International Journal of Crashworthiness

ISSN: 1358-8265 (Print) 1754-2111 (Online) Journal homepage: http://www.tandfonline.com/loi/tcrs20

Crashworthiness of guardrail posts embedded in

cohesionless soils: a parametric study
Abdelmonaam Sassi & Faouzi Ghrib
To cite this article: Abdelmonaam Sassi & Faouzi Ghrib (2016): Crashworthiness of
guardrail posts embedded in cohesionless soils: a parametric study, International Journal of
Crashworthiness, DOI: 10.1080/13588265.2016.1167390
To link to this article: http://dx.doi.org/10.1080/13588265.2016.1167390

Published online: 24 Apr 2016.

Submit your article to this journal

Article views: 7

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tcrs20
Download by: [University of California Santa Barbara]

Date: 03 May 2016, At: 21:10

INTERNATIONAL JOURNAL OF CRASHWORTHINESS, 2016

http://dx.doi.org/10.1080/13588265.2016.1167390

Crashworthiness of guardrail posts embedded in cohesionless soils:

a parametric study
Abdelmonaam Sassi and Faouzi Ghrib

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

Department of Civil and Environmental Engineering, University of Windsor, Windsor, Canada

ABSTRACT

ARTICLE HISTORY

Guardrail systems are designed to provide a safe environment for vehicles and to reduce the
severity of occupants injuries. The performance of this safety system is deemed to be vital to
highway users. Guardrail post is probably the most important component of any guardrail system
design. The evaluation of guardrail posts performance usually involves crash tests, which consist of
colliding the post with a bogie. Crashworthiness tests try to cover a range of design parameters
such as the soil resistance, impact velocity and blockout crushability. When reviewing the available
various dynamic tests conducted to date, it is apparent that the range of the considered design
parameters varies widely. Because of the lack of consistency of the diverse test conditions, the
statistical analysis of the test results is not an easy task. In the present paper, the nite element
method has been employed as a tool to conduct a parametric study and generate statistical data.
The generated data are used to establish correlations between the post parameters and the system
performance indicators.

Received 1 September 2014

Accepted 11 March 2016

1. Introduction
Strong post W-beam guardrail systems are widely used
around the world as an essential hardware to ensure the
safety of errant vehicles. It is well established that the
performance of any guardrail system is fundamentally
associated to the interaction of the vehicle with the
embedded post. Since the change of the design cycle of
vehicles is much shorter that of guardrails, it becomes
urgent to review the performance of these systems. For
example, in the context of North America, vehicle eet
has changed signicantly over the last decades. The private vehicle automobiles (sedans) market share
decreased while Minivans, SUVs and small trucks made
a signicant increase to reach 25% of the vehicle eet as
of 2002 [22]. This change requires a re-assessment of the
safety hardware designed earlier for different vehicle
eet composition. In fact, statistical data collected from
highway accidents show that SUVs and small trucks
vehicles are more susceptible to roll in case of impact
with the W-beam guardrail [20]. It was argued that the
increase of rollover risk is due the high impact force magnitude between the vehicle and the guardrail system as
well as the higher centre of gravity of SUVs when compared to sedans. The study of vehicles colliding W-beam
guardrail systems has been an active research area [11].

CONTACT Faouzi Ghrib

fghrib@uwindsor.ca

2016 Informa UK Limited, trading as Taylor & Francis Group

KEYWORDS

Guardrail post; dynamic test;

nite element model;
cohesionless soil

Different procedures exist worldwide to evaluate the

safety performance of highway system. In USA, MASH
which replaced NCHRP 350, are the most used guidelines in use, while EN1317 is the procedure used in
Europe to provide criteria and standards for evaluating
new safety hardware devices [8]. Dreznes [6], Hubbel
[12] and Anghileri [2] compared the three testing procedures EN 1317, NCHRP 350 and MASH. Anghileri [2]
concluded that many technical aspects of the three procedures were similar. Dreznes [6] explained that any
design that meet one of these testing requirements could
be used in a country which has no established testing
requirements.
Previous research ndings suggested that the crashworthiness of highway guardrail systems is dominated
mainly by the soil
post interaction. Reid and co-authors
have shown that the capacity to contain and redirect
light trucks and SUVs is strongly correlated to the post
stiffness and soil resistance [20]. In fact, the optimum
level of soil reaction is dependent on many parameters
such as the height of the guardrail mounting, the depth
of the post, the nature of the soil and the design features
of the post. During impact, guardrail systems dissipate
the impact energy mainly through deformations in the
vehicle, the soil and the guardrail. Poor soil
post interaction may cause the guardrail system to fail performing

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

A. SASSI AND F. GHRIB

the intended role and might lead to fatal accidents [32].

For the case of strong posts, the reaction is controlled by
the soil resistance, and it was found that the barrier performance begins to degrade when the reaction force
reaches the level of 50 kN.
The crashworthiness evaluation of guardrail systems
involved experimental tests to assess the performance of
guardrail barriers under various case scenarios. Sultani
et al. [28] presented a performance study of guardrail
systems based on crash test data. The study shows a
wide variability of the responses of the 33 designs considered. Experimental studies are always expensive and
time consuming; they usually cover only limited case
scenarios. Finite element simulations are an alternative
to testing particularly at the early stages of the design or
when conducting parametric study to cover the design
space. In fact, this technique has become a fundamental
tool in the analysis of vehicle safety [26]. Ideally, combinations of nite element simulations, carefully designed
tests and the study real crash accidents data lead to a better insight of the performance of the complex responses
of guardrail systems.
The crash analysis of guardrail systems involves
highly nonlinear material behaviour associated with
large deformations due to the impact load. The nite element analysis method appears to be the preferred tool
that offers full control over the range of the involved
parameters. The different design parameters can be evaluated through an optimisation process in order to propose new guardrail design proposals or improve existing
one that can reduce the risk of truck rollover [29]. Due
to the complexity of the involved mechanical phenomena in the impact of a vehicle to a guardrail system, any
nite element model should be confronted with available
experimental results to validate the analysis process.
Consequently, a correlation of any proposed nite element model is fundamental to gain condence about its
capability and efciency. Once validated, nite element
models become a reliable design tool and relatively inexpensive to evaluate guardrails crashworthiness responses
under a wide range of inputs [29]. The performance of
the whole guardrail system during an impact is closely
dependent on the behaviour of individual pole-soil subsystem. Therefore, the present work will concentrate on
the analysis of the pole-soil subsystem.
The objective of the present paper is to conduct a
parametric study on the major design parameters on an
isolated post
soil-guardrail in sandy soils. Five parameters were selected; they are (i) the impactor speed, (ii)
the impactor mass, (iii) the post embedment depth and
(iv) the blockout crushability. A nite element model of
impactor subjected to an initial velocity colliding a
guardrail post was developed and calibrated to the

impact tests conducted by Coon et al. [3]. The model

was used to investigate the effects of the different test
parameters on the interaction of the post and soil to
compare the load
deection curves with the baseline
results. A regression analysis of the data generated by
the parametric study allowed the development of series
of correlations between the base design parameters and
the system performances.
The paper is organised in two sections: in the rst section, the development of the baseline nite element
model and its validation is presented whereas in the second section, a parametric study of the effects of the
embedment depth, the soil resistance, the impactor mass
and the blockout compressibility is discussed.

2. Baseline model development and validation

The major issue associated to the development of an efcient nite element model for the simulation of embedded structures is the proper selection of the
soil
structure interaction model. Soil
structure interaction modelling has been extensively investigated for
structures subjected to quasi-static and dynamic loadings, but very few studies dealt with impact loading. The
available models for soil-structure interaction vary in
complexity from continuum to spring-dashpot-mass
(subgrade method or macro-element) elements. Sassi
and Ghrib [24] compared the subgrade and continuum
approach and showed that the subgrade approach could
be used to simulate the interaction between the soil and
the post and represents a good trade-off between simplicity and accuracy. The use of the continuum approach
is accurate; however, it is computationally demanding
and the identication of the model parameters of the
constitutive law representing the continuum is very
difcult.
In the present study, we propose to use the horizontal
subgrade method to simulate the soil
pole interaction.
This method has been extensively used in the area of
geotechnical engineering to simulate the soil lateral resistance against buried structures. It consists of modelling
the soil surrounding the structure by a set of nonlinear
springs attached to the embedded part of the structure.
The horizontal subgrade modulus at a given point is
dened as the ratio of the horizontal subgrade reaction
force and the corresponding average produced displacement at that point. Today, various empirical expressions
are available in the literature for the identication of the
modulus of subgrade reaction. In the specic case of
roadside safety, the method developed by Habibagahi
and Lancer [10], based on the concept of the bearing
capacity, is widely used. The subgrade method is primarily used for its computational efciency. However, the

INTERNATIONAL JOURNAL OF CRASHWORTHINESS

Impactor
Post
Lumped soil
mass

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

Figure 1. Proposed dynamic model for lateral post response.

method suffers from many shortcomings where the most

signicant is that it does not account for the inertia effect
as it treats the soils as a massless medium. Previous studies have shown that the soil inertia plays a major role
especially at the initial contact of the post
soil during
the impact [14]. Most of the studies conducted in the
area of roadside impact do not consider the effect of the
soil mass neither the energy dissipated through in the
soil mass.
In the present paper, we propose a simple method to
improve the subgrade model by dening a macro-element comprising a spring, a mass and a dashpot. The
soil is decomposed into independent layers. The behaviour of each layer is simulated by one macro-element as
shown in Figure 1. The overall soil reaction is determined as the resultant of the all the macro-elements
resultant. The lumped mass represents the mass of the
soil being involved during the impact. The role of the
springs is to store the elastic energy and simulate the
non-linear behaviour of the soil during the impact. The
inclusion of dashpots is a simplistic method to account
for the energy dissipation observed in crash tests. As the
soil is deformed during the impact, many dissipative
phenomena including friction, heat and plastic yielding
occur simultaneously. The process of representing all the
dissipative mechanisms can only be a simplication to
represent the energy loss in the soil. To take into account
the soils parameters variability, the soil mass has been
divided into layers of 100 mm; each layer is assumed to

have independent properties of its adjacent soil. The discretisation of the post
soil interface is therefore fully
modelled based on the three mechanical parameters: the
nonlinear stiffness, mass and damper coefcient.
To validate the proposed nite element model, the
results of the dynamic tests conducted by Coon et al. [3]
are selected as baseline. These tests will be used to calibrate the added masses and dashpots coefcients.
Coons tests covered various ranges of impactor speeds
with different types of post material and geometry. The
cart impactor consists of a rigid nose bogie vehicle of
946 kg mass, instrumented with an accelerometer to
measure the lateral deceleration during the impact. Four
tests were conducted with W-beam posts, corresponding
to speeds of 4:6 m=s, 5:4 m=s, 5:9 m=s and 8:9 m=s,
respectively. The soil density ranged from 1980 kg=m3
to 2240 kg=m3 and the tests were conducted in soils
with no signicant moisture. The length of the post was
1830 mm with an embedment of 1100 mm. The impact
point of the bogie with the post was located at 550 mm
above the ground level. The results of the test showed
that for the impact speeds ranging from 4:6 m=s to
5:9 m=s, the impactor rebounds back; however, for
higher speed of 8:9 m=s the impactor slides over the
post. The simulations were conducted using Hyperworks
Finite Element software package, in particular the
impact analysis was conducted using RADIOSS nite
element software.
2.1. Spring stiffness identication
The spring stiffness was calculated by the method of
Habibagahi and Langer [10]. The coefcient of subgrade
reaction was found to increase with the depth and
decrease with the deection. The horizontal stiffness kh
is dened as follows:
0

s
Kh D Nq
y

(1)

where s 0 is the effective overburden stress, y is the lateral

deection and Nq denotes the lateral bearing capacity
and it is dened as function of the deection y and the
depth z, whereas B is the width of the post.
r
z
Nq D A MF
B

(2)

A is a dimensionless parameter which depends on the

deection and the internal friction angle and MF is density-dependent modication factor. Plaxico et al. [17]
proposed a method to extrapolate the value of A using

A. SASSI AND F. GHRIB

45
40

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

Force (kN)

35
30

z = 100 mm

25

z = 300 mm

20

z = 500 mm

15

z = 700 mm

10

z = 900 mm

z = 1100 mm

0
0

10

20

30

40

50

60

70

80

90

100

Deflection (mm)
Figure 2. Load
deection curve of the unidirectional spring calculated by Habibagahi and Langer approach for Coon et al. [3] test.

the results of Habibagahi and Langer [10]. The expression of A as a function of the lateral deection y is
given by:
A D 15; 276
14:09 e0:1245y

(3)

To calculate the soil equivalent stiffness for different

friction angle, Plaxico et al. [17] dened a densitydependent modication factor, MF, following a linear
relationship with the friction angle:
MF D

2

30 C 1
3

coaxial cylindrical blocks having different mesh sizes. At

the vicinity of the post, the mesh was ner to capture the
soil deformation and coarser at the outer portion of the
cylindrical block. The size of the soil block used in the
nite element model was 2.7 m whereas the depth of the
block was 2.0 m. The impactor consisted in cylindrical
rigid noose striking the post with an initial velocity. The
soil is modelled using 8-node hexahedron and the parameters of the soil and the guardrail post were determined

(4)

Figure 2 shows the distribution of the springs stiffness for the test conditions of Coon et al [3] at different
depths.
2.2. Evaluation of the lumped mass
The lumped mass of the macro-element is dened as the
mass per layer involved during the impact. Numerical simulations show that only a limited mass surrounding the
post would be activated during the impact. To determine
the size of the block activated during the impact, the
results of a continuum nite element model were used.
The continuum model consists of a post embedded in a
soil block as shown in Figure 3. The soil surrounding the
post is modelled as a cylinder medium in which the post is
embedded at its centre. The soil was divided into three

Figure 3. Continuum model for the dynamic loading of the post

embedded in cohesionless soil.

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

INTERNATIONAL JOURNAL OF CRASHWORTHINESS

Figure 4. Vertical cross-section of the soil mass mobilised during the impact at 20 ms.

from the literature [25]. The displacement of the post was

measured at the impact point and the force was calculated
at the interface of the post and the impactor. The data collected was ltered using CFC60 lter and the
load
deection curve is determined.
The criterion dening the mass activated during the
impact is based on a displacement threshold df . The
region with a displacement higher than a given displacement df was assumed to be involved in the inertia reaction and it is included in the active mass. The
contribution of the mass exhibiting lower displacement
is neglected. A parametric study was conducted to determine the threshold df , and from a parametric study it
was found that a displacement 2 mm produces accurate
results. The results from the continuum model showed
that the total volume of the soil mobilised in the front of
the post during the impact dened a volume having the
shape of a cone where the apex is located at the point of
rotation of the post and the base at the ground level as
shown in Figure 4 [25]. The soil block mass dened
within the contour was calculated and lumped to the system of macro-element.

2.3. Determination of the damping coefcient

To determine the amount of damping coefcient, Cc, the
damping is assumed viscous and expressed as a percentage of the critical value. Since the effect of the soil inertia
occurs at the early loading stage, the stiffness of the
spring is selected from the initial tangent of the nonlinear load
deection curve as illustrated in Figure 2. The
springs located close to the ground surface exhibited the
maximum lateral displacement and the lowest spring
stiffness. The spring located close the pole rotation centre exhibited less lateral displacement amount but higher
stiffness.
A parametric study has been conducted to determine
the damping ratio using of Coons et al. [3] experimental
ndings. For each simulation, the average load, the peak
load and the maximum impactor displacement have
been computed and compared to dynamic test results
taken as reference. The damping ratio, j, of 12% gave
the optimum results matching the load defection curve
of the post and the deection of the post with time.
As summarised in Table 1, the numerical simulations
results when compared to the dynamic test results show

Table 1. Comparison of peak load, average force and maximum deection of dynamic test with the nite element simulation.

Test #1 (4.6 m/s)

Test #2 (5.4 m/s)
Test #3 (5.9 m/s)
Test #4 (8.9 m/s)

Maximum deection (mm)

Improved
Continuum
Test
subgrade model
method
234
233
240
314
296
312
348
353
338
Over ride
Over ride
Over ride

Note: The post used in test 4 is W150 23.5 instead of W150 13.5.

Test
42.8
43.9
47.3
NA

Average force (kN)

Improved
Continuum
subgrade model
method
43.0
41.1
45.9
42.5
47.9
46.3
56.3
55.3

Test
64.0
66.9
67.0
104.7

Peak force (kN)

Improved
Continuum
subgrade model
method
53.1
50.4
57.8
51.2
64.3
52.3
97.2
85.2

A. SASSI AND F. GHRIB

Table 2. Summary of the simulation results of impactor hitting the post embedded in the soil with different
impact speeds.
Speed (m/s)

Peak load
(kN)
43.3
50.7
53.8
58.1
64.1
77.9
88.7
97.2
106.6

a good agreement. These ndings demonstrate that the

proposed subgrade model including an active masses
and dampers predicts accurately the post behaviour during the side impact similar to the continuum model
while keeping the computational time reasonable.

3. Evaluation of the impactors speed effect

To study the impact speed effect on the guardrail post
response, a velocity range of 3
10 m/s was considered.
A literature survey of the experimental tests showed that
the impactor speed varies in general from 4.6 m/s [3] to
9.4 m/s [18]. It is to be noted that a speed of 9 m/s represents approximately the severity of a vehicle impacting a
guardrail at 100 km/h and an angle of 25 when the
vehicle is aligned with the post. The simulations were
conducted over 150 ms time after the impact onset and
the results are summarised in Table 2.

Average load
(kN)
35.4
40.4
42.9
46.2
48.1
51.9
55.0
57.3
59.6

Observation
in the post
Stopped
Stopped
Stopped
Stopped
Stopped
Stopped
Override
Override
Override

Results of the simulations show that the average load

increases with the speed from 35.4 kN, for the lower
speed of 3 m/s, to 59.6 kN, corresponding to higher
speed of 10 m/s, whereas the peak load increases from
43.3 kN to 106.6 kN, respectively. The load-time history
shows a rst peak for all speeds that occurs between 7
and 9 ms, Figure 5. This initial peak is more pronounced
for speeds higher than 4 m/s. A linear regression shows
that the relation of the peak and average loads are given:
Fpeak D 10:795 V
Faverage D 3:439 V C 26:863

3.0 m/s
4.0 m/s
4.6 m/s
5.4 m/s
5.9 m/s
7.0 m/s
8.0 m/s
8.9 m/s
10.0 m/s

100

80

60

40

20

0
0

20

40

60

(5)

The regression curves of the peak and average forces

are plotted as function of the speed in Figure 6 with
comparison of the test results reported by Coon et al.
[3]. These ndings show that the linear regression captures accurately the variation of the average loads.

120

Impactor Load (kN)

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

3
4
4.6
5.4
5.9
7
8
8.9
10

Maximum
displacement (mm)
130.1
191.4
233.1
294.6
336.0
435.6
540.1
645.5
795.9

80
Time (ms)

Figure 5. Time histories of the impactor load for different impact speed.

100

120

140

Figure 6. Variation of the maximum and average impactor load as function of the speed.

The maximum displacement increases from 130 mm at

an impact speed of 3 m/s to 796 mm at 10.0 m/s, Figure 7.
For a speed higher than 9 m/s, the energy absorbed by the
soil
post system was lower than the initial kinetic energy
of the impact. Thus, the post was not able to stop the
impactor which continued its trajectory and overrode the
post. A nonlinear regression of the relationship between
the maximum displacement dened in (m) and the impactor speed in (m/s) is shown in Figure 8 and given by:
Dmax D 5:2972 V C 25:839 V
2

studies showed that the shear strength parameters of

cohesionless soil, c and , are not signicantly affected
by the tests speed [19,30]. However, it is well established that the steel post behaviour is inuenced by
the strain rate effect. Following Simunovic et al [27],
we propose to include the rate effect by using the
Cowper and Symonds equation [4]:
 1q !
s0
e_
D 1C
d
s

(6)

900
3.0 m/s
4.0 m/s
4.6 m/s
5.4 m/s
5.9 m/s
7.0 m/s
8.0 m/s
8.9 m/s
10.0 m/s

800
700
600
500
400
300
200
100
0
0

20

40

(7)

where s and s 0 are the quasi-static and dynamic stresses,

respectively, e_ is the strain rate, q and D denote the Cowper and Symonds coefcients. For mild steel, the Cowper

The variation of the guardrail post response with

the impactor velocity could be attributed to the strain
rate effect of the soil and the steel post. Different

Impactor Displacement (mm)

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

INTERNATIONAL JOURNAL OF CRASHWORTHINESS

60

80

100

120

Time (ms)

Figure 7. Variation of the impactor displacement as function of time for different impact speed.

140

A. SASSI AND F. GHRIB

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

Max. impactor displacement (mm)

CAE simulation
Dynamic test
Dmax = 5.2972 V2 + 25.839 V
R = 0.9996

Impactor Speed (m/s)

Figure 8. Variation of the impactor displacement as function of the impact speed.

and Symonds coefcients q and D are estimated to be 5

and 40, respectively [13].
In crash tests, the selection of the appropriate speed to
use in testing guardrails components is very crucial. In
fact, the impact of a single guardrail post by full front
cart should simulate the lateral component of the impact
involving a vehicle with a speed of 100 km/h and an
impact angle of 25o which is theoretically 42 km/h
(11.74 m/s). Experimental tests and nite element simulation shows that the maximum lateral speed of contact
is lower than 11.74 m/s. During the full-scale testing, the
vehicle is positioned at the critical impact point located
between the posts. Because of the energy dissipation during the impact and the friction between the guardrail
system and the vehicle, the speed is reduced from
100 km/h to approximately 92 km/h whereas the impact
angle is reduced from 25 to 22 [23]. Moreover, when
testing a single post, the impact energy is localised over a
smaller area of the post which does not reect the fullscale test condition where the energy is distributed over
a large area. Similar observations have been reported by
Gabauer et al. [9] when evaluating the results of pendulum test conducted to assess the crash performance of
longitudinal barrier with minor damage. These conditions suggest that it is appropriate to use an impact
speed of 7.0 to 9.0 m/s for the component testing instead
of the theoretical value of 11.74 m/s.

4. Effect of the post embedment depth

Simulations of the guardrail post embedded at depth
ranging from 800 to 1300 mm were conducted to assess
the soil reaction. Since the passive reaction applied to
the post increases with the depth, the simulation used a

W152 23.8 post (W6 16) instead of the W152

13.4 (W6 9) post to minimise the post yielding and to
provide better comparison of the post response for the
different embedment depths. The two posts cross sections, W152 23.8 and W152 13.4, have the same
width of 100 mm. However, the W152 23.8 post has
higher cross section parameters such as web thickness,
section area and moment of inertia compared to
W152 13.4 post. This approach has also been used by
Kuipers and Reid [15] who used a post (W152 23.8)
to determine the dynamic properties of soil-post at various embedment depths under impact loading conditions. The authors conducted a series of 10 dynamic
tests on the embedded steel posts at different depths
with a speed xed at 9 m/s, a bumper height of 630 mm
and an embedment varying from 864 to 1092 mm [15].
The impactor displacement time history illustrated in
Figure 9 shows that the displacement decreases with the
depth from 400 mm at 1000 mm embedment to
260 mm for 1300 mm embedment. For a post embedded
at 1000 mm or more in the soil, the impactor was
stopped by the post-soil reaction. However, the impactor
overrides the post for the embedment of 800 and
900 mm. The average load increased from 26 kN for a
depth of 800 mm to 60.9 kN for 1300 mm, Table 3.
These simulation results are in line with the experiments
conducted by Kuipers and Reid [15]. A nonlinear regression of the maximum impactor displacement as a function of the post embedment, Z, shows that the
relationship is given by:
Dmax D 0:4088:Z 1:8332
where Z and D are in metres.

(8)

INTERNATIONAL JOURNAL OF CRASHWORTHINESS

600

Impactor Displacement (mm) e

Embedment 800 mm

500

Embedment 900 mm
Embedment 1000 mm
Embedment 1100 mm

400

Embedment 1200 mm
Embedment 1300 mm

300

200

100

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

0
0

20

40

60

80

100

120

140

Time (ms)

Figure 9. Variation of the impactor displacement as function time for different depth embedment.

Table 3. Summary of the simulations results of the impactor with the post located at different embedment depths.
Depth (mm)
800
900
1000
1100
1200
1300

Peak load (kN)

35.98
46.01
56.17
65.1
70.14
81.41

Maximum displacement (mm)

636.4
486.1
400.3
329.0
293.1
259.7

Figure 10 shows that the peak load increases with the

depth from 64.1 kN to 81.4 kN when the embedment
depth increased from 1100 to 1300 mm. The variations
of the peak and average load as function of the depth follow a linear relationship as reported in Figure 11. An
increase of the embedment depth above 1100 mm leads
to higher post-soil interaction and a reaction force that
exceeds the target force of 50 kN. For the case of a
shorter post, where the embedment is less than 900 mm,
the interaction of the post with the soil is low and the
system might lose its efciency and the impactor overrides the barrier.

5. Effect of the impactor mass

Crash testing of steel posts embedded in soil was conducted with different impactor mass to evaluate the post
behaviour. The impactor is rigid and consists, in general,
of a thick steel pipe sometimes lled with concrete and
mounted to the front of a bogie vehicle or a pendulum at
a given height above ground level.
To assess the effect of the impactor mass, simulations
of the guardrail post embedded in the soil and impacted
by a cart with different masses were conducted. The
spring stiffness representing the soil reaction, the

Average load (kN)

26.4
33.7
40.7
48.3
54.5
60.9

Observation in the post

Not Stopped
Not Stopped
Stopped
Stopped
Stopped
Stopped

damping coefcient of the soil and the calculated concentrated masses remain the same for all the simulations.
The impactor mass varies from 500 to 3000 kg to cover
the range of masses used in the literature. Kennedy et al.
[14] used a pendulum with a mass of 878 kg whereas
Dewey [5] used a cart of 2324 kg. The soil condition of
the Coon et al. [3] is used for the current study. The
impact speed of 7 m/s remains the same for all masses.
Table 4 summarises the results of the conducted
simulations.
The load
deection responses of various impactor
masses are shown in Figure 12. It can be seen that the
guardrail post displacement increases with the mass.
The curves show that the initial peak load is slightly sensitive to the impactor mass and occurs approximately at
a displacement of 35 mm. The maximum load increases
Table 4. Results of the energy dissipation for different mass
impactor.
Mass (kg)
500
946
1500
2000
2500
3000

Max energy dissipated (kJ)

8.14
15.39
23.46
29.46
34.43
37.33

Peak acceleration (g)

60.7
64.1
66.4
67.2
67.7
68.0

10

A. SASSI AND F. GHRIB

90
Embedment 800 mm
Embedment 900 mm
Embedment 1000 mm
Embedment 1100 mm
Embedment 1200 mm
Embedment 1300 mm

80

Impactor Load (kN)

70
60
50
40
30
20
10

20

40

60

80

100

120

140

Time (ms)

Figure 10. Variation of the impactor reaction for different depth embedment.

slightly with the impactor mass from 60.7 kN for a mass

of 500 kg to 68.0 kN for the heavier impactor (3000 kg).
Figure 13 shows that the impactor with a mass between
500 kg to 1500 kg is stopped by the soil post reaction
whereas the impactor overrides the post for a mass
above 2000 kg. Linear regression of the maximum displacement as function of the impactor mass shows that
the relationship is given by Dmax D 0:3494 Mimpactor
where Dmax is the maximum displacement of the impactor and Mimpactor is the mass of the impactor (units: mm,
kg). Figure 14 reports the dissipated energy as function
of the impactor displacement and shows that the dissipated energy increases linearly with the mass of the

impactor. The rate of dissipated energy appears to

remain the same for all impactor masses, suggesting that
the post response is independent of the impactor mass.

6. Effect of compressible blockout

To provide the appropriate safety levels for an errant
vehicle impacting the guardrail system, the safety barrier
should be designed to maximise energy absorption
through the soil-guardrail system interaction and maintain its integrity [21]. The energy could be dissipated
through the soil deformation and through the post Wbeam structure. However, it is easier to control and to

90
Pmax = 0.088 Z - 33.347
R = 0.993

80
70
Impactor load (kN)

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

60

Average load
Max load

50
40

Pavg = 0.0697 Z - 28.953

R = 0.9938

30
20
10
0
700

800

900

1000

1100

1200

Post embedment (mm)

Figure 11. Variation of the impactor maximum load as function of the post embedment.

1300

1400

INTERNATIONAL JOURNAL OF CRASHWORTHINESS

11

80
70

Impactor load (kN)

60
50
40
Mass 500 kg

30

Mass 1000 kg
Mass 1500 kg

20

Mass 2000 kg

10

Mass 2500 kg
Mass 3000 kg

100

200

300

400

500

600

700

800

Impactor displacement (mm)

Figure 12. Effect of the impactor mass on the load
deection of the guardrail post.

maximise the energy dissipation by redesigning the

blockout spacer between the post and the guardrail. For
these reason, collapsible blockout system attached to the
post are proposed and implemented in the baseline
model to replace the traditional wood made blockout.
The current practice uses blockout consisting of a
rectangular wood piece (200 150 360) or steel Wcross section that serves only as a spacer with no energy
dissipation capacity. In the present work, six blockout
shapes, consisting of a longitudinal box member that
offers higher degree of energy absorption per unit mass
and guarantees a stable folding pattern during the
impact, are proposed. The design was partially inspired

from a concept widely used in automotive crashworthiness where the front rails are designed to absorb the
maximum energy during the frontal impact. Fundamental theoretical studies in the area of thin-walled structures have been conducted in the past by Wierzbicki
[31], Jones (1983), Abramowicz and Wierzbicki [1] and
Mahmood and Puluzny [16]. The proposed blockout is
designed to absorb the kinetic energy more efciently
which contributes to reduce the vehicle speed and soften
the impact of an errant vehicle with the guardrail system.
Five proposals with different cross sections and thicknesses were considered and integrated in the baseline
guardrail post prototype as shown in Figure 15. The

900
Mass 500 kg

800

Mass 1000 kg

Impactor displacement (mm)

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

700

Mass 1500 kg
Mass 2000 kg

600

Mass 2500 kg

500

Mass 3000 kg

400
300
200
100
0
0

20

40

60

80

100

Time (ms)

Figure 13. Variation of the impactor displacement for different mass impactor.

120

140

160

12

A. SASSI AND F. GHRIB

40
Mass 500 kg

Energy dissipated (kJ)

35

Mass 1000 kg
Mass 1500 kg

30

Mass 2000 kg

25

Mass 2500 kg
Mass 3000 kg

20
15
10
5

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

0
0

100

200

300

400

500

600

700

800

Impactor displacement (mm)

Figure 14. Energy dissipation of the guardrail post for different mass impactor.

dimensions of these models were similar to the original

rectangular wood piece (200 150 360) and made of
HSLA High Strength Low Alloy) having a Youngs modulus E D 205 GPa, a density D 7850 kg/m3 and a yield
stress s0 D 615 MPa. HSLA 340, which is a highstrength steel commonly used in the automotive area,
has been considered for all the design scenarios to fabricate the blockout system because of its good fatigue
strength, its low price, its ductility and its welding capability. HSLA 340 was chosen despite the fact that Elmarakbi et al. [7] showed that aluminium was better than
steel in terms of energy dissipation for their case. The six
schemes were chosen based either on the simplicity of
the manufacturing process involved, ease of installation
or the expected high energy absorption. The six blockout
components considered in this work consisted of a simplied crash box for frontal impact with a square section.
The considered designs are as follows;
Blockout 1 consisted of three square tubes with 80
80 mm cross-section and a wall thickness of 2
mm. The side length of the enclosed square section was 150 mm. The three tubes were sandwiched between two metal plates of 2 mm
thickness from the same material as the square
tube (Figure 15(a)).
Blockout 2 was the same as blockout geometry # 1 with
the exception that the wall box thickness was
reduced to 1.0 mm (Figure 15(b)).
Blockout 3 was similar system as blockout #1 and #2.
To ensure a desirable folding mechanism of the
tubes and reduce the peak load of the impactor,
the design was pre-triggered symmetrically in

both sides of the tube as shown in Figure 15(c).

The triggers consisted of a cut-out material executed on all sides of the tube.
Blockout 4 The blockout in this case was 100 200
360 mm rectangular tube (Figure 15d) with a wall
thickness of 3.5 mm and made from the same
material as systems 1 and 3. The block was
attached to the post using two bolts simulated as
springs.
Blockout 5 was similar to blockout #4 but the wall
thickness was reduced to be 2.5 mm as shown in
Figure 15e.
To minimise the mesh size effect on the crash simulation results, the same mesh is maintained for the ve
proposed designs. The element size was chosen to be
approximately twice that of the sheet thickness [27],
which offers the best trade-off between accuracy and
computational efciency. The triggers used in the blockout design were simulated by removing specic elements
from the model.
The results of the analysis are summarised in Table 5
and shows that the performance of the different blockout
systems is quite different. Blockout #3, designed with
three longitudinal tubes and triggered on both sides, dissipated the maximum energy (5.31 kJ) whereas the same
design Blockout 1 with a 2.0 mm thickness reacts as a
rigid block similar to the baseline: a peak load of
66.8 kN, an average load of 48.4 kN and a maximum displacement of 330 mm. The vertical tube with a thickness
of 2.5 mm (blockout 4) absorbs 3.6 kJ within the rst
50 ms as shown in Figure 16, and then collapsed on the
post, the peak load increased to 80 kN as shown in

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

INTERNATIONAL JOURNAL OF CRASHWORTHINESS

13

Figure 15. Different blockout layout to absorb energy during the impact: (a) Blockout 1, (b) Blockout 2, (c) Blockout 3; (d) Blockout 4; (e)
Blockout 5; and (f) No blockout.

Figure 17. The same vertical tube of 3.5 mm thickness

(blockout 5) absorbs even less energy, 0.71 kJ, because
only a localised deformation was created at the lower
part of the blockout as shown in Figure 18. The average
load remained almost the same and the maximum
impactor displacement remains similar to the baseline
model (336 mm).

The triggers used in blockout #3 helped the development of a stable asymmetric crushing mode, the
role of the triggers is to force the buckling to be initiated at the specic locations during the crash event
which contributes in attenuation of the rst peak
load from 64.1 to 50.3 kN and reduce the load transferred to the guardrail structure. It is known that

Table 5. Results of the energy dissipation for different blockout designs.

Design #
Block 1
Block 2
Block 3
Block 4
Block 5
Baseline

Energy dissipated (kJ)

0.25
4.44
5.31
3.64
0.71
0

Maximum load (kN)

66.8
65.2
50.3
80.0
62.5
64.1

Average load (kN)

48.4
48.2
43.9
38.9
47.9
48.1

Impactor displacement (mm)

329.8
331.6
378.3
456.6
336.3
336.0

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

14

A. SASSI AND F. GHRIB

Figure 16. Internal energy dissipation for the different crushable blockout systems.

triggers reduce the variability of crash mode and

instability due to imperfections associated to material
heterogeneity, geometry or manufacturing processes.
The crushable system also contributes in reducing the
average load to 43.9 kN and absorbing an energy of
5.3 kJ with an impactor displacement of 378.3 mm.
The performance of the crushable blockout #3 suggests that this design capability is superior in transferring the load to the post during the impact of a
vehicle with a guardrail.

7. Conclusions
A macro-element was used to model the post-soil
interaction to conduct an exhaustive parametric nite

element study and investigate the effects of the different design parameters on guardrails post reaction.
The response of the guardrail post was determined
under different loading conditions and the
load
deection curve for the different parameters
was determined and compared to a baseline test
model. Such parameters include sand density, impactor speed and mass, post depth and the crushable
blockout system.
The study shows that that the peak load and the
impactor displacement increase with the speed following
a linear equation. For high speeds, greater than 9 m/s,
the impactor was not stopped by the post reaction. For
lower speeds, less than 3 m/s, the strain rate effect seems
to be limited.

Figure 17. Variation of the impactor load for different crushable blockout systems.

15

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

INTERNATIONAL JOURNAL OF CRASHWORTHINESS

Figure 18. Behaviour of the crushable system during the impact.

The peak load increases with the depth following a

rst-degree equation whereas the impactor displacement
decreases with depth following a power function. For
embedment depths less 1000 mm, the impactor

overrides the post. For depths higher than 1000 mm, the
post did not show any plastic strain higher than 3%.
The results of the simulation with different densities
of cohesionless soil show that the peak load and the

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

16

A. SASSI AND F. GHRIB

impactor displacement increase with the sand density.

The impactor overrode the post for the loose sand and
was stopped for medium and dense sand. These results
show that the medium-density sand offers the best soil
interaction as the load level remains around 50 kN.
The post rail undergoes higher displacement with the
heavier weight than with the lighter one. However, the
maximum reaction remains quasi constant with the
increase of the impactor mass. Impactors with masses of
500 to 1500 kg were stopped; however, for impactor
masses above 2000 kg, the impactor overrode the post.
The maximum displacement of the impactor increases
linearly with the mass. The rate of energy dissipation
remains the same for all impactor masses.
The crushable system implemented as a replacement
of the conventional rigid blockout system was able to
reduce the peak load to 50 kN and the average load to
43.9 kN and absorb an energy of 5.3 kJ with an impactor
displacement of 378.3 mm. This performance suggests
that the crushable blockout is able to reduce the impactor load transferred to the post during a full crash of a
vehicle impacting the guardrail. The triggered design
crushed better than simple blockout boxes. The triggers
helped initiate a stable asymmetric crushing mode.

Acknowledgments
The work reported in the present paper was supported by
grants from the National Sciences and Engineering Research
of Canada (NSERC) and AUTO21.

Disclosure statement
No potential conict of interest was reported by the authors.

ORCID
Faouzi Ghrib

http://orcid.org/0000-0002-0244-0996

References
[1] W. Abramowicz and T. Wierzbicki, Axial crushing of
multicorner sheet metal columns, J Appl Mech. 56
(1989), pp. 113
120.
[2] M. Anghileri, Current status of EN 1317: U.S.
Europe
test result mutual recognition, Roadside Safety Design
and Devices International, Workshop Transportation
Italy 2012. Research circular, Number E-C172, Transportation Research Board, R. Troutbeck, ed., Milan,
2013, pp. 22
27.
[3] B.A. Coon, J.D. Reid, and J.R. Rhode, Dynamic impact
testing of guardrail posts embedded in soil, Res. Rep. No
TRP-03-77-98, Midwest Roadside Safety Facility, University of Nebraska-Lincoln, Lincoln, NE, 1999.

[4] G.R. Cowper and P.S. Symonds, Strain hardening and

strain rate effects in the impact loading of cantilever
beams, Rep. No 28, Brown University, Division of
Applied Mathematics, Providence, RI, 1957.
[5] J.F. Dewey, A study of the soil-structure interaction
behavior of highway guardrail posts, Master thesis, Texas
Transport Institute, Texas A&M University, 1982.
[6] M. Dreznes, MASH Compared to NCHRP 350, Roadside
Safety Design and Devices, International Workshop
Transportation Italy 2012, Research circular, Number EC172, Transportation Research Board, R. Troutbeck, ed.,
Milan, 2013, pp. 7
10.
[7] A. Elmarakbi, M. Vesenjak, and K. Sennah, Finite-element simulation of an improved trafc W-shape guardrail system in vehicle collision, Proceedings, Canadian
Multidisciplinary Road Safety Conference CMRSC-XV,
Fredericton, NB, 2005, pp. 1
8.
[8] European Committee for Standardization, EN 1317-1
and 2: Road Restraint Systems, Comite Europeen de Normalisation, Brussels, 2010.
[9] D.J. Gabauer, K.D. Kusano, D. Marzougui, K. Opiela, M.
Hargrave, and H.C. Gabler, Pendulum testing as a means
of assessing the crash performance of longitudinal barrier
with minor damage, Int J Impact Eng. 37(11) (2010)
November, pp. 1121
1137.
[10] K. Habibagahi and. J.A. Langer, Horizontal subgrade
modulus of granular soils, in Laterally Loaded Deep
Foundations, ASTM Special Technical Publication 835,
Mosely. Langer and Thompson, eds., American Society
for Testing Materials, West Conshohocken, PA, 1984,
pp. 21
34.
[11] C.E. Hampton, H.C. Gabler, Crash performance of
Strong-Post W-beam guardrail with missing blockouts,
Int J Crashworthiness 17(1) (2012), pp. 93
103.
[12] J. Hubbell, Can EN 1317 and NCHRP 350
MASH be used
interchangeably?, Roadside Safety Design and Devices
International, Workshop Transportation Italy 2012,
Research circular, Number E-C172, Transportation
Research Board, R. Troutbeck, ed., Milan, 2013, pp. 11
21.
[13] N. Jones, Structural crashworthiness, Cambridge University Press, Cambridge, UK, 1989.
[14] J.C. Kennedy, C.A. Plaxico and C.R. Miele, Design development and qualication of new guardrail post, Transp.
Res. Rec., 2006, paper No (1984), pp. 69
81.
[15] B.D. Kuipers and J.D. Reid, Testing of W152x23.8
(W6x16) Steel posts-Soil embedment depth study for
the Midwest guardrail system (non-propriety guardrail
system), MwRSF Research Report No.TRP-03-136-03,
Final report to the Midwest States Regional
pooled fund program, Midwest roadside safety facility, University of Nebraska-Lincoln, Lincoln, NE,
2003, June 12.
[16] H. Mahmood and A. Puluzny, Design of thin walled columns for crash energy management-their strength and
mode of collapse, No. 811302. SAE Technical Paper,
1981, pp. 7
18.
[17] C.A. Plaxico, G.S. Patzner, and. M.H.. Ray, Finite element
modeling of guardrail timber posts and the post-soil interaction. Paper No. 980791, Transportation Research
Board, Washington, DC, 1997.
[18] K.A. Polivka, R.K. Faller, D.L. Sicking, J.D. Reid, J.R.
Rohde, J.C. Holloway, B.W. Bielenber, and B.D.

INTERNATIONAL JOURNAL OF CRASHWORTHINESS

[19]

[20]

[21]

Downloaded by [University of California Santa Barbara] at 21:10 03 May 2016

[22]

[23]

[24]

Kuipers, Development of the midwest guardrail system

(MGS) for standard and reduced post spacing and in
combination with curbs, MwRSF, Final Report 19992004, Research Report No. TRP-03-139-04, 2004,
September.
S. Prakash and S. Venkatesan, Characteristics of dense
sand at different rates of strains, J. Sci. Ind. Res. 1960.
19A(5) 1960, pp. 219
223.
J.D. Reid, D.L. Sicking, R.K. Faller and B.G. Pfeifer,
Development of a New Guardrail System, Transport
Research Record, 1599, TRB, National Research Council,
Washington, DC, 1997.
Z. Ren and M. Vesenjak, Computational and experimental crash analysis of the road safety barrier, Eng fail.
Anal. J. 12 (2005), pp. 963
973.
H.E. Ross, R.P. Bligh and K.K. Mak, Evaluation of roadside, features to accommodate Vans, Minivans, Pickup
Trucks, and 4-Wheel drive vehicles, Report 471, Transportation Research Board, National Research Council,
Washington, DC, 2002.
A. Sassi, Analysis of W-beam guardrail systems subjected
to lateral impact, Ph.D. diss., Civil Engineering, University of Windsor, Windsor, 2011.
A. Sassi and F. Ghrib, Development of nite element
model for the analysis of a guardrail post subjected to
dynamic lateral loading, Int J Crashworthiness. 19(5)
(2014), pp. 457
468.

17

[25] A. Sassi, F. Ghrib and S. Battachargee, An improved subgrade model for the crash analysis of guardrail posts, CSCE
2010 General Conference, Winnipeg, (2010), June 9
12.
[26] X. Shen, L. Guo, L. Yang, X. Du, and P. Cao, Numerical
analysis of impact effect on mechanical behaviour of
strong guardrail system, J. Phys. Conf. Ser. 96(1), (2008),
pp. 1
6.
[27] S. Simunovic, V. Kumar, P. Nukala, J. Fekete, D. Meuleman and M. Milititsky, Modeling of strain rate effects in
automotive impact, SAE transactions. 112(5), (2003),
pp. 733
743.
[28] M. Soltani, T.B. Moghaddam, M.R. Karim and N.H.R.
Sulong, The safety of performance of guardrail systems:
Review and analysis of crash tests data. Int. J Crashworthiness. 18(5) (2013), pp. 530
545.
[29] A. Tabiei, and J. Wu, Roadmap for crashworthiness nite
element simulation of roadside safety structures, Int J
Finite Elem. Anal. Des. 34(2) (2000), pp. 145
157.
[30] R.V. Whitman and K.A. Healy, Shear strength of sands
during rapid loading, J Soil Mech. Found. Div. ASCE.
88(SM2) (1962), pp. 99
132.
[31] T. Wierzbicki, Crushing analysis of metal honeycombs,
Int J Impact Eng. 1(2) (1983), pp. 157
174.
[32] W. Wu and R. Thomson, A study of the interaction
between a guardrail post and soil during qquasi-static
and dynamic loading, Int J Impact Eng. 34(5) (2007)
May, pp. 883
898.