3 vues

Transféré par Nav Mehrgan

dynamics

- Scimakelatex.29531.None
- the guide to 4th grade parents scholastic
- Crack Formation in Structural Slabs on Underwater Concrete
- CVEN2301MOSCourseProfile_2014
- Flaws of Case Study and Experimental Method Research in Psychology
- Psychology 7
- eg6
- Writing a Literature Review
- The Impact of Multigrade Teaching Approach in Teaching of Basic Science
- scimakelatex.11303.Kis+Gza.Tamsi+ron.Bna+Bla
- Shell
- brain assignment
- DTHRlecture 2 Psychology as a Science
- Ttr Final Met Rothstein
- Skelton 2001
- sciencereport final pigatutra suichi
- Ferrill Digital Curriculum Pendulum Lab Revised
- newsletter - week 4
- Untitled
- Chapter 1 Vocabulary

Vous êtes sur la page 1sur 13

journal homepage: www.elsevier.com/locate/ijimpeng

subjected to localised blast loading Importance in numerical simulations

D. Bonorchis, G.N. Nurick*

Blast Impact and Survivability Research Unit (BISRU), Department of Mechanical Engineering, University of Cape Town, Private Bag, Rondebosch 7701, South Africa

a r t i c l e i n f o

a b s t r a c t

Article history:

Received 14 November 2007

Received in revised form 3 March 2008

Accepted 3 March 2008

Available online 15 May 2008

A series of localised blast loading experiments are performed in order to understand how the experimental set-up inuences the impulse imparted to a plate. The imparted impulse is measured using

a ballistic pendulum. The experimental results show that for both rigid and deformable plates the impulse measured by a ballistic pendulum increases as the height of the boundary (clamps) increases.

Signicantly, it is found that although the measured impulse varies as a function of the boundary height,

the plate deformation is unchanged. This suggests that not all of the impulse measured by the ballistic

pendulum resulted in plate deformation. Therefore, in numerical and analytical modelling, the total

impulse from the ballistic pendulum should not simply be applied as a centrally localised pressure load.

Numerical simulations of localised blast loading in combination with the aforementioned experimental

results are used to develop a localised blast loading model. The loading model is a simplied pressure

loading model which only imparts the deformation causing impulse as opposed to the total ballistic

pendulum impulse. The model is validated using an independent set of localised blast load experiments

on clamped mild steel plates. Results obtained using a published localised blast load model are also

compared.

2008 Elsevier Ltd. All rights reserved.

Keywords:

Localised blast load

Impulse transfer

ABAQUS

AUTODYN

Simulation validation

\1.

Introduction

been represented by a constant pressure load acting over an area

equivalent to the charge or burn diameter [14]. Under the assumption of impulsive loading, the pressure is applied over a short

duration which is usually taken as the time it takes a detonation

wave to travel the length of the charge. This simple method has

been used with various levels of success to predict the deformation

and failure of locally (and uniformly) blast loaded plates.

A variation of this method was presented by Bimha [5] (see

Section 3.4) where the blast load was represented by a constant

pressure acting over the charge area with the pressure decaying to

the edge of the plate. Based on comparison with experimental plate

deection proles the decay constant was found for variations in

charge radius as a function of plate radius/width. In a further variation of this method, Balden and Nurick [6] used the program

AUTODYN [7] to obtain a pressure prole for a localised blast load.

The pressure prole was based on the prole of maximum pressure

obtained from the simulations. The pressure magnitude for the

prole was adjusted so that the applied impulse corresponded with

the measured experimental ballistic pendulum impulse.

E-mail address: gerald.nurick@uct.ac.za (G.N. Nurick).

0734-743X/$ see front matter 2008 Elsevier Ltd. All rights reserved.

doi:10.1016/j.ijimpeng.2008.03.003

ballistic pendulum impulse to determine the magnitude of the

pressure prole. This ensures that the total measured experimental

impulse is applied impulsively to the plate in the form of a pressure

load. These methods assume that the total impulse measured by

the ballistic pendulum is transferred to the plate as a centrally

localised load. The focus of this paper is to investigate this assumption in order to determine the proportion of the measured

ballistic pendulum impulse responsible for plate deformation. The

results of this investigation are used to develop a localised blast

load model. The loading model is used in a nite element model of

clamped rectangular plates subjected to localised blast loading. A

comparison of the nite element results with an independent set of

experiments on clamped rectangular plates provides a validation of

the loading model.

An alternative is to model the explosive using a program such as

AUTODYN [7] which uses an Eulerian mesh for the explosive and air

and the JonesWilkinsLee (JWL) equation of state [8] for the

explosive. If the experimental set-up including the clamping

arrangement is modelled (Fig. 1), then a priori knowledge of how

the impulse is imparted to the plate is not required. The geometry

in the simulation will allow any physical phenomena to be reproduced and the impulse will be imparted as in the experiment. Axisymmetric or 3-dimensional models can be used but 3-dimensional

models greatly increase the computational expense. Experiments

such as rectangular plate experiments cannot be reduced to

41

Nomenclature

3_ P

s

3p

I*(r,m)

m

P

T

tb

material stress

equivalent plastic strain

impulse density prole as

function of radius and charge mass

charge mass including 1 g leader

pressure

temperature

burn time

axi-symmetric models and therefore, for practical reasons, developing a simple validated pressure loading model is advantageous.

The paper begins by describing an experimental investigation

to determine the effect of the presence of clamps (boundaries)

and clamp height on both rigid and deformable plates for

localised blast loading. The rigid plate tests are used to determine the effect of clamping arrangement on impulse transfer

as measured by a ballistic pendulum. The effect of the varying

impulse transfer is subsequently investigated in terms of plate

deformation. Axi-symmetric numerical simulation results from

AUTODYN are used to develop a pressure loading model which

is applied to rectangular plates and compared with an independent set of experiments. The results of the investigation

into the effect of the clamping arrangement are incorporated in

the loading model. The results of the proposed loading model

are compared with the results obtained using a published [5]

localised loading model.

subjected to localised blast loading.

2.1. Localised blast load experimental investigation

A series of experiments are performed in order to gain further

insight into localised blast loading of rectangular plates. The experiments are as follows:

Fig. 2. Quarter-symmetry diagrams showing the rigid and deformable plate layouts used to determine the effect of clamp height for a constant charge mass. The quarter disc above

the plate is the explosive at a stand-off of 13 mm.

42

Fig. 3. Photograph showing the ground side and the post-blast deformed machined side of a deformable plate.

To determine the effect of the boundary height (clamp

height) on the impulse transferred for a given charge mass.

To determine the impulse transferred to the exposed area of

the plate only with no boundary effects.

Deformable plate tests

To determine the effect of the boundary height (clamp

height) on the deformation of the plate for a given charge

mass.

Fig. 2 shows the layouts used to determine the effect of clamp

height (for a constant charge mass) on impulse transfer and deformation. In all cases with an exposed area the breadth (B) is

120 mm and the length (L) is 200 mm. The total plate dimensions

are 260 mm 300 mm. Tests with charge masses of 8 g, 12 g and

15 g (excluding the 1 g leader) are performed with either no

clamps, one clamp or two clamps on the loaded side of the plate

(additional charge masses are tested with two clamps). Each clamp

is 16 mm in height. The charge diameter is kept constant at 40 mm

which in this case results in a charge diameter to plate breadth ratio

of one-third. As a result of the diameter being held constant the

charge height increases with increasing charge mass. For the 8 g,

12 g and 15 g charge masses the calculated nominal charge heights

are 4.0 mm, 6.0 mm and 7.5 mm, respectively. The clamp assembly

is attached to a ballistic pendulum which measures the total

transferred impulse (see [1] for details of the experimental set-up).

The deformable plates are machined from 10 mm thick mild

steel plates. The plates are rst ground on one side to provide a at

surface and baseline from which to machine the thickness from the

opposite side. The plates were machined with a CNC milling machine to an average thickness of 3.74 mm 0.02 mm. The material

properties are not required for these plates since they are only used

for comparison with each other. Photographs of the ground side

and the post-blast deformed machined side of the deformable

plates are shown in Fig. 3.

The rigid exposed area plate tests are performed with a charge

mass of 15 g (excluding the 1 g leader). The rigid exposed area plate

is of the same size as the exposed area of the deformable plates and

the plate is sufciently offset from the pendulum attachments so as

to ensure no direct blast loading of the pendulum, as shown in

Fig. 4.

2.2. Localised blast load experimental results

2.2.1. Rigid plates

The experimental impulse versus mass of explosive results of

the localised blast load tests performed on rigid plates with varying

clamp heights is shown in Fig. 5. The rigid plate results are compared with previous deformable plate results to show that the

measured ballistic pendulum impulse is independent of plate deformation. The single clamp rigid plate results () correspond to the

previous single clamp deformable plate (A) trends. Another important point is that the various trend lines clearly show that for the

same charge mass the impulse recorded on the pendulum varies as

a function of clamp height. Increasing clamp height results in increased transfer of impulse to the ballistic pendulum and therefore

the effect of clamp height (experimental conguration) needs to be

taken into account.

For the cases where there is no clamp height, both sets of data

show a reduced impulse as compared with clamped plate tests. The

results of the exposed area rigid plate tests (,) are also shown in

Fig. 5 and the trend is below the rigid plate tests (:) with no clamp.

This is due to the reduced area of the exposed area plate tests

compared to the plate tests with no boundary. These tests give an

indication of the impulse transferred to the plate with no clamp

effects and no direct loading on top of the clamps. This information

is necessary to determine how much extra impulse is transferred to

the plates as a result of the clamps.

To get a quantitative idea of the effect of the clamps and clamp

height, the trends of the 15 1 g case are examined. Taking the one

clamp height case as a baseline the increase (decrease) in impulse

for the other cases is given as

Two clamps: 8.5%.

No clamps: 9.6%.

Exposed area only: 10.0%.

Fig. 4. Photograph of the rigid exposed area plate. A stand-off is used to ensure no

direct blast loading of the pendulum.

impulse to apply a load either analytically or numerically, the experimental conguration used to obtain the impulse value needs to

be considered with great care.

43

Fig. 5. Graph showing the rigid and previous deformable (one clamp) plate impulses as a function of clamp height. The 1 g leader is included in the charge mass.

It has been established that for the same charge mass signicantly different impulse measurements are possible depending on

Fig. 6. Graphs showing the effect of varying clamp heights on deformable plate

response. (a) Graph comparing impulse measured using deformable plates versus rigid

plates. (b) Graph of midpoint deection for a constant charge mass of 15 1 g.

determine whether for the same charge mass, the variation in

impulse results in a variation in midpoint deections for deformable plates.

The deformable plate impulses as a result of varying clamp

heights are compared with the rigid plate results at a charge mass

of 15 1 g as shown in Fig. 6(a). The grouping of the deformable

plate results corresponds to the grouping of the rigid plate results

for varying clamp heights. There is no trend of the deformable

plates transferring more impulse than the rigid plates and vice

versa. For the current deformable plate experiments it is again

shown that, within experimental variation, deformable plates do

not produce different impulse transfers when compared with rigid

plates.

The plate midpoint deections versus the measured impulse for

the various clamp heights at a constant charge mass of 15 1 g are

shown in Fig. 6(b). The experimental results indicate that even

though there is a large variation in impulse transferred to the

ballistic pendulum due to the varying clamp heights, the plate

midpoint deections are unaffected. Therefore for a given charge

mass the clamp height greatly affects the measured impulse

transfer but has no effect on the plate midpoint deection. This

suggests that not all the impulse measured by the ballistic pendulum results in plate deformation. This result is extremely important when representing the blast load with a simplied

impulsive pressure load for analytical and numerical work. Even

though the impulse measured by the ballistic pendulum is a true

measure of the impulse being transferred through the plate, the

Fig. 7. Schematic showing the rigid plate model set-up in AUTODYN for the 15 1 g

case.

44

Table 1

Material properties used in the rigid plate simulations obtained from the AUTODYN material library [7]

PE4 (Equation of state JWL)

r0 (g/cm3)

A (kPa)

1.601

6.0977 108

B (kPa)

R1

R2

CJ detonation

velocity (m/s)

CJ energy/

volume (kJ/m3)

CJ pressure (kPa)

Auto-convert to

ideal gas

1.295 107

4.5

1.4

0.25

8193

9.0 106

28.0 106

Yes

r0 (g/cm3)

T0 (K)

cp (J/kg K)

E0 (J/kg)

1.225 103

1.4

288.2

7.176 102

2.068 105

localised impulsive pressure load. As a result of the clamps, the

applied impulse must be reduced to take account of the measured

impulse which does not cause deformation.

2.3. Localised blast load numerical model

Localised blast loading simulations are performed using AUTODYN-2D v6.1, a state of-the-art nonlinear dynamics modelling and

simulation software package [7]. AUTODYN and similar codes are

often referred to as hydrocodes and are particularly suited to

modelling blast, impact and penetration events. Their power comes

from their ability to handle complex problems where a Lagrange

processor and an Eulerian processor can work side by side on the

same problem. The Lagrange processor uses a mesh which deforms

with the material it contains while the Eulerian processor has

a xed mesh in space which allows the material to move through it.

The Lagrange processor is typically used for solid, continuum

structures while the Eulerian processor is used for modelling gases,

liquids or solids where large deformations are likely to occur [9].

The localised blast loading simulations are performed axisymmetrically in AUTODYN-2D. The reason for using an axi-symmetric model as opposed to a 3-dimensional model is to reduce

computational costs. The axi-symmetric model is used to characterise how the impulse density varies as a function of distance from

the blast centre and to gain insight into the effect of the interaction

between the blast wave and the boundary (clamps). A simplied

pressure loading model is sought to replace the 3-dimensional

modelling of the explosive and plate interaction within reasonable

limits of accuracy.

2.3.1. Rigid plate model

The AUTODYN model used to obtain pressure histories from the

rigid plate test is shown in Fig. 7. The rigid plate is modelled by

not specifying a boundary condition at the edge of the air zone. This

reected pressure on the rigid plate is measured using gauge

points in the air cell on the edge of the air zone. Transmission

boundaries are included at the outer edges of the mesh. These

boundaries are sufciently far away from the plate so that the

majority of the loading on the plate is complete before any material

is lost through the transmission boundaries. The majority of the

central load transfer is assumed to be complete within approximately 40 ms (Fig. 10(d)) and the explosive products have not yet

reached the boundary (Fig. 9).

The size of the Eulerian cells is approximately 0.5 mm 0.5 mm.

The air is modelled using the ideal gas law and the explosive

(PE4 C4) is modelled using the JWL equation of state [8]. The

material properties of the air and the explosive are given in Table 1

and are obtained from the material library in AUTODYN [7]. When

lling the air Eulerian mesh the specic internal energy (E0) of the

air is required and details of this calculation can be found in Refs.

[9,10].

The detonation is modelled using a programmed burn where

the detonation front travels at a constant detonation velocity of

8193 m/s for PE4 [7, 11]. The programmed burn assumes that detonation takes place instantaneously. As the detonation front reaches a material point within the un-reacted explosive material, the

explosive material is instantaneously transformed into gaseous

detonation products together with the release of energy associated

with the chemical reaction [9]. The 1 g leader used with the detonator is included in the model as seen in Fig. 7. Detonation is initiated at the end of the leader furthest from the plate and the

detonation front travels as a planar wave along the leader. After

reaching the end of the leader the detonation progresses into the

explosive disc automatically using the programmed burn

algorithm.

2.3.2. Loading simulation results

The rigid plate simulation results for charge masses of 8 1 g,

12 1 g and 15 1 g are shown graphically in Fig. 8 for various

clamp heights. The results follow the same trend as the experimental results with an increase in impulse recorded for an increasing clamp height at a constant charge mass. To get

a quantitative idea of the effect of the clamps and clamp height, the

trends of the 15 1 g case are examined for the numerical model.

Taking the one clamp height case as a baseline, the increase (decrease) in impulse for the other cases is given as (rectangular plate

experimental values are given in parenthesis)

Two clamps: 9.9% (8.5%)

No clamps: 21.6% (9.6%)

Exposed area only: 33.3% (10.0%) (trend not shown in Fig. 8)

Fig. 8. Graph of impulse versus charge mass for axi-symmetric rigid plate simulations.

The exposed area only and the no clamps case have slightly

different impulses because the no clamps case has a larger area

than the exposed area only case. Localised blast loading is not

uniform and therefore the ratio of impulse of the exposed area only

and the no clamps cases is not directly proportional to their

45

Fig. 9. Diagrams showing the pressure contours of the rigid plate simulations with varying clamp heights at different times. Only the regions above the plate and clamp are shown.

The bottom edge of each contour plot is the axis of symmetry.

case with the clamped cases because there is some loading on the

top surfaces of the clamps and this area is included in the no clamps

case.

The increase in impulse for an increase in clamp height is more

severe for the numerical results due to the fact that they are axisymmetric and that the radius is taken as the half-width of the

rectangular plate. The entire boundary of the axi-symmetric plate is

rectangular plate where only the half-width boundaries are close.

The axi-symmetric plate using the half-width as a radius gives an

upper bound of the effect of boundary conditions on impulse

transfer. Since only a trend of the effect of clamp height is sought

the upper bound is sufcient. Direct comparisons between rectangular and axi-symmetric plates should not be made from this

choice.

46

Fig. 10. Graphs showing the effect of clamp height on the impulse density prole for a charge mass of 15 1 g.

products expand radially they are met in all directions by

a clamp face perpendicular to the radial expansion. This constraint is equal in all directions and does not allow a relief of

pressure as is the case with the rectangular plates used in the

experiments. For the rectangular plates, only the explosive

products interacting with the closest point on the shortest and

longest boundaries do so at right angles. Explosive products

interacting with other points along the boundaries do so at

therefore lower impulse transfer.

Fig. 9 shows the pressure build-up due to the clamped boundary. The pressure build-up acts on the plate near the boundary after

most of the pressure has receded above the plate with no clamp.

This lingering pressure, although of relatively low magnitude, acts

over a long duration and therefore increases the total impulse

transferred to the plate. Graphs of the intermediate impulse density

proles are shown in Fig. 10. The impulse density is dened as the

Fig. 11. Contour of the mapped impulse density prole onto a quarter-symmetric

rectangular grid.

impulse per unit area. The impulse density at the centre of the plate

has almost reached a maximum by the time the impulse density

begins to increase at the boundary as shown in Fig. 10(b) and (c).

The reected pressure from the clamp travels back and forth across

the plate and gradually adds more impulse. This is seen in the nal

impulse density prole in Fig. 10(f) where the plate with two

clamps has a slightly higher impulse density than the plate with

a single clamp which in turn has a higher impulse density than the

plate with no clamp.

2.4. Experimental versus numerical impulse results

In order to compare the rectangular experimental impulse

results with the axi-symmetric numerical impulse results, the

axi-symmetric impulse density results are mapped onto the quarter-rectangular layout as shown in Fig. 11. The impulse is calculated

by creating a rectangular grid of 1 mm 1 mm square elements

and multiplying the area of each element by its impulse density.

The impulse density of each element is determined by calculating

its distance from the plate centre and using the distance to nd the

impulse density from the AUTODYN impulse density prole with

no clamps. An example of the impulse density prole obtained

from AUTODYN is shown in Fig. 13(a).

Once the impulse acting on each element is known the impulses

are summed over the rectangle to obtain the total impulse (which is

multiplied by 4). Fig. 12 shows a graph comparing the

47

Fig. 12. Graph of experimental versus numerical impulses for the exposed area rigid

plate tests.

impulses. The trend of the numerical impulses obtained from

AUTODYN is below the experimental impulses. Given the necessary

simplications and approximations required to model this complicated event the numerical trend is nevertheless acceptable.

Halving the element size in the region adjacent to the plate and

using double precision as opposed to the usual single precision

AUTODYN processor resulted in no change in the simulation results.

2.5. Final form of simplied pressure loading

Axi-symmetric impulse density proles obtained from the

8 1 g, 12 1 g and 15 1 g AUTODYN simulations with no clamps

are each t (using a least-squares minimisation method) with

a function of the form given in Eq. (1). The function is purely a curvet and no physical meaning is attached. Any function which can t the

impulse density prole could be used. The 15 1 g case is shown in

Fig.13(a). The AUTODYN ts are shifted up slightly so that the impulse

they produce over the quarter-rectangular layout (Fig. 11) matches

their respective experimental impulses (exposed area only). The shift

to match the experimental impulse is shown in Fig. 13(b) for the

15 1 g case (the shift is barely noticeable but nevertheless increases

the impulse by approximately 7.7%). The 15 1 g case is chosen as

a baseline impulse density function. The baseline function is multiplied by a scaling factor which is varied until the difference between

the scaled baseline function and the 12 1 g experimentally

matched function is minimised. The process is repeated for the

Fig. 13. Graphs showing how the AUTODYN impulse density distribution is matched to the experimental impulse for the 15 1 g case. (a) A function is t to the AUTODYN impulse

density distribution. (b) The AUTODYN t is shifted up to matched the experimental impulse.

48

Table 2

Constants for baseline 15 1 g impulse density function

I*h (N s/m2)

I*l (N s/m2)

I*0 (N s/m2)

I*1 (N s/m2)

I*2 (N s/m2)

14500

5657.6

68.17

5743.1

153.67

s1 (mm1)

s2 (mm1)

s3

s4

rinI (mm)

0.10288

0.16348

1.22759

0.91143

20.25

scaling factor, the impulse density distribution from the experimentally corrected AUTODYN simulations can be matched.

3. Loading model validation

Fig. 14. Graph of the optimised specic impulse scaling factor and its trend.

(including the 1 g leader) in Fig. 14 together with a linear trend.

The linear trend allows the scale factor to be determined for any

charge mass in the range 8 1 g to 15 1 g. It is not known how far

the linear trend can be extrapolated on either side of the range; this

needs to be further investigated. The scale factor trend and the

baseline function allow the impulse density distribution to be

determined for any charge mass in the specied range. This means

that by simply specifying the charge mass, the impulse density

distribution which corresponds to the correct experimental impulse

can be found. The baseline function (the 15 1 g case) is given as

I * r

I * r

Ih* Il*

1 exp s1 rinI rs3 I1*

Il* I0* exp s2 r rinI s4 I2*

for r rinI

where r (mm) is the distance from the centre and the constants

used to t the function are given in Table 2. The scaled impulse

density function, using the trend from Fig. 14, is given as

I * r; m 0:059m 0:068I * r

(2)

shown in Fig. 15 for the 8 1 g and 12 1 g cases, respectively. The

results show that by multiplying the baseline function by a simple

In order to validate the scaled impulse density function, an independent set of experiments are performed on 3 mm thick,

monolithic single-clamped mild steel plates with an exposed area

of 120 mm 200 mm. The chemical composition of the commercial quality mild steel plates is given in Table 3. The plate material

mechanical properties are characterised based on tensile tests and

Split Hopkinson Pressure Bar (SHPB) tests. The behaviour at quasistatic (8 104 s1), intermediate (8 102 s1) and dynamic

(2.6 103 s1) strain rates is characterised.

The quasi-static and intermediate results are obtained using an

iterative experimental/numerical method in which the tensile tests

are simulated and the material properties adjusted until the experimental forcedeection curves are matched. A more detailed

description of this method is given by Bao and Wierzbicki [12]. The

result of the quasi-static strain hardening characterisation is given

in Fig. 16(a). The high temperature response is based on high strain

rate, high temperature results of a similar mild steel as presented

by Gilat and Wu [13]. The JohnsonCook [14] temperature model is

characterised based on these results as shown in Fig. 16(b).

The strain rate effects are characterised based on a strain of 0.1.

The strain rate effects are additive as opposed to multiplicative and

therefore the strain rate effect causes a parallel shift of the ow

stress curve. With a multiplicative strain rate model, such as

JohnsonCook [14], the work-hardening rate increases as the strain

rate increases which is not always true for BCC metals as discussed

by Liang and Khan [15]. The temperature effects are taken into

account when characterising the high strain rate response because

Fig. 15. Graph showing how the scaled 8 1 g and 12 1 g curves match the shifted AUTODYN ts.

Table 3

Plate material chemical composition specication (ladle analysis, percent) [19]

Thickness, t (mm)

C (max)

Mn (max)

P (max)

S (max)

Si (max)

t < 4.5

t 4.5

0.15

0.25

1.00

1.60

0.035

0.040

0.035

0.035

0.30

0.50

up to 0.1. The effect of including temperature in the strain rate

characterisation is shown in Fig. 16(c).

The nal form of the material model is similar to that of Zhao

[16] except that the coupling term between strain hardening and

strain rate is omitted. The material model as a function of strain

hardening, strain rate and temperature is given as

i

s s3p s 3_ p

f T

(3)

where

p g

a

"

!

!1=C2 #

p

3_ p

3_ p

s 3_ sref C ln p C1 p

3_ 0

3_ 0

f T 1

T Tref

Tmelt Tref

!M

B3p

49

Table 4. The coefcients cp and h are the specic heat and inelastic

heat fraction (TaylorQuinney coefcient), respectively. 100% of the

plastic work is assumed to be converted into heat.

A graphical summary of the material response at various strain

rates is given in Fig. 17. The tensile test results at quasi-static and

intermediate strain rates are only presented up to the Ultimate

Tensile Stress (UTS).

The nite element program ABAQUS/Explicit v6.56 [17] is

used for the numerical model of the monolithic plates. The

numerical model consists of a 1/4 scale model of the clamped

plate as shown in Fig. 18(a). Appropriate symmetry boundary

conditions are included on the symmetry edges. A rigid clamp is

placed at the boundary and a clamping force is applied. The

plate is meshed using 3-dimensional continuum 8-node linear

brick elements with reduced integration and hourglass control

(C3D8R). The clamps are meshed using 4-node 3-dimensional

discrete rigid brick elements (R3D4). Hard contact with separation allowed is dened between the clamp and the plate. Tangential behaviour is included with a friction coefcient of 0.5.

Simulating an actual clamp as opposed to simply constraining

the nodes at the boundary allows slight pull-in of the plate

which is more realistic [18]. The scaled impulse density prole is

applied to the exposed area of the plate by converting it to

a pressure prole. The pressure is calculated as

Fig. 16. Graphs showing the strain hardening, temperature and strain rate response of the plate material. (a) Quasi-static strain hardening response. (b) Temperature response (at

high strain rate from Gilat et al [13]). (c) High strain rate response including temperature effects.

50

Table 4

Final optimised material constants for the material model

A (MPa)

B (MPa)

ssat (MPa)

278.1

438.4

1.0

349.6

19.12

0.274

2253.3

Tref (K)

Tmelt (K)

cp (J/kg K)

0.669

300

1811

452

sref (MPa)

C1

C2

3_ 0 s1

387

0.01

9.81 105

1.7

8.33 104

[5]. The model consists of a constant pressure acting over the

charge radius with a decaying pressure prole to the plate

boundary. The pressure prole is given as

P P0 ;

ra

P P0 ekra ;

I*

tb

(4)

burn time, tb, which is calculated as the charge radius divided by

the detonation velocity. The burn time equals 2.44 ms in this case.

The pressure load is applied using a Fortran coded VDLOAD usersubroutine in ABAQUS/Explicit v6.56. The extent of the deformation for a 15 1 g charge mass is shown in Fig. 18(b) where

the nal midpoint deection is more than 10 times the plate

thickness.

The experimental results for midpoint deection are shown

in Fig. 19 together with a selection of simulation results and

good correlation is achieved. Since the midpoint deections are

a necessary but insufcient condition of validation the deection

prole is also veried. Fig. 20 shows the experimental and numerical deection proles for a 13 1 g and a 16 1 g charge

mass. The simulation results, which correlate well with the experimental midpoint deection trends and deection proles,

validate the pressure loading prole and the scaled applied

impulse.

a<rR

radius and R plate outer radius. The decay constant, k, obtained

using a combined experimental/numerical method is given by

a

k 130 261

a 2

948

0:15 <

a

< 0:6

R

(6)

The units for the decay constant given in Eq. (6) are m1. For an a/R

ratio of one-third the decay constant is 148.3 m1. The central

pressure is found by substituting Eq. (5) into the impulse equation

I 2p

ZZ

Pr; tr dr dt

(7)

distribution as a function of radius and time (in this case the

pressure is constant in time and the time is taken as the burn time,

tb).

The model developed by Bimha [5] is used herein to obtain the

plate deection prole for the 13 1 g case using both the measured ballistic pendulum impulse and the scaled impulse (from

exposed area only plate tests). A comparison of the Bimha models

as well as the resulting plate deection proles is shown in Fig. 21.

The Bimha model and results are compared with the proposed

impulse function and the proposed function shows better correlation in terms of plate deection prole. It is interesting to note that

good correlation is only achieved for the Bimha model in which the

applied impulse has been scaled to the exposed area only impulses.

The results show that the simple Bimha model provides a relatively

Fig. 17. Graphs showing a summary of experimental and numerical material curves for quasi-static (8 104 s1), intermediate (8 102 s1) and dynamic (2.6 103 s1) strain

rate tests.

51

Fig. 18. Diagrams of the 1/4 scale monolithic plate model. (a) Monolithic plate mesh. (b) Monolithic plate deformation for 15 1 g charge mass.

therefore be used as a localised pressure loading model on condition that the impulse is scaled correctly to remove nondeformation causing impulse.

4. Conclusions

A series of experiments have been performed to determine

the effect of the presence of clamps and clamp height on

impulse transfer for both rigid and deformable plates under

localised blast loading. An increase in clamp height in the

experiments was found to increase the impulse measured by

the ballistic pendulum. Deformable plate experiments with

varying clamp height showed that the additional impulse due

to the clamps did not inuence the deformation of the plate.

This implies that not all the impulse measured by a ballistic

pendulum necessarily contributes to plate deformation. This is

an extremely important result and has implications for localised load representation in numerical simulations and analytical modelling. Previous empirical results in terms of

impulse and corresponding damage numbers need to be

interpreted in terms of the experimental set-up in which they

were obtained. In other words, previous empirical results

implicitly contain the effect of using a single clamp (used in

the majority of cases).

Axi-symmetric simulations with AUTODYN were used to develop a pressure loading model which only takes into account the

impulse which results in plate deformation and not the entire

impulse as measured by a ballistic pendulum. The loading model

was applied to rectangular plates and excellent correlation with

experiments was achieved for plate midpoint deection and plate

deection prole. The results of the loading model were compared

Fig. 19. Graph of the experimental and numerical midpoint deections versus impulse

for the monolithic plates.

proposed model was found to correspond better with experimental

results. When used with only the deformation causing impulse, the

loading model by Bimha [5] was found to be a valid simple

approximation.

It is noted that only one experimental conguration (one load

diameter and plate size) has been investigated in this work. The

signicance of the results obtained suggests that the work needs

to be extended to include various charge stand-offs and diameters and various plate congurations. The effect of clamping

arrangement also needs to be investigated for uniform blast

loads.

Acknowledgements

The authors thank Glen Newins and the workshop staff of

the Department of Mechanical Engineering for their technical

support. Special acknowledgement is due to Trevor Cloete for

his insight into high strain rate material modelling. Financial

Fig. 20. Graphs comparing experimental and numerical deection proles using the

proposed impulse density model. (a) 13 1 g charge mass. (b) 16 1 g charge mass.

52

References

Fig. 21. Graphs comparing the Bimha [5] pressure proles and resulting plate deection proles for the full impulse and scaled impulse for a 13 1 g charge mass. The

results are compared with the proposed impulse function. (a) Pressure proles. (b)

Long and short boundary plate deection proles.

(NRF) of South Africa and the CSIR (Defence, Peace, Safety and

Security).

[1] Jacob N, Chung Kim Yuen S, Nurick GN, Bonorchis D, Desai SA, Tait D. Scaling

aspects of quadrangular plates subjected to localised blast loads experiments

and predictions. Int J Impact Eng 2004;30:1179208.

[2] Langdon GS, Chung Kim Yuen S, Nurick GN. Experimental and numerical

studies on the response of quadrangular stiffened plates. Part II: localised blast

loading. Int J Impact Eng 2005;31:85111.

[3] Lee Y-W, Wierzbicki T. Fracture prediction of thin plates under localized impulsive loading. Part I: dishing. Int J Impact Eng 2005;31:125376.

[4] Lee Y-W, Wierzbicki T. Fracture prediction of thin plates under localized impulsive loading. Part II: discing and petalling. Int J Impact Eng 2005;31:1277

308.

[5] Bimha R. Response of thin circular plates to blast loading. MSc thesis; 1996.

[6] Balden VH, Nurick GN. Numerical simulation of the post-failure motion of

steel plates subjected to blast loading. Int J Impact Eng 2005;32:1434.

[7] AUTODYN-2D and 3D v6.1 user documentation. Horsham, United Kingdom:

Century Dynamics Inc.; 2005.

[8] Lee EL, Hornig HC, Kury JW. Adiabatic expansion of high explosive detonation

products. Technical report UCRL-50422. Lawrence Radiation Laboratory, University of California; 1968.

[9] Grujicic M, Pandurangan B, Cheeseman BA. A computational analysis of

detonation of buried mines. Multidiscipline Model Mater Struct 2006;2:

36387.

[10] Grujicic M, Pandurangan B, Huang Y, Cheeseman BA, Roy WN, Skaggs RR.

Impulse loading resulting from shallow buried explosives in water-saturated

sand. J Mater Des Appl 2006;220:115.

[11] Dobratz BM, Crawford PC. LLNL explosives handbook. Lawrence Livermore

National Laboratory; January 1985 [UCRL-52997].

[12] Bao Y, Wierzbicki T. A comparative study on various ductile crack formation

criteria. J Eng Mater Technol 2004;126:31424. Transactions of the ASME.

[13] Gilat A, Wu X. Plastic deformation of 1020 steel over a wide range of strain

rates and temperatures. Int J Plast 1997;13:61132.

[14] Johnson GR, Cook WH. A constitutive model and data for metals subjected to

large strains, high strain rates and high temperatures. In: Proceedings of the

7th international symposium on ballistics, The Hague, Netherlands; 1983. p.

5417.

[15] Liang R, Khan AS. A critical review of experimental results and constitutive

models for BCC and FCC metals over a wide range of strain rates and temperatures. Int J Plasticity 1999;15:96380.

[16] Zhao H. A constitutive model for metals over a large range of strain rates.

Identication for mild-steel and aluminium sheets. Mater Sci Eng 1997;A230:

959.

[17] ABAQUS/Explicit v6.5 user documentation. Providence, Rhode Island: ABAQUS, Inc.; 2004.

[18] Bonorchis D, Nurick GN. The effect of welded boundaries on the response of

rectangular hot-rolled mild steel plates subjected to localised blast loading. Int

J Impact Eng 2007;34:172938.

[19] Iscor data sheet. General purpose and commercial steels: hot rolled steel plate

and strip for general applications. Technical report le reference A2.1, Iscor

at steel; 2004.

- Scimakelatex.29531.NoneTransféré par51p
- the guide to 4th grade parents scholasticTransféré parapi-378350740
- Crack Formation in Structural Slabs on Underwater ConcreteTransféré parGonzalo Diz Lois
- CVEN2301MOSCourseProfile_2014Transféré parJnrules123
- Flaws of Case Study and Experimental Method Research in PsychologyTransféré parBoban Lempic
- Psychology 7Transféré parJoe
- eg6Transféré pardeces
- Writing a Literature ReviewTransféré parGiridhari Chandrabansi
- The Impact of Multigrade Teaching Approach in Teaching of Basic ScienceTransféré parGlobal Research and Development Services
- scimakelatex.11303.Kis+Gza.Tamsi+ron.Bna+BlaTransféré parJuhász Tamás
- ShellTransféré parNesa Stefen
- brain assignmentTransféré parapi-32133818
- DTHRlecture 2 Psychology as a ScienceTransféré partevin_prawl
- Ttr Final Met RothsteinTransféré parNational Education Policy Center
- Skelton 2001Transféré parsidhaji
- sciencereport final pigatutra suichiTransféré parapi-268742364
- Ferrill Digital Curriculum Pendulum Lab RevisedTransféré parAnonymous QvdxO5XTR
- newsletter - week 4Transféré parapi-292190375
- UntitledTransféré parapi-2757236
- Chapter 1 VocabularyTransféré parLailaS17
- Nc 3 Rs Arrive Guidelines 2013Transféré parEcaterina Chiriac
- Critical AnalysisTransféré parTra Lawrence
- x10Transféré parorangramai
- Sample Questions for web.pdfTransféré parJuri Saikia
- Lec34 NotesTransféré parAvinnaash Babylonian
- 0102 2018-1 - task 1Transféré parRossi Andreina
- 1-s2.0-S1877705813001409-mainTransféré parakshay cv
- ps1 - science lesson plan 5Transféré parapi-330883613
- Sesi 3-Research MethodTransféré parKezia N. Aprilia
- lesson 30f5Transféré parapi-319867689

- Conditionals Exercises (Gold Redux)Transféré parCheetoman666
- 2006-01 Junior Mentoring PaperTransféré parNav Mehrgan
- Tuition Fees e FeTransféré parNav Mehrgan
- BWT- GBPTransféré parNav Mehrgan
- Biologyunit 20 2016 (1)Transféré parNav Mehrgan
- 336400-336400-2019-specimen-paper-1Transféré parNav Mehrgan
- GCSE Mathematics 2015 SAMTransféré parHafsa Jalisi
- Nbc 01 SolutionTransféré parNav Mehrgan
- Regent QuizTransféré parNav Mehrgan
- Crack Width With ImagesTransféré parNav Mehrgan
- MicrobesTransféré parNav Mehrgan
- Wave Equation CrosswordTransféré parNav Mehrgan
- Respiration QuestionsTransféré parAnonymous Azxx3Kp9
- Acts 13 QuestionsTransféré parNav Mehrgan
- EHTC2011_ARRIGONITransféré parNav Mehrgan
- VocabTransféré parNav Mehrgan
- English Vocabulary - Compound Adjectives2Transféré parNav Mehrgan
- complex sentenceTransféré parVelmurugan Subramaniam
- P2-1Transféré parNav Mehrgan
- Solve 20Transféré parNav Mehrgan
- PuzzleTransféré parNav Mehrgan
- Faith and Fitness Bible Study-Acts12Transféré parNav Mehrgan
- MAths Y9Transféré parNav Mehrgan
- Mathematics Test-cover Year 9Transféré parNav Mehrgan
- Gcse Maths ProbmTransféré parNav Mehrgan
- KS3 Maths Quiz Number StatisticsTransféré parNav Mehrgan
- 1.1 Assessed HomeworkTransféré parŤhîŁíŃî GÛňÅwãŘĐâńě
- Quadratic sTransféré parNav Mehrgan

- 9 Seismic Study at Subba Oil Field.pdfTransféré parIJAERS JOURNAL
- LcdTransféré parqweawe
- Dhakal, P. Et al (2018). GC-MOSCED a Group Contribution Method for Predicting MOSCED Parameters. Fluid Phase Equilibria. 470, 232-240Transféré parMarco Ravelo
- PVC Pressure FittingsTransféré parFluidra Group
- Ci2010Jul011724424296629 USGS Risk Reduction USGSC 200Transféré parBhavana Valeti
- Adsorption HerbicideTransféré parUJBUK
- boiler description description boiler description description boiler description description boiler description description boiler description descriptionTransféré parAbdulRahimShaikh
- Beam Expanders.docTransféré parVaBi Left
- R-501 Uneven Bed TemperaturesTransféré parTalal Ashraf
- ExtractionTransféré parMujahid Haddad
- Water Ring PumpsTransféré parHari Mohan Gupta
- Tube to Tube Sheet Mockup Qualification NewTransféré paralokbdas
- Nut & Bolt Connection DesignTransféré parChandana Kumara
- Lubrication TheoryTransféré parjane
- How Do Intermolecular Forces Affect Boiling Points and Melting PointsTransféré partrudy
- Will Coherent Take OverTransféré parHatem Abdali
- Chapter 6 - Beam DeflectionsTransféré parJovy Notorio
- k4-Fatigue Fracture AppearancesTransféré parWibowo Chandra
- Problem Solving on High Unburned Carbon Losses Used Dirty Air Test and Isokinetic Coal SamplingTransféré parAnonymous a19X9GHZ
- Finite Element Modeling Strategies for Sandwich Composite Laminates UnderTransféré pargemagdy
- student_solutions_ch10.pdfTransféré parSergio Magalhaes Ferreira
- PIANC 2003 Guidelines for Managing Wake Wash WG41Transféré parcalky117
- MechanismsTransféré parDiala Durubi
- Seminar on Rocket EngineTransféré parSam Samrat
- Fluid Mechanics sample problemsTransféré parDenver Banlasan
- Thesis - Shellac Film StabilityTransféré parSatish Vaidya
- el niñoTransféré parCarlos Javier Tejada Diaz
- Removal of Nitrate From Aqueous Solution by Modified SugarcaneTransféré parJorge Fitzgerald Vergara Rojas
- Ch2-3Transféré parZia Ur Rehman
- ManholeTransféré parRajAnand