Академический Документы
Профессиональный Документы
Культура Документы
Ceramic Target
Fig. 7.2 Methodology of add-on armour design optimisation (Glvez and Paradela,
2009)
eventually may be fractured, pulverized, and ejected depending on induced impulse intensity. Another interesting feature is tensile spalling near the opposite surface owing to
reflected tensile wave. These phenomenological characteristics can be studied with insightful observation if a suitable computational framework can accommodate the related
constitutive behaviour with reasonable accuracy. The pseudo-spring SPH framework is
explored in this regard in the following sections after adapting a established damage
model chosen from the varieties reported in literature.
Ceramic Target
7.2.3 Micro-mechanics
Ceramics with strong ionic/covalent bonding, the microstructural parameters such as especially the grain size, shape, and orientation play a key role in determining its basic
mechanical properties like tensile and compressive strengths, hardness, toughness, and
wear properties. With similarity to Hall Petch relationship in metals, ceramic compressive strength and hardness is inversely proportional to the square root of the grain size.
The tensile or flexural strength vary in a similar trend (non-linear decrease in the tensile
strength with increasing grain size). Whereas the fracture toughness is a strong function
of bonding and microstructure. And in addition to grain size, it is influenced by grain
shape, presence of second phases, grain orientation, and porosity. At high loading rate,
failure fronts propagate at speeds approaching to sound speed (more precisely, Rayleigh
wave speed) in the medium and tune according to grain size. But for heavy projectile,
the dependency is not so clearly understood.
Ceramic manufacturing and processing include pressing, sintering, hot isostatic pressing with or without additives. Each of the processing steps can tap flaws potentially and
hence the structural behaviour of the finished product are sensitive to synthesis and fabrication techniques with large heterogeneity and less reliability. Because of low tensile
and fracture toughness (0.5 5MPa m), ceramics are used generally in front of armour
system to primarily absorb the initial kinetic energy from the projectile by carrying
compressive loads predominantly. The damage pattern observed under such loadings
include - extensively fractured but still interlocked debris with high micro-crack density (commonly known as comminuted or mescall zone) in high pressure-high shear
zone, frontal surface with visible macro radial cracks due to profuse dislocations, lateral
cracking owing to material spallation at backside from reflected tensile pulse. Studies
(Ashby and Hallam, 1986; Vekinis et al., 1991) have shown that micro-structural defects
such as pores, cracks, inclusion, inhomogeneities and stress concentration regions like
secondary phases and elasticity mismatch at grain boundaries, triple points and even
sub-grain features like twin, stacking faults may induce tensile micro-crack nucleation
under external loadings by transforming global compression into a population of local
tensile zones. But the effects of such defects distribution on dynamic failure behaviour
are not established well yet and needs more careful investigation. Confining effect inhibits the nucleation and propagation of those micro-flaws - increasing the materials
load carrying capacity, and even may cause the brittle material nominally yield due to
plasticity (Lankford, 1977) when high confinement effect is present in materials with
low flaw density.
Microplasticity is the failure mechanism for the finer grain sizes. But for the larger
grain sizes, where the flaw dimensions are comparable to or greater than the grain sizes,
the most anticipated failure mechanism is Griffith flaw failure.
140
Ceramic Target
7.3.2.1 Simha model
Fahrenthold (1991) developed a continuum model based on the idea of complimentary
energy density that used a second-order tensor to represent anisotropic damage. Simha
et al. (2002) developed a damage model similar to Fahrentholds model, inferred from
bar and plate impact tests on AD 99.5 alumina. In their model, it was assumed that
ceramics comminute at the Hugoniot Elastic Limit (HEL) and yield strength of material
is weighted sum of intact and damaged strength.
y = intact (1 D) + f ailed D +
3
2
(7.1)
0 , 1 , , max and PHEL are material parameters. The term controls the contribution of the effective deviatoric strain rate . This rate dependency is the phenomenological contribution of micro-crack sliding, dislocation activity and grain boundary sliding.
7.3.2.2 John-Holmquist models
Pressure-dependent strength, damage and fracture based on accumulated plastic strain,
significant strength reduction and bulking after fracture, and strain rate effects are incorporated in models developed by Johnson and his co-workers. The first version (Johnson
and Holmquist, 1992), JH-1 model, employs piecewise linear strength envelope with
sudden strength reduction and pressure increases due to bulking just when damage variable D reaches 1. Improvisation (JH-2) was made on this by using analytic smooth
function as the strength envelope to avoid sudden change in strength profile at juncture
of piecewise linear fits (Johnson and Holmquist, 1994). And gradual softening, bulking
due to incremental damage accumulation was employed here. But the actual behaviour
of brittle ceramics is perceived to be more realistically represented by sudden change
in strength after complete damage (D = 1), particularly in plate impact scenario and
demonstrating dwell-penetration transition (Simha et al., 2002). Hence Johnson et al.
(2003) introduced another variety (JHB) by keeping sudden change of strength as per
JH-1 but approximating the strength profile by smoothed analytic function as in JH-2.
Additionally phase-change and hysteresis during unloading were also incorporated. The
pressure estimation independent of internal energy (except bulking pressure), accumulation of damage and effect of strain rate were represented similarly in all three versions.
Pressure is computed by polynomial fit (similar to Hugoniot fit) equation of state
(EOS) with =
1 as,
K + K 2 + K 3 + P
2
3
1
P=
K
1
142
, if, > 1
, otherwise
(7.2)
eq = eq,0 1 +Cln
(7.3)
0
Damage (D) is perceived to be accumulating with the increments in effective plastic
ef f
strain ( pl ) as,
ef f
D=
pl
(7.4)
plf
i,1 PP+T
1 T
i,2 i,1 )
eq,i = i,1 + (PP1P)(P
2
1
i,2
eq, f =
c f P
, T P < P1
,P1 P < P2
,P P2
,0 < P <
f ,max
,P
(7.5)
f ,max
cf
f ,max
cf
(7.6)
The fracture strength is employed only when D reaches 1 with growth law as per equa143
Ceramic Target
tion 7.4, where the fracture strain is estimated as,
plf =
P+T f
Pmax + T max
(7.7)
As evident from the figure 7.4, only input state variable in above equation is pressure
P, all other parameters being material constants determined from combinations of test
procedures.
After complete damage (D = 1), in addition to strength reduction, an increase in
terms of bulking (P) due to volume increase with increased free surfaces in damaged
ceramic is added to P. This pressure increment is determined from decrease in internal
2
ev
elastic energy (U = 6G
, ev and G being von-Mises effective stress and shear modulus
respectively) due to damage as,
P = K1 +
(K1 + P)2 + 2 f K1 U
(7.8)
where, U = Ui U f is the difference in internal elastic energy before and after damage
and f is the fraction (0 f 1) of the internal deviatoric energy loss converted to
potential hydrostatic energy.
eq = eq,i D(eq,i eq, f ) where each of the strength envelope is represented as,
eq,i = A
eq, f = B
P
PHEL
P
PHEL
T
PHEL
M
N
HEL
HEL
(7.9)
(7.10)
144
HEL
4
2
3
HEL = K1 HEL + K2 HEL
+ K3 HEL
+ G
3 1 + HEL
(7.11)
2
3
and HEL = 23 (HEL PHEL )
+ K3 HEL
Then PHEL = K1 HEL + K2 HEL
plf
= D1
P
PHEL
T
PHEL
D2
(0 D 1.0)
(7.12)
And similarly like gradual strength reduction, bulking pressure is added incrementally
as,
Pt+t = K1 t+t +
q
(K1 t+t + Pt )2 + 2 f K1 (U |D(t) U |D(t+t) )
(7.13)
In this model the intact and failure strengths are computed through the following
smooth function of pressure P,
eq,i =
i PP+T
i T
, T P Pi
(PPi )i
,P Pi
i + (i,max i ) 1 e (Pi+T )(i,max i )
145
(7.14)
Ceramic Target
and,
P P
f
eq, f =
"
(PP )
P ( f f )
f
f f ,max
f + ( f ,max f ) 1 e
,0 P Pf
,P Pf
(7.15)
The sudden application of bulking pressure is quantified similar to the procedure adopted
in JH-1 version (equation 7.8), and the fracture strain is estimated as,
plf
= D1
i,max
i,max
n
f
max
(7.16)
Now the additional features include incorporation of phase change and hysteresis
during unloading. At the second phase after threshold compressibility 2 , the pressure
is computed with modified coefficients of EOS as,
P = K 1 ( 0 ) + K 2 ( 0 )2 + K 3 ( 0 )3
(7.17)
ef f
ef f
ef f
(K1 f )2 + 2 f K1 U
(7.18)
ef f
ef f
1
where, K1 = K1 (1 ) + K1 , and f = f 0 with = max
. The effects
2 1
of phase change and that of hysteresis during unloading is shown in the figure 7.7.
Fig. 7.8 Loading and unloading under uni-axial compression (Johnson et al., 2003)
veloped with primary focus on this characteristics. But model parameter identification
being sensitive to reference test configurations, only uni-axial test scenario as in flyer
plate test or shock wave testing is not always preferable. Lankford (1977) tried the identification via multi-axial compression in split Hopkinson pressure bar (SHPB) test but
with expensive features.
Now among the mostly used models with larger share of test data available in literature, Johnson-Holmquist models (JH-1, JH-2 and JHB) hold potential of a realistic
representation. Among them the JH-2 version employs gradual strength reduction with
increments in damage and hence is not chosen to represent sudden brittle behaviour
in the test cases to follow (flyer plate spall test, distal boundary effect in deep target
and conoid formation when backed by metallic plate). JHB version uses an improved
147
Ceramic Target
analytic funtion in computing strength envelope compared to piecewise linear fits in
JH-1. But JHB model also requires additional material parameters due to considerations of phase change and hysteresis effects during unloading. As mentioned earlier,
identification of these parameters requires special attention and even may require tuning
to represent the actual real life behaviour. Avoiding those complexities, here the JH-1
version is chosen to represent the particular test cases considered.
The material considered in all test cases is Silicon Carbide (SiC with reference
density 0 = 3215kg/m3, elastic modulus E = 449GPa, Poissons ratio = 0.16 and
HEL= 11.6GPa), parameters for which are taken from (Holmquist and Johnson, 2002)
and reproduced in table 7.1
Table 7.1 JH-1 model parameters for silicon carbide (SiC) (Holmquist and Johnson,
2002)
Equation of State
K1 =220 GPa
K2 =361 GPa
K3 =0
f =1.0
Damage
Pmax =99.75 GPa
f ,max =1.2
Fig. 7.9 Wave propagation in flyer and plate due to the impact (Hiermaier, 2007)
Now, the two release front will meet at the middle of the target plate resulting in
amplification. Depending on the material and initial impulse this amplified release may
exceed the tensile strength. For most ceramics, the tensile limit is very less ( 1GPa).
Hence even at velocity 150-200 m/s, the ceramic target plate will demonstrate a distinct spall plane perpendicular to the wave motion. The remaining shock and rarefaction
waves are reflected from that new spall plane and subsequently travel back and forth
within the spalled out material (see figure 7.10, as zoomed). This type of failure is common in brittle materials with small tensile to compressive strength ratio and generally
under dynamic loading like impulse on rocks.
Ceramic Target
when the first compressive front reaches the free surface. The signal will then remain at
that amplitude (vmax ) when the first release wave travels towards spall zone. During the
spallation, there will be characteristic drop (vsp ) in the signal (see figure 7.11) until it
rises again as the reflected wave from spall plane reaches this free surface. From this
velocity pullback vsp , the spall stress is computed as 0.50Cl vsp . Though, to correct
for pre-deformation during pre-spall wave propagation corrections was suggested by
Stepanov (1976).
150
100
Dandekar and Bartowski (ARL-TR-2430)
Holmquist and Johnson (JAP 02)
Quan et al (IJIE 06)
Simulation Output
50
0
0
0.5
1.5
2
Time (s)
2.5
3.5
Fig. 7.11 Free surface velocity measured for SiC with a flyer velocity of 148 m/s
The obtained spall signal (with average inter-particle spacing of 0.1955 mm) is compared with that of experiment by Dandekar and Bartkowski (2001), prescribed profile
by Holmquist and Johnson (2002) and the numerical counterpart via AUTODYN by
Quan et al. (2006) in figure 7.11. The pseudo-spring SPH version successfully produces
the conformal characteristic drop (vsp ) by adaptation of piecewise linear JH-1 model.
A point is to note that for the mollifying effect of the SPH parameter h (smoothing
length) and for the used artificial viscosity, the jump in the signal is bit smoothened over
a finite small time-scale as compared to that of experimental evidence.
Apart from that, the SPH framework as a collocation method uses information from
a finite sub-domain bounded by the kernel support (here 2h) and hence is bound to
yield a non-local effect. But brittle materials like ceramic is experimentally evidenced
to demonstrate very small fracture processing zone. A possible way-out is reducing the
h/p ratio (where p: average particle spacing) within the acceptable range of robust
computation.
This type of fine tuning was required in this particular simulation despite reduced
sensitivity of h when using gradient correction (see figure 3.5). But this tuning does not
opposes the claim of pseudo-spring version, that material strength/damage behaviour
would solely be represented by damage algorithm at pseudo-spring level, irrespective of
150
Fig. 7.12 Velocity signal with and without pseudo-spring level damage
These observations imply that the classical SPH without any fracture algorithm can
capture the propagation of waves but can not simulate the failure behaviour even with
adaptive resolution. This proves once again the necessity of using explicit damage in
material-independent kernel based particle simulation. Although the tuning related to h
is believed to be replaceable with adjustment of the failure parameter as was identified
by Quan et al. (2006).
The minute difference between experimental and model output signal (figure 7.11)
is perceived to be due to difference of real-life test scenario and ideal surrounding of numerical test. Free-free condition is hard to achieve completely in real-life experiments
but easily modelled in numerical test environment. Air pressure resistance in front of
the moving plate may be another factor contributing. Further the attenuation of the spall
signal may also happen after ceramic micro-structure changes as the stress wave passes
through it. But this air-resistance or micro-structural changes are not included in the
present simulation regime. When incorporating the macro-level ceramic material model
(Johnson-Holmquist model is taken here), the definition of phase changes and consideration of hysteresis (as in Johnson-Holmquist-Biessel Model) may improve this close
151
Ceramic Target
agreement. However that is subjected to availability of suitable material parameters for
SiC and otherwise more physically-based material model for ceramic.
(a) @ 5 s
(b) @ 8 s
(c) @ 10 s
(d) @ 20 s
Ceramic Target
(a) Phase 1
(b) Phase 2
1
0.8
0.6
0.4
0.2
0
1
0.8
0.6
0.4
0.2
0
1
0.8
0.6
0.4
0.2
0
1
0.8
0.6
0.4
0.2
0
7.7 Closure
1
0.8
0.6
0.4
0.2
0
1
0.8
0.6
0.4
0.2
0
Fig. 7.17 Bulging at back plate @ 22 s with thick and thin Backing
long damage development time within the ceramic. Without any penetration, this phenomenon is called interface defeat. Shockey et al. (1990) found that confined ceramics
were much more efficient in defeating the penetrator this way redistributing the impact
load to a larger area on the surface of the backing plate. Also flow and abrasive properties of the finely fragmented material govern the penetration resistance of confined
ceramics.
And the thickness of backing plate prevents overall bending by providing higher
resistance with higher thickness. The time before significant bulging in the thicker back
plate also influences the conoid angle. Simultaneously, by dissipating larger share of
imparted energy in plastic deformation (figure 7.18a), the thicker back plate improves
the ballistic efficiency of the overall target assembly. The lower residual kinetic energy
in case of thicker backing (figure 7.18b) provides evidence of of that improved efficiency
in this particular configuration
400000
450000
Thick Backing
Thin Backing
400000
Thick Backing
Thin Backing
350000
350000
KE (J)
PW (J)
300000
250000
200000
300000
150000
100000
50000
0
250000
0
5E-06
1E-05
1.5E-05
2E-05
2.5E-05
3E-05
3.5E-05
5E-06
1E-05
1.5E-05
2E-05
2.5E-05
3E-05
3.5E-05
Time (sec)
Time (sec)
7.7 Closure
The localised fracture process zone and subsequent crack interaction in confined ceramic
is investigated with the help of pseudo-spring SPH simulation framework after adapting
the constitutive models by Johnson and Holmquist (1992) (JH-1). The damage model
155
Ceramic Target
uses pressure dependent intact and failure envelope. The failure strength only to be
followed after complete damage (D = 1) accumulation with increments in plastic stain.
Strain rate effect and bulking after complete damage are also fused in the model.
The spall signal from a flyer plate test on SiC disc is investigated to validate the
implementation aspects of the simulation framework. The smoothening effect of SPH
kernel is found to be introducing a non-local effect in the output signal. Hence the
smoothing length (h) is fine tuned for better performance. But the explicit damage algorithm through pseudo-spring analogy is indispensable as that tuned kernel support is
found to be incapable of producing a conformal signal with earlier experimental and
numerical evidences from literature. Moreover the spall plane remained dormant. Another perspective proved that the tuning does not polluted the simulation with un-desired
numerical instabilities.
After successful reproduction of stress wave propagation in spall test scenario, the
distal boundary efffect in a deep ceramic target impacted upon by deformable cylindrical
projectile is investigated. Multi-axial interaction among discrete cracks and eventual
spalling from opposite face with delayed energy dissipation time in an deep target is
simulated with reasonable robustness.
Finally the conoid formation and fragment identification in a ceramic plate backed
by ductile metal plate is successfully simulated. The influence of relative thickness is
qualitatively studied to indicate influence on conoid angle deviation and energy dissipation characteristics of the target assembly and the co-existing deformation in the
deformable steel projectile. This whole strategy has potential to create a virtual experimental paradigm for efficient designing of target system harnessing the advantages of
both brittle and ductile materials by a parametric variation of design parameters and
converging towards the desired performance response.
The inter-layer contact was modelled with springs which are active only in tension
and the strength is guided by that of weaker material. More sophisticated interface definition and use of adhesive layers poses a scope of improvement in the current framework
other than the obvious requirement of accurate constitutive parameter characterization.
Optimum adhesive layer thickness for the best performance as was demonstrated by
Lpez-Puente et al. (2005) to show various amount of spalling with different adhesive
layer thickness is to be investigated in near future.
156