Professional Documents
Culture Documents
Finite Elements in Analysis and Design: D. Mundo, R. Hadjit, S. Donders, M. Brughmans, P. Mas, W. Desmet
Finite Elements in Analysis and Design: D. Mundo, R. Hadjit, S. Donders, M. Brughmans, P. Mas, W. Desmet
A R T I C L E I N F O A B S T R A C T
Article history: The paper proposes an engineering approach for the replacement of beam-like structures and joints in
Received 3 March 2008 a vehicle model. The final goal is to provide the designer with an effective methodology for creating a
Received in revised form 3 December 2008 concept model of such automotive components, so that an NVH optimization of the body in white (BIW)
Accepted 10 December 2008
can be performed at the earliest phases of the vehicle design process. The proposed replacement method-
Available online 7 February 2009
ology is based on the reduced beam and joint modelling approach, which involves a geometric analysis of
Keywords:
beam-member cross-sections and a static analysis of joints. The first analysis aims at identifying the beam
Beam center nodes and computing the equivalent beam properties. The second analysis produces a simplified
Joint model of a joint that connects three or more beam-members through a static reduction of the detailed
Conceptual design joint FE model.
NVH In order to validate the proposed approach, an industrial case-study is presented, where beams and joints
Vehicle body of the upper region of a vehicle's BIW are replaced by simplified models. Two static load-cases are defined
to compare the original and the simplified model by evaluating the stiffness of the full vehicle under
torsion and bending in accordance with the standards used by automotive original equipment manufac-
turer (OEM) companies. A dynamic comparison between the two models, based on global frequencies
and modal shapes of the full vehicle, is presented as well.
© 2009 Elsevier B.V. All rights reserved.
0168-874X/$ - see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.finel.2008.12.003
D. Mundo et al. / Finite Elements in Analysis and Design 45 (2009) 456 -- 462 457
where K is the stiffness matrix, F and x are the force and the dis-
placement vectors, respectively. By identifying nt boundary degrees
of freedom (DOFs), which must be retained in the solution, and no
internal DOFs, which are to be removed by static condensation, the
system of Eq. (1) can be partitioned as follows:
Koo Kot x F
· o = o (2)
Kto Ktt xt Ft
where subscripts t and o are used to designate the boundary and the
internal DOFs, respectively. From the first line of Eq. (2), the internal
displacement vector can be determined as
−1
xo = Koo (Fo − Kot · xt ) (3)
−1
By introducing the static reduction matrix Got =−Koo Kot and substi-
tuting Eq. (3) into the second line of Eq. (2), the following equation
is obtained:
3. Case-study
where L and W denote the wheelbase and the width of the car,
respectively, measured at the front suspension points.
Based on the torsional deflection t , the torsional stiffness Kt is
determined as
M
Kt = (7)
Fig. 4. Original FE model of the BIW. t
where M = F · W is the moment applied at the front suspension,
resulting from two oppositely oriented forces F.
Similarly, the bending stiffness Kb is determined from b as
2FL
Kb = (8)
b
where F is the vertical force applied at the frontal suspension loca-
tion.
The stiffness properties of the BIW are estimated for both the
models in Figs. 4 and 5, by performing a static FE analysis (Nastran-
Sol 101) with both models. In Table 1, the torsional and bending
stiffness indicators are listed, as well as the approximation involved
by the simplified model w.r.t. the original model. The results show
that the bending stiffness of the original vehicle model is accurately
Fig. 5. Conceptual FE model of the BIW. The original meshes of 10 beam-members predicted by the model with the replaced simplified beams and
and four joints are replaced by simplified beam elements and joint superelements. joints, while a significant discrepancy between the original and the
460 D. Mundo et al. / Finite Elements in Analysis and Design 45 (2009) 456 -- 462
simplified models is obtained for torsion. In the latter case, the stiff- free–free conditions is performed, the first six modes are rigid body
ness of the full vehicle body is overestimated by 10.15%, which sug- modes. From mode seven onwards, the natural modes are found. The
gests that the definition of correction factors is indeed required. This original model of the vehicle body has 10 non-rigid modes in the
will be evaluated and assessed in Section 4. frequency range of interest. Only five out of them are global modes
of the BIW, since the remaining modes involve a local deformation
3.3. Dynamic comparison of the structure. The eigenfrequency values of the five global modes
are used as dynamic indicators to evaluate the correlation between
In order to compare the simplified and the original model in terms the original and the conceptual BIW model. In Table 2, the global
of dynamic behavior, the frequencies and mode shapes of the BIW frequencies of both models are listed. Also the dynamic comparison
are estimated through an FE modal analysis (Nastran-Sol 103) in the shows that the conceptual model overestimates the stiffness of the
low-frequency range of 0–50 Hz. When a normal mode analysis in full structure. It can be seen that the natural frequencies are shifted
upwards by an amount in the range 0.15–6.70%.
A further comparison between the two models is also performed
Table 1 in terms of modal shapes by using the modal assurance criterion
Torsional and bending stiffness of the original and of the conceptual FE model.
(MAC) [20]. Let V1 and V2 be the modal matrices of the original and
Torsion Bending the modified model, respectively. The correlation index between two
generic modes {V1 }i belonging to the first matrix and {V2 }j belonging
Original model Concept model Original model Concept model
to the second matrix can be evaluated as [21]
Stiffness
(Nm/rad) 1.456E+05 1.603E+05 5.013E+04 5.036E+04 ({V1 }H
i
{V2 }j )2
(%) – 10.15 – 0.45 MACi,j = (9)
({V1 }i {V1 }j )({V2 }H
H
i
{V2 }j )
0.9
0.8
1
0.8 0.7
0.6 0.6
0.4 0.5
0.2 0.4
0 0.3
Id 10-47.9
Id 7-47.9
Id 8-41.7
Id 5-42.0
Id 7-39.4
Id 4-40.0
0.2
Id 3-26.1
Id 2-27.9
Id 1-19.3
Id 1-18.2
ues o del
val nal m
Mod Discrete cre
t e
rigi
0.1
e Se v
t.2 - alues Dis .1 - O
t
unco Se
rrec
ted M ode <0
Fig. 7. MAC matrix between the original and the conceptual model.
D. Mundo et al. / Finite Elements in Analysis and Design 45 (2009) 456 -- 462 461
beams, as computed by means of a geometric approach, involve an Fig. 9. History of the goal function as defined in Eq. (13).
overestimation of the global stiffness of the vehicle. Such a conclu-
sion is in line with the results reported in other research papers
[16,17], which suggest a correction of the beam properties in order Table 3
to take into account section variations and discontinuities (holes, Optimal values of the correction factors for bending stiffness parameters.
spot-welds, stiffeners). For this reason, proper correction factors are Correction factors
defined as the ratio between the actual value of each stiffness pa-
rameter and the nominal value. For instance, according to such a B1 B2 B3 B4 B5
definition, the equivalent bending stiffness parameters of a beam Ixx 0.260 0.157 0.311 0.007 0.001
end-section are given by Iyy 0.933 0.391 0.771 0.079 0.011
Izz 0.761 0.455 0.893 0.192 0.219
Iyy = cyy · Iyy,nom (10) Ixy 0.809 0.371 0.812 0.113 0.001
where Iyy,nom and Izz,nom are the stiffness parameters computed where the inequality constraint ensures that the actual set of cor-
through the geometric approach described in Section 2.1, while cyy rection factors defines a feasible solution.
and czz are the corresponding correction factors. An optimal and feasible solution can be searched for by means of
A set of correction factors is then defined for each beam-member a genetic algorithm [22]. For this purpose, the constrained problem
and estimated by means of a model updating procedure, which is defined above is transformed into an unconstrained problem by us-
described in the following sub-section. ing a penalty formulation: a large cost-value is added to the objec-
tive function in case that the constraint is violated. Such a procedure
ensures that an unfeasible solution has a larger goal function than
4.1. Estimation of correction factors
any feasible solution. This enables the convergence of the algorithm
towards a global optimum, which fulfils all constraints. The original
In order to estimate proper correction factors by means of a model
constrained optimization problem is then replaced by the following
updating procedure, the virtual test-case represented in Fig. 8 is used.
unconstrained problem:
The vehicle body is clamped at the rear suspensions (points C and
D) and statically loaded at the front suspensions (points A and B). Minimize f (c1 , . . . , cN ) + k(c1 , . . . , cN ) · C (13)
The displacements of 10 control points Pi , (i = 1, . . . ,10) are estimated
and used to compare the static stiffness of the original model, used where C is a penalty cost-value, while k(c1 , . . . , cN ) is a Boolean func-
as reference, and the simplified model. The following function is tion defined as
defined as a measure of the difference between the simplified and
0 if I1 ∗ I2 > I12
the original model in terms of static response to the load-case shown k(c1 , . . . , cN ) = (14)
in the Fig. 8 1 otherwise
yi,A − yi,R zi,A − zi,R The optimization problem is solved by a genetic algorithm imple-
f (c1 , . . . , cN ) = + (11)
y z mented in Matlab. In Fig. 9 the history of the goal function defined
i,R i,R
i
by Eq. (13) is shown.
where c1 , . . . , cN are the actual values of the correction factors, while Table 3 lists the optimal values of the correction factors evalu-
subscripts A and R refer to the actual and the reference model, re- ated for both bending stiffness-parameters of all the replaced beam-
spectively. members.
An optimization problem can be defined as the search for the The geometric approach used to compute the equivalent beam
vector of correction factors that minimizes the difference between parameters considers each cross-section as closed, even for the roof
the two models, that is cross members. This is the main reason why correction factors are
quite low (significantly lower than one), especially for these mem-
Minimize f (c1 , . . . , cN ) bers. Roof cross members, in fact, are formed by two panels con-
Subject to I1 ∗ I2 > I12 for all beams (12) nected to each other by glue connections, which are much more
462 D. Mundo et al. / Finite Elements in Analysis and Design 45 (2009) 456 -- 462
Stiffness
Acknowledgment
(Nm/rad) 1.456E+05 1.462E+05 5.013E+04 5.024E+04
(%) – 0.46 – 0.22
The work presented in this paper has been performed in
the framework of the research project “Analysis Leads Design-
Table 5 Frontloading Digital Functional Performance Engineering”, which is
Dynamic comparison between the original and the final conceptual FE model in supported by I.W.T. Vlaanderen.
terms of global frequencies and modal shapes.