biofidelity evaluation of a restrained under ... - ircobi

Abstract The present study evaluated the kinematics of the Global Human Body Model Consortium

(GHBMC) model, under frontal impact. Resultant acceleration data from simulations at head center of gravity

(CG), T1, sternum, T12, and sacrum were compared with sled test responses. The objectives of the present

study were to normalize the experimental data, qualitatively compare the simulation kinematics to that of the

PMHS response, and quantify the goodness‐of‐fit using correlation analyses (CORA). Kinematics data from

restrained eight PMHS tests – low (3.3 m/s), and medium (6.7 m/s) – were used to evaluate the biofidelity of

the GHMBC model under frontal impact. The experimental data were normalized to represent 50th percentile

male population, to minimize response variations due to demographics. The restrained GHBMC model was

positioned with its standard posture on a rigid seat. Simulations were performed using two different input

speeds – low, and medium ‐ under frontal impact conditions. Qualitative observation of the kinematics between

simulations and experiments indicated a good match. CORA ratings ranged from 0.51 to 0.78, and 0.53 to 0.75

for the low, and medium speed cases respectively, indicating acceptable kinematic biofidelity of the model

under frontal impact.

Keywords Finite element model, GHBMC, Validation, Biofidelity, Frontal impact.

I. INTRODUCTION

Delineating injury mechanisms in impact biomechanics is critical to prevent or attenuate injury‐causing factors

through counter‐measures. Traditionally, this delineation is done using post mortem human specimens (PMHS)

[1, 2], computer models [3], and anthropometric test devices (ATD, also known as “dummies”) [4, 5]. However,

compared to real life humans, the stiffness and biofidelity of these devices are very different, as they are

designed for repeated testing. Computer models constructed using rigid bodies, were available for crash

investigations since early 1960’s [6], however these models were limited in their capabilities, mainly due to the

lack, and cost of computational hardware at that time. Several finite element (FE) based human body models

(HBM) were recently developed due to the exponential increase in cost‐efficient powerful computational

hardware [7, 8]. In addition, rigorously formulated FE explicit theories [9, 10], and solvers – like Ls‐Dyna, Pam‐

Crash etc., ‐ accelerated the developments of these detailed FE models. Nevertheless, in order to obtain

trustworthy data from these HBMs, they must be validated against PMHS tests, for different modes, and rates

of loadings. Anthropometric variations are unavoidable in PMHS experiments. Generally, these variations

influence experimental responses [11]. To minimize these variations, individual PMHS responses can be

normalized to a predetermined reference population, such as mid‐sized male [12‐14], to be used to validate

HBMs.

One such detailed FE HBM was created – by the Global Human Body Model Consortium (GHBMC) ‐ to represent

50th percentile American male, with a standard seating posture. The GHBMC HBM was built with 1.95 million

elements, and 1.3 million nodes to represent a detailed anatomical geometry of the whole body parts. Past

studies have evaluated various biofidelity aspects of the model, under different modes, and rates of impact

loading. Park et al. (2013) evaluated the biofidelity of the model under lateral impact, using kinematics data

from sled tests. The model was impacted laterally with a rigid impactor to validate responses to that of PMHS

responses [15]. Hayes et al. (2014) used chestband data to validate the thoracic response of the GHBMC model,

Mike W.J. Arun ([email protected]) is a Post‐Doctoral Fellow, John R. Humm is a Research Engineer, Narayan Yoganandan is a Professor, Frank A. Pintar is a Professor, in the Department of Neurosurgery at Medical College of Wisconsin in the United States of America.

Biofidelity Evaluation of a Restrained Whole Body Finite Element Model under Frontal Impact using Kinematics Data from PMHS Sled Tests

1Mike W.J. Arun, John R. Humm, Narayan Yoganandan, Frank A. Pintar

IRC-15-69 IRCOBI Conference 2015

- 622 -

un

to

th

ki

ev

m

ac

w

qu

us

Ex

Re

di

(2

[1

he

m

m

se

(F

ac

FM

pa

lim

sy

nder frontal,

o validate the

he head, an

nematics co

valuating the

model. The p

cceleration d

with sled tes

ualitatively c

sing correlat

xperimental

estrained (th

ifferent velo

2010) were u

18], for com

eight: 169.5±

m/s) velocitie

medium the s

eat pan at 1

Figure 1.a).

ccommodate

MVSS‐208 te

aper honeyc

mb, such tha

Figure 1 (a

ystem) (b) A

, and lateral

e biofidelity

d spine has

ould directly

e kinematics

present stud

data from sim

st responses

compare the

tion analyses

setup:

hree‐point b

cities, utilizi

used for this

pleteness, o

±7.0 cm; we

es, using two

speed pulse,

15 degrees f

The seat wa

ed fixtures

esting protoc

comb (205 kP

at the distanc

a) Schematic

natomical lo

impact cond

of chest, sho

s not been

influences k

s of the mod

dy evaluated

mulations at

s. The objec

simulation k

s (CORA).

belt and kne

ng a custom

s validation

only a brief d

eight: 59.2±2

o input acce

resulting in

from the ho

as rigidly att

to attach th

col. The kne

Pa) was fixed

ce between

c representa

ocations of m

ditions [16].

oulder, thora

validated fo

kinetics – for

del, using PM

d the kinema

head center

ctives of the

kinematics t

I

ee restraints)

‐made rigid

study. A det

description i

21.5 kg) were

eleration pul

a total of eig

orizontal, and

ached to a s

hree‐point s

e restraint a

d to the plat

the limb, and

ation of expe

measured ac

Vavalle et a

ax, and abdo

or the whol

r example, a

MHS data ma

atics of the

r of gravity (C

e present s

o that of the

I. METHOD

) unembalm

seat setup.

tailed descr

s presented

e tested on

ses. Each PM

ght tests. Ea

d the seat b

sled platform

seatbelt and

assembly con

te’s outer su

d the surface

erimental se

celerations

al. (2015) use

ominal respo

e GHBMC m

accelerations

ay be a cruci

GHBMC mo

CG), T1, ster

tudy were

e PMHS resp

DS

med PMHSs w

Data from e

iption of the

here. Four

a sled syste

MHS was tes

ch PMHS wa

back was 25

m that applie

d knee restr

nsisted of a

rface. The re

e of the hone

etup (blue do

ed region spe

onses [17]. H

model under

s directly inf

ial step to e

odel, under

num, T12, an

to normaliz

onse, and qu

were tested

experiments

e experimen

unembalme

m, at low (3

sted with th

as seated on

degrees wi

ed the input

raints. The

plate at an a

estraint was

eycomb was

ot shows the

ecific data fr

However, the

r frontal im

fluence force

nsure the re

frontal imp

nd sacrum w

e the exper

uantify the g

on a sled s

performed b

ntal setup ca

ed PMHSs (a

3.3 m/s), and

he low speed

a custom rig

ith respect t

t pulses. The

belt positio

angle of 45 d

fixed in fron

s 25 mm.

e origin of t

rom the mod

e kinematics

pact. Becau

es – therefor

eliability of t

pact. Resulta

were compar

rimental dat

goodness‐of‐

system, at tw

by Pintar et

an be found

ge: 60.0±22

d medium (6

d, followed

gid seat with

to the vertic

e platform al

oning follow

degrees, and

nt of the low

the coordina

del

of

use

re,

he

ant

ed

ta,

‐fit

wo

al.

in

.7;

6.7

by

h a

cal

lso

ed

d a

wer

ate


- 623 -

In

Ac

fil

at

tr

m

si

th

vi

qu

si

D

po

Ex

Th

as

id

eq

m

no

W

w

in

Ta

Si

Th

si

se

el

(F

in

to

de

re

si

ls

G

su

nstrumentat

cceleration d

ltered and t

ttached to t

ransformed t

mounts, linea

de of the sp

he specimen

deos were u

ualitatively c

des of the h

ata from the

osture betwe

xperimental

he normaliza

ssumes linea

dentical tissu

quations sho

multiplied by

ormalization

Where, the su

which is give

ndividual PM

able 1 Norm

imulation se

he restraine

mulate expe

eat. Because

lements wer

Figure 2.a).

ncluded in th

o the seatbe

eceleration,

estraint did

mulation set

‐dyna keywo

HBMC mode

urface. At

ion and data

data – collec

ime‐zeroed

the surface

to estimate

r tri‐axial acc

ine. In addit

(Figure 1.b)

used to captu

compare the

head and T1

ese markers

een experim

data norma

ation proced

ar relation

ue density a

ow the nor

the normal

n factors used

ubscripts n

n by the ra

HS mass.

malization fac

etup:

ed model wa

erimental lo

e the seat w

re also used t

A generic s

he setup, to

elt model to

cantilever‐li

not signific

tup. The sea

ord – to avo

el and the rig

the start o

a processing

cted (using a

using accept

of the spec

acceleration

celerometer

tion, a tri‐axi

. The accele

ure the expe

e kinematics

spine, these

were transf

ments and mo

alization:

dure conside

between len

and elastic m

rmalization f

ization facto

d in this stud

and i denotio between

ctors used fo

as placed on

ading condit

was not expe

to model a f

seatbelt mod

mimic the ex

mimic real

ike high mag

antly engag

at, B‐pillar, a

oid relative

gid seat was

of the simu

:

cceleromete

ted methods

cimen’s hea

ns at the ce

packages w

al package w

eration data w

riment at 10

of the GHB

e markers w

formed to h

odel (see disc

red in this st

ngth, mass,

modulus bet

factors for

ors to derive

dy are given

ote normaliz

n the refere

or this study

n a rigid sea

tions. The G

ected to def

ootrest, and

del – made

xperimental

world seatb

gnitude vibr

ge the lowe

nd the footr

motions bet

defined as 0

ulations, in

er packages)

s. A pyramid

d. Using kin

enter of grav

were attached

was attached

were filtered

000 frames p

BMC. Retro r

were tracked

ead CG and

cussion).

tudy used a

and time u

tween indivi

acceleration

e the norma

in Table 1.

;

;ed and origi

nce mass (7

at and the d

GHBMC mod

form signific

a B‐pillar w

with 550 se

boundary co

belts. From

ration of kne

r limb. Hen

rest were rig

tween the st

0.249 [15], to

itial velocit

at head CG

d‐shaped nin

nematic equ

vity of the s

d at the T1, T

d to the ster

d using the S

per second; f

reflective ma

using Vicon

T1 vertebra

basic appro

units [12]. F

dual PMHS

n (a ) and t

lized signals

inal signals.

76 kg) that

deceleration

el with its s

antly, it was

here an anch

eatbelt, and

onditions. A

experimenta

ee restraint

nce, the kne

gidly constra

tructures. Th

o mimic the

ies were a

, T1, sternum

ne‐accelerom

uations, data

specimen’s h

T12, and sacr

num, betwe

SAE CFC 60 fi

rames from

arkers were

n cameras at

al body and

ach propose

Furthermore

and referen

time ( t ). T, using the e

Lambda (λ)

represent th

n pulses wer

tandard pos

s modeled u

hor point of t

60 one‐dim

fabric mate

al high fram

fixtures was

ee restraint

ained – using

he coefficien

contact betw

applied to

m, T12, and

meter packag

a from the

head. Using

rum levels to

een third and

ilter standard

these videos

attached to

t 1000 frame

used to com

ed by Epping

e, the proce

nce PMHS.

The original

equations (1

) is the norm

he target po

re applied t

sture was po

using rigid e

the seat belt

mensional el

rial property

me‐rate video

s observed;

was not in

g the *RIGID

nt of friction

ween neopre

the whole

sacrum ‐ we

ge (PNAP) w

package we

custom‐ma

o the posteri

d fourth ribs

ds. High‐spe

s were used

o the posteri

es per secon

mpare head‐

ger (1976) th

edure assum

The followi

l signals we

1) and (2). T

(

(

malizing facto

opulation, a

o the seat,

ositioned on

lements. Rig

t was attach

ements – w

y was assign

os, during sl

therefore t

ncluded in t

D_CONSTRAIN

n between t

ene and Teflo

model usi

ere

was

ere

de

ior

of

ed

to

ior

nd.

T1

hat

mes

ng

ere

he

(1)

(2)

or,

nd

to

n a

gid

ed

was

ed

ed

he

he

NT

he

on

ng


- 624 -

*I

sl

lo

rig

bi

no

bo

Th

Fi

Re

ex

ac

no

an

re

go

pr

ac

co

INITIAL_VELO

ed pulses (F

ower spine b

gid bodies, t

iofidelity. H

odes were c

odies (Figure

he joints wer

Figure 2 (a

igure 3 (a) Jo

esultant nod

xperimental

ccelerations

odes. Becau

nd z) veloci

esultant acce

oodness‐of‐f

rocessing ap

cceleration c

omputationa

OCITY keywo

Figure 2.b) w

iofidelity wa

the joint stif

However, in

reated using

e 3.a). The s

re defined pe

a) Simulation

oint definitio

dal accelera

data. In or

from head

se the noda

ity data we

elerations we

fit. In addit

pplication (LS

components.

al cluster, usi

ord, to mimi

were applied

s evaluated,

ffness betwe

the original

g *CONSTRAI

stiffness data

er SAE J211 s

n setup (b) In

on between

tion data fr

rder to com

CG, T1, ster

l acceleratio

ere different

ere then com

ion, it was

S‐PREPOST) w

. All simulati

ing 64 cores.

ic the stored

to the rigid

and becaus

een the verte

model, join

INED_EXTRA

a of the T11

sign convent

nput sled pu

rigid vertebr

rom simulat

pare the re

rnum, T12, a

on data were

tiated, and

mpared with

observed t

were not ac

ons were so

.

d kinetic ene

seat to load

e the verteb

ebral bodies

nts were not

A_NODES and

1‐T12 joint w

tion (Figure 3

ulses used in

ral bodies (b

tion results

esponses fro

and sacrum

e too noisy,

operated to

experiment

that the res

ccurate, as th

olved using L

ergies in exp

d the GHBM

bral bodies in

s may play a

t defined in

d spherical jo

was used for

3.b).

this study

b) Sign conve

were comp

om the mod

from the m

even after f

o obtain re

tal data, usin

sultant acce

hey were ca

s‐Dyna MPP

0 20-5

0

5

10

15

20

Sle

d P

uls

e (g

)

erimentation

MC model, m

n the lower s

significant r

the lower s

oints were d

the defined

ention used i

ared with t

el, and exp

model were e

filtering, hen

sultant acce

ng correlatio

elerations di

lculated inte

version 7.0

0 40 60

Ti

ns. Then, th

imicking sled

spine was co

role in overa

spine. There

defined, betw

d joints in th

in joint defin

the respectiv

periments, re

extracted fr

nce compon

elerations. T

n methods t

irectly given

ernally befor

solver on a

80 100

me (ms)

LOW MEDI

e decelerati

d tests. As t

nstructed wi

all lower spi

efore commo

ween verteb

e lower spin

nition

ve normaliz

esultant nod

om respecti

ent wise (x,

The calculat

to quantify t

n by the po

re filtering t

Linux RHEL 5

120 140

UM

on

he

ith

ne

on

ral

ne.

ed

dal

ve

y,

ed

he

ost

he

5.4


- 625 -

Co

Co

re

pe

co

be

ex

w

st

ca

w

ra

0

ca

si

ti

by

gr

Th

he

si

sh

st

W

co

st

CO

orrelation an

orrelation an

esponses. Th

erfect match

orrelations b

etween the

xperimental

with a consta

tudy, the inn

alculated for

within the inn

ating of 0 is g

and 1 is giv

alculated by

mplicity; how

me shift bet

y the corrido

Figure 4 C

reen curves

he cross‐cor

ere, a more

ze of the re

hifting the m

tate. All the t

Weighted sum

orridor, and

tudy are liste

ORA executi

nalyses:

nd analyses (

he ratings in

h. CORA use

between the

responses u

responses. A

nt y‐offset. T

er corridors

r 50 percent

ner corridor

given. Howev

ven, based o

averaging a

wever, the m

tween the re

or method.

Corridor met

show 5 perc

relation met

detailed des

sponses, usi

mean curve b

three metric

m of these t

cross‐correl

ed in Table

on directory

(CORA) was

CORA range

es two meth

e responses.

using user‐d

Along the m

This offset is

were calcula

of the peak

then a ratin

ver, if the re

on the user

all ratings fr

major disadv

esponses. A

thod: black c

cent (of peak

thod is more

scription can

ing three res

by a small ti

cs are calcula

three metric

ation metho

2. These par

y.

used to quan

e from 0 to

hods – that a

The first m

defined corri

ean trace, tw

computed b

ated for 5 pe

mean trace

ng of 1 is giv

esponse lies i

defined ble

rom each tim

vantage of t

cross‐correla

curve shows

k mean) inne

e involved a

n be found in

spective me

me step, an

ated using th

cs give the

ods give the

rameters can

ntify the goo

1, where 0 r

are fundame

method is the

idors. This m

wo corridors

by the user‐d

ercent of the

e (Figure 4).

ven, wherea

in between t

nding functi

me step. Th

his method

ation metho

s the mean

er corridor; a

and complex

n [19]. This m

etrics (Figure

d a cross co

he time‐shift

rating for c

overall corr

n be found

odness‐of‐fit

representing

entally very

e corridor m

method calc

s – an inner,

defined perc

e peak mean

For each tim

s if the resp

the inner and

on. The ove

e advantage

is that the r

od is used to

trace; red c

and blue cur

, for brevity

method quan

e 5). The cro

orrelation va

ted data for

orrelation m

relation ratin

and edited p

between sim

g poor corre

different –

method that

culates a me

and an oute

entage of th

trace, wher

me step, if th

onse lies ou

d the outer c

erall rating fo

e of this me

rating could

o compensat

urve shows

rves show 50

, only a brie

ntifies the p

oss correlatio

lue is calcula

the maximu

method. Equ

ng. The COR

per need in

mulation and

elation, and 1

to quantify

calculates t

ean trace us

er corridor –

he peak mea

reas the oute

he simulation

utside the ou

corridors a r

for the corrid

ethod is its d

become poo

te this disadv

the simulat

0 percent ou

ef descriptio

rogression, t

on method s

ated for eac

um cross cor

ually weighte

RA paramete

the “*.cps”

d experiment

1 representi

the degree

the correlati

sing individu

are generat

n trace. In th

er corridor w

n response li

uter corridor

rating betwe

dor method

directness a

or if there is

vantage pos

tion respons

uter corridor

n is present

time‐shift, a

starts by tim

ch time‐shift

relation valu

ed sum of t

rs used in th

file inside t

tal

ng

of

on

ual

ed

his

was

ies

r a

en

is

nd

s a

ed

se;

ed

nd

me‐

ed

ue.

he

his

he


- 626 -

Ta

Ki

Ea

ki

re

de

se

bo

of

sp

Figure 5 Pl

able 2 CORA

inematics co

ach simulatio

nematics co

espectively,

ecelerate, d

eatbelt, henc

ody model c

f the PMHS a

Figure 6 M

peed video, a

lots showing

A parameters

omparison:

on took app

omparison b

overall, the

ue the initia

ce loading th

changed dire

about the z‐a

Motion comp

and bottom

Method

Corridor

Cross correlat

g (a) progres

s used in this

proximately

between th

kinematics

al kinetic en

he model. In

ctions and r

axis due to s

parison at lo

row: motio

Para

A_

B_

D

D_

INT

tion

sion rating (

s study

20 hours to

e experime

showed goo

nergy stored

n both low a

ebounded b

houlder belt

ow speed:

n obtained f

ameters Value

K 2

G_1 0.5

A_0 0.05

B_0 0.5

_SIGMA 0

_SIGMA 0

D_MIN 0.01

_MAX 0.12

T_MIN 0.8

K_V 10

K_G 1

K_P 1

G_V 0.5

G_G 0.25

G_P 0.25

G_2 0.5

(b) time‐shif

III. RESULTS

solve on th

ent and sim

od agreemen

in the hum

and medium

ack to the se

t engagemen

Top row: Si

from simulat

es

Transit

5

1

2

Transitional or

Transition

Transitiona

5

5

We

t rating (c) s

S

he cluster. Fi

mulation for

nt. In simula

man body m

m speed expe

eats at appro

nt were also

ilhouette re

tion at (a) t=

De

tional order of funct

Weighting factor

Width of th

Width of th

Parameter to wi

Parameter to wid

Minimum int

Maximum int

Minimum

rder of function bet

al order of function

l order of function

Weighting factors

Weighting fact

Weighting factors

eighting factors of th

ize rating

igure 6 and

the low a

ations, when

odel, it mov

eriments, an

oximately 75

captured in s

ndering of m

=0 (b) t=75, a

scription

tion between rating

of the corridor met

he inner corridor

he outer corridor

den the inner corrid

den the outer corrid

erval of evaluation

terval of evaluation

interval overlap

tween ratings of 1 a

between ratings of

between ratings of

of the progression r

tors of the size ratin

of the phase shift ra

he cross correlation

Figure 7 sh

nd medium

n the rigid s

ved forward

nd simulation

5 and 60 ms

simulations.

motion capt

and t=135 m

gs of 1 and 0

hod

dor

dor

nd 0 for progression

f 1 and 0 for size

1 and 0 for phase

rating

g

ating

method

ow qualitati

m speed cas

seat started

d engaging t

ns, the hum

. The rotatio

tured by hig

s

n

ve

ses

to

he

an

ons

gh‐


- 627 -

sp

Figure 7 M

peed video, a

Figure 8 No

00

5

10

15

20

25

Hea

d C

G a

ccel

erat

ions

(g)

00

5

10

15

20

25

Ste

rnum

acc

eler

atio

n (g

)

Motion comp

and bottom

ormalized re

20 40

20 40

parison at me

row: motio

esultant acce

60 80 100

Time (ms)

GHB EXP

60 80 100

Time (ms)

GHBM EXPE

Sac

rum

acc

eler

atio

ns (

g)

edium speed

n obtained f

eleration dat

0 120 140

BMCPERIMENT

0 120 140

MCERIMENTS

0 200

5

10

15

20

25

d: Top row:

from simulat

ta from expe

40 60 80

Time (ms)

Silhouette r

tion at (a) t=

eriments, an

0 20

5

10

15

20

25

T1

acce

lera

tion

(g)

0 20

5

10

15

20

25

T12

acc

ele

ratio

n (

g)

100 120

GHBMC EXPERIMENT

rendering of

=0 (b) t=75, a

nd simulatio

20 40 60

Tim

20 40 60

Tim

140

f motion cap

and t=135 m

n for low sp

80 100

me (ms)

GHBMC EXPERIME

80 100

me (ms)

GHB EXP

ptured by hig

s

eed case

120 140

ENT

120 140

BMCPERIMENT

gh‐


- 628 -

Data Normalization:

The normalized low, and medium speed data are presented in Figure 8 and Figure 9. For low, and medium

speed cases, experimental peak head CG accelerations ranged from 3 to 9, and 16 to 27 g, whereas the

simulations showed approximately 10 and 23 g. Normalized experimental T1 accelerations ranged from 6 to 9,

and 18 to 20 g, whereas for simulations the accelerations were 9 and 25 g. For sternum the experimental

accelerations were between 9 and 20, and 16 and 23 g, although simulations showed 14 and 31 g. Experimental

T12 accelerations indicated a range from 10 to 13, and 19 to 26 g, whereas simulations indicated 19 and 42 g.

Finally, for sacrum the experimental accelerations were between 9 and 22, and 20 and 35 g, with simulations

indicating 12 and 26 g. Also it should be noted that the PMHS did not show stiffness degradation after low

speed tests, therefore the response of the medium speed tests were not affected by low speed tests, more

information on this can be found in Pintar et al. (2010).

Figure 9 Normalized resultant acceleration data from experiments, and simulation for medium speed case

Correlation analyses:

The results from the correlation analyses are shown in Table 3. For the low, and medium speed cases, the

combined rating was highest for the sacrum, whereas the lowest rating was to the head CG, and T12

0 20 40 60 80 100 120 1400

5

10

15

20

25

30

35

40

45

Hea

d C

G a

ccel

erat

ion

(g)

Time (ms)

GHBMC EXPERIMENTS

0 20 40 60 80 100 120 1400

5

10

15

20

25

30

35

40

45

T1

acce

lera

tion

(g)

Time (ms)

GHBMC EXPERIMENT

0 20 40 60 80 100 120 1400

5

10

15

20

25

30

35

40

45

Ste

rnum

acc

eler

atio

n (g

)

Time(ms)

GHBMC EXPERIMENT

0 20 40 60 80 100 120 1400

5

10

15

20

25

30

35

40

45

T12

acc

ele

ratio

n (

g)

Time (ms)

GHBMC EXPERIMENT

0 20 40 60 80 100 120 1400

5

10

15

20

25

30

35

40

45

Sac

rum

acc

eler

atio

ns (

g)

Time (ms)

GHBMC EXPERIMENT


- 629 -

re

sp

Th

its

st

no

du

re

no

ve

in

W

re

of

re

m

G

no

ar

di

an

us

an

ou

ac

0.

in

re

K

ho

te

espectively. I

peed increas

Table 3 CO

he objective

s kinematics

ternum, T12,

odal acceler

ue to variati

esponses, co

odal accelera

erified using

ntroduce sign

When the kin

esponse show

f the PMHS,

espect to the

may have infl

HBMC mode

ot be obtain

rtificial boun

ifference in

nd Figure 13

sed only at t

nd T1 ‐ had m

ut of the co

cceleration d

.6 and 0.8 fo

nitial posture

esponse som

Kinematics v

owever, Park

ests and use

It was intere

ed.

ORA ratings f

of the prese

s to that of

, and sacrum

ation data f

ions in demo

rrelation ana

ations were

trial problem

nificant error

nematics be

wed a good

and when t

e z‐ axis. The

uenced the

el was derive

ned during P

ndary condit

initial postu

3. Head and

hese two reg

minimum dif

onsidered 8

data showed

or T1 under

e similar to t

mewhat impro

alidation of

k et al. (2013

ed CORA to

sting to obse

or different

ent study wa

PMHS sled

m from eight

rom the GH

ographics. To

alyses were

obtained by

ms with know

rs in the mod

etween the

agreement.

the shoulder

e GHBMC m

response co

ed using a 2

MHS testing

tions, and a

re between

T1 coordina

gions in the t

fference and

test cases.

ratings of 0

low and me

the ideal GH

oving biofide

the whole

3) evaluated

o rate the g

erve the incr

body regions

IV

s to evaluate

tests, under

t PMHS tests

BMC model.

o quantify th

performed u

y differentiat

w acceleratio

del response

experiment

In experime

r belt loaded

model also ca

rridors, and

6‐year‐old (y

g mainly due

availability o

the PMHS a

ates were co

tests. Figure

Figure 13 sh

Despite the

.5 and 0.7 fo

edium speed

HBMC postu

elity ratings.

body GHBM

the biofidel

good‐of‐fit [

reasing trend

s

V. DISCUSSIO

e the biofide

r frontal imp

s – four low

. The experi

he goodness

using CORA m

ting and ope

ons and velo

es.

s and simul

ents, becaus

d the specim

aptured this

in turn the b

young) living

e to the age

of test speci

and GHBMC

onsidered in

e 12 shows th

hows the cas

ese variation

or head unde

d impact res

re may have

MC model w

ity of the mo

[15]. Using

d in average

ON

elity of the re

pact. Resulta

, and four m

mental data

s‐of‐fit betw

methodology

erating on co

ocities, there

lations were

se the seatbe

men, it rotate

complex mo

biofidelity ra

g subject; ho

of the PMH

mens. In or

, two extrem

this compa

he case in wh

se in which t

ns in initial

er low and m

pectively. Ho

e resulted in

as not repo

odel using ki

an impact

corridor, and

estrained GH

ant accelerat

medium spee

were norm

ween the sim

y. It should b

omponent ve

efore, this pr

e qualitative

elt was anch

ed in the co

otion. The in

tings of the

owever, this

S, absence o

rder to give

me cases we

rison as retr

hich the post

the postures

posture, co

medium spee

owever, per

n a mean cu

rted in liter

nematic dat

velocity of

d combined

HBMC model

tion data at

eds – were c

malized to mi

mulation, and

be noted tha

elocities. Th

rocess was n

ely compare

hored to the

ounter‐clock

nitial posture

model. The

kind of idea

of muscle ac

a quantifie

ere presente

ro reflective

tures – betw

s had maxim

rrelation an

ed impact res

rforming exp

urve closer t

rature for fr

ta from late

4.3 m/s, th

ratings, as t

l by compari

t head CG, T

compared wi

inimize effec

d experiment

t the resulta

is process w

ot expected

d, the over

left‐hand si

direction wi

e of the PM

posture of t

al posture m

ctivity, forcef

ed idea of t

d in Figure

trackers we

ween head Co

um differenc

alyses for t

spectively; a

periments wi

to the mode

ontal impac

ral impact sl

hey compar

he

ng

T1,

ith

cts

tal

ant

was

to

rall

de

ith

HS

he

ay

ful

he

12

ere

oG

ce,

he

nd

ith

el’s

ts;

ed

ed


- 630 -

accelerations of the model at T1, T6, and pelvis with experimental responses, and the CORA ratings were

reported as 0.27, 0.37, and 0.68 respectively. However, it is not clear whether the reported CORA ratings were

using the corridor method or correlation method, or the weighted sum of both the methods.

Figure 10 Head CoG and T1 posture with minimum difference for (a) low speed, and (b) medium speed

Figure 11 Head CoG and T1 posture with maximum difference for (a) low speed, and (b) medium speed

As indicated in the methods section, in the present study, for CORA ratings with the corridor method, 50% of

the peak mean values were used to construct the outer corridor. However, the ratings – both corridor, and

overall rating ‐ could differ if the width of the outer corridor is varied. In order to observe the difference, the

ratings were calculated for corridors constructed using 25% of the peak mean values. Figure 12 and Figure 13

shows the comparison of corridor, and overall ratings between 50%, and 25% outer corridors for low, and

medium speeds. In the corridor ratings the highest variations were observed in T12, in which the ratings

between 50, and 25 percent outer corridors dropped by 48, and 43 percent, for low, and medium speeds.

However, due to the contribution from the correlation method, for T12 the overall ratings only dropped by 16,

and 18 percent, for low and medium speeds. Therefore these construction parameters in CORA should be

chosen appropriately depending on specific requirements, and applications.

One of the limitations of the present study is the usage of a generic seatbelt material property, instead of a

characterized material property, used in the specific experimental sled tests. Considering the high stiffness of

these seatbelt materials, using a generic model should not make a significant difference, however minimum

variations could occur in simulation responses. The other limitation is the vertebral bodies of the thoracic, and

lower spine were modeled using rigid body elements, and the intervertebral discs were constructed using a

beam element surrounded by shell elements. Though these types of constructions may only estimate the real

world kinematics, and kinetics of the lower spine. In order to closely approximate the motions and loads, a

more realistic deformable model is preferred over rigid bodies, hence may further increase the GHBMC model’s

correlation ratings.

-200 -150 -100 -50 0 50 100 150 200 250550

600

650

700

750

800

850

900

t=130 ms

Experiment head CG Experiment T1 GHBMC head CG GHBMC T1

-z c

ordi

nate

(m

m)

x coordinate (mm)

t=0

-200 -150 -100 -50 0 50 100 150 200 250 300 350300

400

500

600

700

800

900

t=130 ms


-z c

ordi

nate

(m

m)

x corordinate (mm)

t=0

-200 -150 -100 -50 0 50 100 150 200 250 300 350 400550

600

650

700

750

800

850

900 Experiment head CG Experiment T1 GHBMC head CG GHBMC T1

-z c

ordi

nate

(m

m)

x coordinate (mm)

t=130 ms

t=0

-150 -100 -50 0 50 100 150 200 250 300 350 400300

400

500

600

700

800

900

t=0

t=130 ms


-z c

oror

dina

te (

mm

)

x coordinate (mm)


- 631 -

M

M

Theiki0.of

Th

re

W

M

N

fe

re

1.

Figure 12

Medium spee

Figure 13

Medium spee

his study evaight sled tesnematics be.51 to 0.78, af the model u

he study wa

esult of work

Wisconsin an

Model Conso

ational Labo

eedback on t

epresentative

. Lopezand RTraffic

Comparison

ed

Comparison

ed

aluated the sts with twoetween simuand 0.53 to 0under fronta

as supported

k supported

d the Medic

ortium for p

oratory for p

this manusc

e of the fund

z‐Valdes, F.J.R. Kent, The sc Inj Prev, 20

n of corrido

n of overall

biofidelity ofo different lations and 0.75 for the al impact.

d by the US

with resourc

cal College

roviding the

providing clu

ript. Any vie

ding organiza

, P.O. Riley, Dsix degrees of014. 15(3): p

r ratings be

ratings bet

V.

f the GHBMCspeeds, undexperimentslow, and me

VI. A

Department

ces and the u

of Wisconsin

e model for

uster resourc

ews expresse

ations.

V

D.J. Lessley, f freedom m. 294‐301.

etween 50%

tween 50%,

. CONCLUSIO

C finite elemder frontal is showed a gedium speed

ACKNOWLEDG

t of Transpo

use of facilit

n. The auth

r this study.

ces. The aut

ed in this art

VII. REFERENC

K.B. Arbogasmotion of the

, and 25% o

and 25% o

ONS

ment model, impact condgood match. cases, indic

EMENT

ortation DTN

ies at the Za

ors would li

The autho

thors thank

ticle are tho

CES

st, T. Seacrishuman head

outer corrid

outer corrido

using normadition. Quali. Combined Cating accept

NH22‐13‐D‐0

ablocki VA M

ike to thank

rs would als

Dr. Rodney

ose of the au

t, S. Balasubd, spine, and

dors for (a)

ors for (a) l

alized kinemitative obserCORA ratingtable kinema

00290. This m

Medical Cente

k the Global

so like to t

Rudd from

uthors and n

bramanian, Md pelvis in a f

low speed (

low speed (

matic data frorvation of tgs ranged froatics biofidel

material is t

er, Milwauke

l Human Bo

hank Argon

NHTSA for h

not necessar

M. Maltese, frontal impac

(b)

(b)

om he om ity

he

ee,

dy

ne

his

rily

ct.


- 632 -

2. Yoganandan, N., M.W. Arun, D.E. Halloway, F.A. Pintar, D.J. Maiman, A. Szabo, and R.W. Rudd, Crash characteristics and injury patterns of restrained front seat occupants in far‐side impacts. Traffic Inj Prev, 2014. 15 Suppl 1: p. S27‐34.

3. Yang, K.H., J. Hu, N.A. White, A.I. King, C.C. Chou, and P. Prasad, Development of numerical models for injury biomechanics research: a review of 50 years of publications in the Stapp Car Crash Conference. Stapp Car Crash J, 2006. 50: p. 429‐90.

4. Klinich, K.D., M.P. Reed, M.A. Manary, and N.R. Orton, Development and testing of a more realistic pelvis for the Hybrid III 6‐year‐old ATD. Traffic Inj Prev, 2010. 11(6): p. 606‐12.

5. Newby, N., J.T. Somers, E.E. Caldwell, C. Perry, J. Littell, and M. Gernhardt, Assessing biofidelity of the Test Device for Human Occupant Restraint (THOR) against historic human volunteer data. Stapp Car Crash J, 2013. 57: p. 469‐505.

6. McHenry, Analysis of the dynamics of automobile passenger restraint system, in Proc of 7th Stapp Car Crash Conference. 1963. p. 207‐249.

7. Schinkel‐Ivy, A., W.J. Altenhof, and D.M. Andrews, Validation of a full body finite element model (THUMS) for running‐type impacts to the lower extremity. Comput Methods Biomech Biomed Engin, 2014. 17(2): p. 137‐48.

8. Sun, J., A. Rojas, P. Bertrand, Y. Petit, R. Kraenzler, and P.J. Arnoux, Investigation of motorcyclist cervical spine trauma using HUMOS model. Traffic Inj Prev, 2012. 13(5): p. 519‐28.

9. Bathe, K.‐J., Finite element procedures. 1996, London: Prentice Hall. 10. Hallquist, J.O., LS‐Dyna theory manual. 2006: Livermore software technology corporation. 11. Yoganandan, N., M.W. Arun, and F.A. Pintar, Normalizing and scaling of data to derive human response

corridors from impact tests. J Biomech, 2014. 47(8): p. 1749‐56. 12. Eppinger, R.H. Prediction of thoracic injury using measurable experimental parameters. in Proceedings

of the International Conference of Experimental Safety Vehicles, NHTSA, Washington, DC. 1976. 13. Mertz, H.J., A procedure for normalizing impact response data. 1984, SAE Technical Paper. 14. Moorhouse, K. An improved normalization methodology for developing mean human response curves. in

Proceedings of the International Conference of Enhanced Safety of Vehicles, Seoul, Korea, Paper. 2013.

15. Park, G., T. Kim, J.R. Crandall, C. Arregui‐Dalmases, and J. Luzon‐Narro, Comparison of Kinematics of

GHBMC to PMHS on the Side Impact Condition, in IRCOBI. 2013: Gothenburg. 16. Hayes, A.R., N.A. Vavalle, D.P. Moreno, J.D. Stitzel, and F.S. Gayzik, Validation of simulated chestband

data in frontal and lateral loading using a human body finite element model. Traffic Inj Prev, 2014. 15(2): p. 181‐6.

17. Vavalle, N.A., M.L. Davis, J.D. Stitzel, and F.S. Gayzik, Quantitative Validation of a Human Body Finite Element Model Using Rigid Body Impacts. Ann Biomed Eng, 2015.

18. Pintar, F.A., N. Yoganandan, and D.J. Maiman, Lower cervical spine loading in frontal sled tests using inverse dynamics: potential applications for lower neck injury criteria. Stapp Car Crash J, 2010. 54: p. 133‐66.

19. Gehre, C., H. Gades, and P. Wernicke. Objective rating of signals using test and simulation responses. in 21st International Technical Conference on the Enhanced Safety of Vehicles, Stuttgart, Germany. 2009.


- 633 -

biofidelity evaluation of a restrained under ... - ircobi

Documents