Documente Academic
Documente Profesional
Documente Cultură
To cite this article: Cho-Chung Liang & Giang-Nam Le (2009) Bus rollover crashworthiness under European standard:
an optimal analysis of superstructure strength using successive response surface method, International Journal of
Crashworthiness, 14:6, 623-639, DOI: 10.1080/13588260902920670
Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
International Journal of Crashworthiness
Vol. 14, No. 6, December 2009, 623639
Bus rollover is one of the most serious of accidents. Strengthening bus frames to maintain survivor space and reduce occupant
injury is necessary following the issue of Regulation No. 66 by the Economic Commission for Europe (ECE R66). Whilst
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
increasing bus weight is unlikely because of production costs and fuel economy, this paper presents an optimal method of
bus rollover crashworthiness design. In this study a full-scale, validated, nite element (FE) model of the vehicle was used.
Optimisation was performed by the successive response surface method (SRSM) with LS-OPT, a design variable analysis
method based on the parameterisation of the energy absorption ability of bus frame components. LS-DYNA was used as the
FE solver. An optimal prototype of the vehicle was obtained with crashworthiness following ECE R66 and vehicle weight
at the existing level was maintained with the improvement in the lower displacement by 49.2% and upper displacement by
39.4% of bus frames versus bus survivor space. This paper presents a procedure for bus rollover crashworthiness design
related to vehicle weight, with a robust and effective method using an optimal technique combining LS-DYNA and LS-OPT.
Keywords: bus rollover; energy absorption; survivor space; LS-DYNA; LS-OPT; ECE R66
Corresponding author. Email: ccliang@mail.dyu.edu.tw
ISSN: 1358-8265
Copyright
C 2009 Taylor & Francis
DOI: 10.1080/13588260902920670
http://www.informaworld.com
624 C.-C. Liang and G.-N. Le
Table 1. Motorcoach crashes and fatalities by most harmful 3-D, dynamic FE computer code LS-DYNA. The calcu-
event of 48 crashes and 146 fatalities (FARS 19962005). lation technique has been checked by the verication of
calculation tests applied to a breast knot of the side-body
Motorcoach crashes Motorcoach fatalities
and on a roof-edge knot of the vehicle, and subsequent
Event Quantity Rate (%) Quantity Rate (%) numerical simulations were performed. A high degree of
theoretical and experimental correlation is obtained, which
Rollover 14 29 49 34
Roadside 15 31 53 36 conrms its validity [6]. With an assessed method in the
Multi-vehicle 17 36 20 14 ECE R66, a complete vehicle rollover test simulation was
Other 2 4 24 16 carried out and the deformation results with respect to the
residual space were observed. It infers the structural status
of the bus following the required regulations. This numer-
strength for lightweight vehicles, only a few studies have in- ical study also inferred the absorbed energy distribution
vestigated the optimum design of bus superstructure based throughout the whole bus, and then found and focused on
on ECE R66. Only Lin and Nian [17] have carried out re- the highest energy absorption region. The investigation of
search relating to ECE R66. That study, however, is still the absorbed energy distribution there and reorganisation
limited to the body section. Thus, this paper will present an of the energy absorption ability of that region produce the
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
optimal structure for the whole bus based on the ECE R66 set of bus frame thickness parameters for optimal analysis
process and SRSM using LS-DYNA for the FE solver and of the design. Optimal analysis with LS-OPT, a prototype
LS-OPT for optimisation analysis. of bus with optimised thickness, was found to satisfy ECE
In this paper, an ECE R66 calculation procedure per- R66.
formed for a certain bus is described. This bus is 12.6 m This paper presents an optimisation technique based
long (see Figure 1) and is constructed in a steel material on SRSM and LS-OPT procedure where design variables
with special reinforcements of bars and places on the roof- are obtained from the reorganisation manner of the en-
side and side-oor joins where large displacements happen ergy absorption ability on the highest energy absorption
[20]. The FE modelling is done by the pre-processing Finite region of the bus superstructure. We found improvement of
Element Builder (FEMB) of LS-DYNA software, and cal- bus frames in the upper displacement by 39.4% and lower
culations are made by the means of a non-linear, explicit, displacement by 49.2% versus the bus survivor space while
maintaining the bus weight at the existing level. This is a above the oor under the seat, 150 mm and 250 mm from
contribution to the automotive industry in an effort to re- the inside surface of the side wall [11, 12].
duce the production cost and the occupant injury through
passive safety of the rollover event.
2.2. Rollover test
This regulation is continuously updated on the basis of
2. ECE R66 regulation actual requirements and it is used as an international bus
The ECE R66 regulation was issued on 30 January 1987 rollover regulation. The current version was issued on 22
and enforced by the Economic Commission of Europe, in February 2006. The rollover test is a lateral tilting test, as
response to the serious problem of rollover accidents. It ap- shown in Figure 3. The complete vehicle stands on a tilting
plies to all single-decked vehicles constructed for carrying platform with blocked suspension and is tilted slowly to
more than 22 passengers, whether seated or standing, in ad- an unstable equilibrium position. If the vehicle type is not
dition to the driver and the crew. Superstructure refers to tted with occupant restraints, it will be tested at an unladen
the parts of a vehicle structure that contribute to the strength curb mass. If the vehicle is tted with occupant restraints, it
of the vehicle in the event of rollover accidents. will be tested at the total effective vehicle mass. The rollover
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
Figure 2. The residual space, measured in millimetres (mm) [22]. Reprinted with permission.
626 C.-C. Liang and G.-N. Le
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
r The asymmetry of the residual space scenario until they obtain a safe and optimal design while
r The different asymmetrical construction features of the saving time and money in developing costly prototypes.
two sides of the vehicle, and the support given by the The analysis processes are as follows:
partition or inner boxes (e.g. wardrobe, toilet and kitch-
enette) 1. Constructing the testing model from a full-scale bus
model and tilting platform model;
The side with less support is chosen as the direction of 2. On the basis of testing conditions, setting up the material
the rollover test. card, boundary condition card, contact card and applying
load on this testing model;
3. Determination of the already tilting angle to reduce com-
2.3. Test methods puting time;
The latest version of ECE R66, version 2006, detailing the 4. Use of numerical analysis software to carry out the sim-
above requirements, describes a test to be chosen from these ulation of a rollover accident of this testing model;
ve different methods: 5. Evaluation of the status of the testing model after nish-
ing rollover based on the obtained simulation results.
1. Complete vehicle rollover test
2. Body section rollover test
3. Body section test with quasi-static load 3. Computational model
4. Component testing based on quasi-static calculation The FE vehicle model used for simulation is based on the
5. Complete vehicle rollover test based on computer sim- full-scale bus model developed at Da-Yeh University, Tai-
ulation wan, for rollover crashworthiness investigation and evalu-
ation of reinforcement structures [46]. It included 69,155
Method (1) was accepted as the standard method. Others elements. Those comprise 69,056 of two types of element
are equivalent methods. Amongst them, methods (3) and shells, 68,142 quadrilateral elements, 914 triangular ele-
(4) are new methods in ECE R66, version 2006. Methods ments and 99 mass. All deformable parts were modelled
(1)(3) are experimental methods based on the real test. with the 4-noded Belytschko-Tsay shell elements with
Method (5) is ofcially accepted with full-scale computer three integration points through the shell thickness. The
simulation [12]. In this paper, method (5) is used to perform shell elements formulation is based on Belytschko-Lin-
the numerical analysis. Tsay formulation with reduced integration available in LS-
DYNA [9, 18]. This element is generally considered to be
computationally efcient and accurate. The shell element
2.4. Computer simulation of a rollover test that has been, and still remains, the basis of all crashwor-
Computer simulation of a rollover test on a complete vehicle thiness simulation is the 4-noded Belytschko-Tsay shell.
is an equivalent approval method. It allows manufacturers The centre of gravity (CG) of the vehicle was measured
to test designs and safety features virtually in the crash by means of a test platform at the Automotive Research &
International Journal of Crashworthiness 627
Testing Centre, Taiwan, R.O.C (ARTC). The measured val- pillar knot (Figure 4c), which were extracted from the ve-
ues were in good agreement with the ones coming from hicle superstructure. Those three separate specimens were
the FEA model. To match the measures and calculated subjected to certain boundary conditions and quasi-static
CG exactly, the CG of the engine, gearbox and axles was loads at ARTC. The same test scenarios were simulated by
ne-tuned in the FEA model. The unloaded vehicle weight the use of LS-DYNA. Forcedeection curves for both the
is 7716.5 kg (7.7165 tons) and its capacity is 49 passen- experiment and simulation were compared and a good cor-
gers. The vehicle size and its position of CG are shown in relation was seen between the experiment and simulation
Figure 1. results (Figure 4).
The FE modelling was done by the FEMB of LS-DYNA,
and calculations were made by means of a non-linear, ex-
plicit, 3-D, dynamic FE computer code LS-DYNA. The 3.1. Survivor space denition
calculation technique was checked by the verication of For estimation of ECE R66 requirements, the survivor
calculation tests applied to a breast knot of the side-body space was specied in FEMB in line with the statement
(Figure 4a), on a roof-edge knot (Figure 4b) and on a oor- in ECE R66. Throughout the whole vehicle, the SR points
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
Figure 4. Tests, simulations and results of the body knot for model validation.
628 C.-C. Liang and G.-N. Le
are located on the seatback of each outer forward or rear- the surface of another body. This contact type is easy to
ward facing seat, 500 mm above the oor under the seat use because no contact or target surface denitions are
and 150 mm from the inside surfaces of the side walls of required. It is efcient for self-contacting problems or
the vehicle. The model of the survivor space consists of large deformation problems where general areas of con-
a rigid shell frame in each section along the vehicle inte- tact are not known beforehand. This contact was used for
rior (Figure 2), rigidly mounted in the stiff region under simulating the contacting mode of the vehicle and the
the oor. There is no stiff connection between these rigid tilting plate in rollover tests. The contact called CON-
shell frames because these shell elements are modelled with TACT AUTOMATIC SURFACE TO SURFACE is estab-
NUL MATERIAL for visualisation only. lished when a surface of one body penetrates the surface
of another body. This is commonly used for arbitrary bod-
ies that have large contact areas and is very efcient for
3.2. Material and contact models bodies that experience large amounts of relative sliding
To obtain the material data, tension tests were applied with friction, such as block sliding on a plane. This con-
to several specimens at ARTC facilities [6]. The true tact was used for simulating the contacting mode of the
stressstrain curves were obtained and imposed in LS- bus frame with some reinforcements. The contact called
CONTACT RIGID WALL PLANAR was used to set the
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
models; the large library of element types; and the special divided into two categories: topology optimisation tools
implementation for the automobile industry [9, 18]. and parameter optimisation tools. Topology optimisation
LS-OPT is a stand-alone optimisation software and can determines an optimal distribution of material with respect
be linked to any simulation code. It is perfectly suitable to given package space, loads and boundary conditions.
in combination with LS-DYNA, and provides a clearly Parameter optimisation tools nd optima for scalar values,
arranged graphical user interface (GUI) for denition of e.g. sizing values like diameter or sheet metal thickness.
optimisation and stochastic analysis problems. LS-OPT is Changes of the shape of a mechanical structure can also be
available for Linux, Windows and some UNIX platforms quantied with scalar values. Hence parameter optimisation
[21]. is suitable for sizing and shape optimisation. LS-OPT is a
Numerous design optimisation software tools have been favourite tool for this problem.
introduced during the last few years. These tools can be
4.2. LS-DYNA solution procedure For non-linear problems solved by explicit solutions,
The numerical study is based on the LS-DYNA FEA. The a lumped mass matrix is required for simple inversion.
LS-DYNA software contains a pre-processing FE builder, The equations become uncoupled and can be solved di-
an LS-DYNA solver and a post-processing LS-PREPOST rectly (explicitly). No inversion of the stiffness matrix is
[9, 18]. With LS-DYNA, the standard input such as geom- required. All nonlinearities (including contact) are included
etry, mesh density, materials, element properties, boundary in the internal force vector. The major computational ex-
conditions and contact modes can be used. The LS-DYNA pense is in calculating the internal forces. No conver-
solver will perform solutions. The output results such as gence checks are needed since the equations are uncoupled.
stress and strain of elements, displacement, velocity and Very small time-steps are required to maintain the stability
acceleration of nodes, and energy distribution, etc., can limit.
be shown clearly through the user interface [9,18]. The
main solution is based on explicit time integration. The
LS-DYNA solution procedure is shown in Figure 5. 4.3. ECE R66 simulations
The explicit method was originally developed, and is The testing model is established by the full-scale bus model
primarily used, to solve dynamic problems involving de- as shown in Figure 1 and the tilting platform model as shown
formable bodies. Accelerations and velocities at a particu- in Figure 6. According to ECE R66, the initial condition
lar point in time are assumed to be constant during a time is that the angular velocity of the tilting platform must not
increment and are used to solve for the next point in time.
For the explicit method, a central difference time integra-
tion method is used. The word implicit here refers to the
method by which the state of an FE model is updated from
time t to t + t. A fully implicit procedure means that the
state at t + t is determined according to information at
time t + t, while the explicit method solves for t + t
according to information at time t [1, 9, 18].
With the explicit FE method, the solver equations can
be solved directly to determine the solution without iter-
ation, thus providing an alternative, more robust method.
Harewood and McHugh [10] showed that for simpler load-
ing conditions the implicit method had a shorter solution
time. In the case of loading conditions involving contact,
the explicit method proved to be the preferred choice.
Moreover, the explicit method displayed constantly high
levels of parallelisation efciency compared with the im-
plicit method for analyses solved with multiple processors. Figure 9. Principle of creating an approximation with RSM.
International Journal of Crashworthiness 631
Table 3. Energy absorption of the vehicle and vehicle section. Table 4. Energy absorption of the side wall section and its
components.
Followed ECE R66
Absorbed energy, Weight, Ability,
Vehicle and section Ea (kJ) Rate (%) Components Ea (KJ) m (kg) A (kJ/kg)
Whole vehicle 63.08 100 Side wall section 37.86 445.10 85.06E-3
Front section 5.69 9.02 WDP group 9.47 37.35 253.55E-3
Rear section 7.10 11.26 SWP group 9.90 50.68 195.30E-3
Roof section 12.12 19.21 Others 18.49 357.07 51.79E-3
Side wall section 37.86 60.02
Others 0.31 0.49
Figure 11. Relevant points on the bus frame for ECE R66 test.
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
energy to stop the vehicle rollover. The side wall section wall section are the major components and have the highest
and the roof section are the most important regions of the energy absorption ability as shown in Table 4. Therefore,
superstructure that will absorb those energies, as shown the thicknesses of the WDPs and SWPs are the objects
in Figure 7. The absorbed energy rate of each component considered for design variable analysis of optimal design.
is displayed in Table 3, where the side wall section will
be especially considered because of its energy absorption
behaviour which obtained the biggest energy absorption 5. Optimisation of superstructure strength
(60.02%) of the whole vehicle. following ECE R66
5.1. Successive response surface method
Response surface method (RSM) is a statistical method
4.4.2. Distortion conguration of side wall section fol- for constructing smooth approximations of the objective
lowing absorbed energy function in the multi-dimensional parameter space. For
The side wall sections were constructed by window pil- illustration purposes, see Figure 9. So-called experimen-
lars (WDPs), and the side truss included side wall pillars tal design points (variable sets) are selected within a pre-
(SWPs), diagonal bars and longitudinal bars, as seen in dened design space. For these points, the dened model
Figure 8. The absorbed energies of the side wall sections responses are calculated. In a subsequent step, polynomial
major components under rollover conditions are displayed functions are tted to these experimental design points in or-
in Table 4, where the energy absorption ability (A) is the der to substitute for the original response. The tting of the
components absorbed energy over its component weight, polynomial function is done by the use of regression anal-
Ai = Eai /mi . The high A caused high sensibility of the ysis. Least-squares approximations are commonly used for
energy absorption. Thus, the WDPs and SWPs of the side this purpose. In the last step, an optimum point (minimum
where
xi is the vector of design variables including thicknesses of
seven WDPs and nine SWPs,
W0 is the unvarying weight of vehicle while considering 16
design variables,
Wi is the weight constant of each considered part and
y1k , y2k are constraint functions of the side wall displace-
ments for upper and lower space versus survivor space
at concerned points, as shown in Figure 11.
Part i Eai /mi Ai /Amin Ai /Ki Eai /Ai mi /p Si Vi /Si ti /ti DV ki
i = 1n = Eai /(Ai /Ki )
= (Eai /Ai )Ki
= mi Ki
Note: Ea , absorbed energy; m, weight of component part; A, energy absorption ability; K, relation factor; , material density; k, thickness parameter;
Amin , Min {Ai , i = 1n}; n, number of concerned parts in each group; m , weight of component part after reorganisation; A , energy absorption
ability after reorganisation; DV, design variable for concerned parts group; V , S , t volume, square area and thickness for re-design.
energy absorption ability of each member so that all group tion ability and the way to parameterise to obtain de-
members, after reorganisation, would be able to achieve sign variables for the relevant groups are summarised in
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
the same energy absorption ability as the lowest one. The Table 5.
review of geometric conguration and characteristics of With p as design variable for the WDP group and i =
the shell element displayed are shown in Figure 12, and 1, 2, . . . , 7 for seven WDPs, the results obtained for the
the methodology of the reorganisation of energy absorp- WDP group are summarised in Table 6. With q as design
Reorganisation Design
Pre-design Energy absorption of energy variable (p) and
(Bi , Li , Hi , ti ) ability (A) absorption ability Re-design parameterisation
WDP- Vi mi Ai mi
group (m3 ) Si (m )2
ti (m) Eai (kJ) (kg) (kJ/kg) Ki Ai (kg) Vi (m3 ) Si (m2 ) ti (m) ki ti (m)
WDP1 6.70E-4 3.35E-1 2E-3 8.76E-1 5.25 0.167 1.00 0.167 5.24 6.70E-4 3.35E-1 2.00E-3 1.00 p k1
WDP2 6.78E-4 3.39E-1 2E-3 9.84E-1 5.31 0.185 1.11 0.167 5.89 7.53E-4 3.39E-1 2.22E-3 1.11 p k2
WDP3 6.78E-4 3.39E-1 2E-3 1.18E+0 5.31 0.221 1.33 0.167 7.04 8.99E-4 3.39E-1 2.65E-3 1.33 p k3
WDP4 6.78E-4 3.39E-1 2E-3 1.19E+0 5.31 0.224 1.34 0.167 7.13 9.10E-4 3.39E-1 2.68E-3 1.34 p k4
WDP5 7.16E-4 3.58E-1 2E-3 1.67E+0 5.61 0.298 1.79 0.167 10.00 1.28E-3 3.58E-1 3.57E-3 1.79 p k5
WDP6 6.78E-4 3.39E-1 2E-3 2.40E+0 5.31 0.451 2.70 0.167 14.40 1.83E-3 3.39E-1 5.41E-3 2.70 p k6
WDP7 6.70E-4 3.35E-1 2E-3 1.17E+0 5.25 0.223 1.34 0.167 7.02 8.97E-4 3.35E-1 2.68E-3 1.34 p k7
Re-organisation
Pre-design Energy absorption of energy Design
(Bj , Lj , Hj , tj ) ability (A) absorption ability Re-design variable (q) and
parameterisation
SWP- Vj mj Aj mj Vj
group (m3 ) Sj (m2 ) ti (m) Eaj (kJ) (kg) (kJ/kg) Kj Aj (kg) (m3 ) Sj (m2 ) tj (m) kj tj (m)
SWP1 1.17E-3 5.87E-1 2E-3 01.50E+0 09.19 0.163 1.37 0.119 12.59 16.08E-4 5.87E-1 2.73 1.36 q k1
SWP2 4.52E-4 2.26E-1 2E-3 01.32E+0 03.54 0.373 3.14 0.119 11.11 14.18E-4 2.26E-1 6.25 3.13 q k2
SWP3 4.30E-4 2.15E-1 2E-3 88.43E-2 03.36 0.263 2.21 0.119 7.43 09.49 E-4 2.15E-1 4.40 2.20 q k3
SWP4 4.30E-4 2.15E-1 2E-3 97.21E-2 03.36 0.289 2.43 0.119 8.17 10.43E-4 2.15E-1 4.84 2.42 q k4
SWP5 4.30E-4 2.15E-1 2E-3 86.40E-2 03.36 0.257 2.16 0.119 7.26 09.27 E-4 2.15E-1 4.30 2.15 q k5
SWP6 1.14E-3 5.68E-1 2E-3 01.26E+0 08.89 0.142 1.19 0.119 10.58 13.51E-4 5.68E-1 2.37 1.18 q k6
SWP7 6.76E-4 3.38E-1 2E-3 01.20E+0 05.29 0.227 1.91 0.119 10.09 12.88E-4 3.38E-1 3.80 1.90 q k7
SWP8 6.26E-4 3.13E-1 2E-3 84.80E-2 04.90 0.173 1.45 0.119 7.13 09.10E-4 3.13E-1 2.90 1.45 q k8
SWP9 11.27E-4 5.61E-1 2E-3 01.05E+0 08.79 0.119 1.00 0.119 8.82 11.27E-4 5.61E-1 2.00 1.00 q k9
variable for the SWP group, and replacement of i dummy Solution was reached by the use of LS-OPT and LS-
by j dummy with j = 1, 2, . . . , 9 for nine SWPs, results DYNA combination for optimisation analysis; the optimi-
obtained for this group are summarised in Table 7. The sation method is SRSM, as presented in Section 5.1.
multi-variable problem is, therefore, made simple and is
now a two-variable optimisation problem, formulated as
5.4. Optimisation process
follows:
The optimisation process of LS-OPT is shown in Figure 13.
Minimise: F (x) = W0 + Wp p + Wq q, Preparation of inputs included design variables, range of de-
Subject to: y1k (p, q) 150 0, k = 1, 2, . . . , 8, sign variables, FE solver, parameterised FE le for gener-
y2k (p, q) 400 0, k = 1, 2, . . . , 8, ating the inputs of simulation jobs automatically inside the
optimisation process as well as sampling type of design of
experiment (DOE), constraint and objective functions, etc.
where
All of them were managed by the LS-OPT main menu, as
shown in Figure 14. Program les for automatic execution
p, q are the design variables for the WDP and the SWP included COM, LSOPT INPUT and LSOPT DB les that
groups, respectively, are automatically generated for standard input of LS-OPT
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
W0 is the unvarying weight of the vehicle while considering processing from preparation data. The output of the opti-
the WDPs and the SWPs, misation process was an LSOPT REPORT le. Other ways
Wp is the weight constant with respect to parts in the WDP of output can be displayed in the viewer tab of LS-OPTs
group, main menu.
Wq is the weight constant with respect to parts in the SWP
group and
y1k , y2k are constraint functions of the side wall displace- 5.5. Verication of design
ments for upper and lower space versus survivor space Figure 15 displays the optimisation history of the objec-
at concerned points, as shown in Figure 11. tive. Optimal value is obtained after the seventh iteration at
The ndings of the optimal design of bus superstructure [16] and Kim [13, 14], investigates the whole bus. Lan et al.
were a set of optimal thicknesses of the WDPs and the SWPs [16] stated the lightest possible weight of bus structure
with the design variable p = 2.15974, q = 1.89493 for a by removal of the supporting bars and optimisation of the
bus prototype satisfying the ECE R66 (Tables 6 and 7). The side frame thickness by maintaining torsional rigidity and
whole side wall section was investigated and multi-design fundamental torsional frequency. Kim [13 ,14] observed
variable problem was simplied by the absorbed energy maximum stiffness involving torsional stiffness and bend-
relevant as shown in Table 5. The different thicknesses of ing stiffness of full structure while maintaining the weight at
each WDP and SWP imply that the optimal design of a the same level as the specic bus. Neither of them, however,
bus frame is relevant not only to the weakest body section stated that the studied bus structure satised the rollover
because the ability of the energy absorption is different in safety standard ECE R66, the strict requirement of a bus
each component. This study, as well as that of Lan et al. superstructure.
638 C.-C. Liang and G.-N. Le
Downloaded by [Kungliga Tekniska Hogskola] at 19:40 08 October 2014
Figure 20. Upper displacement of bus frame: optimisation versus Figure 21. Lower displacement of bus frame: optimisation versus
original. original.
International Journal of Crashworthiness 639