Anda di halaman 1dari 14

Indian Geotech J

DOI 10.1007/s40098-016-0196-4

ORIGINAL PAPER

Evaluation of Appropriate Material Models in LS-DYNA


for MM-ALE Finite Element Simulations of Small-Scale Explosive
Airblast Tests on Clay Soils
Courtney L. Busch1 Rafiqul A. Tarefder1

Received: 13 October 2015 / Accepted: 6 June 2016


Indian Geotechnical Society 2016

Abstract This study conducted finite element simulations cohesive soils. The Pseudo Tensor model compared par-
of a small scale, explosive airblast on clay soil in LS- ticularly well to experimental results and predicted crater
DYNA using multi-material arbitrary LagrangianEulerian dimensions very well. Additional studies with the Geologic
methods. An explosive mass of 100.9 g with a 7.6 cm blast Cap model implementing smoothed particle hydrodynam-
height was simulated in this study under two-dimensional ics are recommended which may resolve the errors asso-
conditions. The study compared four different material ciated with large strain increments.
models to evaluate their ability to simulate explosively-
induced airblast loads on cohesive soils. Material models Keywords Airblast  Clay soil  Cratering  MM-ALE 
included the Soil and Foam pressure dependent strength LS-DYNA  Soil and Foam model  Pseudo Tensor model 
model, the Pseudo Tensor pressure dependent strength FHWA model  Geologic Cap model
model, the FHWA DruckerPrager model and the Two-
Invariant Geologic Cap model. Crater dimensions were
measured from the simulations and compared to the results Introduction
of airblast experiments performed in another study. Three
material models ran to completion with the exception of Physical Response of Soils to Airblast Loading
the Geologic Cap model, which terminated early in the
calculation due to plasticity algorithm errors caused by the An explosive airblast occurs when an explosive is deto-
large strain increments. While the remaining three models nated in air, which generates a shock wave that compresses
somewhat over-predicted crater depth, the diameter the air and travels through the medium until the surface of
dimensions and crater volumes from the Soil and Foam and the ground or a structure is encountered, at which point
particularly the Pseudo Tensor model compared very well energy is transmitted into the structure. Explosive airblast
to the experimental data. The crater shape from the FHWA detonations occurring near earthen materials transmit high-
model did not match particularly well with experimental stress, dynamic loads into the structure that may make
results and was larger than the experimental crater. A earthen structures (slopes walls, etc.) susceptible to failure.
simplified, pressure dependent strength model such as Soil High yield airblast events could induce large scale defor-
and Foam model or the Pseudo Tensor model is therefore mations and potentially result in failure of the earth
recommended for modeling small-scale airblast events on structure. It is therefore of interest to investigate the effects
of airblast loading on earthen materials.
When an explosive airblast occurs over earthen mate-
& Courtney L. Busch rials a crater is usually generated as shown in Fig. 1. A
vallejo@unm.edu
true crater is formed when the initial shock scours and
Rafiqul A. Tarefder compacts the soil [29]. Detonation gases and rarefaction
tarefder@unm.edu
waves are transferred into the soil, ejecting material (which
1
Department of Civil Engineering, University of New Mexico, is termed ejecta) as the gases expand and rarefaction
MSC01 1070, Albuquerque, NM 87131, USA waves reverse direction [6]. A portion of the ejecta falls

123
Indian Geotech J

Fig. 1 Typical crater geometry


from an explosive event

back into the crater and forms the apparent crater, which Most experimental studies that involve blast loading of
is the final crater observed after a blast event. earthen structures have been performed by government
During explosive airblast events on earthen materials, organizations and many of these results are not open source
the soil remains in an undrained condition because loading literature. The majority of explosive airblast data for soils
takes place very rapidly and soil pore pressures cannot in open literature consists of empirically derived relation-
dissipate. The soil response in this high intensity, dynamic ships that relate explosive mass to crater dimensions. Data
loading environment cannot generally be described by for elevated explosive blasts on soils are particularly sparse
conventional soil mechanics, in which loading takes place and data from many different soil types such as gravels,
gradually. For example, the compressibility of the solid sands and clays are typically lumped together without
mineral grains, the soil skeleton and the pore water cannot detailed laboratory characterization tests. One historical
be neglected as is usually assumed for conventional prob- reference from the Defense Nuclear Agency [9] describes
lems [4]. To account for these effects, equations for bulk numerous explosive tests conducted in varying geology to
and constrained compressibility of undrained, saturated study cratering efficiency as a function of geology type.
soils were developed by Blouin and Kwang [4] that Average craters were compared for near-surface bursts in
account for the compressibility of the different soil phases. materials ranging from hard rock to soft sand and clay.
The equation for the fully coupled bulk modulus, which Another well-known, historical study compiled the
accounts for the effects of different soil phases and effec- results of 200 accidental, large magnitude surface explo-
tive stress on volume strain in the soilwater mixture is sions on soils and developed an empirically derived
given in Eqs. (1) through (4). An approach such as this equation that estimated the apparent crater diameter to be
should be used when calculating compressibility parame- equal to eighty percent of the cubed root of the explosive
ters for inputs to numerical modeling of undrained soils charge mass [15]. Predicated crater depth was estimated to
subjected to blast loads. be approximately one quarter of the diameter. Results from
" # high explosive surface blasts in soils were compiled by
Km Ks  KKm Kg s  Kg
Kf Kp Km Ks 1 Vortman [27] and showed that the ratio of apparent crater
Kg2  Km Ks radius to depth increased with increasing explosive energy
Km Ks and varied based on site soil conditions. Ambrosini et al.
Kp Kd  2 [2] performed numerous small scale, elevated explosive
Kg
tests on in situ, fine grained soils and developed an
Kg Kw empirical relationship that showed an increase in crater
Km   3
Kw n Kg  Kw dimensions with increasing explosive mass and decreasing
Kd Km Ks 4 blast height. Ambrosini and Luccioni [3] performed
numerical studies and developed relationships to estimate
where: crater dimensions from spherical surface blasts.
Kf, Kp, Kd, Km = fully coupled, partially coupled,
decoupled, and mixture model undrained bulk modulus, Numerical Modeling of Soils Subjected to Airblast
Ks = bulk modulus of soil skeleton, Load
Kg = bulk modulus of solid grains,
Kw = bulk modulus of water, The soil behavior under blast loading is difficult to capture
n = porosity due to the abrupt increase in pressures and large, sudden

123
Indian Geotech J

 p

strains that develop within the material. Many numerical 1 1 3 3J3
h = lode angle: h 3 sin ,
material models have been developed to attempt to capture 3=2
2J2
the behavior of soils under rapid dynamic loading. A dis- J3 = third invariant of the stress deviator
cussion of a select number of models is provided below.
DruckerPrager Plasticity Model
MohrCoulomb Plasticity Model
The DruckerPrager plasticity model addresses some of the
The classical MohrCoulomb plasticity model is one of the
deficiencies of the MohrCoulomb model by accounting
most commonly used models for describing soil behavior.
for the effects of all principal stresses through the use of
The MohrCoulomb shear strength criterion is described in
stress invariants [10]. This model can provide a good first
terms of maximum and minimum principal stresses and is
approximation of soil failure [7]. The model is smooth
given in Eq. (5). While this criterion is computationally
generalization of the MohrCoulomb model and can cap-
simple, it does not account for the effects of intermediate
ture elastic soil response, failure and elastic unloading after
principal stresses. The three-dimensional MohrCoulomb
yielding. The yield surface is defined by Eq. (7) and plots
yield surface is described in a pressure versus stress devi- p
as a straight line in J2 versus J1 stress space. The surface
ator space and incorporates a lode angle function, K(h) as
plots as a circular cone in three dimensional principal stress
shown in Eq. (6), [1]. The lode angle describes the mag-
space and a circle in the p plane. While this consistent
nitude of the intermediate principal stress, r2 to the mini-
surface is computationally efficient, a circular failure sur-
mum principal stress, r3 and the maximum principal stress,
face in the p plane does not accurately represent some
r1. The lode angle has values of ?30 for triaxial com-
materials such as cohesionless soils [7].
pression since r2 = r3 during this stress state. The lode p
angle has a value of -30 for triaxial extension since ry J2  aJ1  k 7
r2 = r1 in the stress state [20].
where:
The standard MohrCoulomb surface forms a hexagonal
pyramid in principal stress space, which is deficient for ry = yield function,
numerical calculations because of the singularity created at a, k = slope and intercept of yield surface in vs J1 stress
the pressure axis when the shear strength becomes zero. space, respectively,
This area of the pressure axis is important to capture the J = first invariant of the stress tensor,
p1
effects of surface explosions in soils with no confining J2 = square root of the second invariant of the stress
pressure, and having a singularity in the yield surface at deviator tensor
this point can create numerical instabilities and ineffi-
The total strain for the DruckerPrager model is sepa-
ciencies. The standard MohrCoulomb surface also has
rated into elastic and plastic components. Elastic defor-
sharp vertices in the octahedral plane, which can result in
mation occurs when the stress state remains below the yield
additional numerical instabilities.
surface, while plastic deformation accumulates when the
ry s  c  r tan / 0 5 stress state reaches the yield surface. Stress states located
where: above the yield surface are not permitted. Incremental
plastic strain is described through an associated flow rule
ry = yield function, that is normal to the yield surface. The incremental plastic
s = shear stress on the failure plane, strain has a negative volumetric component which allows
r = normal effective stress on the failure plane, dilation to occur at failure in both the DruckerPrager and
c = cohesion, the MohrCoulomb models [10]. This aspect can cause an
u = angle of internal friction unrealistic response when modeling soils, since most soils
p
ry P sin / Kh J2  c cos / 0 6 primarily undergo compaction during shear loading [13].
The Federal Highway Administration (FHWA) soil
where: model adapted the DruckerPrager model for numerical
ry = yield function, simulations to simulate soil-structure interaction of road base
P = pressure (mean stress), soils and safety structures [18]. The model incorporated the
u = angle of internal friction, effects of varying saturation levels, strain hardening, strain
c = cohesion, softening and pore water pressure. A major drawback to the
J2 = second invariant of the stress deviator, FHWA model is that it requires more than twenty input
K(h) = function of angle h in the deviatoric plane: parameters, not all of which can be derived from physical soil
Kh cos h  p13 sin / sin h, testing and must be determined through parametric testing.

123
Indian Geotech J

The FHWA model uses the modified MohrCoulomb D2 = parameter for pore-water pressure before air voids
yield surface [1] shown in Eq. (8). A hyperbola is fit to the are collapsed
standard MohrCoulomb surface, which creates a smooth
surface at the intersection of the pressure axis at zero shear
Simplified Pressure Dependent Strength Models
strength and eliminates the singularity. The AHYP param-
eter controls the shape of the yield surface at the inter-
Pressure dependent strength models can be used to capture
section; if AHYP is set to zero the standard MohrCoulomb
the constitutive response of soils due to the pressure
surface is retained. The shape of the yield surface in the
dependent nature of the soil strength. Strength models of
octahedral plane is corrected through the modification of
this type are also commonly used to capture the response of
the K(h) parameter with a function developed by Klisinski
foam materials. One pressure dependent strength model
[16] as shown in Eq. (9).
q that simulates crushing of soil materials through volumetric
ry P sin / J2 Kh2 AHYP2 sin2 /  c cos / deformations, the Soil and Foam model, is discussed in
Hallquist [13]. While the model is one of the most basic
0
pressure dependent strength models, it is very robust and
8 has been used for a considerable amount of time, thereby
41  e2 cos2 h 2e  12 benefitting and improving from user experience and feed-
Kh back [26].
21  e2 cos h 2e  141  e2 cos2 h 5e2  4e 1=2
The Soil and Foam model is defined by inputs for mass
9
density, shear modulus, and unloading bulk modulus. The
where: model is also defined by three coefficients that define the
shear failure surface of the material and a stressstrain
e = eccentricity parameter describing the ratio of triax-
curve that describes the compressibility. In addition, a
ial extension to triaxial compression strength,
maximum pressure cutoff in tension, which represents
AHYP = fit coefficient to determine how closely the
tensile fracture, is required for the model. If the pressure in
modified surface is fit to the standard MohrCoulomb
an element ever reaches the specified value, tension is no
surface
longer allowed in the element.
The FHWA model modifies the bulk modulus with a The shear failure surface for this material model is
constant, D1, as shown in Eq. (10) to describe the effects of described by the second deviatoric stress invariant, J2 as a
moisture content and air voids in partially saturated soils. function of the mean stress. The failure surface for the
The equation controls the stiffness of the soil before air material can be obtained through triaxial compression
voids are collapsed [21]. The model also simulates the laboratory testing and is defined using the pressure-de-
effects of excess pore water pressure by Eq. (11). An pendent, perfectly plastic yield function shown in Eq. (12),
effective stress is calculated by reducing total pressure by [13]. The yield surface coefficients a0, a1 and a2, are the
pore water pressure and decreasing the soil shear strength. coefficients of the quadratic fit of the J2 versus p yield
Ksk surface curve. The ao term is the y-intercept of the yield
K 10 curve with the J2 axis, a1 is the initial slope of the yield
1 Ksk D1 ncur
curve and a2 is the curvature coefficient of the yield curve
where: [26].
K = bulk modulus modified for moisture content and air
/s J2  a0 a1 p a2 p2 12
voids,
Ksk = skeleton bulk modulus, where:
D1 = parameter that controls the stiffness of the soil
/s = plastic yield function,
before air voids are collapsed,
J2 = second deviatoric stress invariant,
ncur = current porosity
a0, a1, a2 = yield surface coefficients,
Ksk
u ev 11 p = pressure (equal to mean stress, positive in
1 Ksk D2 ncur
compression)
where:
Volumetric deformation of the Soil and Foam model is
u = pore water pressure, described by a material-specific curve of mean stress ver-
Ksk = skeleton bulk modulus, sus volumetric strain. The model requires the stressstrain
ncur = current porosity, curve to be defined by the natural logarithm of volumetric
ev = total volumetric strain, strain as shown in the following equation:

123
Indian Geotech J

 
V Cap plasticity models were advanced to include non-
elog ln 13 linear kinematic hardening [14] and capture the effects of
V0
dynamic loads on geologic materials. The Geologic Cap
where: model is one such inviscid cap plasticity model that
elog = logarithmic strain, incorporates a two-invariant cap and kinematic hardening
V = current volume, for numerical analyses. The model is based on the non-
V0 = initial volume before loading linear kinematic hardening models and the formulations of
Simo et al. [24, 25] and Sandler and Rubin [22]. The failure
Another simplified pressure dependent strength model, surface and the cap are defined by Eq. (14). The total
the Pseudo Tensor Model, is discussed in Hallquist [13]. volume strain is separated into elastic and plastic compo-
This model is commonly used to model soils and other nents with a hardening function related to the plastic vol-
geologic material subjected to impulse loadings and is a ume strain by Eq. (15). More recent advances of this type
robust model for simulating high-strain displacements. The of model include the incorporation of the third stress
model is capable of running in two different modes to invariant, strain rate effects and damage modeling [23].
define the shear failure surface: a simplified pressure-de- p
pendent failure surface and a more complex mode that J2 a  c expbJ1 hJ1 14
utilizes two failure surface functions to describe intact and where:
damaged material. Mode one is generally more suitable to p
capture the response of soils, while the second mode is J2 = square root of the second invariant of the stress
typically used to describe the response of rigid materials deviator tensor,
such as concrete or rock. J1 = first invariant of the stress tensor,
Mode one of the Pseudo Tensor model can be used to a, b, h, and c = failure envelope coefficients
implement a MohrCoulomb yield surface with a Tresca epv W f1  expDX j  X0 g 15
limit. The shear failure surface is defined as a curve of where:
stress difference, (r1 - r3)/2 versus mean stress, p. The
yield surface coefficients a0, a1 and a2, are the coefficients epv = plastic volume strain,
of the quadratic fit to the yield surface curve. The model is W = hardening law coefficient,
also defined by inputs for mass density, shear modulus, and D = hardening law exponent,
Poissons ratio. X0 = hardening law exponent (assumed zero for soil),
The Pseudo Tensor model must be coupled with an j = J1 coordinate of the intersection of cap surface and
equation of state to describe the compaction of the soil. failure surface
This generally takes the form of material-specific, tabu-
lated values of pressure versus volumetric strain. Strain is
defined in terms of the natural logarithm of volumetric Objective and Scope
strain as shown in Eq. (13).
The objective of this study was to compare four different
Cap Plasticity Models material models available in the finite element simulation
software LS-DYNA (LSTC [19] and evaluate their ability
Two-invariant cap plasticity models were initially pro- to simulate explosively-induced airblast loads on cohesive
posed by Drucker et al. [11] to control the plastic dilatency soils. Crater size and shape from the finite element calcu-
and account for hysteresis during hydrostatic loading and lations were compared to a small-scale airblast test per-
unloading [7]. Cap models treat soil and rock as elastic formed on clay soils as described in Busch et al. [5]. Finite
plastic, work-hardening materials. The cap model describes element simulations of the field experiment were per-
the plastic strain that occurs almost immediately after a soil formed and crater geometry and stress response were
is loaded through the use of successive yield surfaces until compared to data obtained from the experimental blasts.
the stress state reaches the ultimate yield surface. The yield Material models compared in this study included:
surfaces intersect the J1 axis to account for plastic behavior
under hydrostatic loading, forming a convex end cap. As 1. Soil and Foam pressure dependent strength model (LS-
the soil continues to yield it exhibits work hardening and a DYNA Model 5)
new yield surface is defined when the stress state moves 2. Pseudo Tensor pressure dependent strength model (LS-
beyond the current yield surface. The cap surface contracts DYNA Model 16)
through a hardening law and controls the dilatency of the 3. Federal Highway Administration (FHWA) Drucker
material [13]. Prager model (LS-DYNA Model 147)

123
Indian Geotech J

4. Two-Invariant Geologic Cap model (LS-DYNA Model flammable liquid nitromethane [12]. Constant blast sus-
25) pension heights of 2.5 and 7.6 cm above the clay surface
were used in the tests and the explosive mass was varied
from 0.9 to 100.9 g. The crater dimensions in the clay
surface pad were measured after each blast using a profil-
Methodology
ing tool. Crater profiles were digitized into an electronic
format using computer aided drafting software [8] and
Finite element calculations were compared to a small-scale
electronically rotated to estimate the three-dimensional
airblast test performed on homogeneous clay soils. The
crater shape and calculate volumes.
experimental program as described in Busch et al. [5]
included laboratory geotechnical characterization tests,
undrained triaxial compression tests, ultrasonic wave Finite Element Simulations
velocity tests, and direct shear tests.
Small-scale explosive airblast tests were conducted on In this study, finite element simulations were performed of
clay soils in the experiments. Crater dimensions and a single airblast test with an explosive mass of 100.9 g and
ground motions were recorded for the experimental blasts. a blast height of 7.6 cm above the clay surface as sum-
The experiment test bed was constructed by excavating a marized in Table 1. Simulations were performed using the
pit in in situ, sandy soils as shown in Fig. 2. A clay column software LS-DYNA (LSTC [19] using a two-dimensional,
was constructed within the excavation using a cylindrical axisymmetric calculation with non-reflecting boundaries as
cardboard form backfilled with homogeneous clay soil. shown in Fig. 3. The finite element mesh used in the cal-
Triaxial geophones were embedded in the clay column to culations is shown in Fig. 4 and was constructed using two-
measure ground vibrations; airblast sensors were arranged dimensional elements that were 5 by 5 mm in size near the
on the ground to measure air overpressure. The annular explosive transitioning to 25 by 25 mm in size near the
space surrounding the clay column was backfilled with the geometry edges. Element sizes were selected to be small
site sandy soils placed in a loosely compacted condition to enough to sufficiently capture crater geometries.
minimize vibration reflections back into the clay column. A Explicit time integration with a multi-material arbitrary
square clay pad was constructed near the ground surface to LagrangianEulerian (MM-ALE) formulation was used to
capture blast-induced crater geometries as shown in Fig. 2. simulate the explosive blasts. The MM-ALE method
The experimental explosive blasts were conducted with allows the simulated materials to flow across elements
spherical explosive charges suspended above the clay and the computational mesh to move in a manner which is
surface. The explosive was a two-part binary explosive not equal to the movement of the material. The moving
composed of solid ammonium nitrate oxidizer and mesh results in fewer instances where multiple materials

Fig. 2 Explosive test bed configuration: a profile view, b plan view

123
Indian Geotech J

Table 1 Experimental explosive blast parameters modeled in LS- occupy a single element, thus reducing computational
DYNA errors and improving accuracy. During an MM-ALE
Parameter Value analysis, material interfaces are reconstructed for each time
step based on volumes of a material occupying an element.
Explosive mass, W 100.9 g
The explosive was modeled using the Jones Wilkins and
Blast height above clay surface, h 7.6 cm Lee (JWL) equation of state (Eq. 16), which is an empir-
ically derived model developed from cylindrical explosive
test data and is widely utilized due to its simplicity and
Non- computational efficiency. The LS-DYNA high explosive
reflecting burn (HEB) material model was used with the JWL
boundary
equation of state. The HEB model describes the pressure in
Explosive Air the high explosive material and utilizes a burn fraction
during detonation simulations to calculate the amount of
0.46 m 0.3 m
reacted explosive material and control the release of
Clay 0.14 m
chemical energy during the detonation [13]. An LS-DYNA
null material and an ideal gas linear polynomial equation of
Symmetry state were used to model the air in the calculations.
plane Sand    
Non-
x R1 V x xE
reflecting pA 1 e B 1 eR2 V
1.22 m
boundary R1 V R2 V V
16
where:
p = blast pressure,
V = current relative volume (volume of material at
pressure p divided by the initial volume of un-reacted
0.15 m
0.15 m explosive),
E = specific internal energy (internal energy per initial
Fig. 3 Geometry and boundary conditions for finite element simu-
lations (not to scale) volume),

Fig. 4 Finite element mesh

Air

Explosive

Clay Sand

123
Indian Geotech J

A, B, R1, R2, x = JWL coefficients based on cylindrical 2. Pseudo Tensor pressure dependent strength model (LS-
explosive testing DYNA Model 16)
3. Federal Highway Administration (FHWA) Drucker-
Four material models available in LS-DYNA were
Prager model (LS-DYNA Model 147)
selected to model the clay and compare the results:
4. Two-Invariant Geologic Cap model (LS-DYNA Model
1. Soil and Foam pressure dependent strength model (LS- 25)
DYNA Model 5)
Input parameters for the material models compared in
this study are presented in Tables 2, 3, 4, 5, 6 and 7. Inputs
were obtained from laboratory test results as described in
Table 2 Geotechnical properties of clayey soils used in experiment
Busch et al. [5]. A few soil properties were estimated based
Parameter Result on engineering judgment by substituting typical properties
Dry density 1.963 g/cm3 based on soil type. The strength inputs for the FHWA
Saturation 98.0 % model deviated slightly from the laboratory test results in
Moisture content 29.0 % order to properly fit the strength envelope by the equations
Liquid limit (LL) 35 used in the model. In addition, some of the more obscure
Plasticity index (PI) 15
input parameters for the FHWA model were obtained from
C0.075 (% passing 0.075 mm) 92 %
an in-depth parametric study performed by Lee [17]. Linear
mm
compressibility was assumed for all models due to the

Table 3 Strength and stiffness properties of clayey soils used in experiment


Parameter Result Obtained from

Cohesion, c 0.021 MPa Triaxial lab testing


Friction angle, u 4.0 Triaxial lab testing
Elastic modulus, E 7.57 MPa Triaxial lab testing
Poissons ratio, m 0.49973 Calculated from wavespeed measurements
Shear modulus, G 2.524 MPa Calculated from E and m
Saturated bulk modulus, K 4673 MPa Calculated by Eq. (1)
Soil skeleton bulk modulus, Ks 355 MPa Estimated
Mineral grain bulk modulus (Kaolinite), Kg 49,000 MPa Wang et al. [28]

Table 4 Input parameters for Soil and Foam material model (LS-DYNA Model 5)
Parameter Input Obtained from

Yield function constant, a0 0.001 MPa2 Fit to strength envelope


Yield function constant, a1 0.0049 MPa Fit to strength envelope
Yield function constant, a2 0.0079 Fit to strength envelope
Tensile pressure cutoff, Pc -0.05 MPa Estimated
Compressibility curve inputs
Pressure, P (MPa) Volume strain, elog

0.0 0.0000 Linear compressibility assumed


100.0 -0.0216
200.0 -0.0437
400.0 -0.0895
600.0 -0.1374
800.0 -0.1878
1000.0 -0.2408
2000.0 -0.5586
3000.0 -1.0272
4000.0 -1.9380

123
Indian Geotech J

Table 5 Input parameters for Pseudo Tensor material model (LS-DYNA Model 16)
Parameter Input Obtained from

Tensile pressure cutoff, Pc -0.15 MPa Estimated


Compressibility curve inputs See Table 4
Strength envelope inputs
Pressure, P (MPa) Yield stress, (r1 - r3) (MPa)

0.0 0.05 Fit to strength envelope


1.0 0.20
2.0 0.36
4.0 0.66
6.0 0.97
10.0 1.59
100.0 15.47
200.0 30.90
300.0 46.32
400.0 61.75
500.0 77.17
600.0 92.60
700.0 108.02
800.0 123.44
2000.0 308.54

Table 6 Input parameters for FHWA material model (LS-DYNA Model 147)
Parameter Input Obtained from

Skeletal bulk modulus, Ksk 0.4673 MPa Estimated as 10 % of K [17]


Specific gravity, csp 2.78 Assumed
Cohesion, c 0.038 MPa Fit to strength envelope
Friction angle, u 5.0 Fit to strength envelope
Residual friction angle, ures 5.0 Assumed
DruckerPrager coeff., AHYP 0.38 MPa Fit to strength envelope

almost fully-saturated, undrained condition of the clay. The possibility could be incompatibility of the material model
sand backfill material was modeled with the Pseudo Tensor with the MM-ALE solver. The material model has
pressure dependent strength model for all simulations. numerous history variables that perhaps could not be
advected by volume fraction averaging over the mesh
elements.
Results and Discussion The simulations utilizing the Soil and Foam, Pseudo
Tensor, and FHWA material models ran successfully to
Numerical simulations with the Soil and Foam, Pseudo completion. The finite element simulations resulted in the
Tensor, and FHWA material models ran to completion and creation of craters in the clay soil models. An example of
generated craters in the clay soils. The Geologic Cap the resulting crater using the Pseudo Tensor model is
material model did not run to completion but produced a shown in Fig. 5a. The approximately hemispherical-shaped
not a number error (NaN) and terminated approximately crater is clearly visible in the figure, with ejected material
0.3 ms into the calculation. This error was likely generated pushed up along the side of the crater to form the crater lip.
due to errors in the plasticity algorithm caused by the large The ejected clay from the blast event is clearly visible
strain increments in the simulation. The time step was along the boundaries of the crater and displaced up above
reduced numerous times and the simulations were re-run to the crater. True crater dimensions were measured by gen-
try and resolve this error with no success. Another erating density contours of the post-blast geometries as

123
Indian Geotech J

Table 7 Input parameters for Geologic Cap material model (LS-DYNA Model 25)
Parameter Input Obtained from

Failure envelope coefficient, a 0.035 MPa Fit to strength envelope


Failure envelope coefficient, b -0.014 MPa-1 Fit to strength envelope
Failure envelope coefficient, h 0.0044 Fit to strength envelope
Failure envelope coefficient, c 0.02 MPa Fit to strength envelope
Tensile pressure cutoff, Pc -0.15 MPa Estimated
Cap surface axis ratio, R 2.0 Assumed from Chen and Baladi [7]
Hardening coefficients W, D, X0 0.0 Linear compressibility assumed
Tensile pressure cutoff, Pc -0.05 MPa Estimated

Fig. 5 Pseudo Tensor model (a) (b)


a blast-induced crater, b density
(g/cm3) contours Displaced
Air 1.96
1.77
Explosive & 1.57
Air 1.37
1.18
0.98
0.79
0.59
Ejecta 0.39
0.20
0.0

Clay
Crater

Sand

shown in Fig. 5b. In the figure the low density, ejected experiment, with DT values 2.2 cm larger than the exper-
material does not appear due to the contour range selection imental DT value for the Soil and Foam Model and only
and the crater shape can be clearly distinguished. 0.2 cm larger for the Pseudo Tensor model. These values
Crater geometries obtained from the numerical simula- correspond to a percent difference of 9.6 and 0.9 % com-
tions were compared to the geometry from the experi- pared to experimental values. The crater depths from the
mental explosive blast. Plots of the numerically-simulated Soil and Foam model and the Pseudo Tensor model were
craters are shown in Figs. 6, 7, 8 and 9. The crater mea- larger than experimental results. The Soil and Foam model
sured from the experimental blast was overlain on the plots crater was 1.9 cm deeper (44.2 % difference) and the
to compare how well the numerically-simulated crater Pseudo Tensor crater was 1.6 cm deeper (37.2 % differ-
shape matched the experimental shape. Measurements of ence) than experimental depths. However, both models
true diameter (DT), apparent diameter (DA) and depth predicted experimental volume very well due to the sim-
(d) were obtained for the simulation results and compared ulated crater shape. The Soil and Foam crater volume was
to experimental results as summarized in Table 8. Percent 10.8 % greater than experimental volumes, while the
difference between experimental and simulated dimensions Pseudo Tensor model volume compared exceptionally well
is shown in Table 9. with only a 0.6 % difference.
The crater shapes from both the Soil and Foam model The crater shape from the FHWA model did not match
and the Pseudo Tensor model compared well to the particularly well with experimental results as shown in
experimental crater as shown in Figs. 6 and 7. The diam- Fig. 8. The crater from the simulation had a central
eters from these models compared particularly well to the depression with a steep slope transitioning to a more

123
Indian Geotech J

Fig. 6 Soil and Foam model 10


crater profile compared to Soil and Foam Model
experimental crater 7.5 Experimental Crater
5

Distance along Y (cm)


2.5 DA = 29.8 cm

DT = 25.0 cm
0

-2.5 d = 6.2 cm

-5

-7.5

-10

-12.5

-15
-25 -20 -15 -10 -5 0 5 10 15 20 25
Distance along X (cm)

Fig. 7 Pseudo Tensor model 10


crater profile compared to Pseudo Tensor Model
experimental crater 7.5 Experimental Crater
5
Distance along Y (cm)

2.5 DA = 29.2 cm

DT = 23.0 cm
0

-2.5 d = 5.9 cm

-5

-7.5

-10

-12.5

-15
-25 -20 -15 -10 -5 0 5 10 15 20 25
Distance along X (cm)

Fig. 8 FHWA model crater 10


profile compared to FHWA Model
experimental crater 7.5 Experimental Crater

5
Distance along Y (cm)

2.5
DT = 33.0 cm
0
d = 5.4 cm
-2.5

-5

-7.5

-10

-12.5

-15
-25 -20 -15 -10 -5 0 5 10 15 20 25
Distance along X (cm)

123
Indian Geotech J

Fig. 9 Geologic model crater 10


profile at time = 0.3 ms Geologic Cap Model
compared to experimental crater 7.5
Experimental Crater
(model did not successfully run 5
to completion)

Distance along Y (cm)


2.5
DT = 6.0 cm
0
d = 1.1 cm
-2.5

-5

-7.5

-10

-12.5

-15
-25 -20 -15 -10 -5 0 5 10 15 20 25
Distance along X (cm)

Table 8 Simulated crater dimensions compared to experimental crater dimensions


Material model True diameter, DT (cm) Apparent diameter, DA (cm) Depth, d (cm) Volume, V (cm3)

Experiment 22.8 25.5 4.3 930.1


Soil and Foam 25.0 29.8 6.2 1030.6
Pseudo Tensor 23.0 29.2 5.9 936.1
FHWA 33.0 n/a 5.4 741.2
Geologic Cap n/a (calculation did not run to completion)

Table 9 Difference between simulated crater and experimental crater dimensions


Material model Percent difference experimental versus simulated
True diameter, DT (%) Depth, d (%) Volume, V (%)

Soil and Foam 9.6 44.2 10.8


Pseudo Tensor 0.9 37.2 0.6
FHWA 44.7 25.6 20.3
Geologic Cap n/a (calculation did not run to completion)

tapered crater slope as shown in the figure. This result Crater dimensions from the Geologic Cap model are not
deviated from the experimental crater, which was more discussed or presented in Tables 8 or 9 since the model did
hemispherical in shape. The tapered outer slopes of the not successfully run to completion. However, the crater
crater from the FHWA model resulted in a larger diameter shape at the time of the simulation crash (t = 0.3 ms) is
than the experimental crater, with DT values 10.2 cm larger shown in Fig. 9. In the figure the crater is beginning to
than experimental values and a corresponding difference of form and appears to be hemispherical in shape, but it is not
44.7 %. The simulated crater was also 1.1 cm deeper possible to conclude what the final crater dimensions
(25.6 % difference) than the experimental crater. The would be.
volume of the FHWA crater was 20.3 % larger than Based on the discussion above, the Pseudo Tensor soil
experimental values. The FHWA model did not form ejecta model matched the experimental crater shape best and
during the simulations and there was no visible crater lip. had the lowest percent difference as compared to exper-
Therefore, apparent diameter (DA) values are not provided imental crater dimensions. To investigate the acceptability
in Table 8. of the stress propagation from the simulated blast using

123
Indian Geotech J

Air 0.03 ms 0.04 ms 0.08 ms 0.13 ms

Explosive

Clay surface pad

Fig. 10 Pressure contours from the Pseudo Tensor model simulation

this soil model, pressure (mean stress) contour plots from material model produced a not a number error (NaN)
the simulation were created as shown in Fig. 10. The and terminated approximately 0.3 ms to the calculation.
initial part of the blast can be seen at time equal to This error was likely generated due to the large strain
0.03 ms with a spherical pressure pulse propagating away increments causing errors in the plasticity algorithm. Due
from explosive. The pressure wave can be seen interact- to this early termination, results from this model could not
ing with the clay surface at 0.04 ms and propagating into be used in the material model comparison.
the clay surface pad at 0.08 and 0.13 ms. The pressure While the remaining three models somewhat over-pre-
pulse remains approximately spherical in shape as it dicted crater depth, the diameter and volume values from
propagates into the soil, which is what was expected due the Soil and Foam and Pseudo Tensor models compared
to the hemispherical shape of the experimental crater. very well to the experimental data. The crater depth from
Ground surface interaction can be seen in the later plots at the Soil and Foam model had a 44.2 % difference in depth
0.08 and 0.13 ms as the air pressure changes shape at the but had only a 9.6 % difference in diameter and 10.8 %
soil/air boundary. difference in overall crater volume as compared to exper-
imental results. The crater depth generated by Pseudo
Tensor model had a somewhat lower difference in depth
Conclusions and Recommendations than the Soil and Foam model with a percent difference of
37.2 % from experimental crater depth. The model had
This study conducted finite element simulations of a small extremely good predictions of crater diameter and depth,
scale, explosive airblast on clay soil in LS-DYNA using with a difference of only 0.2 % for diameter and 0.6 % in
multi-material arbitrary LagrangianEulerian (MM-ALE) volume compared to experimental dimensions. Pressure
methods. The study compared four different material contours were constructed for the Pseudo tensor model and
models to evaluate their ability to simulate explosively- were found to compare well to the stress pulse shape
induced airblast loads on cohesive soils. Material models anticipated for the experimental conditions based on the
included the Soil and Foam pressure dependent strength hemispherical craters observed from the field experiment.
model, the Pseudo Tensor pressure dependent strength The crater shape from the FHWA model did not match
model, the FHWA DruckerPrager model and the Two- well with experimental results and had a central depression
Invariant Geologic Cap model. Crater dimensions were with a steep slope transitioning to a more tapered crater
measured from the simulations and compared to the results slope, which resulted in an overall larger crater than pre-
of airblast experiments performed in another study [5]. dicted. This result did not compare well to the experimental
The MM-ALE simulation method predicted the crater, which was more hemispherical in shape. Percent
response of explosive airblast loading on clay soils for the differences for the FHWA model compared to experi-
Soil and Foam model and Pseudo Tensor model as shown mental results were 25.6 % for depth, 44.7 % for diameter
by the formation of hemispherical craters and the ejecta and 20.3 % for volume.
pushed away from the crater and along the crater walls. Based on the results of this study, a simplified, pressure
The FHWA model produced a crater shape that was not dependent strength model such as Soil and Foam model or
similar to the experimental shape and also did not create the Pseudo Tensor model is recommended for modeling
ejecta to form an apparent crater. The Geologic Cap small-scale, explosive airblast loads on cohesive soils. The

123
Indian Geotech J

Pseudo Tensor model compared particularly well to 11. Drucker DC, Gibson RE, Henkel DJ (1957) Soil mechanics and
experimental results and predicted crater dimensions very work hardening theories of plasticity. Trans Am Soc Civ Eng
122(1):338346
well. Furthermore, the simplified, robust nature of these 12. Manufacturing Hallowell (2009) KinepakTM technical data sheet.
types of models along with the associated historical user Hallowell Manufacturing, LLC, Columbus
feedback and improvement make these models well-suited 13. Hallquist JO (2012) LS-DYNA theoretical manual. Livermore
for simulating soil response to blast-induced dynamic Software Technology Corporation, Livermore
14. Isenberg J, Vaughan DK, Sandler IS (1978) Nonlinear soil
loading. structure interaction. Electric Power Research Institute, Report
Recommended future work includes the simulation of EPRI NP-945, Weidlinger Associates
the additional experimental blasts from Busch et al. [5] 15. Kinney GF, Graham KJ (1985) Explosive shocks in air, 2nd edn.
with a pressure dependent strength model to determine if Springer, Berlin
16. Klisinski M (1985) Degradation and plastic deformation of con-
the material model captures the effects of a wider range of crete. Ph.D. dissertation, Institute of Fundamental Technology
explosive masses and offset heights. Additional studies Research (IFTR) Report 38, Polish Academy of Sciences
with the Geologic Cap model are also recommended to 17. Lee WY (2006) Numerical modeling of blast-induced liquefac-
determine if the model can be used to describe the response tion. Ph.D. dissertation, Brigham Young University, Provo, UT
18. Lewis BA (2004) Manual for LS-DYNA soil material model 147.
of cohesive soil to airblast loading. In particular, an FHWA-HRT-04-095, Federal Highway Administration, Research
approach implementing smoothed particle hydrodynamics and Development, McLean
(SPH) should be investigated as this formulation type is 19. LSTC (2013) LS-DYNA Version SMP D R7.0.0, Livermore
compatible with the Geologic Cap model. Utilizing the Software Technology Corporation, Livermore
20. Mellegard KD, DeVries KL, Callahan GD (2005) Lode angle
SPH method would eliminate the need for a mesh grid and effects on the deformational properties of natural rock salt. In:
may resolve the errors associated with large strain incre- Proceedings of the 40th U.S. symposium on rock mechanics
ments seen in the MM-ALE calculations of this study. (USRMS): rock mechanics for energy, mineral and infrastructure
development in the Northern Regions, Anchorage, Alaska
21. Reid JD, Coon BA, Lewis BA, Murray YD (2004) Evaluation of
LS-DYNA soil material model 147. FHWA-HRT-04-094, Fed-
eral Highway Administration, Research and Development,
References McLean
22. Sandler IS, Rubin D (1979) An algorithm and a modular sub-
1. Abbo AJ, Sloan SW (1995) A smooth hyperbolic approximation routine for the cap model. Int J Numer Anal Methods Geomech
to the MohrCoulomb yield criterion. Comput Struct 3:173186
54(3):427441 23. Schwer LE, Murray YD (1994) A three-invariant smooth cap
2. Ambrosini RD, Luccioni BM, Danesi RF, Riera JD, Rocha MM model with mixed hardening. Int J Numer Anal Methods Geo-
(2002) Size of craters produced by explosive charges on or above mech 18:657688
the ground surface. Shock Waves 12:6978 24. Simo JC, Ju JW, Pister KS, Taylor RL (1988) An assessment of
3. Ambrosini RD, Luccioni BM (2006) Craters produced by the cap model: consistent return algorithms and rate-dependent
explosions on the soil surface. Trans ASME 73:890900 extension. J Eng Mech Eng 114(2):191218
4. Blouin SE, Kwang JK (1984) Undrained compressibility of sat- 25. Simo JC, Ju JW, Pister KS, Taylor RL (1990) Softening response,
urated soil. DNA-TR-87-42, technical report prepared for the completeness condition, and numerical algorithms for the cap
Defense Nuclear Agency, Washington, DC model. Int J Numer Anal Methods Eng (accepted)
5. Busch CL, Aimone-Martin CT, Tarefder RA (2015) Experi- 26. Thomas MA, Chitty DE, Gildea ML, TKindt CM (2008) Con-
mental evaluation of cratering and ground vibration in clay soils stitutive soil properties for Cuddeback Lake, California and
subjected to airblast loading. J Test Eval 43(2):111 Carson Sink, Nevada. CR-2008-215345, National Aeronautics
6. Cooper PW (1996) Explosives engineering. Wiley-VCH Publi- and Space Administration, Langley Research Center, Hampton
cations, New York 27. Vortman LJ (1977) Craters from surface explosions and energy
7. Chen WF, Baladi GY (1985) Soil plasticity theory and imple- dependencea retrospective view. In: Roddy DJ, Pepin RO,
mentation, developments in geotechnical engineering, vol 38. Merrill RB (eds) Impact and explosion cratering. Pergamon
Elsevier Science, New York Press, New York, pp 12151229
8. Syste`mes Dassault (2013) SolidWorks student edition. Dassault 28. Wang ZZ, Wang H, Cates ME (2001) Effective elastic properties
Syste`mes SolidWorks Corporation, Velizy-Villacoublay of solid clays. Geophysics 66(2):428440
9. Defense Nuclear Agency (1979) Nuclear geoplosics sourcebook, 29. Zimmie TF, Abdoun T, Tessari A (2010) Physical modeling of
Volume IV, Part II, empirical analysis of nuclear and high explosive effects on tunnels. In: Fourth international symposium
explosive cratering and ejecta. DNA 6501H-4-2, Washington, DC on tunnel safety and security, Frankfurt, Germany, pp 159168
10. Desai CS, Siriwardane HJ (1984) Constitutive laws for engi-
neering materials with emphasis on geologic materials. Prentice
Hall, Englewood Cliffs

123

Anda mungkin juga menyukai