Content uploaded by Caroline Deck
Author content
All content in this area was uploaded by Caroline Deck on Aug 25, 2015
Content may be subject to copyright.
Special Issue Article
Proc IMechE Part P:
J Sports Engineering and Technology
226(3/4) 282–289
ÓIMechE 2012
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1754337112442326
pip.sagepub.com
Head impact conditions in the case of
cyclist falls
Nicolas Bourdet
1
, Caroline Deck
1
, Rui P Carreira
2
and Re
´my Willinger
1
Abstract
In 2009 approximately half of the French population owned a bicycle. However, the cyclist’s accident rate is the highest
of all road users. Hence, it is necessary to set up a protection system for cyclists, especially for the cephalic segment.
Currently, relatively little literature has dealt with the head impact condition for this kind of accident, especially for cases
of cyclist falls. Therefore, the objective of this work was to identify the initial conditions for head impact in cases of
cycling fall accidents. The present paper proposes a parametric study by simulating a lot of accident scenarios. A total of
1024 simulations have been automatically carried out using Madymo
Ò
’s software and a specially designed program. Two
situations of cycling falls have been investigated according to real accident configuration: cyclist falls due to skidding and
after hitting a curb. The TNO (Dutch Organization for Applied Scientific Research, Nederlandse Organisatie voor
Toegepast Natuurwetenschappelijk Onderzoek) pedestrian 50th percentile human model was coupled to a city bicycle
model. The analysed outputs are head impact area and head velocity before impact. The work has provided a solid infor-
mation base for future work on cyclist accidents. The parametric analysis was also used to study the effects of poorly
known environmental parameters, such as speed or torso inclination. The results provided estimates of the impact area
and speed of the head, which helps to improve the design of safer helmets and of helmet certification standards tests.
Keywords
Cycling fall, head impact, multibody simulation, parametric study
Date received: 15 December 2011; accepted: 24 February 2012
Introduction
The cyclist accident rate remains high among all road
users. According to Linn et al.’s study in 1998,
1
the
three segments that are most often hit during cycling
accident are the arms, the legs and the head, with 50%
of the cases. Moreover, Otte
2
showed that 70% of fatal
cyclist accident cases are due to head injuries.
Maimaris et al.
3
and Eilert-Petersson and Schelp
4
reported that most of the cycling accident configura-
tions were single falls. In contrast, Yang and Otte
5
and
Got and Got
6
have shown that most of the cyclist acci-
dents happened against antagonist vehicles. Finally,
Depreitere et al.
7
showed that cycling accidents resulted
from falls and from collisions with vehicles in the same
proportion. According to Ricard and Thelot
8
victims
from cycling falls make up the majority of the injured
population received in emergency (89%).
Currently, there is little information available on the
head impact conditions for this kind of accident, which
mainly concerns windscreen impact. In 2000 and 2003,
Maki et al.
9
and Maki and Kajzer
10
studied the pedes-
trian and cyclist kinematics according to the vehicle
geometry. It appeared that the impact zones, the
impact velocity and angle were different between
cycling accidents and pedestrian accidents. Hence, the
authors concluded that the existing pedestrian tests
were not applicable to cyclists. In 2001, Werner et al.
11
studied the human–bicycle decoupling. He showed that
for off-centre impact, or a bicycle velocity over 5 m/s,
the cyclist was often thrown to the ground without
touching the car. Several authors, such as Serre et al.,
12
have reconstructed many accidents between car and
cyclists in order to identify the cyclist kinematics during
impact. However, cyclist falls that occur on their own
have not been studied because of the wide range of pos-
sible scenarios and the difficulty of categorising them.
Therefore, the objective of the present work is to
1
Strasbourg University, France
2
Oxylane Research Villeneuve d’Ascq, France
Corresponding author:
Nicolas Bourdet, Strasbourg University, 2 rue Boussingault, Strasbourg
F67000, France.
Email: nicolas.bourdet@unistra.fr
evaluate the initial conditions of head impact for cases
of cycling falls by simulating a great number of possible
accident situations.
Materials and methods
The present paper proposes a methodology based on
an automatic parametric study to evaluate the head
impact area and velocity for single cyclist falls. A total
of 1024 simulations have been automatically carried
out using MadymoÒsoftware coupled with a special
designed program allowing to compute and to analyse
the simulated scenarios. The analysed outputs are, suc-
cessively, head initial orientation and velocity just
before head contact. The cyclist’s kinematics were
reconstructed based on multi-body computing with
Madymo software. The principle of solving a multi-
body system is to define a set of rigid bodies repre-
sented by ellipsoids and connected by joints. Unlike
finite elements, contact between two bodies is not com-
puted by deformable surfaces but by a penetration
force defined by a function. The computational time of
this multi-body approach is strongly reduced compared
to finite element simulation. The models are developed
using ellipsoids in such a way that the geometry, mass
and inertia are respected. Hence, the human model
used was the TNO pedestrian model implemented in
the Madymo package and validated against post mor-
tem human tests, as depicted by van Hoof et al.
13
and
De Lange et al.
14
In the present study, the model used
to represent the cyclist had a mass of 75 kg and a height
of 1.75 m. The bicycle was modelled in accordance with
the geometry and mass of a city bicycle, as illustrated
in Figure 1(a). Each part of the frame was modelled by
a rigid body including mass and inertia computed as
steel pipes and connected by ‘bracket’ joints. The frame
and the front fork as well as the frame and the pedal-
board were connected by a revolute joint. Figure 1(b)
illustrates the bicycle model.
Two situations of cycling fall have been studied
according to a hypothetical configuration of real
frequent-accident cases. One configuration consists of a
fall alone after skidding, as depicted in Figure 2(a), and
the second consists of a fall after hitting a curb, as illu-
strated in Figure 2(b). For each configuration a set of
parameters was selected to be changed in the model
such as bicycle linear and angular velocity, torso pos-
ture, hip position, pedals position, handlebars angle
and bicycle incline. For pedalboard, four positions
were selected (0°,90°,180°and 270°) while for handle-
bars only two positions were studied (–30°and 30°)
with the final objective to study some extreme positions
without adding too much to the simulations. The out-
put of the simulations focused on three major results:
head velocity, head orientation and head impact area,
in order to define the conditions of head impact just
before head contact.
Table 1 reports the selected parameters considered
for the parametric study carried out for each configura-
tion with the associated values. The design of the simu-
lation table established for the present parametric study
is a full factorial table with 512 combinations. Hence,
the total number of simulations is 1024.
One of the outputs is the head impact area, so this
point had to be further defined. The head impact point
coordinates are extracted from the simulation and
plotted on the EN960 headform geometry by matching
the height and depth of both head geometries. The
choice of partitioning the cyclist’s head into latitude
and longitude zones was made in order to define easily
and quickly the impact point on the head, extracted
from simulation. Figure 3 illustrates the defined head
regions as follows: around the vertical axis, the head
has been divided into 12 separate regions, with a 30°
interval, leading to six symmetric regions called
Longitude 1 to Longitude 6. In the horizontal plan,
above the reference plan AA#definedintheEN960
standard,
15
the headform has been divided into five
equidistant horizontal sections (Latitude 1 to Latitude
5) and a section located below line AA#with an equiv-
alent width (Latitude 6). The last region, called
Latitude 7, describes the remaining part. The metho-
dology applied for the analysis of the computation
results was conducted in two steps. First, a global
Figure 1. Representation of a real city bicycle (a) and its model under MadymoÒsoftware (b).
Bourdet et al. 283
distribution of the contact area according to the previ-
ous head partition, as well as a mean value of velocity
in terms of resultant, normal and tangential compo-
nents, was evaluated. In a second step, the effect of
each parameter on the impact area and velocity
responses was studied. For this purpose, it was neces-
sary to define a response model associated with the
parametric analysis.
Figure 3. Representation of EN960 headform geometry partition: (a) in latitude and (b) in longitude. The reference line AA#
defined in EN960 is also reported.
Figure 2. Representation of the cyclist model coupled to the bicycle model used for the simulations of fall alone: (a) after
‘skidding’, (b) after ‘hitting a curb’.
Table 1. Display of the selected factors for the parametric study and the associated values.
Factors Bicycle parameters Human parameters
F1 F2 F3 F4 F5 F6 F7 F8
Pedalboard
rotation (°)
Handlebars
rotation (°)
Linear
velocity (m/s)
Angular
velocity (rad/s)
Bicycle
roll (°)
Torso
inclination (°)Hip position (mm)
Vert. Horiz.
Level 1 0 –30 5.5 0 0 21 0 0
Level 2 90 30 11.1 3.14 30 37 200 200
Level 3 180
Level 4 270
284 Proc IMechE Part P: J Sports Engineering and Technology 226(3/4)
The response model was a linear combination of fac-
tors and their interactions, as described in equation (1).
In this study, only the effect values of factors are evalu-
ated, and the effect values of interactions are neglected.
Y=m+X
N
i=1
Fi+X
N1
i=1
Iið1Þ
where mis the total average of the studied response from
all simulations (Nsimulations), F
i2[1,7]
are the effect val-
ues corresponding to the selected factors and I
i2[1,6]
are
the sum of the effect values from interactions between
several factors with order i,asdepictedinequation(2)
I1=P
N1
k=1 P
N
l=k+1
FkFl
...
I6=F1F2F3F4F5F6F76 ð2Þ
The effect value of a factor/interaction at a given level
was calculated as the mean value between the responses
when the factor/interaction had the level uand the aver-
age m.
Results
A total of 1024 simulations of cycling falls were carried
out with Madymo software. Figure 4(a) and (b) shows
the distribution of head impact area for both the ‘skid-
ding’ and ‘after hitting a curb’ fall configurations in
terms of latitude and longitude. It can be observed that
in the ‘skidding’ configuration the latitude areas most
often impacted are Latitude 6 and Latitude 5, with
90% of the total impacts. However, for the fall config-
uration after hitting a curb, the area is more spread out,
including Latitudes 4 to 6, with 65% of total impacts.
Considering longitude distribution, both fall config-
urations present a large range of impact areas. 80% of
these impacts are located from Longitude 2 to
Longitude 5 for falls after skidding, with an average of
20 61%, and from Longitude 1 to Longitude 4 for falls
after hitting a curb, with an average of 17 62%.
Tables 2 and 3 report the factors influence on the
corresponding response. The significance of the calcu-
lated influence value of the factors according to the
ANOVA test is also calculated. Hence, the non-
significant values are shaded in the tables. It is clear
that the resultant and tangential velocities are strongly
influenced by the bicycle speed in both configurations.
However, the normal component is also influenced by
the bicycle roll angle and the vertical position of the
hip. This can be explained by the fact that these factors
define the vertical position of the head and, hence, the
head drop height.
The previous analysis can further be divided into
two categories according to the bicycle speed, as
follows.
For a cyclist moving with a speed of 5.5 m/s (20 km/h)
and falling alone, as simulated in this study, the resultant
head velocity just before ground contact is, respectively,
6.9 61.2 m/s and 6.4 60.9 m/s in the cases of ‘skidding’
and ‘after hitting a curb’ configurations. Considering the
normal components of the velocities, results are
5.7 61.3 m/s and 5.2 61.0 m/s for ‘skidding’ and ‘after
hitting a curb’, respectively. Coming to the tangential
velocity, both fall configurations give about 3.7 m/s, as
reported in Table 4. In terms of impact angle versus the
normal direction a value of about 35°can be established.
Figure 5 represents the head impact area distribution
extracted from both configurations of falls with a bicycle
speed of 5.5 m/s. The latitude distribution is similar for
both configurations, with a majority of impacts located
at Latitude 5 and Latitude 6 (80% and 65% in the cases
of ‘skidding’ and ‘after hitting a curb’, respectively). On
the other hand, and as for the global distribution, the
impacts among the longitude partitions are spread in the
range from Longitude 1 to Longitude 4.
Considering a bicycle speed of 11.1 m/s (40 km/h),
the resultant head velocity just before ground contact is
11.3 61.1 m/s (6.2 61.0 m/s and 9.4 61.0 m/s for nor-
mal and tangential components, respectively) in the
case of falls after ‘skidding’, and 9.1 62.1 m/s
(4.8 61.3 m/s and 7.7 61.9 m/s for normal and tangen-
tial components, respectively) in the case of falls ‘after
hitting a curb’, as reported in Table 5. It can be
Figure 4. Histogram of the head impact area distribution obtained for the complete set of 1024 simulations in terms of: (a)
latitude; and (b) longitude for both falling configurations.
Bourdet et al. 285
observed that in case of this bicycle speed, the tangential
component is higher than the normal one for both fall
configurations. Indeed, the impact angle is shifted from
35°to 57°. Considering the distribution of head impact
area, it was observed that in the case of falls after
‘skidding’, latitude distribution remains identical to the
configuration with a bicycle speed of 5.5 m/s. On the
other hand, in the case of falls ‘after hitting a curb’, the
distribution becomes more spread out, as shown in
Figure 6.
Figure 5. Histogram of the head impact area distribution in terms of: (a) latitude; and (b) longitude for both falling configurations
with a bicycle speed of 5.5 m/s.
Table 2. Influence of factors on simulated responses in case of falls after ‘skidding’. The non significant influences are in bold.
Factors Bicycle parameters Human parameters
F1 F2 F3 F4 F5 F6 F7 F8
Pedalboard
rotation (°)
Handlebars
rotation (°)
Linear
velocity (m/s)
Angular
velocity (rad/s)
Bicycle
roll (°)
Torso
inclination (°)Hip position (mm)
Vert. Horiz.
V
Resultant
0% 0% 88% 5% 6% 0% 0% 0%
V
Normal
2% 7% 9% 6% 62% 5% 10% 1%
V
Tangential
0% 0% 95% 4% 1% 0% 0% 0%
Latitude 2% 3% 1% 26% 51% 6% 1% 9%
Longitude 1% 2% 0% 74% 19% 0% 1% 3%
Table 3. Influence of factors on simulated responses in case of falls after ‘hitting a curb’. The non significant influences are in bold.
Factors Bicycle parameters Human parameters
F1 F2 F3 F4 F5 F6 F7 F8
Pedalboard
rotation (°)
Handlebars
rotation (°)
Linear
velocity (m/s)
Angular
velocity (rad/s)
Bicycle
roll (°)
Torso
inclination (°)Hip position (mm)
Vert. Horiz.
V
Resultant
0% 2% 62% 3% 16% 1% 16% 0%
V
Normal
2% 7% 7% 0% 31% 5% 47% 0%
V
Tangential
0% 0% 85% 3% 5% 0% 5% 0%
Latitude 1% 6% 31% 1% 47% 0% 12% 1%
Longitude 16% 6% 0% 32% 44% 0% 1% 1%
Table 4. Computed head impact velocities and impact angle for a bicycle speed of 5 .5 m/s.
V
Resultant
(m/s) V
Normal
(m/s) V
Tangential
(m/s) Impact angle (°)
Skidding fall 6.9 61.2 5.7 61.3 3.7 60.9 33.5 68.7
Curb hitting 6.4 60.9 5.2 61.0 3.7 60.8 36 67.7
286 Proc IMechE Part P: J Sports Engineering and Technology 226(3/4)
Figure 7 represents the head impact points extracted
from all 1024 simulations and plotted on the EN960
headform geometry. In order to superimpose this result
with the impact area prescribed in the standard, the test
line defined in the EN1078 standard was plotted. It can
be observed that an important proportion of impact
points are located under the helmet test line.
Discussion
Although cycling falls occur often, it is very difficult to
get real statistics about the head injuries caused to the
victim, especially in the case of accidents without any
antagonist. Due to the large number of possible
scenarios, it is unrealistic to consider all accident types.
In the present study, two realistic and classical accident
scenarios have been considered: one caused after ‘skid-
ding’ and the other ‘after hitting a curb’. Moreover, the
same bicycle and the same human model size were used
for both accident configurations. These constitute a
first limit of this work. Indeed, some further bicycles
with different masses and sizes as well as further heights
and weights for the human body model should be con-
sidered in further research. For the parametric study,
four pedalboard positions were selected, while for han-
dlebars only two positions were studied. This choice
was made in order to study some extreme positions
but also in order to keep a reasonable number of
Figure 7. Representation of the head impact points: (a) for falls after skidding; (b) for falls after hitting a curb.
Figure 6. Histogram of the head impact area distribution in terms of: (a) latitude; and (b) longitude for both falling configurations
with a bicycle speed of 11.1 m/s.
Table 5. Computed head impact velocities and impact angle for a bicycle speed of 11.1 m/s.
V
Resultant
(m/s) V
Normal
(m/s) V
Tangential
(m/s) Impact angle (°)
Skidding fall 11.3 61.1 6.2 61.0 9.4 61.0 56.6 65.1
Curb hitting 9.1 62.1 4.8 61.3 7.7 61.9 58 66.5
Bourdet et al. 287
simulations. It should also be mentioned that for the
zero degree handlebars rotation, the time for the bicycle
fall was too long and caused simulation instability
problems. Nevertheless, it should be integrated into
future work. However, the results obtained with a
bicycle model coupled to a human model with the
height and weight of a 50th percentile man, provided
reasonable new knowledge about cycling falls, dis-
cussed further below.
For each fall configuration, seven parameters linked
to the bicycle and human characteristics have been con-
sidered for a parametric study with 512 simulations.
The analysis of impact locations at head level showed
that the impacts are very often (42%) close to the hel-
met rim line (Latitude 6) for both fall configurations;
more precisely, 57% in the case of falls after ‘skidding’,
but more spread out in the case of falls after ‘hitting a
curb’ (28%). These results are consistent with other
studies presented in the literature, such as Ching et al.,
16
who noted that 49.3% of impact locations on 311 hel-
mets are on the helmet rim. More recently, in 2009,
Serre
17
showed that among a collection of 862 helmets,
the majority of impacts occurred at the border line.
For both fall configurations, the impact in longitude
area is fronto-parietal, with more than 50%. Coming
to the discussion concerning the head speed at the time
just before impact, the results showed that for the nor-
mal components, the obtained values are in the velocity
range recommended by testing standards for the certifi-
cation of helmets (5.42 m/s). Moreover, during head
impact against ground surface, a significant tangential
component of the velocity is observed. Indeed, for a
bicycle speed configuration of 5.5 m/s the head resul-
tant velocity presents a 35°incline versus normal axis,
whereas for a bicycle speed configuration of 11.1 m/s
this angle is about 57°. This tangential component
of the head velocity generates a rotational acceleration
of the head and thus increases the brain injury risk,
as demonstrated by Deck et al.
18
and Deck and
Willinger.
19
Therefore, the standard test for bicycle hel-
met should include this tangential component in future
improved test procedures.
Conclusion
The present study focuses on cyclist head impact condi-
tion in the case of an accident without an antagonist,
an aspect poorly reported in the literature. A multi-
body bicycle and cyclist model has been developed and
applied for extensive parametric study (1024 simula-
tions). Two fall conditions have been considered (‘skid-
ding’ and ‘hitting a curb’) at two bicycle initial speeds
(5.5 m/s and 11.1 m/s). Results demonstrate that head
impact points are very often located around and under
the helmet rim, the normal head initial speed is close
to 5.5 m/s and the head velocity presents a significant
tangential component. It is considered that this infor-
mation should be taken into consideration both in test
procedure design and helmet development.
Funding statement
This research received no specific grant from any funding
agency in the public, commercial, or not-for-profit sectors.
Acknowledgments
This work has been developed within the ANR
PREDIT GO4 project BicyTeˆ te (French Department
of Transport).
References
1. Linn S, Smith D and Sheps S. Epidemiology of bicycle
injury, head injury, and helmet use among children in
British Columbia: a five year descriptive study. Inj Prev
1998; 4(2): 122–125.
2. Otte D. Injury mechanism and crash kinematic of cyclists
in accidents – an analysis of real accidents. SAE paper
1989-892425, 1989.
3. Maimaris C, Summers CL, Browning C, et al. Injury pat-
terns in cyclists attending an accident and emergency
department: a comparison of helmet wearers and non-
wearers. BMJ 1994; 308(6943): 1537–1540.
4. Eilert-Petersson E and Schelp L. An epidemiological
study of bicycle-related injuries. Accident Analysis & Pre-
vention 1997; 29(3): 363–372.
5. Yang J and Otte D. A comparison study on vehicle traf-
fic accident and injuries of vulnerable road users in
China and Germany. In: 20th international technical con-
ference on the enhanced safety of vehicles, Lyon, France,
June 2007. US Department of Transportation National
Highway Traffic Safety Administration.
6. Got C and Got CM. Accidents mortels de cyclistes: cir-
constances, configurations types. Rapport UMRESTTE
0908 journe
´e spe
´cialise
´eve
´lo et casque, mai 2009, p.9.
7. Depreitere B, Van Lierde C, Maene S, et al. Bicycle-
related head injury: a study of 86 cases. Accident Analysis
& Prevention 2004; 36(4): 561–567.
8. Ricard C and The
´lot B. Description e
´pide
´miologique des
accidents de ve
´lo d’apre
`s l’Enqueˆ te permanente sur les
accidents de la vie courante 2004-2007, Rapport
UMRESTTE 0908 journe
´e spe
´cialise
´eve
´lo et casque,
mai 2009, p.4.
9. Maki T, Kajzer J, Mizuno K, et al. Comparative
analysis of vehicle–cyclist and vehicle–pedestrian
accidents in Japan. Accident Analysis & Prevention
2003; 35(6): 927–940.
10. Maki T and Kajzer J. The behavior of cyclists in frontal
and rear crash accidents with cars. JSAE Review 2001;
22(3): 357–363.
11. Werner S, Newberry W, Fijan R, et al. Modeling of
bicycle rider collision kinematics. SAE paper 2001-01-
0765, 2001.
12. Serre T, Masson C, Perrin C, et al. Real accidents involv-
ing vulnerable road users: in-depth investigation, numeri-
cal simulation and experimental reconstitution with
PMHS. Int J Crashworthiness 2007; 12(3): 227–234.
13. Van Hoof J, De Lange R and Wismans JSHM. Improv-
ing pedestrian safety using numerical human models.
Stapp car crash J 2003; 47(October): 401–436.
14. De Lange R, Happee R and Liu X. Validation and appli-
cation of human pedestrian models. In: Madymo China
user’s meeting, Shanghai, PR China, 2005.
288 Proc IMechE Part P: J Sports Engineering and Technology 226(3/4)
15. BS EN 960: 2006. Headforms for use in the testing of pro-
tective helmets.
16. Ching RP, Thompson DC, Thompson RS, et al. Damage
to bicycle helmets involved with crashes. Accident Analy-
sis and Prevention 1997; 29(5): 555–562.
17. Serre T. Analyse d’une base de donne
´es de 800 casques
cyclistes endommage
´s. Rapport UMRESTTE 0908 Jour-
ne
´e spe
´cialise
´e casque et ve
´lo, mai 2009, pp.36–39.
18. Deck C, Baumgartner D and Willinger R. Influence of
rotational acceleration on intracranial mechanical para-
meters under accidental circumstances. In: Proceeding of
IRCOBI conference, Maastricht, The Netherlands, 2007,
pp.185–197.
19. Deck C and Willinger R. Evaluation of a new protective
helmet solution to reduce head injury for tangential
impact. In: Proceeding of the international conference of
crashworthiness, Washington, DC, 2010.
Bourdet et al. 289