You are on page 1of 11

International Journal of Automotive Technology, Vol. 14, No. 5, pp.

817−827 (2013) Copyright © 2013 KSAE/ 073−18


DOI 10.1007/s12239−013−0090−3 pISSN 1229−9138/ eISSN 1976−3832

METHODOLOGY FOR THE ACCURACY IMPROVEMENT OF FEM


BEAM TYPE T-JUNCTIONS OF BUSES AND COACHES STRUCTURES

E. ALCALÁ, F. BADEA*, Á. MARTIN and F. APARICIO


Development, University Institute of Automobile Research of the Madrid Community, Madrid 28031, Spain

(Received 19 September 2012; Revised 2 February 2013; Accepted 12 February 2013)

ABSTRACT−The analysis of structures with finite elements methods (FEM) represents a widely spread technique. For large
tubular structures similar to the buses and coaches upper structures, beam type elements are utilized due to the fact that these
elements provide satisfactory results at relatively reduced computational performances. However the beam type elements have
a main disadvantage determined by the fact that the modeled joints have an infinite rigid behavior. This shortcoming
determines a stiffer behavior of the modeled structures which translates into an error source for the structural simulations (up
to 45%). To overcome this problem, a simple methodology was conceived and an alternative optimized equivalent beam
model obtained. The methodology studies the behavioral characteristics of beam modeled T-junctions determining their
limitations and comparing them to equivalent T-junctions modeled with shell and volume elements. This way an improved T-
junction has been obtained, in which the behavioral error was reduced to less than 5%. Furthermore the FEM obtained results
were validated with real T-junctions.

KEY WORDS : Structural optimization, Bus structural analysis, FEM (Finite Elements Methods)

1. INTRODUCTION another important characteristic that makes them preferred


when modeling large tubular structures is given by the fact
In the last decades the study and improvement of structures that they do not require large amounts of computational
through Finite Element Analysis (FEA) has become capabilities (processor speed, memory, etc.) which offers
extremely important (Adams and Askenazi, 1998). This engineers the possibility of creating more complex and
fact has been determined by the fast evolution of detailed full models for more complex simulations.
computers, along with the FEA software. Also, the The beam type elements have a simplified formulation,
constant necessity to build better and improved products at they are uniaxial elements that can be defined by two nodes
faster rates and with lower costs has determined the general (I, J), in a tridimensional coordinate system. They are
acceptance of the FEM in most of the industry fields. largely utilized and accepted for modeling large tubular
On most of the FEA software for structural analysis structures (buses and coaches upper structures) with
there are three main categories of elements: Beam, Shell satisfactory results in a variety of analysis, structural,
and Volume. modal, transitory, fatigue, impact, etc. (Gauchia et al.,
Choosing the appropriate element for a certain 2010; Balázs, 2005; Lan et al., 2004, Liang and Le, 2010).
simulation represents a complicated process in which it is Despite their wide spread acceptance and use, the beam
necessary to take into consideration a series of characteristics: type elements have some important limitations determined
the features and complexity of structures (models), the by their simple formulation. Their main disadvantage is
computational resources available, the type of simulation determined by their incapability to reproduce the characteristics
and the desired quality of the results (Lee et al., 2000, and the localized behavior at the joint level.
2003; Eriksson, 2001; Kim et al., 2001a, 2001b). In fact, for any modeled junction with beam type
This process is extremely important and it is generally elements, the behavior of the joint is infinitely rigid (Badea
determined by the experience of the analyst. et al., 2010; Garcia and Vicente, 2002). This is determined
For the modeling of large tubular structures, similar to by the fact that the joint is reduced into a rigid node,
the buses and coaches upper structure, the use of beam type therefore when performing structural analysis this feature
elements is very common (Gauchia et al., 2010; Lan et al., introduces a series of estimations errors (Arribas et al.,
2004; Balázs et al., 2005; Milojevic and Pesic, 2011). 2012; García and Vicente, 2002).
Besides the simplicity of modeling with these elements, By modeling tubular junctions with shell and volume
elements, more realistic models can be obtained (Adams
*Corresponding author. e-mail: Francisco.badea@upm.es and Askenazi, 1998; Bosch, 2007), because the topological

817
818 E. ALCALÁ, F. BADEA, Á. MARTIN and F. APARICIO

characteristics of the junction at the joint level can be


reproduced more accurately. This way, the shortcoming
that the beam type elements present can be solved.
The upper buses and coaches structures have large
dimensions and are considerably complex. Modeling these
structures with shell and volume elements represents an
extremely complex and time consuming process (Zienkiewicz
and Taylor, 2002; Adams and Askenazi, 1998).
Even though the results that can be achieved with the
shell and volume type elements are more realistic than the Figure 1. Finite element model of an urban bus.
results that can be obtained with the beam type elements,
the complexity of the modeling process along with the vast
computational performances required, make them unattractive their tridimensional coordinates and the beam element has
for this type of structural simulations. to be associated with these nodes.
Hence, there is a present necessity of conceiving a Also the beam type elements have a great flexibility
methodology to improve the behavior of tubular structures when modifications have to be done within the models.
modeled with beam type elements. For example, if a certain square profile needs to be
This article presents a complete methodology in which changed, the modification can be easily done by modifying
various analysis are performed characterizing the behavior the parameters (properties) of that profile.
of FEM modeled T-junctions, along with a complete Figure 1 presents the FEM of an urban bus. The upper
experimental validation resulting in the development and structure has been entirely modeled with BEAM 4
verification of an improved alternative beam T-junction elements.
model. These characteristics make this type of element
In earlier investigations, mainly conducted in the attractive and widely used for analyzing large tubular
European research project Optivirtest (TRA200914513), it structures (buses and coaches upper structures).
has been seen that between the beam type element
junctions compared, with the shell and volume junctions, 2.2. Limitations of the Beam Type Elements
there are differences in rigidity that can vary between 5 to One of the most important disadvantages that the beam
45% (Badea et al., 2010; Arribas et al., 2012). type elements have is determined by their incapability to
The level of error depends on a series of parameters on reproduce the localized behavior of structures at the joint
which the most important are the dimensions of the profiles level.
(height, width and thickness) and the complexity of the In fact, due to their simple formulation for any modeled
structures. As the structures are more complex the higher junction, the joint it is concentrated at a node having an
the difference in rigidity between models are. infinitely rigid behavior.
With the proposed alternative beam junction model an Also, with the beam modeled junction, the topological
improvement of more than 90% can be obtained, leading to characteristics of different joints cannot be reproduced.
more realistic models providing improved results for Figure 2 presents a comparison between two different T-
structural simulations with important applications for buses junctions (a, b) and their corresponding unique FEM model
and coaches analysis.

2. DESCRIPTION AND DETAILS OF THE


STUDIED PROBLEM

For the development of this study, the FEA software


ANSYS was used, among the variety of beam type
elements available, the beam type element named “BEAM
4” which utilizes the formulation of the Euler-Bernoulli
beam has been selected.

2.1. Advantages of the Beam Type Elements


As mentioned in the previous paragraphs, the main
advantage of the beam elements is related to their simple
formulation due to which big tubular models can be solved
fast without the need of large computational performances.
Also, the geometrical modeling with these elements it is Figure 2. Equivalent BEAM model (c) for two different
extremely simple. Two nodes I and J, have to be defined by simple T-junctions (a), (b).
METHODOLOGY FOR THE ACCURACY IMPROVEMENT OF FEM BEAM TYPE T-JUNCTIONS 819

with beam elements (c). the T-junction show in Figure 2 (b) will be references as T2.
Between the two real T-junctions (a, b) there are In order to understand and characterize the behavior of
important differences at the joint level, these differences the analyzed models two main categories of simulations
determine important behavioral variations. Despite this were necessary.
fact, the equivalent beam model is the same. The first category was intended to analyze and
As a general characteristic, for any T-junctions, having determine if there were any differences when simple
the same dimensional characteristics but with a different structures (cantilever beams) were modeled with beam,
configurations at the joint level, the equivalent FEM model shell and volume type elements.
with beam type elements will always be the same (c) and The second category was conceived to analyze and
will be composed by three beam elements (1-3), sharing a quantify the behavioral differences that exist when
common rigid node (4). different T-junctions were modeled with beam, shell and
Because all of the topological details of the joints are volume type elements.
reduced into the common node (4), none of the behavioral
characteristics determined by the interactions that take 3.1. Behavioral Analysis of Simple Cantilever Beams,
place at the joint level are being taken into consideration. Modeled with Beam, Shell and Volume Type Elements
Furthermore, because the common node has an infinitely In order to evaluate if there were any significant differences
rigid behavior (unrealistic situation) the modeled determined exclusively by the elements formulation, a
structures, with beam type elements, are considerably more simple methodology was conceived in which simple
rigid than the real structures. tubular structures were modeled with beam, shell and
In earlier investigations it has been observed that the volume type elements, and their results compared.
difference in rigidity have a direct relation with the In this scope, simple cantilever beams with all of the
characteristics of the tubular profiles and the complexity of mentioned element types had been consecutively modeled.
the structure, in some cases rigidity differences of up to For each one of these models one of the ends was
45% have been seen (Badea et al., 2010). embedded and a four load cases were consecutively
applied at the free end.
2.3. Alternatives to the Beam Type Elements Figure 3 presents the dimensional characteristics and the
By modeling tubular structures with shell and volume type load cases for the simple cantilever beam model analyzed.
elements, more complex models can be obtained in which Along the main axial direction (X) a total of six sections
the topological characteristics of the joints are being taken (S1 – S6) were selected at an equal distance of 1/5 of the
into consideration, providing more realistic results. total length of the cantilever beam (L).
Although there is a clear advantage on the quality of the Among the load cases defined (C1 – C4) two of them
results, modeling tubular structures with shell or volume had a flexural effect on the structure (C1-C2), one was
type elements represents an extremely complicated task purely torsional (C3) and the last one had a purely traction
that also leads to models with a high number of elements axial effect (C4).
that require vast amounts of computational performances, For each one of the load cases the average nodal
the shell element T-junction has at least 16 times more displacements and/or average rotations at each one of the
elements than equivalent beam model. defined sections were obtained. This way any differences
Studies performed on buses and coaches upper structures that exist between models can be quantified.
throughout FEM modeling usually require structural Because the embedment in all of the three cases was
modifications or simply changing a certain profile for completely rigid, any behavioral differences between
another. If the model is made with beam type elements, this models were to be determined mainly by the element type
task is resumed to a simple modification of the structure or formulation.
the properties of certain profiles. After a series of commonly utilized rectangular and
This is not the case with shell and volume type elements
in which the modifications are much more complicated and
time consuming. In most of the situations it is necessary to
model a complete new profile with all its adjacent connec-
tions to the main structure.

3. METHODOLOGY FOR THE ANALYSIS AND


ESTIMATION OF THE BEHAVIORAL
DIFFERENCES BETWEEN BEAM, SHELL AND
VOLUME T-JUNCTIONS

For simplicity purposes, in the next paragraphs the T- Figure 3. Dimensional characteristics and load cases of the
junction shown in Figure 2 (a) will be referenced as T1, and simple cantilever beam model analyzed.
820 E. ALCALÁ, F. BADEA, Á. MARTIN and F. APARICIO

Figure 4 presents the T-junction model with its


dimensional characteristics and the load cases that have
been utilized.
After the T1 and T2 junctions were analyzed with a
series of different dimensional configurations for the T-
junction and the cross-section of the rectangular profile, it
was determined that in these situations there are significant
differences between the beam modeled T-junctions results
compared to those obtained with the shell and volume
models.
As was expected, the beam models have a more rigid
behavior, regardless of the load case analyzed, the junction
dimensional characteristics and the cross-sectional profile
Fiugre 4. Dimensional characteristics and load cases for the (Figure 10).
T-junctions analyzed.
4. EXPERIMENTAL VALIDATION TROUGH
MECHANICAL ANALYSIS
square hollow profiles had been analyzed, it has been
determined that there are no significant differences in the For the evaluation of the results obtained for the T1 and T2
results. junctions modeled with shell and volume type elements
Figure 8 presents the nodal displacements obtained (reference models), a series of experimental measurements
along the Y axis corresponding to the C1 load case of the were carried out.
cantilever beam with a square hollow profile of 40×40×3 These experiments were conceived in order to
mm. reproduce, in the most accurate way possible, the
Figure 9 presents the nodal rotations along the Y axis characteristics of the FEM models (Figure 4).
corresponding to the C2 load case of the cantilever beam The square hollow profile with 40 mm by side and a
with a square hollow profile of 40×40×3 mm. thickness of 3 mm is one of the most utilized profiles in the
construction of the upper structures for buses and coaches,
3.2. Behavioral Analysis for the Analyzed T1 and T2 because of that, this profile was utilized for the
Junctions Modeled with Beam, Shell and Volume Type experimental validation.
Elements Both of the T1 and T2 junctions were manufactured with
Once it has been determined that there are no significant S235J steel, commonly utilized in the buses and coaches
behavioral differences given by the element formulations upper structures. The dimensions of the junctions (Figure
when simple tubular structures are modeled with beam 4) had been selected as L1 = 1000 mm and L = 1000 mm,
shell or volume type elements, an analogous methodology with a manufacturing tolerance under of ± 0.5%.
was applied for the study of the T1 and T2 junctions. Figure 5 presents the experimental setup for the
The scope in this situation was to determine and quantify mechanical validation.
the behavioral differences determined by the fact that the All of the experiments were performed on a universal t-
beam type elements cannot reproduce the characteristics of slotted floor plate mounted on vibration reducer polymer
the joints in an accurate way.
Similar to the cantilever beam situation, a total of six
sections (S1-S6) were defined on which the nodal
displacements and/or rotations were obtained on each one
of the modeled T-junctions.
Also in order to characterize the stiffness of the joint, all
of the degrees of freedom for the T-junction, where applied
the same four load cases (C1-C4).
From the C1 and C2 load cases, two displacements (Y-
axis, Z-axis) and two rotations (Z-axis, Y-axis) can be
characterized, the C3 load case provides the rotation along
the X-axis and the C4 load case the displacements along
the X-axis, this way from the four load cases the six
degrees of freedom can be obtained.
In this case the extremities of the vertical beam were
embedded and the load cases applied at the free end of the Figure 5. Experimental setup for the mechanical validation
horizontal beam. though mechanical analysis.
METHODOLOGY FOR THE ACCURACY IMPROVEMENT OF FEM BEAM TYPE T-JUNCTIONS 821

silent blocks (Figure 5, point 3) of the loads


The embedment of the FEM models is infinitely rigid, The load cases applying device, is formed of two main
because this type of embedment cannot be completely components (Figure 6, points 1, 4), that are assembled
reproduced in real life, a series of evaluations and together with four M6 screws (Figure 6, point 3).
experiments had been carried out in order to obtain a In the center part of the first component there is a
technical solution that would approach the infinitely rigid protruding region (Figure 6, point 2), that fits tightly on the
embedment model in the most accurate way. The best interior of the utilized profile. Moreover, the central area of
results were achieved utilizing a 100 × 100 × 10 mm L the second component presents a machined region that
profile, reinforced in the middle with a welded steel corresponds to the exterior of the utilized profile.
triangle of 10 mm thickness (Figure 5, point 1). This way, when both of the components are mounted, in
Each one of the free ends of the transversal beam were the central region remains a gap that corresponds to the
welded to the L profile supports. shape of the selected profile.
All of the manufacturing and welding processes were Once the device is mounted on the T-junction, is secured
performed under laboratory conditions in order to obtain to the profile with four special M10 screws with
high quality results suitable for the validation process. elastomeric discs at the end, providing a tight fit on the
The L profiles were mounted on top of two custom made profile without creating local deformations of the material
H-profile supports and bolted together with a total of four (Figure 6, point 5).
M20 screws (Figure 5, point 2). In order to avoid any errors that might result from
Each one of the H-profile supports were fixed to the t- differences in the applied forces, were utilized a series of
slotted floor plate with four M20 screws, bolted into steel 30 kg calibrated weights that were attached to the load case
threaded blocks. device utilizing nylon strings and two symmetrical
To avoid any relative movement between the supports, positioned eyebolts.
all of the M20 screws were tightened to 70% of their For the C1 load case (Figure 5) the calibrated weigh was
resistance limit. The resulting force was considerably hung from a string running through the eyebolts, ensuring
higher than the necessary force to restrict any relative that the weight was positioned in the middle of the profile.
movement without affecting the characteristics of the For the C2 load case, the arrangement was similar
screws. complemented with a pulley mounted on a custom made
The three axial forces and the torsional moment utilized support in order to obtain a lateral force. The position of the
in the FEM models, represent pure loads that are extremely pulley was selected utilizing an auto leveling laser level to
complicated to reproduce in real live. ensure a straight force.
In order to obtain load situations approaching the most to The most problematic load case situation was the C3 due
this theoretical situation, a custom made CNC machined to the complexity of obtaining a pure torsional moment
device was utilized (Figure 5, point 6). avoiding any flexional forces.
Due to the necessity of obtaining highly reliable results, In order to achieve this, a custom made support with a
the utilization of this device had an important role in the ball bearing mounted on the load application device was
validation experiments, by allowing us to obtain load cases utilized.
relatively close to the theoretical ones. For the application of the torsional moment a modified
Figure 6 presents a detailed image with the two M24 key was employed, on which the calibrated weights
constitutive parts of the device utilized for the applications were hung. Once again, all of the components positioning
was done with the laser level previously mentioned.
Finally for the C4 load case, an eye bolt mounted on the
center bolt of the load case application device was used,
along with a pulley system in order to obtain an axial force.
For both of the T1 and T2 junctions and each one of the
load cases, the main interest displacements were measured
utilizing mechanical dial gauges. A total of four dial gauges
were utilized: two of them for small displacements from 0-
5 mm with a resolution of 0.001 mm and the other two for
larger displacements from 0-25 mm with a resolution of
0.01 mm.
To avoid errors due to reading or measuring problems, a
semiautomatic reading system was implemented.
Also a quality criteria was defined, in which the typical
deviation of a series of measurements had to be less than
Figure 6. Device utilized for the application of the load 0.005 mm for the 0.001 mm resolution gauges and less
cases on the mechanical validation analysis. then 0.05mm for the 0.01 mm resolution gauges.
822 E. ALCALÁ, F. BADEA, Á. MARTIN and F. APARICIO

Figure 9. Rotations along the Y axis corresponding to the


C2 load case for the cantilever beam.
Figure 7. Alternative-optimized T-junction model.

structures.
On average for each interest point 12 measurements Therefore the necessity to conceive an improved model
were performed. for FEM modeled T-junctions with beam type elements
Complementary, on the surrounding areas around the represents an important issue.
junction, the displacements were measured with a digital The theoretical model proposed it is presented in Figure 7.
image correlation system. This system was utilized to This model is composed by three beam elements (1-3).
verify the gauges measurements and verify that the applied Two of the beam elements (1, 3) share a common node (4).
loads were accurate (pure axial loads). Between the last beam element (2) and the common node
Analyzing the data it has been determined that the real (4) there are a total of six elastic elements (k1…k6).
junctions have a more flexible behavior compared to the Three of the elastic elements (k1…k3), behave like linear
shell and volume models with about 2-5% (Figure 10, springs, working on the axial directions x, y, z.
Figure 11). These elastic elements behavior is given by Hooke’s law,
More detailed analysis will be presented in the results the spring constants are determined by the Equation (1).
chapter.
ki = ⎛⎝ ----i⎞⎠ ---- , i = 1…3
F N
(1)
di m
5. ALTERNATIVE (OPTIMIZED) T-JUNCTION
MODEL Where:
Fi –elastic element axial force corresponding to the x, y,
Comparing the behavioral characteristics of the beam z directions
modeled T-junctions with more complex shell and volume di – nodal displacement on the x, y, z directions.
models differences in rigidity up to 40% were seen (Figure The other three elastic elements (k4…k6), behave like
10). torsional springs and their constant are determined by the
Furthermore, between the shell and volume FEM Equation (2).
models and the experimental validation T-junctions rigidity
ki = ⎛⎝ ------i⎞⎠ -------- , i = 4…6
M Nm
differences up to 9% were observed. (2)
ϕi rad
Thus, the beam type element tubular structures can be up
to a significant 49% more rigid than the equivalent real Where:
Mi – spring torsional moment corresponding to the x, y,
z directions
ϕi – angle of rotation.
With this optimized model it is expected that, if the
estimations of the elastic element constants (k1…k6) are
made correctly, the behavior of the modeled T-junction will
be more realistic, and most of the influences and
characteristics of the junction will be described by the six
elastic elements.

5.1. Estimation of the Elastic Element Values


For the beam, shell and volume models, the average nodal
Figure 8. Displacement along the Y axis corresponding to displacements and rotations were obtained for each one of
the C1 load case for the cantilever beam. the six sections S1-S6 (Figure 3-4). A similar procedure
METHODOLOGY FOR THE ACCURACY IMPROVEMENT OF FEM BEAM TYPE T-JUNCTIONS 823

has been followed in the experimental validation in which graphics as a suffix on the element type legend.
the displacements were measured with the dial gauges in Figure 8 presents the displacement along the Y axis
the corresponding section. obtained for the cantilever beam with the C1 load case.
The reference force Fi and torsional moment Mi values Figure 9 presents the nodal rotations around the Y axis
had been selected taking into consideration practical issues obtained for the cantilever beam with the C2 load case.
from the mechanical validation. The values of the axial The behavioral differences that have been seen for the
forces and the torsional moment have to be big enough to nodal displacements and rotations between the different
produce measurable displacements in the T-junctions and type elements were around a 0.2%.
small enough to avoid the concentration of stresses above By analyzing the influence of the meshing dimensions, it
the elasticity limit. has been determined that for element sizes smaller than the
For the C1, C2 and C4 load cases, a force of Fi = 294.3 thickness of the profile, there are no significant differences
N (30 kg) was selected. For the C3 load case a torsional in the results.
moment Mi = 132.44 [Nm] (30 kg with a 0.45 m lever arm)
was selected. 6.2. Results Obtained for the T1 and T2 Junctions, through
By comparing the behavioral differences that exist FEM Simulations and Experimental Validation
between the beam models with the shell and volume The simulation performed for the T1 and T2 junction
models, the rigidity differences that exist between them can models had the objective of determining the differences
be quantified. If these differences approximate to zero that exist between the beam models compared to the shell
(similar behavior between models) the values of the elastic and volume models.
elements will tend to infinity. On the other hand, if the For the simulated situation, the variables L and L1
differences in displacements and rotations tend to infinity (Figure 4) have been selected equal to 1000 mm.
the elastic elements will tend to zero. Figure 10 presents the results obtained for the
Further details on the results and characteristics of the displacements along the Y axis corresponding to the T1
simulations for the cantilever beam and the T1, T2 junction and the C1 load case.
junctions are presented in the results chapter. On all of the mechanical experiments the relative
displacements of the embedment system had been
6. RESULTS measured and taken into consideration, this way all of the
presented results for the experimental validation represent
In the next paragraphs the results obtained for the the isolated behavior of the T-junctions.
cantilever beam simulations and the results obtained for the As was expected, the beam type element model has a
T1 and T2 junctions through FEM simulations and considerably more rigid behavior compared to the real T-
experimental validation will be presented separately. junction. For this particular configuration and direction, the
beam model is 27 % more rigid than the real T-junction.
6.1. Results Obtained for the Cantilever Beam Simulations Furthermore, by being able to reproduce the topological
The cantilever beam simulations were utilized to highlight characteristics of the joint, the shell and volume models are
any differences that might exist between beam, shell and able to provide more realistic results, in this situation they
volume type elements. are 9% more rigid than the real T-junction.
Because the thickness of the selected profile was equal Figure 11 presents the results obtained for the
to 3 mm, in order to obtain a good quality meshing displacements along the Y axis corresponding to the T1
simulations utilizing two meshing dimensions equal to 3 junction and the C1 load case.
and 1.5 mm were performed. The T1 and T2 junctions have different topologies at the
These dimensions are references in the presented joint level, by analyzing a wide variety of profiles

Fiugre 10. Displacement along the Y axis for the T1 Figure 11. Displacements along the Y axis for the T2
junction corresponding to the C1 load case. junctions corresponding to the C1 load case.
824 E. ALCALÁ, F. BADEA, Á. MARTIN and F. APARICIO

Table 1. Elastic element values corresponding to the


alternative (optimized) T-junction that adapts to the T1
junction characteristics.
Real Shell
Nr Type Axis Unit
junction model
1 Linear X 86515323 92037578 N/m
2 Linear Y 6971415 7416398 N/m
3 Linear Z 3076332 3272693 N/m
4 Torsional X 278242 296002 N.m/rad
5 Torsional Y 623586 663389 N.m/rad
Figure 12. Rotations about the Z axis for the T1 junction
corresponding to the C1 load case. 6 Torsional Z 112385 119559 N.m/rad

Table 2. Elastic element values corresponding to the


combinations for the junctions, was observed that generally alternative (optimized) T-junction that adapts to the T2
the T2 junction has a more rigid behavior than the T1 junction characteristics.
junction. These differences present a wide variation
Real Shell
depending on the sections of the profiles and the load case Nr Type Axis
junction model
Unit
situation.
Particularly, for the C1 load case, on the shell and 1 Linear X 15145173 14539366 N/m
volume models, there are rigidity differences of about 14%, 2 Linear Y 7416399 7119743 N/m
and about 16% for the real T-junctions. 3 Linear Z 5075966 4872927 N/m
The beam model has similar behavioral characteristics to
the T2 shell and volume models for this particular 4 Torsional X 72737 69827 N.m/rad
configuration and load case. Even then, these models are 5 Torsional Y 1013641 973095 N.m/rad
considerably more rigid than the real junction. 6 Torsional Z 3167989 3041269 N.m/rad
These characteristics repeat for the rest of analyzed load
cases, on all of the situations the FEM models have a more
rigid behavior than the real T-junctions. seen that for the alternative beam model there are
Figure 12 presents the rotations along the Z axis for the significant rigidity differences between the adaptation to
T1 junction corresponding to the C1 load case. the T1 and T2 junctions, particularly the value of the
Similar to the displacements results, the beam model has torsional Z-axis torsional elastic element for the T2
a more rigid behavior (21%), also there are no significant junction is 25 times more rigid than the corresponding one
differences between the shell and volume models and the for the T1 junction.
real T-junction present a less rigid behavior. Figure 13 presents the nodal displacements along the Y-
On each of the studied load cases, differences exist on axis for the T1 junction corresponding to the C1 load case,
the rigidity between the T1 and T2 junctions modeled with with the improved beam model.
shell or volume type elements similar to the experimental Fiugre 14 presents the rotations along the Z-axis for the
analysis. T1 junction along with the optimized T-junction model. In
By utilizing the alternative optimized T-junction model
presented in , the rigidity characteristics of the junction can
be adapted to a reference model by modifying the elastic
element rigidity values.
Therefore, the adaptation of the alternative beam
modeled T-junction can be made to either the shell and
volume models or to the real T-junctions.
Table 1 presents the values obtained for the elastic
elements that reproduce the behavioral characteristics of
the T1 junction adapted to the shell/volume models and to
the real junction behavior.
Table 2 presents the values obtained for the elastic
elements that reproduce the behavioral characteristics of
the T2 junction adapted to the shell/volume models and the Fiugre 13. Displacements along the Y axis for the T1
real junction behavior. junctions with the alternative optimized beam T-junction
Analyzing the results presented in tables 1, 2 it can be for the C1 load case.
METHODOLOGY FOR THE ACCURACY IMPROVEMENT OF FEM BEAM TYPE T-JUNCTIONS 825

Table 3. Numbers of elements obtained for a 1 m beam


profile with beam and shell elements.
Profile Mesh size 10 [mm] 5 [mm]
40×40×3 Beam 100 200
[mm] Shell 1600 6400

elements obtained for a 1 m beam profile modeled with


beam and shell elements.
On average the upper structure of an urban bus modeled
Fiugre 14. Rotations along the Z axis for the T1 junction entirely with shell type elements would have around
with the alternative optimized beam T-junction for the C1 912000 elements with a 10 mm meshing and 3648000
load case. elements with a 5 mm meshing. In contrast to the 57000
elements for a 10 mm meshing, and 114000 elements for a
5 mm meshing resulting from the beam type elements.
order to evaluate the computational cost derived from the Although it is possible to utilize meshing techniques
implementations of the six elastic elements at the joint allowing concentration of elements around the junctions
level, multiple simulations were performed. and regions of interest and larger elements elsewhere, the
resulting models still have large numbers of elements,
6.3. Comparative Analysis for the Modeling Alternatives affecting the utilization of these models. Also the utilization
With the methodology proposed, there are two main of these meshing techniques require supplementary efforts.
alternatives for the estimation of the elastic element values. On the other hand, with the beam type elements meshing
The fist one is based on mechanical experiments with real techniques with one or a few elements per modeled beam
T-junctions, and the second one is based on FEM modeling can be utilized, which will drastically reduce the total
with more detailed shell models. number of elements without compromising the result.
By estimating the elastic element values utilizing The simulation and postprocesing (3) aspects are closely
mechanical experiments more accurate results than those related to the number of elements that the models have.
provided by the more complex shell or volume models can Table 4 presents a comparison of the solution times for a T-
be obtained. The utilization of mechanical experiments junction modeled with shell, regular beam and the
require considerable amounts of time and high costs which alternative beam elements
make them unattractive as a widely utilized methodology. The simulation time required for the T-junction modeled
By utilizing the shell or volume models, the regular beam with shell elements is 53 times higher than the required
T-junctions can still be importantly improved, this process time for beam model.
requires significantly less time and costs. Utilizing the alternative beam element implies a 17%
Adapting the alternative beam model to the more increase on the simulation time. This time increase
detailed shell models presents a series of advantages represents a minor shortcoming compared to the potential
compared to the option of modeling the buses and coaches benefits that can be obtained from the alternative optimized
upper structures entirely with shell elements. In order to beam model.
evaluate these advantages it is necessary to take into When optimizing buses and coaches upper structures,
consideration the time and cost of a series of processes the analysis commonly requires the modification (4) of
during the analysis of these structures: modeling (1), certain profiles in order to evaluate various characteristics
meshing (2), simulation-postrocesing (3) and modification and configurations of the structures. When the structures
of the models (4). are modeled with shell type elements these modification
The modeling process of tubular structures (1) with shell are more complex and time consuming to perform.
type elements compared to the beam elements, presents a The implementation of the alternative beam model for
series of inconveniences due to the fact that it requires buses and coaches upper structures allows the creations of
modeling the complete topology and adjacent connections
for every modeled junction. Also, taking into consideration Table 4. Solving time comparison between the shell,
the fact that on the buses and coaches upper structures can regular beam and alternative optimized beam T1 junctions
be found on average 520 junctions, the amount of for all of the C1-C4 load cases.
necessary work multiplies significantly.
Reg. Alt. Beam
Related to the meshing (2) and simulation aspects, an Shell
beam beam
Shell Beam
Alt Beam
important factor is given by the number of elements that [s] Diff [%]
[s] [s] Diff [%]
result from modeling these structures with shell elements.
76.12 12.04 14.10 532 17
Table 3 presents a comparison between the numbers of
826 E. ALCALÁ, F. BADEA, Á. MARTIN and F. APARICIO

by the rigid behavior of the joint on the beam models


altogether with incapability of the beam models to reproduce
the characteristics of the joints.
Also, from the experiments and simulations performed it
has been observed that volume and shell elements have the
possibility to reproduce the topology and behavioral
characteristics of T-junctions in a more realistic way.
Still, the use of these elements as an alternative for the
modeling of tubular structures implies a series of important
shortcoming mainly determined by the complexity of the
modeling processes and the high computational resources
required.
By adapting the alternative beam model to the real T-
junctions, a realistic and accurate model can be obtained,
although the complexity and the derived costs of the
mechanical experiments make this methodology unfeasible,
especially considering that there are a wide variety of square
and rectangular hollow profiles available, commonly utilized
in the buses and coaches fabrication industry.
However, adapting the alternative beam T-junctions to
the shell and volume models represents a more feasible
situation, in which a beam model with a more realistic
behavior at a relatively reduced cost can be obtained.
Although the adaptation to the shell and volume models
Figure 15. Block diagram of the semi-automated process
still presents some shortcomings compared to the real
for the estimation of the elastic element values.
models, it provides an important improvement for the
simplified beam junctions, for some configurations these
rigidity differences proven to be as high as 40%.
a database with different profile configurations. The Another important advantage derived from the
automation of the process presents multiple advantages and utilization of the shell and volume type elements is
can be easily done by parameterizing the dimensional determined by the simplicity of automating the process
characteristics of the T-junctions and implementing a series which will offer the possibility of obtaining a database for
of routines. Block diagram of the semi-automated process the alternative beam T-junctions.
for the estimation of the elastic element values. Figure 15 The implementation of the optimized junctions on an
presents the block diagram for the estimation of the elastic already existing beam modeled structure can be easily done
element values with an automated process. by a semi-automate process, in which the node of a
The authors are currently working on the characteriza- determined junction is replaced by a group of six elastic
tion of the behavioral characteristics for the commonly elements. This process does not require a topological
utilized profiles in the buses and coaches upper structures. modification of the existing model, therefore can be
The final intention being the development of a completed in a reduced amount of time.
mathematical model for the elastic elements based on the Finally, it can be concluded that this methodology
dimensional characteristics of the profiles. This way the represents an acceptable alternative to obtain better and
elastic element values will be obtained without the improved models for large tubular structures.
necessity of utilizing modeling techniques. Obtaining a better or worse equivalent optimized beam
T-junction model depends entirely in the quality and
7. CONCLUSION capacity of the estimation for the six elastic elements.
For the estimation of the values of the elastic elements
Modeling large tubular structures with beam type elements other methodologies can be utilized.
represents an extremely utilized methodology among FEM For future applications the estimations of the elastic
analyst mainly determined by the simplicity of the elements will be performed through modal vibration
modeling processes along with the reduced computational analysis.
performances for relatively large models.
Analyzing the results obtained for the cantilever beam and ACKNOWLEDGEMENT−To the “Consejería de Educación y
the T1 and T2 junctions, it has been concluded that the Empleo” of the Community of Madrid for all of their support and
behavioral differences that appear between the beam model involvement in the development of all of the research activities
and the equivalent shell and volume models, are determined that lead to the development of this article. To the European
METHODOLOGY FOR THE ACCURACY IMPROVEMENT OF FEM BEAM TYPE T-JUNCTIONS 827

Research Project OPTIVIRTEST (TRA200914513) for the structure. Int. J. Automotive Technology 11, 1, 41−47.
support among the research activities of this article. Kim, H., Hwang, Y. and Yoon, H. (2001). Dynamic Stress
Analysis of a Bus Systems. Commercial Vehicle
REFERENCES Engineering & Research Center Hyundai Motor
Company.
Adams, V. and Askenazi, A. (1998). Building Better Kim, M., Suh, H. and Bae, D. (2001) Development of an
Products with Finite Element Analysis. OnWord Press optimum design technique for the bus window pillar.
Santa Fe. ISBN 1-56690-160X. Proc. Instn. Mech. Engrs, Part D: J. Automotive
Arribas, D., Badea, F. and Pérez, J. (2012). Análisis y Engineering 215, 1, 11−20.
optimización estructural de autobuses mediante modelos Lan, F., Chen, J. and J. Lin (2004). Comparative analysis
matemáticos. X Congreso de Ingeniería del Transporte. for bus side structures and lightweight optimization.
ISBN 978-84-338-5402-5. Proc. Instn. Mech. Engrs, Part D: J. Automotive
Badea, F., Alcalá, E., Grimaldi, R. and Aparicio, F. (2010). Engineering, 218, 1067−1075.
Optimización de uniones y estructuras de autobuses. Lee, S., Baik, B. and Yim, S. (2003) Optimal reliability
XVII Congreso Nacional de Ingeniería Mecánica, ISSN design for thin-walled beam of vehicle structure
0212-2072. considering vibration. Int. J. Automotive Technology 4,
Balázs, G. (2005). Dynamic analysis of a bus body frame: 3, 135−140.
Determination of the loads and stresses. Vehicle System Lee, M., Pine, T. and Jones, T. (2000). Automotive box
Dynamics 43, 11, 807−822. section design under torsion. Part 1: Nite element
Bosch R. GmbH (2007). Automotive Handbook. 7th edn. modelling strategy. Proc. Instn. Mech. Engrs, Part D: J.
Bentley Publishers Massachusets. ISBN 978-0-8376- Automotive Engineering 214, 4, 347−359.
1540-0. Liang, C. and Le, G. (2010). Analysis of bus rollover
Dezso, S. and Sandor, P. (2010). Stress calculation and protection under legislated standards using LS-DYNA
estimation of bus structure in case of transient software simulation techniques. Int. J. Automotive
displacement excitation. FISITA World Automotive Technology 11, 4, 495−506.
Cong. 2010, Budapest, Hungary. Milojevic, S. and Pesic, R. (2011). Theoretical and
Eriksson, P. (2001). Optimization of a bus body structure. experimental analysis of a CNG cylinder rack connection
Heavy Vehicle Systems 8, 1, 1−16. to a bus roof. Int. J. Automotive Technology 13, 3, 497−
García, A. and Vicente, T. (2002). Caracterización e Influencia 503.
de las Uniones Semi-rígidas en el Comportamiento Zienkiewicz, C. and Taylor, L. (2002). The Finite Element
Estructural de Autobuses y Autocares. Cádiz. ISBN 84 607 Method for Solid and Structural Mechanics. Butterworth-
6264. heinemann. New York. ISBN 0-7506-6321-9.
Gauchia, A., Diaz, V., Boada, L. and Boada, B. (2010).
Torsional stiffness and weight optimization of a real bus

You might also like