A is the transformation matrix from the body fixed to global coordinates and xk denotes
the position of node k. Vectors Cδ and Cθ are the reaction forces and moment of the flexible
part of the body over the rigid part is given by

Cδ = gδ − Fδ − (KL + KNL )δδ δ  − (KL + KNL )δθ θ  (11)


Cδ = gθ − Fθ − (KL + KNL )θ δ δ  − (KL + KNL )θ θ θ  . (12)

In these equations, the subscripts δ  and θ  refer to the partition of the vectors and matrices with
respect to the translational and rotational nodal degrees of freedom. The underlined subscripts
refer to nodal displacements of the nodes fixed to the rigid part.
The equations of the flexible bodies are included in the equations for the constrained multi-
body system using the Lagrange multiplier method when a kinematic constraint involving
the nodal coordinates has to be set. The mechanical joints of the vehicles, modelled using
this formulation, are described by such kinematic constraints. For more information on the
definition of the kinematic constraints, the interested reader is referred to Ambrósio [27].

3.4 Contact detection

The numerical procedures for contact detection in crash applications are similar to method-
ologies used to detect contact in other type of applications in vehicle dynamics. Let a body of
the system get close to a surface during the motion of the multibody system, as represented
in figure 11. Without lack of generality, let the impacting surface be described by a mesh of
triangle patches. In particular, let the triangular patch, where node k of the body will impact,
be defined by points i, j and l. Note that node in this context means either a point of a rigid
body or a nodal point of the finite element mesh of a flexible body. The normal to the outside
surface of the contact patch is defined as n = rij × rj l /rij × rj l .
Let the position of the structural node k with respect to point i of the surface be

rik = rk − ri . (13)
Crash analysis and dynamical behaviour of vehicles 395

Figure 11. Detection of contact between a multibody component and a triangular patch.

This vector is decomposed in its tangential component, which locates point k ∗ in the patch
surface, and a normal component, given, respectively, by
t
rik = rik − (rik
T
n)n (14)
n
rik = T
(rik n)n. (15)

A necessary condition for contact is that node k penetrates the surface of the patch, i.e.
T
rik n ≤ 0. (16)

In order to ensure that a node does not penetrate the surface through its ‘interior’ face, a
thickness e must be associated to the patch. The thickness penetration condition is

−rik
T
n ≤ e. (17)

The condition described by equation (17) prevents that penetration is detected when the flexible
body is far away, behind the contact surface. The remaining necessary conditions for contact
requires that the node is inside of the triangular patch. These three extra conditions are
t
(r̃ik rij )T n ≤ 0; (r̃jt k rj l )T n ≤ 0 and t
(r̃lk rli )T n ≤ 0. (18)

Equations (13)–(18) are necessary conditions for contact. However, depending on the con-
tact force model actually used, they may not be sufficient to ensure effective contact. Note
that when point k ∗ is on the boundary of the triangular patch the equality in one or more of
relations (18) hold true being the node k considered to be inside such patch.

3.5 Continuous contact force model

A model for the contact force must consider the material and geometric properties of the
surfaces, contribute to a stable integration and account for some level of energy dissipation.
396 J. Ambrósio

On the basis of the Hertzian description of the contact forces between two solids, Lankarani
et al. [28] propose a continuous force contact model that accounts for energy dissipation during
impact. The procedure is used for both rigid body and nodal contact.
Let the contact force between two bodies or a system component and an external object be
a function of the pseudo-penetration δ and pseudo-velocity of penetration δ̇ given by

fs,i = (Kδ n + D δ̇) u (19)

where K is the equivalent stiffness, D a damping coefficient and u, in this context, a unit
vector normal to the impacting surfaces. The hysteresis dissipation is introduced in equation
(19) by D δ̇, being the damping coefficient written as
3K(1 − e2 ) n
D= δ . (20)
4δ̇ (−)
This coefficient is a function of the impact velocity δ̇ (−) , stiffness of the contacting surfaces
and restitution coefficient e. For a fully elastic contact e = 1, whereas for a fully plastic contact
e = 0. The generalized stiffness coefficient K depends on the geometry material properties of
the surfaces in contact. For the contact between a sphere and a flat surface, the stiffness is [28]
−1
√ 1 − νi2 1 − νj2
K = 0.424 r + (21)
πEi π Ej

where νl and El are the Poisson’s ratio and the Young’s modulus associated to each surface,
respectively, and r is the radius of the impacting sphere.
The nonlinear contact force is obtained substituting equation (20) into equation (19)
 
3(1 − e2 ) δ̇
fs,i = K δ 1 +
n
u (22)
4 δ̇ (−)
This equation is valid for impact conditions, in which the contacting
√ velocities are much lower
than the propagation speed of elastic waves, i.e. δ̇ − ≤ 10−5 E/ρ.
The contact forces between the node and the surface include friction forces modelled using
the Coulomb friction model. The dynamic friction forces in the presence of sliding are
n 
|fs,i |
f friction = −µd fd q̇k (23)
|q̇k |
where µd is the dynamic friction coefficient and q̇k is the velocity of point k. The dynamic
correction coefficient fd is expressed as


 0 if |q̇k | ≤ v0
 (|q̇ | − v )
k 0
fd = if v0 ≤ |q̇k | ≤ v1 (24)

 (v − v0 )
 1
1 if |q̇k | ≥ v1 .

The dynamic correction factor prevents the friction force from changing direction for almost
null values of the nodal tangential velocity, which would be perceived by the integration
algorithm as a response with high frequency contents, forcing it to dramatically reduce the time
step size. The friction model represented by equation (24) does not account for the adherence
between the node and the contact surface. The interested reader is referred to the work of Wu
et al. [29] for a comprehensive discussion on the topics of friction and sliding in multibody
dynamics.
Crash analysis and dynamical behaviour of vehicles 397

4. Applications to crash analysis of road and rail vehicles

The applications used in this work to demonstrate the procedures proposed address both
structural and biomechanical crashworthiness of road and rail vehicles in crash scenarios
for which the general dynamics of the vehicles, associated to the suspension systems and to
the wheel–rail or road interaction, play an important role on the system response. The first
application, to the design of the interface between rail vehicles of the same train, shows the
importance of having an accurate model for the vehicle pitching. In the second application, to
the side impact of a road vehicle, it is the roll and side displacement of the car that influences
the outcome of the crash. In the third and final application, the study of the rollover of an all-
terrain vehicle, not only the complex interaction between the vehicle suspension systems and
the ground is emphasized but also the occupants’ behaviour during ejection is characterized.

4.1 Railway crashworthiness

The design of railway vehicles for crashworthiness requires that not only their ends are able
to deform in a controlled manner, therefore absorbing energy during a crash, but also that all
the devices that connect the different cars in the same train also deform and absorb energy by
plastic deformations. However, in order for the structural components and connection devices
to work properly, it is necessary that the vehicles remain aligned during the crash. The anti-
climbers are the devices that, being located at the ends of each car, ensure that such alignment is
maintained. In their design, it is necessary to know the shear forces that they have to withstand
during the train crash. The methodology described in this work is demonstrated in the design
of these rail vehicle components.
A typical arrangement of a train set with eight individual car-bodies is presented in table 1.
The length and the mass of each individual car are also shown in table 1. The model of each
individual car, shown in figure 12, is composed of five rigid bodies, B1 to B5 , which represent
the passenger compartment, bogie chassis and deformable end extremities. The relative motion
between the multibody components is restricted by two revolute joints, R1 and R2 , and by two
translation joints, T1 and T2 . The vehicles are assumed to be stiff and, therefore, no bending
flexibility is included in the models. The inertia and mechanical properties of the system
components are described in Milho et al. [30].
The first simulation scenarios are characterized by a moving train, travelling from right to
left, which collides with another train parked with brakes applied. The trains are guided along
the same rails, the collision velocities being 30, 40 and 55 km/h for each simulation. Of special
importance to the anti-climber design are the simulation results for the contact forces and the
relative displacements between car-body extremities [30].
The vertical relative displacement between the points of the contact surfaces defining the
anti-climber devices is described by the distance g measured along the contact surface, between
points A and B, which are initially leveled, as shown in figure 13. This displacement is
calculated when contact between the end extremities of the car-bodies occurs. The vertical
relative displacement obtained in the interfaces between car-bodies is illustrated in figure 14.

Table 1. Train set configuration.

Length (m) 20 26 26 26 26 26 26 26 26 26
Mass (103 kg) 68 51 34 34 34 34 34 34 34 51
398 J. Ambrósio

Figure 12. Car-body model for a single car.

Figure 13. Anti-climber device contact geometry.

The vertical relative displacement, which results from the pitch motion of adjacent cars, tends
to increase for the interfaces away from the high-energy interface, reaching maximum levels in
the colliding train. High-energy zones (HE) mean the extremities of the train set in the frontal
zone of the motor car-body and in the opposing back zone of the last car-body. The HE are
potential impact extremities between two train sets. The low-energy zones (LE) are located in
the remaining extremities of the train car-bodies and correspond to regions of contact between
cars of the same train set.
The tangential force in the anti-climber device is described as the tangential component of
the contact force between the end extremities of the car-bodies. The maximum values of the
tangential force are lower than half of the weight of the passenger compartment. The higher

Figure 14. Vertical relative displacement between car-bodies at the interfaces.


Crash analysis and dynamical behaviour of vehicles 399

Figure 15. Maximum tangential force along the interfaces.

levels for the tangential force occur at the interfaces away from the HE, where the vertical gap
between adjacent cars reaches higher levels and are located predominantly in the colliding
train, as illustrated in figure 15. It is observed that the tangential force at the interfaces tends
to increase both in magnitude and in frequency in the final stage of the trains impact.
In a second design stage, the multibody railway vehicle is simulated in a train crash scenario
similar to that of an experimental test performed to validate a low-energy end design developed
within the framework of the Brite/Euram III project SAFETRAIN [31]. The experimental test
consists of a vehicle moving with a velocity of 54 km/h toward a composition with two vehicles
stopped on the railroad, as depicted in figure 16. The two stationary vehicles are equipped
with low-energy ends and connected by a coupler device. See Milho et al. [32] for a more
detailed description of the model.
The force–time history of the buffers of wagon A is displayed in figure 17 for both the
simulation and the experimental test. Note that the experimental test results are plotted for
a single buffer, whereas the expected force resulting from the simulation is shown for the
cumulative effects of the two buffers of wagon A.
The velocities of the three cars during the simulation are plotted in figure 18. It can be
observed that the velocities predicted by the model are very similar to those observed in the
experimental test.
The contact between wagons A and C is predicted to happen with no initial vertical gap
between the buffers, as shown in figure 19. However, as the buffers approach each other, the
vertical gap between the wagons increases, reaching a maximum of 15 mm. The vertical forces
that the buffers have to support in order to prevent overriding, presented in figure 19, oscillate
with a maximum peak of 15 ton.
Table 2 presents the amount of energy dissipated by the different components of the energy
ends and the energy breakdown for the experimental test. Note that the energies predicted in
the simulation and in the test have a difference of 10%, mainly due to the honeycomb. Another

Figure 16. Collision scenario used in the numerical simulations and in the experimental test.
400 J. Ambrósio

Figure 17. Force–time history of the buffers of wagon A. Half of the force magnitude on the buffers is compared
with the force measured in the left buffer of wagon A.

relevant observation is that the total energy absorbed by the buffers and coupler is similar for
the test and for the simulation. However, its breakdown is completely different. This is due
to the force displacement curves used for the coupler and buffers in the model. Some of the
resistance of the coupler is included in the buffers curve instead of these two systems being
modelled independently.

4.2 Side impact of a road vehicle

The appraisal of new strategies for protection of occupants of road vehicles during side impact
is the aim of the study described here, which is part of the work developed in projectAPROSYS-
SP6 [33]. A multibody model of a vehicle Chrysler Neon with two US-DOT SID dummies
inside, based on the original model developed by TNO Automotive, uses the methodologies
described in this work. The analysis of a crash scenario described in the norm FMVSS 214 is
shown in figure 20. The dummy and impact barrier models correspond to the MADYMO 5.2

Figure 18. Time history for the wagons in the simulation and experimental test.
Crash analysis and dynamical behaviour of vehicles 401

Figure 19. Time history of the gap and vertical contact forces in the buffers.

database US-DOT SID dummy and FMVSS 214 barrier. For the simulations carried out and
reported in this document, the multibody dynamics simulation code MADYMO [12] is used.
The model of the Chrysler Neon for side impact is made of eight subsystems represented in
figure 20. Besides the side structure of the vehicle and the seats all remaining structural parts
are considered rigid, as they are supposed to play no role in the side impact crash scenario.
Each subsystem is made by rigid bodies constrained by kinematic joints. For each part of the
side structure of the vehicle, the bodies that make it up are presented in figure 21.

Table 2. Energy dissipation distribution in the components of the train.

Component Absorption (kJ) Remarks Test (kJ)

Buffers Wagon C: 624 This result does not include structural 280
Wagon A: 373 deformation behind the buffers
Total: 997 No structural deformation occurs in the simulation
Coupler 300 Test data includes structural deformation
behind the coupler 835
Low energy end 1 297 Test data includes structural deformation
behind the coupler and buffers 1 435
Front honeycomb 2 780 3 016
Total energy absorption 4 077 4 451

Figure 20. Multibody model for the Chrysler Neon in the side impact test defined by the norm FMVSS 214.
402 J. Ambrósio

Figure 21. Rigid bodies defining the side structure of the vehicle.

The kinematic joints represented in figure 22 are set according to the expected mechanisms of
deformation that the vehicle can experience for the side crash tests. Caution must be exercised
if the model is to be used in any other type of side impact test besides the ones prescribed in
the FMVSS or ECE regulations.
In order for the plastic hinge definition of the side structure to be complete, it is necessary to
define the constitutive relations that relate the moments developed in the kinematic joints and
the angles associated to each degree-of-freedom. Such constitutive relations, for a selected
number of plastic hinges, are illustrated in figure 23.

Figure 22. Kinematic joints for the side structure representation.


Crash analysis and dynamical behaviour of vehicles 403

Figure 23. Moment–angle relations for plastic hinges of the side structure.

The energy absorption of the vehicle side structure in the test configuration can only be
realized when it is included in the complete vehicle. Because all structural plastic deformations
are contained in the domain of the side structure, the remaining of the chassis can be considered
rigid. The attachments between the rigid bodies of the side structure and the rigid body of
the chassis are realized by the crushable elements. The deformation of these elements must
be within prescribed limits. The plastic deformation energy that these elements account for is
due to the deformation of the structure surrounding the roof, sill, A-pillar and C-pillar. If the
deformation of these crushable elements exceeds a given limit, it can be argued that the plastic
deformations of the side structure exceed the modelled region of the vehicle and, therefore,
the validity of the model can be questioned.
The multibody model of the vehicle Chrysler Neon is simulated in the same crash test
scenario with which prototypes of the real vehicle have been experimentally tested by NHTSA.
The quantities measured during the experimental tests are used as target responses that the
model of the vehicle must meet in order to be considered validated. The vehicle responses
in time used for the model validation are: velocity of the vehicle center of mass; velocity
of the barrier; velocity of the front door; velocity of the rear door; velocity of the rear floor;
acceleration of the rear door in the Y -direction; velocity of the sill; acceleration of the US-DOT
driver rib; acceleration of the driver dummy torso; acceleration of the driver dummy pelvis;
acceleration of the passenger dummy rib; acceleration of the passenger dummy torso and
acceleration of the passenger dummy pelvis. The validation procedure, carried out by TNO
Automotive, included the fine tuning of the plastic hinges data that leads to a better correlation
between simulation and experimental responses.
Figure 24 shows the results obtained for the simulation of the FMVSS 214 side crash
test in terms of head, pelvis, thoracic vertebras and mid-rib accelerations. The US-DOT SID
injury criteria corresponding to this simulation are presented in table 3. The value of TTI is
above the maximum acceptable value specified in FMVSS 214 and the value for the pelvis
lateral acceleration is also quite high, even though below the limit specified in the norm. The
velocities of different points of the structure and dummies are similar to those obtained in the
experimental test.
The contact forces for the dummy presented in figure 25 have not been measured in the
experimental test. It is observed that the peak forces for the pelvis, thorax and head occur
simultaneously. The pelvis contact force predicted by the simulation is clearly larger than the
forces carried by the head and by the thorax. This is mainly due to the proximity of the seat
and occupant driver pelvis to the B-pillar, which is directly struck by the barrier.
Another type of data that is not recorded in the experimental test is the kinematics of the
B-pillar and of the occupant. Such kinematics is sketched in figure 26 for different instants
of time. It is clear that the B-pillar intrudes the vehicle passenger compartment, with special
incidence in its lower part. The intrusion of the B-pillar leads to contact with the driver’s seat
and pushes it to the vehicle interior.
404 J. Ambrósio

Figure 24. Y -component accelerations for parts of the dummy measured in the global coordinate system.

The results of the simulation also show the importance of modelling the suspension of
the vehicle and the contact between the tires and the ground. The vehicle roll and its lateral
displacement play an important role in the progression of the crash event. Moreover, the
amount of kinetic energy that is dissipated by the work of the suspension systems and by the
friction between tires and ground cannot be neglected for lower impact speeds.
The results reported for the side impact demonstrate the reliability of the multibody model
of the vehicle and occupants used in this crash scenario. Though not presented here, several
modifications on the structural components of the side of the vehicle, interior furbishing and
equipment arrangement have been tested. On the basis of the outcome of this simulations,
described by parameters such as relative displacements between structural components and
occupant anatomical segments, accelerations of points on the dummies and on the structure,
contact forces or injury indexes, it is possible to obtain vehicle designs that optimize passenger
protection. However, the extensive number of re-analysis of the complete vehicle is only
possible when the computation time for each analysis is acceptable. In order to appraise
the computational efficiency of the methodology described here, the time required for each
simulation of the vehicle side impact with the multibody formulation is measured in terms of
minutes, whereas for the equivalent model in finite elements, in the same test conditions and
period of analysis, the time required is measured in terms of days.

Table 3. Injury criteria values: pelvis accelerations


and thoracic trauma index.

Simulation ID PLA (g) TTI (g)

FMVSS 214 limits 130 85


Chrysler Neon performance 110.23 87.83
Crash analysis and dynamical behaviour of vehicles 405

Figure 25. Contact forces in the pelvis region, thorax region and head.

4.3 Road vehicle rollover

With the purpose of showing the performance of the formulation in situations where multiple
contacts are important issues, such as in the case of vehicle rollover, a model for an all-terrain
vehicle, a M151A2 Jeep represented in figure 27, is presented. The full vehicle, which includes
a rollbar cage for occupant protection, has a total mass of 1470 kg. The location of the vehicle’s
center of mass is 1.232 m behind the front axle and 0.607 m above the ground, when parked
on a flat horizontal surface.
The utility vehicle is modelled with 13 rigid bodies, corresponding to the chassis, double
A-arm front suspension systems and trailing arm rear suspension systems. The interested
reader will find the full set of data for the utility vehicle in Ambrósio [34]. A rollbar cage is
installed in the truck in order to protect the vehicle’s occupants in case of rollover. This is a
flexible frame mounted over the chassis, as depicted in figure 28, and is made of 1025–1030

Figure 26. Sequence of images for the side crash.


406 J. Ambrósio

Figure 27. General dimensions of the all-terrain vehicle.

steel. The cross sectional area of each bar is annular with an outside radius of 2.54 cm. A
model of the rollbar cage with 13 beam elements is used here.
The interaction of the vehicle and/or the rollbars with the ground is described by controlling
the coordinates of six points in the rollbar cage P1 through P6 and eight points (C1 through C8 )
for possible ground contact. The model for the deformation of the vehicle chassis is valid only
if all deformations occur on the rollbar. This model cannot be used, as it is, to describe the
deformation of other parts of the chassis. However, some of the energy dissipation involved
in the impact of the rigid chassis is still described by using the continuous force contact
model.
The all-terrain vehicle model is simulated here in a rollover situation with the initial
conditions described in figure 29. The initial conditions of the simulations correspond to
experimental conditions where the vehicle moves on a cart with a lateral velocity of 13.41 m/s
until the impact with a water-filled decelerator system occurs. The vehicle is ejected with a roll
angle of 23◦ . The initial velocity of the vehicle, when ejected, is 11.75 m/s in the Y -direction,
while the angular roll velocity is 1.5 rad/s.
Three occupants, with a 50th percentile, are modeled and integrated with the vehicle. The
two occupants in the front of the vehicle have shoulder and lap seatbelts, whereas the occupant
seated in the back of the vehicle has no seatbelt. The initial positions of the occupants corre-
spond to a normally seated driver, a front passenger bent to check out the ‘glove compartment’
and a rear occupant with a ‘relaxed’ position.
This setup and the simulation outcome are compared with that of two experimental tests of
the vehicle with three Hybrid III dummies that have been carried at the Transportation Research

Figure 28. Computational model of the rollbar cage.


Crash analysis and dynamical behaviour of vehicles 407

Figure 29. Rollover simulation scenario: (a) initial conditions; (b) position of the vehicle occupants.

Figure 30. View of the experimental test for the truck rollover.

Center of Ohio [35, 36]. An overview of the footage obtained in one of the experimental tests
is shown in figure 30.
The first 2 s of the simulations, presented in figure 31, show that the first contact of the
wheels with the ground occurs at 0.3 s, for all simulations, causing the vehicle to bounce with
an increasing roll velocity. At 0.75 s, the vehicle impacts the ground with the rollbar cage and
continues its rolling motion with contact by different points of the structure. The occupants
in the front of the vehicle are hold in place by the seatbelts. Upon continuing its roll motion,
the vehicle impacts the ground with its rollbar cage, while the ejection of the rear occupant
is complete. The results do not seem to be very sensitive to different contact models or to the
complexity of the finite element mesh applied in the rollbar, but they are extremely sensitive to
the values of the friction between vehicle structure and ground. Owing to the highly nonlinear
nature of the problem, after relatively large periods of analysis, the behaviours of different
models diverge.

Figure 31. View of the outcome of the rollover simulation of a vehicle with three occupants.
408 J. Ambrósio

Figure 32. Deformations of the rollbar cage, represented by node P1 .

In figure 32, the permanent deformations of the front nodal points of the rollbar cage,
which impacted the ground first, are displayed. The permanent deformation of 13 cm in
the lateral direction and 20 cm in the frontal direction obtained in the simulations is sim-
ilar to the permanent deformations observed in experimental tests. Models with 12 and
24 beam finite elements for the rollbar cage show similar results for the first part of the
analysis. The differences between the responses in the second part of the analysis must be
attributed to the nonlinearity of the problem rather than to the models used for the rollbar
cage.

Figure 33. Severity index for the vehicle occupants.


Crash analysis and dynamical behaviour of vehicles 409

The severity index observed for the occupants in figure 33 indicates a very high probability of
fatal injuries in the conditions simulated. Note that the model has rigid seats, interior trimming
for the dashboard, side and floor panels, and that the ground is also considered to be rigid. If
some compliance is included in the vehicle interior, it is expected that the head accelerations
are lower.
The kinematics of the biomechanical models of the occupants, and in particular that of
the ejected occupant, are similar to the kinematics of the crash test dummies used in the
experimental tests. Several simulations of the vehicle rollover with occupants seating with
different postures have been performed. These simulations show that regardless of the rear
occupant seating posture the ejection and post-ejection occupant kinematics remains basically
unchanged.

5. Conclusions

This work demonstrated a multibody dynamics-based formulation that effectively promotes the
simultaneous analysis of the vehicle stability and manoeuvrability dynamics with structural
and occupant crashworthiness. As the formalisms used are common to all disciplines that
are relevant to vehicle design, including structural crashworthiness, it is possible to recycle
vehicle models for different applications, to account for the different mechanisms of energy
absorption for a wider range of crash velocities and to have a better appraisal for the occupant’s
mechanisms of injury and injury criteria. In the process, it was also demonstrated how nonlinear
finite elements can be integrated with conventional rigid multibody descriptions in order to
build better general vehicle models. Though not demonstrated through the applications used
here, the reliability of the prediction of certain crash mechanisms allows devising driving
support systems that assist the driver in crash avoidance manoeuvres. Through the railway
crashworthiness application, it was demonstrated that the design of the train systems that
ensure the train alignment during the crash requires that the relative motion between the cars
of the same train is well described, and therefore, that the suspension models, developed in
the framework of typical vehicle dynamics applications, are used to account for the relative
pitching between adjacent vehicles. The side impact of the road vehicle confirmed these
conclusions, now in what the vehicle roll motion is concerned. The rollover simulations showed
how the vehicles can withstand multiple contacts when the simulation is being preformed. All
these applications are run in a reasonable short amount of time, measured in minutes, which
demonstrates the suitability of the methods for studies that require extensive re-analysis.

Acknowledgements

The support of the Portuguese Foundation for Science and Technology, FCT, through project
2/2.1/TPAR/2041/95, to study the vehicle rollover is gratefully acknowledged. The support of
EC through projects BE-96-3092 (SAFETRAIN), with the partners ADTRANZ/SOREFAME
(Pt), ERRI (Nl), SNCF (Fr), DB (Ge), PKP (Pl), FMH (Pt), CIC (UK), GEC/MC (UK), AEA
Technology/BRR (UK), U. Valencienne (Fr), GAC/CIMT (Fr), ALS/DDF (Fr), IFS (Ge),
TU Dresden (Ge), DUE (Ge), enabled to develop the studies on train crashworthiness and
through project TIP3-CT-2004-506503 (APROSYS-SP6), with the partners SIEMENS VDO
(Ge), Faurecia (Ge), DaimlerChrysler (Ge), CIDAUT (Sp), Fh G (Ge), Warsaw University of
Technology (Pl) and TNO (Nl) provided the results of the road vehicle side impact studies.
The EC support is also greatly appreciated.
410 J. Ambrósio

References

[1] Wilson, R.A., 1970, A review of vehicle impact testing: how it began and what is being done. SAE Transactions,
SAE paper 700403.
[2] Tidbury, G.H., 1980, Vehicle structural analysis: a survey. International Journal of Vehicle Design, 1, 165–172.
[3] Kamal, M.M. and Wolf, J.A. (Eds), 1982, Modern Automotive Structural Analysis (New York: Van Nostrand
Reinhold Company).
[4] McHenry, R.R., 1963, Analysis of the dynamics of automobile passenger restraint systems. In Proceedings of
the 7th Stapp Car Crash Conference, Playa del Rey, California, 11–13 November, pp. 207–249.
[5] Lobdell, T.E., 1973, Impact response of the human thorax. In: Human Impact Response: Measurement and
Simulation (New York: Plenum Press), pp. 201–245.
[6] Nikravesh, P.E., Chung, I.S. and Benedict, R.L., 1983, Plastic hinge approach to vehicle crash simulation,
Computers and Structures, 16(1–4), 395–400.
[7] Ambrósio, J., Pereira, M.S. and Dias, J., 1996, Distributed and discrete nonlinear deformations in multibody
dynamics, Nonlinear Dynamics, 10(4), 359–379.
[8] Milho, J., Ambrósio, J. and Pereira, M., 2004, Design of train crash experimental tests by optimization
procedures. International Journal of Crashworthiness, 9(5), 483–493.
[9] Robbins, D.H., Bowman, B.M. and Bennett, R.O., 1974, The MVMA two-dimensional crash victim simulation.
In Proceedings of the 18th Stapp Car Crash Conference, Ann Arbor, Michigan, pp. 657–678.
[10] Fleck, J.T., Butler, F.E. and Vogel, S.L., 1974, An improved three-dimensional computer simulation of motor
vehicle crash victims. Technical Report no. ZQ-5180-L-1, Calspan Corporation, Buffalo, New York.
[11] Laananen, D.H., Bolukbasi, A.O. and Coltman, J.W., 1983, Computer simulation of an aircraft seat and occupant
in a crash environment: volume 1, Technical Report DOT/FAA/CT-82/33-I, US Department of Transportation,
Federal Aviation Administration, Atlantic City, New Jersey.
[12] TNO Automotive, 1999, MADYMO Theory Manual vs. 5.4 (Delft: TNO Automotive).
[13] Thompson, J.E., 1973, Vehicle crush prediction using finite element techniques, SAE Transactions, SAE paper
no. 730157.
[14] Belytschko, T. and Hsieh, B.J., 1973, Nonlinear transient finite element analysis with convected coordinates.
International Journal of Numerical Methods in Engineering, 7, 255–271.
[15] Hallquist, J.O., 1976, Preliminary users manual for DYNA3D and DYNAP (nonlinear dynamic analysis of solids
in three dimensions). Report UCID-17268, Berkley, California, Lawrence Livermore National Laboratory.
[16] Belytschko, T. and Kenedy, J.M., 1986, WHAMS-3D, An Explicit 3D Finite Element Program (Willow Springs,
Illinois: KBS2 Inc).
[17] Hallquist, J.O., 1998, LS-DYNA Theoretical Manual: Nonlinear Dynamics Analysis of Structures (Livermore,
California: Livermore Software Technology Corporation).
[18] Haug, E., 1989, The PAM-CRASH code as an efficient tool for crashworthiness simulation and design, In
Second European Cars/trucks Simulation Symposium, Schliersee, Germany, 22–24 May.
[19] Mecalog, 2000, RADIOSS Input Manual, Version 4.2 (Anthony, France: Mecalog).
[20] Nikravesh, P.E., 1988, Computer-Aided Analysis of Mechanical Systems (Englewood-Cliffs, New Jersey:
Prentice-Hall).
[21] Petzold, L., 1994, Computational challenges in mechanical systems simulation. In: M. Pereira and J. Ambrósio
(Eds) Computer-Aided Analysis of Rigid and Flexible Mechanical Systems (Dordrecht, The Netherlands: Kluwer
Academic Publishers), pp. 483–499.
[22] Augusta Neto, M. and Ambrósio, J., 2003, Stabilization methods for the integration of differential-algebraic
equations in the presence of redundant constraints. Multibody Systems Dynamics, 10(1), 81–105.
[23] Huston, R.L. and Wang, Y., 1994, Flexibility effects in multibody systems. In: M. Pereira and J. Ambrósio (Eds)
Computer-Aided Analysis of Rigid and Flexible Mechanical Systems (Dordrecht, The Netherlands: Kluwer
Academic Publishers), pp. 351–376.
[24] Winmer, A., 1977, Einfluss der lelastungsgeschwindigkeit auf das festigkeitsund verformungsverhalten am
beispiel von kraftfarhzeugen. ATZ, 77(10), 281–286.
[25] Ambrósio, J., 1996, Dynamics of structures undergoing gross motion and nonlinear deformations: a multibody
approach. Computers and Structures, 59(6), 1001–1012.
[26] Ambrósio, J. and Nikravesh, P., 1992, Elastic–plastic deformations in multibody dynamics. Nonlinear Dynamics,
3, 85–104.
[27] Ambrósio, J., 2003, Efficient kinematic joints descriptions for flexible multibody systems experiencing linear
and non-linear deformations. International Journal of Numerical Methods in Engineering, 56, 1771–1793.
[28] Lankarani, H.M., Ma, D. and Menon, R., 1995, Impact dynamics of multibody mechanical systems and appli-
cation to crash responses of aircraft occupant/structure. In: M.S. Pereira and J. Ambrósio (Eds) Computational
Dynamics in Multibody Systems (Dordrecht, The Netherlands: Kluwer Academic Publishers), pp. 239–265.
[29] Wu, S., Yang, S. and Haug, E., 1984, Dynamics of mechanical systems with coulomb friction, stiction, impact
and constraint addition–deletion. Technical Report 84-19, Center for Computer Aided Design, University of
Iowa, Iowa City, Iowa.
[30] Milho, J., Ambrósio, J. and Pereira, M., 2002, A multibody methodology for the design of anti-climber devices
for train crashworthiness simulation. International Journal of Crashworthiness, 7(1), 7–20.
[31] SAFETRAIN, 2001, BRITE/EURAM Project n.BE-309, Dynamic Tests, SAFETRAIN Technical Report T8.2-F,
Deutsche Bann, Berlin, Germany.
Crash analysis and dynamical behaviour of vehicles 411

[32] Milho, J., Ambrósio, J. and Pereira, M., 2003, Validated multibody model for train crash analysis. International
Journal of Crashworthiness, 8(4), 339–352.
[33] Dias, J, and Ambrósio, J., 2005, APROSYS-SP6: Actuator requirements definition multibody simulations,
Technical Report AP-SP61-D614.
[34] Ambrósio, J., 2002, Contact and impact models for vehicle crashworthiness simulation. In: E. Chirwa
(Ed.), Proceedings of ICRASH2002 International Crashworthiness Conference, Melbourne, Australia,
25–27 February.
[35] TRCO, 1985, 30 mph Rollover Test of an AM General Model M151-A2 1/4 Ton Jeep, The Transportation
Research Center of Ohio, Test Report, November.
[36] TRCO, 1986, 30 mph Rollover Test of an AM General Model M151-A2 1/4 Ton Jeep, The Transportation
Research Center of Ohio, Test Report, January.

You might also like