Vol. 15. No. 3. pp. 201-218.

1994
Copyright
( 1994 Elsevier Science Lrd
Printed
in Grw
Britain.
All rights reserved

Inr. J. In~parr

Pergamon

Enyny

0734.743x/94

57.00+0.00

MODELLING
OF THE IMPACT
RESPONSE
OF MOTORCYCLE
HELMETS
A.
School

of Metallurgy

(Recriced

GILCHRIST

and N. J.

MILLS

and Materials.
The University
Birmingham
B15 ZTT. U.K.

29 Decrndwr

1993; itI reciserl~/om

or Birmingham,

IO Augusr

1993)

SummaryPExperimental
measurements
have been made of the shell deformation
during
of a motorcycle
helmet, and the compressive
stress-strain
relationship
ol the loam
information
was used to construct
a computer
model or the performance
of helmets, that
effects of diRerent shell materials and different foam densities to be predicted. The main
the deformation
of an impact helmet were identified and the optimisation
or the design [or
types of impact is discussed.

the impact
liner. The
allow the
events in
particular

NOTATION
A
;:
D
E
F
9
k
,,I
n
P
R

0
ap.go
Q

contact area
acceleration
constant in equation (8)
relative density
Youngs modulus
force
acceleration
0r gravity
spring constant
mass
damper constant
pressure
radius of curvature
radial position
thickness, time
velocity
position coordinate
strain
stress
yield stress of polymer,
loam
resonant kequency

I. INTRODUCTION

The authors [l] analysed the impact performance of industrial helmets, and made a
preliminary analysis of the performance of motorcycle helmets [23. At that time there was
a paucity of data for the deformation of the helmet shell (Fig. l), and only approximate
data for the performance of the liner foams at high strains. Consequently the development
of the analysis was incomplete. Subsequently we have made measurements both of the
foam impact performance [3] and of the helmet shell deformation [4]. This has allowed
a re-examination of the modelling of the overall helmet performance.
The reason for such modelling is to assist in the design of improved helmets, and to be
able to analyse acceleration-time
traces, collected from tests on many designs of helmets,
in terms of the performance of the components. Manufacturers design helmets to meet the
performance requirements in national standards, rather than considering the range of
impacts that will occur to motorcyclists and trying to minimise the number of injuries.
The helmet designer empirically selects the density and thickness of the polystyrene foam
to meet the impact tests in the standard, which are at one velocity (5-8 m s-l depending
201

A. GILCHRIST

202

and

N. J. MILLS

rear
FIG.

1. The

components

of a Jet style

helmet.

seen

in fore-and-aft

section.

on the standard) onto rigid flat and convex anvils, with the impact sites not including the
front edge ofthe shell [S]. The shell thickness is determined by the need to pass a penetration
test, in which a conical steel indentor with a 90 J impact energy (in BS 6658, or 30 J in
prEN 398 [6]) must not penetrate to touch the headform In real crashes the impacts are
at a range of velocities; the sites struck are mainly the front and sides of fullface helmets
[7]; the objects struck are mainly flat and rigid [S], with smaller numbers being flat and
deformable, round and deformable, and round and rigid. There are very few impacts with
rigid spikes.
It is possible to carry out helmet tests over a range of impact velocities, onto different
types of object, with a range of impact sites, but this is an expensive method of assessing
the overall performance of a helmet. An understanding of the main deformation mechanisms
and the causes of mechanical vibration in an impact would allow the test programme to
be reduced, and the limitations of each design of helmet to be evaluated. For a direct
impact onto an immovable anvil a helmet design will only keep the peak head acceleration
below the biomechanical limit for impact velocities below some limit. The value of this
critical velocity depends on the impact site and the shape of the object struck. There is a
range of crash outcomes for the helmet wearer, ranging through concussion, short term
brain function impairment,
permanent brain damage, to death. The tolerance of the
individual to head accelerations or localised forces is not exactly known. The biomechanics
of head injury is a complex subject [9] and the strain in the brain tissue is likely to play
a major part in the injury causation. The almost universal use of rigid metal headforms
for motorcycle helmet testing is arbitrary; however, the headforms are difficult to damage,
they do not absorb energy, and the tests results are reproducible. It is assumed that there
is a correlation between the acceleration of the rigid metal headform and the strains in the
human brain if both undergo the same impact while wearing a helmet. This means that
the pass/fail limit in national standards (usually 300 g maximum acceleration [5,6,10]-this
corresponds to a maximum 15 kN force if the headform mass is 5 kg) may not be the
optimum to minimise the number of head injuries. It is possible that the limit is too high
and that avoidable head injuries occur in moderate velocity impacts. The value of helmets
in reducing injury must in the end be evaluated by epidemiological
surveys.
A ring-element computer model, using the axial symmetry of the structure at the crown
of a helmet, was used to analyse impacts on the crown of motorcycle helmets [l I]. This
model did not consider the force oscillations due to dynamic effects of the masses of the
helmet components. It will be shown later that these dynamic effects play a major part in
the response of helmets for impact velocities in the range 2-10 m s- . There appear to be
no published finite-element analyses of helmet impacts. Despite the ready availability of

Modelling

of the impact

response

of motorcycle

helmets

203

such models for elastic deformation, there are several reasons why the helmet impact
problem is beyond the capacity of current programmes--(a)
the material properties are
complex, including viscoelasticity, strain hardening and (for fibre reinforced shells)
anisotropic elasticity (b) there are major changes to the shape of the helmet during the
impact (c) there are dynamic effects that need to be modelled. Hence the simple model
presented here is a first attempt at covering all the significant phenomena in a helmet impact.
2. MODELLING

2.1. Components of the helmet

The main impact absorbing component is the foam liner (Fig. l), which is usually moulded
from polystyrene beads. The reasons for this choice of material are, the low foam density,
the economic moulding process, and the good one-impact energy absorption. There is
comfort-foam inside the helmet to keep out draughts, and to allow a limited range of
helmet sizes to fit a great range of head sizes [12]. This open-cell foam is, in contrast with
polystyrene foam, capable of deforming to the shape of the head, without exerting a pressure
exceeding 10 kN m-* which would lead to headaches [12]. The shell of the helmets is
either injection moulded from ABS (Acrylonitrile
Butadiene Styrene copolymer) or
rubber-toughened
polycarbonate thermoplastics, or moulded from polyester thermoset
resin reinforced with glass fibre cloth/mat/rovings
(GRP). The shell can absorb a significant
proportion of the impact energy especially for impacts onto rigid convex objects.
Table 1 summarises the main deformation mechanisms in a helmet. The details of the
data for each of these will be given, and these four mechanisms will appear as the four
spring/dashpot elements in the model.
2.2. Structure

of the model

Figure 2a shows the load paths that exist between the head and the object struck, and
Fig. 2b the model used. There are 4 masses involved:
of the
m, of the
m, of the
m, of the
headform
m,

steel striker or anvil. This can be infinite if it represents the road surface;
helmet shell;
helmet liner foam;
headform. This is 5 kg in BS 6658, and it varies from 3.1 to 6.1 kg with the
size in the draft European Standard prEN 398.

In Fig. 2b the equivalent paths through the model are shown. In common with viscoelastic
models the convention is used that horizontal members remain horizontal when the springs
or dashpots deform. Thus single position variables x1 to xq can be allocated to the masses
m, to m4. The variables x3 and xq do not represent single points on the inner surface of
the liner and the headform respectively; rather they represent the x component of the
general point on these constant shape surfaces. Hence the difference x3-x* represents the

TABLE

1. HELMET

COMPONENTS

Component

AND
Data

blocks

Comfort
foam compression
Shell deformation

Impact tests on rectangular

Impacts on shells with
accelerometers
at sides

blocks

Elastic
foam

Force
striker

foam crushing

deformation

of polystyrene

difference

between

IMPACT

BEHAVIOR
Data

source

tests on rectangular

Polystyrene

Impact

DATA FOR THEIR

head and

treatment

Curve fit with Eqn (1). Integrate

across contact area.
Curve fit with Eqn (9).
Interpret
force-deflection
data to
give loading stiffness and force to
buckle/delaminate.
Measure resonant frequency
and
infer stiffness.

A. GILCHRIST

and

N. J. MILLS
,

Load

Paths

Wiker

Load

Paths

I Anvil

M,

&I

k,
I

I
FIG.

?a.

The

load

paths

between

a rigid
Rat surface
(the
mass. spring
and damper

road)
model.

Headform

and

the

head.

4
M.

(b)

the

equivalent

vertical component of the gap between the liner and the headform. The masses are only
permitted to move along the vertical x axis, so this is a one-dimensional
model. As such
it is incapable of predicting the variation of shell stiffness with the impact site (the relevant
shell stiffness must be input at the start of the modelling), or simulating the rotation of a
helmet on a headform.
The loading paths are either through the crushed foam liner, or through the outer regions
of the shell and the uncrushed liner, to the comfort foam, then to the headform. Load path 1
involves the bending of the shell (parameters k, and n,) in series with the elastic
deformation of the liner (parameters k, and nz). Load path 2 contains the forcedeflection
relationship of the crushed polystyrene foam, and that of the comfort foam (parameters
k, and n,). We will see later that the very low spring constant of the comfort foam means
that it will be the first to compress, and that its compressive strain will remain close to
100% throughout the majority of the impact sequence. The parallel connection of load
paths 1 and 2 implies that the flat or hemispherical anvil deforms the polystyrene foam
from the outside (Fig. 3a). The area of crushed polystyrene is too small to allow the rigid
headform to penetrate the inner surface of the liner. Consequently the position variable
xj for the inner surface of the liner is the same for load paths 1 and 2. Experimental
evidence [4] shows that there is considerable crushing of the inside of the liner by rigid
metal headforms, especially for impacts with rigid flat surfaces. This is due to the lack of
an exact fit between the inside of the liner and the surface of the headform. In real life the
deformable heads of motorcyclists make a good fit to the inside of their helmet liners, and
this interior crushing mechanism rarely occurs. The forces on the head would have to
exceed 20 kN for the average compressive stress on the liner to exceed the yield stress.
Consequently the model was made to reflect the real world situation rather than the
artificial conditions of helmet testing.
2.3. Crushing of the polystyrenefoam

The compressive stress-strain behavior of foams under impact conditions has been
measured [3]. Rectangular foam blocks of thickness ~20 mm are supported on a flat
steel anvil and impacted from heights in the range 0.25-l m with a flat steel striker. The
acceleration of the striker is integrated twice to obtain the deflection of the top of the foam
block, and hence the strain in the foam. For foam such as polystyrene, where there is
permanent buckling of the cell faces, the loading parts of different impact energy stress-strain
curves superimpose to form a mastercurve. For foam such as high density polyethylene,
where the cell element deformation is viscoelastic, there is no mastercurve as the stress
falls when the strain rate drops as the maximum strain in the impact approaches. We have

Modelling

of the

impact

response

rlgld

of motorcycle

helmets

205

headform

40

radial

FIG. 3a. The assumed

centre
of the contact

pattern
of liner
area. The identical

the headform
cannot
penetrate
a crushed
density
56 kg mm3
liner

used a gas pressure hardening

of polystyrene foams:

deformation
shapes

60

distance

mm

in an impact.
of the headform

the liner from Ihe inside.

density
liner,
for central
outer

radius

of 140 mm,

expression

a=a,+-

r is the radial
and the liner

distance
interior

from the
mean
that

fb) the radial

pressure
distribution
liner strains
of IO. 30, 50. 70. 90%
hitting

across
and a

a Rat anvil.

[13] to curve fit the stress-strain

mastercurve

POE
l-&-D

where E is the compressive strain and cs is the stress, while ~~ is the initial yield stress of
the foam, D is the relative density of the foam compared with that of the bulk polymer.
p,, should be the pressure of the gas in the cells in the undeformed foam (1 bar) but it is
used as a disposable constant for the strain hardening of the foam. For a polystyrene foam
of relative density D there are two constants that describe the compressive loading
stress-strain
curve (for D =0.030, cre =0.315 MN me2 and p. =O.lO MN mP2). The
relationship between o. and D was found experimentally to be
a 0 za P D.5

(2)

where ap is the yield stress of the solid polymer. The constant p,zO.l MN m- for relative
densities D ~0.035, then it rises towards 0.2 MN mm2 at higher densities. The reason for
the use of values higher than 1 bar in Eqn (1) is that the contribution of cell face deformation
mechanisms to the yield stress is equivalent to a higher gas pressure [13]. For the initial
small strains less than 6% an elastic loading expression was used that rises to the yield
stress a0 of Eqn (1). The viscoelastic contribution
to the compressive yield stress of

206

A. GILCHRIST

and

N. J. MILLS

polystyrene is relatively small; for instance the yield stress of 56 kg me3 density foam in
an impact lasting 5 ms is 0.73 MN m- whereas in a slow test lasting 50 s it is 0.53 MN m-.
Since nearly all of the helmet impact occurs at a high strain rate, the rate-dependent
contribution
to the compressive yield stress was ignored.
The next problem is to calculate the force (F)-deformation
(x) relationship for the foam
in the crushed contact area of foam. This needs a description of the geometry of the
headform and of the object struck by the helmet [3]. The approximation
used is that the
headform is locally spherical and rigid. Real heads are not rigid, but we are trying to model
laboratory impact tests using metal or other solid headforms. The object struck is taken
either to be flat and rigid, or a rigid hemispherical anvil of radius R,=50 mm. For the
impact on the flat anvil the geometrical parameter of importance is R,, the local radius
of curvature of the helmet exterior. In [3] we used the approximation
that the yield stress
(TVof the foam was a constant and derived the following relationship for the contact force
F as a function of the central deflection x of the foam:
F z 271Rhup.

(3)

This approximation
does not allow for the strain hardening of the foam and it is assumed
that the helmet shell is not stiff enough to increase the contact area above that for a
shell-less helmet. Equation (3) was found to give a good description of the behaviour of
thin-shelled bicycle helmets that were not impacted beyond their protectivecapacity
[14].
For an impact on a hemispherical anvil, consideration of the geometry shows that the
parameter R, in Eqn (3) should be replaced by R where
A=f+L.

(4)
u

Rh

This equation predicts that for R, = 50 and R,= 140 mm then R = 37 mm. Experimental
studies of the contact area [IS] show that the 4 mm thick ABS shell of a motorcycle helmet
considerably increases the contact area for impacts with hemispherical anvils. Hence in
the later modelling the value of R was set at 70 mm. It was also found that the strain in
the crushed polystyrene foam is uniform through the thickness of the foam, so the stress
in the crushed foam (Fig. 3a) is only a function of the radial distance r from the centre of
the helmet/anvil contact area. The total contact force is the sum of the forces on concentric
annuli, with a different compressive strain on each. Figure 3b shows how the contact
pressure 0, calculated from Eqn (1) varies with radial position, and how the contact area
increases with the central deflection of the foam. It was found that the use of 10 annuli
was sufficient to give a reasonable approximation to the force integral, using
F= f

ai27rriAr.

(5)

i=l

This integral was calculated as a function of the deflection value x at the centre of the
contact area, before the main modelling of the impact. Figure 4 shows the prediction for
a 25 mm thickness of polystyrene foam of density 56 kg me3 (a,=0.73 MN m- and
p,=O.156 MN m -2) in a helmet of shell radius 140 mm. In spite of the considerable
non-linearity of the stress-strain curve for the polystyrene, the force on the contact area
rises nearly linearly with the central compression. This is a result of the integral being
dominated by forces on the outer annuli. When the central strain exceeds 96% there will
be the equivalent of metal to metal contact between the headform and the anvil, which is
treated in section 2.7. The simpl:, relationship of Eqn (3) would have been a reasonable
approximation to the data in Fig. 4.
If the liner deflection at the centre of the contact area falls during the impact the total
force is assumed to fall along a line of slope 5 MN m-i. This is an approximation to the

Modelling

of the

impact

response

10
deflection
FIG. 4. Predicted
mm,

thickness

FIG. 5. The

contact
25 mm

location

force

between

a flat

and

density

56 kg m-l,

Eqn

(5). An

unloading

of the

3 accelerometers

surface

of motorcycle

and

207

25

22

as a function
path is shown

used

helmets

a polystyrene

foam

liner

of the central
deflection,
with an arrow.

to measure

shell

deformation

of radius

140

according

IO

in an impact.

slight positive curvature on the experimental [16] unloading curves for polystyrene liners.
Any reloading will also follow this line until the curve in Fig. 4 is met again. This is an
approximation to the reloading stress-strain curves of pre-impacted polystyrene foam [ 133
which have a slight negative curvature and join up smoothly with the loading mastercurve
when the previous maximum strain has been exceeded. These approximations have little
effect for single impacts but they cause abrupt slope changes in the force-deflection modelling
of second impacts.
2.4. Shell deformation

It has not been possible until recently to measure the shell deformation in a rapid way.
One slow method is to analyse high speed photographs of the impact. We found [2] that
debris, generated when composite shells were impacted, often obscured the position of the
shell. It was tedious to measure a series of photographs, and the data could be in error if
the helmet rotated during the impact. Some of the boundary conditions on the shell during
a helmet impact were established by using pressure sensitive paper at the interface between
the foam liner and the inside surface of the shell [15]. There was a high pressure in the
region where the foam crushed permanently, but the crushed area was relatively small in
extent.
To measure the shell deformation the helmet is supported on a rigid aluminium headform,
which is mounted via a load cell to a massive steel plate. Two miniature accelerometers
mounted on aluminium brackets were fixed to the exterior of the shell with cyanoacrylate
adhesive. Their position (Fig. 5) is remote from the impact site so they monitor the position
of the undeformed part of the shell. The 5 kg striker is guided by two cables so that it can
only move along the vertical x axis, and the accelerometer mounted on the top of the

A. GILCHRIST

208
TABLE

UK
UK
UK

l-f
IT
I-T

UK
UK
Italy

fi
t-T
IT

Italy IT
UK Jet
UK

Jet
Helmet

styles:-ff

2. DATA

FROM

THE LOADING

N. J. MILLS
CURVES

ABS

hemi

front

ABS
ABS
ABS

kerb
Rat

crown
front
crown

ABS
glass
glass

fibre
fibre

glass
glass

fibre
libre

= fullface,

flat
nat
hemi
Rat
hemi

iace

with

the force-deformation
the model
shell

DEFORMATION

690
660

3.0 B
none < 7
none < 9

700
720
720

n*ne < 6
none < IO
3.5 D

1350
2250

front
rronl
front

flat
Jet = open

FOR SHELL

crown
front

10
FIG. 6. striker:
---

and

none<
10
4.5 D

2440
3280

7.0 D

no chinbar.

deflection

mm

trace for rhe ABS shell

response
with parameters
3 kN: loading
and unloading

20
of a fullface
helmet
k,,=700.
li,,,= 200
at 5 m SC.

30

hit by a hemispherical
kN rn- , buckling

at

striker is also used to calculate the force on the striker. The mean of the left and right
shell accelerations is subtracted from the acceleration of the striker and the resulting
quantity integrated twice with respect to time to calculate the deformation of the contact
point relative to the remote parts of the shell.
The results of the tests (Table 2) show that thermoplastic helmet shells are less stiff than
GRP ones and they rebound more. The radius of curvature of the shell at the impact point
varied from 120 mm in the left to right direction to 140 mm in the fore and aft direction,
and the centre of the impact was 120 mm above the top of the visor aperture. Buckling
of the thermoplastic shells was only observed for the impacts with a hemispherical anvil
of radius 50 mm. The interpretation of the data in Table 2 depends on the assumptions
made. We assume that the shell is loaded while it is accelerating away from the striker,
so the liner is not in firm contact with the headform, and consequently is not significantly
loaded. Hence the loading slopes in Table 2 can be assumed to be approximately equal
to the initial stiffness k,, of the shell. One confirmation of this is that the values are very
close to the static stiffnesses measured on sections of shells supported at the rim on a steel
plate [3]. There will be a minor contribution to the loading slope from the liner. One way
to determine the latter for an ABS shell (which is not damaged by the impact) would be
to repeat the impacts at the same site, when the permanently crushed liner would not
contribute to the initial slope, and measure the decrease in the loading slope.
The form of the experimental force-deformation
relation is compared in Fig. 6 with the
idealisation used in the model. On loading the slope of the graph falls from the initial high

Modelling

of the impact

response

of motorcycle

helmets

209

value k ,0 z 700 kN m- to the lower post-buckled value k,, z 200 kN m- when the force
exceeds the buckle force F,. If unloading occurs it is as at the high slope k,, until the force
has fallen by F,. On subsequent reloading the force increases at the high slope k,, until
it rejoins the line of the earlier loading. If there is complete unloading to a zero deflection
any new loading will follow the path of the initial loading. The value of the dashpot
constant n, was chosen so that the ratio n,!k,, which would be a retardation time for
creep loading, is 0.2 ms; this ensures that there is a realistic amount of hysteresis in the
loading/unloading
sequence in the absence of buckling. Since the model will not be used
to predict the response on time scales other than those for impact, there is no need to
have a retardation time spectrum; such arrays of viscoelastic elements are only necessary
if the modelling of creep over a wide range of time is being attempted.
The buckling process does not damage the thermoplastic, rather it is an example of high
strain (up to 10% on the surface) single fatigue cycle. When GRP shells hit a hemispherical
anvil there is delamination between layers of the glass matt/roving/cloth
reinforcement,
and on a microscale at the glass fibre polyester resin interface. This behaviour can be
modelled in a similar way to the buckling of the ABS. The damage in GRP is permanent,
but there is negligible viscoelasticity, which means that on reloading the stiffness of the
shell will have decreased.
2.5. Bending and shear of the elastic region of the polystyrene joam

Ideally a complete elastic stress analysis should be carried out of the deformation in the
elastic part of the liner (Fig. 3a). The boundary conditions imposed by the rigid headform
and the elastically deforming helmet shell are complex. It would be necessary to use finite
element analysis with the inclusion of dynamic effects to solve this problem.
For the present we rely on experimental data for the difference between the headform
force F, and the striker force F, versus time. When the crown of a Jet style helmet, containing
a polystyrene foam of density 60 kg rnm3, was impacted the difference F,-- F, was found
to oscillate at a frequency of 400 Hz [16]. As the resonant angular frequency o is related
to the spring constant k and to the oscillating mass 111by:
w=

then for a shell mass of 0.6 kg the spring

used for the stiffness k, of a foam liner
foam are considered use was made of the
k of a block of foam of area A, thickness

k
-,
m

(6)

constant must be 4.0 MN m-i. This value was

of density 56 kg rne3. When other densities of
relationship for the compressive spring constant
t and Youngs modulus E:
(7)

The liner foam is loaded in a mixture of shear and bending, but there is a similar relationship

to Eqn (7) for the shear spring constant. The Youngs modulus E of closed cell foams is
given to a good approximation
by:
E = E&D,

(8)

where E, is the modulus of the solid polymer (3.0 GN rnA2 for polystyrene), the constant
C is equal to 1.0, D is the relative density of the foam compared with the bulk polymer,
and the exponent n falls in the range 1.5-2 [17]. A similar relationship applies for the
shear modulus of the foam. The use of Eqns (6) and (8), with the exponent n = 2, leads to
the rule that the liner stiffness k, is proportional to the square of the relative density of
the foam. There is a viscous dashpot in parallel with the spring and the value of the dashpot
constant n2 is chosen so that the characteristic time n,/k,=0.25
ms.

A. GILCHRIST

210

2.5. Compression

and N. J. MILLS

of the open-cell comfort foam

The comfort foams used are polyurethane or PVC foams often faced with a cloth layer.
The uncompressed foam thickness of ~5 mm is not a good estimate of the gap between
the inside of the polystyrene liner and the skull of the wearer. In a head size survey [12]
this gap was measured while a large helmet was being worn. At local high points on the
head the foam was compressed, while there were air gaps at other locations. Computer
Tomography scans of motorcycle helmets on a headform showed that there can also be
gaps between the outer surface of the liner and the interior surface of the shell [18].
It was difficult to measure the impact stresses on comfort foams at low strains because
the compressive stresses were orders of magnitude below those for polystyrene foam, and
the accelerometer system on the falling striker test rig was insufficiently sensitive to detect
the low strain behaviour. Therefore the initial part of the stress-strain
curve was measured
under slow speed compression tests on an Instron mechanical testing machine. Figure 7
shows that the force rises rapidly above a strain of 80%. For the modelling the contact
area was taken as being constant and equivalent to a disc of radius 75 mm. The experimental
curve was fitted by:
if ~~0.6,

(9)

and if the strain E in the foam exceeds 0.6, the factor k, is multiplied

by

exp[40(s - 0.6)2]
to approximate the bottoming out of the foam. There is a viscous dashpot in parallel with
the spring in the model, and the ratio n3/k3 = 1 ms. This produces a significant viscoelastic
effect when the loading velocity is 5 m s- (Fig. 7). The constant n3 may be unrealistically
large for the stress-strain
behaviour of polyurethane foam, but such a value was required
to prevent overlarge force oscillations when the liner hits the headform in the modelling
process.
2.7. Contact between the striker

and the headform

If the deflection of the liner and comfort foam exceeds their total thickness there would
be a direct contact between the headform and the striker, cushioned only by the thickness
of the shell and the bottomed-out
foam. A term was placed in the analysis to allow for
this eventuality, with a warning message ifthis happened. In such a case the head acceleration
will exceed 500 g.

0.8

40

FIG. 7. The force-strain

Eqn (9) using k,=40

strain %

60

curve of a comfort
foam, for a contact
kN m-l,
n,=40
Ns m-.
The curves
compression
in m s- .

80

100

area of radius 75 mm, fitted with

are labelled
with the velocity
of

Modelling

ol the impact
3. THE

response

COMPUTER

of motorcycle

helmets

211

MODEL

3.1. Basis of the model

The model can cope with the two main types of helmet tests, those with a fixed headform
and a moving striker, and those with a fixed anvil onto which the helmeted headform falls.
The core of the programme is simple, with the following sequence repeated at time interval
of 2 /S.
(i) The old positions si and velocities K of the four masses are used.
(ii) The forces acting on the masses are calculated using the constants for the elements.
For instance the force F,, acting between masses 2 and 3 is calculated using
F,, = k&-x2)+

n2( V,-V,).

(10)

The calculation of the forces for shell bending and for the crushing of the polystyrene
foam is more complex because it depends on whether there has been any unloading.
The details are given in sections 2.3 and 2.4 above. The forces on masses m, and
UI~ are not allowed to be negative so that the helmet can move freely away from
the headform and the anvil.
(iii) The new accelerations of the masses are calculated using Newtons 2nd law. For
instance.
a =h,+L+b
I

+Yl.

(11)

m 1

The term g1 is included to deal with the alternatives of a freely falling headform or
a freely falling striker. This term is equal to 9.81 m s - if the striker is freely falling,
otherwise it is zero.
(iv) The new velocities of the masses are calculated after a time interval At from the old
velocities using
(12)
(v)

The new positions of the masses are calculated

of the velocities over the time interval A.t.

by a similar numerical

The output of the model can be shown in various ways-as

the forces on
as a function of time, as the deflection of all or part of the helmet, or
function of the energy input to the helmet. It is possible to model the
impacts as the programme can be instructed to reuse the maximum
deflections used in the simulation of the first impact.

integration

one of the masses

as the force as a
effects of second
polystyrene liner

3.2. Predictions
The deformations of the helmet components and the forces as a function of time are
shown in Figs 8 and 9 for an impact of a flat striker on a region of the shell with a radius
of curvature of 140 mm. The comfort foam compression (x3-x4) rises rapidly to more than
90% of the comfort foam thickness in the early stages of the impact event (Fig. 8). The
striker force begins to rise to an initial peak during the comfort foam compression. The
implication is that the comfort foam is too soft to impede the acceleration of the complete
helmet away from the striker. Once the comfort foam is highly compressed it becomes
very stiff and plays very little part in the subsequent events.
The elastic bending of the polystyrene liner (x3-x2) initially moves in a negative direction
while the comfort foam is being compressed. This is interpreted as the inner surface of the
liner below the impact site approaching the headform, while the outer regions of the liner
have not moved. At this stage the elastic part of the liner has flattened in the region

A. GILCHRIST

and N. J. MILLS

20

100
/----

___----.

--+_

;
15 -

--__
-.

- 80

,-

...................~..

-60
\
\

,
-40

8
.E
!!
z
:
Gi
z
2
8

- 20

10

time ms
FIG. 8. Compression
of the comfort
foam ---;
liner crush -:
and
(zero shifted up by IO mm) versus time for an impact of a flat striker
with a 56 kg mm3 density liner.

(4

elastic deflection
of the liner
at 7.0 m s- on an ABS shell

2o4

(b)

head
f\

FIG. 9a. Predicted headform

---;
and striker force for the same conditions
as Fig. 8; A is the
striker force peak as the headform
accelerates, B is the headform
force peak when the helmet hits
the headform.
C and D are oscillations
as the sides of the liner flex away from and towards
the
headform,
and E is the striker force peak. (b) Experimental
data for a flat striker impact on the
crown of a Jet style helmet at 7 m s-.

Modelling

of the impact

response

of motorcycle

helmets

213

surrounding
the impact site. The part of the liner-bending curve after 2 ms shows a large
positive peak followed by damped oscillations. In this stage the outer parts of the liner
are moving more than the contact region, so the outer parts of the liner are becoming
more curved and/or shearing. Since the sides of the liner can move without contacting the
headform, this motion is only limited by the bending/shear stiffness of the liner. Therefore
the shell mass tt13 can oscillate on the elastic part of the liner.
The liner crushing (.Y-.~i) increases to a peak of 5 mm after 1 ms; this may represent
the crushing of the outer layers of the foam under the striker, because the local pressure
on the comfort foam side of the liner is below the yield stress of the polystyrene at this
stage. The liner crushing then rises to a peak of 12 mm after 4 ms. In this test with 123 J
of impact energy the maximum in the liner compression at the centre of the contact area
is only z 50% of the 25 mm liner thickness. In a falling headform test at the same impact
velocity of 7 m s- the impact energy would be ~20% higher because of the extra 1 kg
mass of the helmet, and therefore the maximum liner compression would be z 20% higher.
The shell bending (-Y~-.Y~) curve has the same general shape as the liner compression, and
is not shown in the figure to avoid confusion.
Figure 9a shows the predicted variation of the striker force and the headform force with
time. The force on the headform shows two distinct peaks, the first being due to the impact
of the inside of the rapidly moving helmet with the headform. The first peak is the higher
in this test, which suggests that a change in the characteristic
of the comfort foam would
be able to reduce the peak acceleration of the whole test. A stiffer comfort foam would
not allow such a sudden impact of the interior of the polystyrene foam on the headform.
The predicted forces exceed the BS 6658 test criteria; this is a result of the headform not
being allowed to penetrate the liner from the inside. Comparison with the equivalent graphs
of the component deformation (Fig. 8) shows that the peak A in the striker force occurs
while the comfort foam is being compressed and the helmet is accelerating away from the
anvil. The headform force does not rise rapidly to peak B until this process is complete.
The relative magnitude of the peaks can be changed by altering the comfort foam thickness.
The subsequent minimum in the headform force at C is due to the oscillation in the value
of the elastic liner bending. The second peak at D is again a result in the large scale
oscillation of the mass of the helmet shell on the elastic part of the liner. The peak in the
striker force at E is smaller than either of the peaks in the headform force and it occurs
close the point where the liner compression
is a maximum, In other simulations the
striker/anvil
force maximum is higher than the headform force by amounts up to 20%,
which shows why it is necessary to measure the headform force in helmet test standards.
Figure 9b shows an experimental result for an equal energy impact on the crown of a
Jet style helmet. The same features are shown as in Fig. 9a but the magnitudes of the peaks
are smaller. This is due to there being some extra crushing of the polystyrene liner from
the inside by the headform. The amount of filtering of the signal for the striker force is
less than that for the headform trace, so the slight ringing of the striker trace in the initial
peak should be ignored.
The model can be used to evaluate a certain design over a range of impact conditions.
Figure 10a shows how the peak headform force in a falling headform test onto a flat surface
increases with the impact velocity until it reaches the failure limit of 15 kN (equivalent to
the 300 g acceleration limit in BS 6658:1985, as the assumed headform mass is 5 kg). The
increase of the peak force is a nearly linear function of the velocity, and the force limit is
reached at a velocity of 6 m s-l. It can be seen from the superimposed graph of the peak
liner crush that the force limit is reached when the liner crush is only 11 mm; this is only
44% of the liner thickness and shows that the 56 kg m -3 density of the liner is too high
to be optimal for this test. Figure lob shows the performance of the same design for impacts
into a hemispherical anvil; in this case the 15 kN limit is not reached until the velocity is
8.8 m s-l and the liner is 98% crushed. For any higher velocities the headform force rises
immediately to injurious levels. In this case the foam density is optimal. There are impact
sites on the front of the helmet where the average radius of curvature of the shell is about
130 mm (the 140 mm value used in these calculations is close to the median value of the

A.

and N. J. MILLS

GILCHRIST

flat

anvil

telocity

m/s6

Velocity

m/s

10

b
FIG.

10. Predicted peak headform

lorce versus impact velocity for ABS shell and 56 kg me3 liner
on (a). flat, (b) hemispherical
anvils. The impact site has a radius of curvature
ol 140 mm.

shell), and where the shell stiffness is less due to the nearness of a free edge. For these sites
the slopes of the predicted maximum headform force versus velocity graphs will be less,
which will mean that a 56 kg mS3 density polystyrene foam will be of lower than optimal
yield stress for the hemispherical impact and nearer to optimal for the flat impact. Conversely
at the side of the helmet the average radius of curvature is 170 mm which means that the
high density foam is even less ideal for impacts into a flat surface.
If a polystyrene foam of low 32 kg mm3 density is used in the simulated impacts into a
flat anvil it is possible to reduce the peak headform accelerations by 27% over the range
of impact velocities up to 8 m s-r, and to increase the velocity to cause a peak headform
force of 15 kN by 25%, compared with the results for a foam of density of 56 kg mm3.
When the high loading stiffnesses for GRP shells from Table 2 were used in the simulations,
the effect shown in Fig. lOa, of the foam yield stress being too high to optimise impacts
with a flat surface, was larger. As GRP shells are of varied construction, and there are
differences in thickness from point to point on the same shell, it was felt to be more useful
to present results for the more consistent thermoplastic shells.

Modelling

of the impact

10

response

deflecI&

of motorcycle

mm

20

helmets

25

215

30

FIG. Il. Simulated

prEN 398 impacts for ABS shell and 56 kg m- liner, showing the anvil force
versus the total foam deflection for the first impact at 7 m s- onto a Rat anvil, --the second
impact at 6 m se onto a hemispherical
anvil.

The predicted results of second impacts in the test standards are instructive as they
reveal which impact is critical in the design. In BS 6658 there is a second impact at a
reduced velocity using the same anvil at exactly the same site. The velocities specified differ
according to the grade of the helmet (A or B) and the anvil shape. Type A is intended for
competitive events and it affords a higher level of protection, whereas Type B is intended
for the ordinary motorcycle rider. We have simulated the impact of Type A helmets onto
a hemispherical anvil, with successive velocities of 7 and 5 m s- . It was found that the
second impact, which has 5 1% of the kinetic energy of the first, does not cause a 56 kg mm3
density liner to deform to any greater strain than the 70% in the first impact. Hence the
first impact is the critical one. It would only be possible for the second impact to be critical
if the test conditions in the first impact caused the liner strain to be close to 100%.
In the prEN 398 impact tests there is a first impact at 7 m s- onto a flat anvil followed
by a second impact, at a site 15 mm away from the first, at 6 m s- onto a hemispherical
anvil. The spacing of the impact sites is insufficient for the second impact to avoid the
damage from the first impact and it is difficult to allow for the considerable amount of
overlap. Hence we have assumed that the second impact is at the same site. Figure 1 shows
the force on the anvil versus the total deflection of the liner and comfort foams. The second
impact has 73% of the kinetic energy of the first impact and it causes the total deflection
to increase by 4 mm. It is necessary to keep some liner crush distance in reserve in the
first impact, for the critical second test to be passed.
4. DISCUSSION

The main limitation of using a one-dimensional model for impacts is that the stiffnesses
of the shell and the elastic part of the liner cannot be calculated from their dimensions
and material properties. To take the analysis of the vibrations of the liner and the shell
stiffness further requires the use of finite element analysis. The details of how the components
of the model are connected affect the simulation. It is an approximation
that the inner
surface shape of the liner does not change; in reality the liner will bend elastically under
the shell contact patch when the helmet first contacts the striker, before the comfort foam
is fully compressed. This means that the predicted through-thickness liner crushing in the
first 1 ms of the impact is unrealistic. It is possible to recast the model to simulate the
crushing of the liner from the inside by a rigid headform later in the impact, but it is
difficult to change the model connections during the test. Hence we have concentrated on
the arrangement that is most relevant to assessing the performance of helmets on deformable
human heads.

16

A. GILCHRIST

and

N. J. MILLS

The model uses experimental data for nearly all the components of the helmet, whereas
the preliminary model [2] used disposable parameters for all but the polystyrene foam
yield stress. There need to be more measurements of the impact response of comfort foams
to confirm the high level of damping assumed in the model. The general level of viscoelastic
damping in the model is now similar to that observed in helmet impact traces. If the
comfort foam damping parameter is reduced by a fact of 10 then there are predicted to
be high frequency force oscillations when the comfort foam bottoms out and the liner
interior impacts the headform. There is now greater confidence that the model is valid,
and the predictions of the performance of helmets are reliable. The shell type and the foam
density can be varied to optimize the helmet design for certain impact test conditions.
It has been shown that current designs of motorcycle helmets are not optimised for
impacts on to rigid flat surfaces, when the impact site is away from the front edge of the
helmet. There are several reasons why this is SO. One reason for the use of high foam
densities is that there are also impacts onto hemispherical anvils, and there are impact
sites towards the front of the helmet where the shell is less stiff than for the site used in
the model (we intend to evaluate the shell stiffnesses for impact site on the front and sides).
It is necessary for the helmet to pass all the tests, and it is unusual for a polystyrene foam
liner of variable density or with large variations in thickness to be used. Hence the density
used must be the best compromise for the range of impact tests. The foam in the crown
of helmets has been modified either by reducing the area by drilling holes, or by using
inserts of lower yield stress foam. For experimental impacts of the crown of a helmet at
6.7 m SK onto a flat surface the lowest Head Injury Criterion values (a weighted average
of the headform acceleration raised to the 2.5th power) were found to be obtained [19]
with a liner density of 25 kg me3. However it is not justified to conclude, on the basis of
these experiments alone, that the design of the helmets is wrong;
the design is nearly
optimal for the range of impact tests in the British Standard, so it is necessary to review
the tests in the standard. We have argued that the hemispherical anvil impact should be
abandoned or replaced by an impact with a kerbstone, as accident statistics show that
rigid hemispherical objects are rarely hit by motorcyclists.
Another reason for the use of high density foam is that the helmet must pass two impact
tests. The second test is critical in the prEN 398 standard but not in BS 6658, because of
the higher relative energy level of the second impact in prEN 398. The first impact must
not cause a liner peak strain in excessof 80% otherwise there will not be reserve energy
absorption capacity for the secondimpact. There is no statistical evidence [20] that multiple
impacts occur on the same spot in crashes-there is evidence of multiple impacts but these
involve different objects and, as there is always some rotation of the motorcyclists head,
the probability of two major impacts on the same site is small. The damaged area for a
major impact is z 100 mm diameter, which is 5% of the total protected area on a helmet
shell. If the second impact is at a randomly selected site on the shell then the probability
that its centre is within 80 mm of the centre of the first impact is only ~5%. We conclude
that the second impact test is unnecessary, and that it can lead to the use of foams of
higher than optimal yield stressfor the first impact. If the second impact is performed to
reject helmets where the she11fractures in the first impact, then there are other ways of
achieving this.
It has been shown that the comfort foam thickness and type used in helmets can lead
to the first peak in the headform force-time trace being the largest. If the peak acceleration
from a test on a prototype helmet exceeds the limit at the first peak, the test performance
could be improved by increasing the stiffness of the comfort foam. The scalp and hair of
the average motorcyclist provide additional layers which may be more energy-absorbing
than the comfort foam. Hence it is unlikely that the first force peak on the riders head is
the maximum one. We feel that it is unnecessary to optimise the comfort foam design of
helmets to remove the first headform force peak in laboratory tests.
The requirement for a penetration test in the motorcycle helmet standards has lead to
the shells being relatively thick. This is noticeable by comparing the designs of bicycle
helmets and horse riding helmets where the relevant standards do not contain penetration

Modelling

of the impact

response

of motorcycle

helmels

217

tests. There are some designs of bicycle helmet that do not have a shell, but this is risky
as the polystyrene foam has a low fracture toughness and can break up in a crash; in
general there is a thin shell to protect the foam and to minimise sliding friction. It is
common with BS 4472 Jockey skull caps to have a very thin (< 1 mm) fibreglass shell that
delaminates easily in an impact. The high energy penetration test in BS 6658 is particularly
severe and it has lead to GRP shells that do not delaminate for high energy impacts into
a flat rigid surface. Such helmets pass the BS impact tests but our measurements of shell
deformation [4] show that this is partly due to the rigid headform penetrating the liner
from the inside, which is not observed with helmets from crashes. The surveys by Pedder
[ZO.Zl] show that shell penetration by sharp objects is extremely rare, so we feel that the
penetration test should be dropped altogether.
It is impossible to optimise the design of helmets for all impact velocities and sites. The
compromise design should attempt to minimise the total harm to helmet wearers. The
injuries predicted for a specific impact velocity and impact object should be weighted
according to the frequency distribution
found in accident surveys.
CONCLUSIONS

(1) The deformation events in the laboratory testing of motorcycle helmets have been
successfully modelled, and this will be of benefit in the design of helmets. The forces
on the head have been predicted as a function of the impact velocity and the type
of object struck, and this may assist in the evaluation of the risks of head injury.
(2) Current motorcycle helmet tests standards are arbitrary in the use of rigid headforms,
tests at a single velocity, and a peak acceleration test criterion.
(3) Helmets are designed to meet the mix of impact tests in Standards, so are not
optimized for impacts with a flat rigid surface, which is the most common object
struck. Such an optimisation would only be possible with a reduction of the impact
velocity onto the hemispherical anvil used in standards.
(4) It is recommended that the second impact test, on a site close to or identical to the
first site, is dropped from standards because it leads to polystyrene foam liners with
unnecessarily high yield stresses. The risk ofsuch a second impact in a crash is low.
(5) It is recommended that the penetration test is dropped from test standards because
it leads to over-stiff shells.
A~.Lno\~,/edr~r,tlr,,rs-The
granr F29936.

authors

are graleful

IO the Science

and Engineering

Research

Council

for support

under

REFERENCES
I. A. GILCHRIST and N. .I. MILLS. Construction-site
workers
helmets. J. occup. Arcid.
8. 199-211
(1987).
2. N J. MILLS, A. GILCHRIST and F. J. ROWLAND. Mathematical
modelling
of the effectiveness
of helmets in
head protection.
IRCOBI
(lnterrmrionul
Research
Council
on the Biokinerics
q/ Iinpacrs) Cot<, Bergisch
Gladbach.
215-226 (1988).
3. N .I. MILLS and A. GILCHRIST. The effectiveness of foams in bicycle and motorcycle
helmets. Accident
A,la/y.sis
cold Prrcmtiorl
23. 153-63 (1991).
4. N. J. Mt~ts and A. GILCHRIST.
Motorcycle
helmet shell optimisation.
Pro<. .4ssoc. f;>r Adc. o$ Auromotice
Mudiciple Con\:. Portland.
OR 149-162 (1992).
5. BS 6658:1985. Profecfice
I~elmrrsjor
I,e/ril-/u u.scr.s. British Standards
Instilution.
London.
6. prEN 398, Prorrctil:e
/&~ers jiir dricrrs n,ld possrrlyrrs
q/ nrororcyles
and nwpuds.
Cornit& European
de
Normalisation.
DIN, Berlin (1991).
7. D. OTTE, P. JESSL and E. G. SUREN. Impact points and resultant injuries to the head of motorcyclists
involved
in accidents, with and without
crash helmets. IRCOBI
Con/:,
Delft, 47-64 (1984).
8. H. VALLEE. F. HARTEMANN, C. THOMAS, C. TARRIERE. A. PATEL and C. GOT, The fracturing
of helmet shells.
IRCOBf
Co,!/:. Delft, 99-109 (1984).
9. D. C. VIANO. Biomechanics
of head injury. Sot. Auto. Enyny. Trans., 97(4). 107&1089
(1988).
IO. Srmdordfor
protecritw
/wadyear.for
~4s~rvirh moforcycles.
Snell Memorial
Foundation.
Wakefield.
RI (1984).
11. H. KOSTNER and U. W. STOCKER, Improvements
of the protective
effects of motorcycle
helmets based on a
mathematical
approach.
IRCOBI
ConJ. 192-213. Bergisch Gladbach
(1988).

A. GILCHRIST

218

and N. J. MILLS

12. A. GILCHRIST,
helmet design.
13.

14.
15.
16.
17.
18.

N. J. MILLS and T. KHAN, Survey of head, helmet and headform

sizes related to motorcycle
Ergonomics
31. 1395-1412
(1988).
N. J. MILLS,
Impact properties.
In: Physics
oj loo detuiry
Ce/lu/nr
P/asks
(Edited by N. WLYARD and
N. CUNNINGHAM),
Elsevier. to be published (1994).
N. J. MILLS. The protective
capacity of bicycle helmets. &iris/l
J. Sports
Med. 24. 55-60
(1990).
S. CHANDLER. A. GILCHRIST and N. J. MILLS. Motorcycle
helmet load spreading
performance
for impacts
into rigid and deformable
objects. IRCOBI
Con/:.
Berlin, 249-261 (1991).
A. GALE and N. J. MILLS, EIfect of polystyrene
foam density on motorcycle
helmet shock absorption.
Plastics
& Rubber
Processiny
& Applic.
5, 101-108 (1985).
L. GIBSON and M. J. ASHBY. Cellular
solids.
Pergamon
Press. Oxford
(1988).
R. D. COOTER. Computer
tomography
in the assessment of protective
helmet deformation.
J. Trauma
30,
55-68

(1990).

P. D. HOPES and B. P. CHINN, Helmets: a new look at design and possible protection.
IRCOBI
Con&,
Stockholm,
39-54 (1989).
20. J. B. PEDDER, Ph.D. thesis, University
of Birmingham
(1993).
21. J. B. PEDDER, S. B. HAGUES and J. M. MACKAY, Head protection
for road users with particular
reference to
helmets for motorcyclists.
Paper 32. AGARD
Con/:
Proc.
No. 322 on Impact
Injury
Caused
by Linear
Accelerutiou.
Cologne (1982).
19.