scieee Science in your language
[en] (orig)

A Stable Hybrid Potential–SPH Technique to Enforce the Fluid Incompressibility

Abstract

The SPH method has extensively used in fluid flow simulation. Through SPH the fluid is modelled by a particles system whose mutual interaction is weighed by a function, named kernel function, whose limits define the neighbouring of each particle. In spite of the high capabilities of SPH for simulate complex environments, it shows shortcomings specially if the fluid is subject to high changes in the pressure, the velocity and the density as occur in phenomena such as shock–tube, blast–wave or in the boundary and discontinuities where the number of neighbour particles is relatively low. In this case, the pressure gradient is inaccurate. As consequence, the simulation is instable with an erratic behaviour of particles. To avoid this problem, we propose a hybrid technique. This one consists in formulating the pressure gradient from a potential defined on each particles pair. Thus, the pressure gradient is immune to the low number of neighbour particles. Also, our proposal allows enforcing the fluid incompressibility. To show the improvements obtained we will carry out a set of simulation

Read accessible full text

A Stable Hybrid Potential–SPH Technique to Enforce the Fluid Incompressibility

Author: Perea Rodríguez, Juan José; Cordero Valle, Juan Manuel
Publisher: World Scientific and Engineering Academy and Society (WSEAS)
Year: 2018
Source: https://idus.us.es/bitstreams/08c08e2c-7882-4b77-b918-2bc899982fb4/download
A S able Hyb id Po en ial–SPH Technique o En o ce he Fluid
Incomp essibili y
JUAN J. PEREA
Uni e si y o Se ille
School o Compu e Enginee ing
A da. Reina Me cedes, s/n, 41012 Se illa
SPAIN
[email p o ec ed]
JUAN M. CORDERO
Uni e si y o Se ille
School o Compu e Enginee ing
A da. Reina Me cedes, s/n, 41012 Se illa
SPAIN
[email p o ec ed]
Abs ac : The SPH me hod has ex ensi ely used in luid low simula ion. Th ough SPH he luid is modelled by
a pa icles sys em whose mu ual in e ac ion is weighed by a unc ion, named ke nel unc ion, whose limi s de ine
he neighbou ing o each pa icle. In spi e o he high capabili ies o SPH o simula e complex en i onmen s, i
shows sho comings specially i he luid is subjec o high changes in he p essu e, he eloci y and he densi y
as occu in phenomena such as shock– ube, blas –wa e o in he bounda y and discon inui ies whe e he numbe
o neighbou pa icles is ela i ely low. In his case, he p essu e g adien is inaccu a e. As consequence, he
simula ion is ins able wi h an e a ic beha iou o pa icles. To a oid his p oblem, we p opose a hyb id echnique.
This one consis s in o mula ing he p essu e g adien om a po en ial de ined on each pa icles pai . Thus, he
p essu e g adien is immune o he low numbe o neighbou pa icles. Also, ou p oposal allows en o cing he
luid incomp essibili y. To show he imp o emen s ob ained we will ca y ou a se o simula ions.
Key–Wo ds: luid simula ion, SPH, s abili y, accu acy, po en ial–based.
1 In oduc ion
The luid low simula ion plays an essen ial ole in
se e al disciplines bo h he scien i ic and echnical.
F om an analy ic poin o iew, i is o mula ed by
Fluid Mechanics ha se he pa ial di e en ial equa-
ions desc ibing he luid low. Howe e , i s com-
plexi y only allows ob aining analy ic esul s o a
es ic ed se o cases wi h high spa ial symme y.
In mos cases, he nume ical me hods mus be used.
These ones a e s udied by Compu a ional Fluid Dy-
namic (CFD). Th ough CFD, he pa ial di e en ial
equa ions, a e ans o med in o a se o algeb aic
equa ions, whose solu ion allow simula ing he luid
low. One o hese me hods is Smoo hed Pa icle Hy-
d odynamics (SPH) which is widely used o i s adap -
abili y and e sa ili y [21, 32].
SPH ope a es on luid disc e iza ion h ough pa -
icles ha in e ac among hemsel es. Thus, i is based
on he Lag angian o mula ion. Desc ip i ely, each
pa icle depic s a disc e e po ion o he con inuum
luid and mo es wi h he low. In his me hodology,
he dynamic quan i ies a e i on each pa icle and i s
mu ual in e ac ion is weighed by a scala unc ion de-
penden on dis ance. This unc ion, known as ke -
nel unc ion, mus sa is y a se o analy ical p ope -
ies such as i mus be mono onically dec easing, i
mus be no malized and i s domain, named suppo ed
domain, should be limi ed. Each pa icle has go an
associa ed ke nel unc ion, whose suppo ed domain
limi s he numbe o neighbou pa icles.
The SPH me hod has highligh ea u es such as
he conse a ion o mass [13, 19, 21], i s adap abili y
and e sa ili y o simula e complex en i onmen s [19]
and he simpli ica ion o he o mula ion o sol e he
luid dynamic equa ions [9, 13]. Ne e heless, despi e
hese ad an ages, he simula ion by SPH can show
s abili y p oblems ha a ec he ealism o simula-
ion [3, 6, 28]. The e a e se e al ins abili ies ypes:
he local mixing ins abili y (LMI) [25], he hea ing
ins abili y [25], bu he mos ema kable is ela ed o
p essu e g adien [20, 24, 25]. The inaccu acy o p es-
su e g adien induces a ensile ins abili y [14, 18, 21]
o pai ing o pa icles [2, 3, 24].
The e is a consensus ha ela es he ins abil-
i y o p essu e g adien and he numbe o neigh-
bou ing pa icles, specially when his numbe is el-
a i ely low o he pa icles dis ibu ion is no smoo h
[3, 10, 21, 24, 28]. Acco ding he numbe o neigh-
bou ing pa icles dec eases, he inaccu acy o p es-
su e o ce is inc eased. Consequen ly, he pa icles
a e sca e ed and show e a ic beha iou . Usually, i
has been ega ded ha inc easing he adius o he
suppo ed domain he inaccu acy is educed. Ne -
e heless, he inc ease in he suppo ed domain e-
WSEAS TRANSACTIONS on FLUID MECHANICS
Juan J. Pe ea, Juan M. Co de o
E-ISSN: 2224-347X
50
Volume 13, 2018
duces he esolu ion and, as consequence, he accu-
acy [3, 21, 32]. Fu he mo e, he e a e egions, such
as he bounda ies o luid o i s discon inui ies, whe e
he numbe o neighbou pa icles canno be inc eased
[28]. O he me hodologies ha e been de eloped o
a oid he p oblems associa ed o he low numbe o
neighbou ing. These echniques ei he e o mula e
he dynamic equa ions modelled by SPH [21, 25] o
posi ioning con ol poin s o educe he e ec s o he
ex e nal loads on he p essu e g adien [4, 33]. Ne e -
heless, hese p oposals show sho comings since can
iola e he conse a ion o momen um [3, 16], pa ic-
ula ly in bounda ies and discon inui y egions. Also,
hey a e only e ec i e unde speci ic condi ions, i
hese condi ions a e no sa is ied, an e oneous o e -
damping is induced [3].
To a oid hese sho comings ela ed o he p es-
su e o ce and he incomp essibili y en o cemen ,
we p opose a hyb id echnique po en ial–SPH. Suc-
cinc ly, ou p oposal is based on a o mula ion o a
conse a i e po en ial ha allow us o de i e he p es-
su e o ce. The cons ain s o ob ain his o ce has
he aim he imp o ing he s abili y and accu acy spe-
cially in he discon inui ies and bounda y egions. On
he o he hand, he o mula ed p essu e o ce allow us
o en o ce incomp essibili y wi hou he need o use
a p edic o –co ec o p ocess wha imp o es he e i-
ciency. To show he imp o emen s ha ou p oposal
allows ob aining, we will ca y ou a se o es usually
used o quan i y he s abili y and accu acy.
2 SPH Me hod
The SPH me hod, ini ially de eloped by Gingold and
Monaghan [5] and Lucy [11], was o mula ed om
he in eg al o mula ion o he del a Di ac unc ion
[13].
( ) = Z ( 0)δ( − 0)d 0,(1)
whe e depic s any scala unc ion and δis he del a
Di ac unc ion.
Acco ding o he in e pola ion heo y, he unc-
ion δcan be eplaced by a unc ion wi h spa ial ex-
ension. Thus, he in eg al exp ession (1) can be e-
o mula ed by he below equa ion.
( ) = ZΩ
( 0)W( − 0, h)d 0+O(h2),(2)
whe e Wis named he ke nel unc ion, Ω e e s o he
olume o he de ini ion domain o Wand his he
adius o he suppo ed domain.
Rega ding he ke nel unc ion, his one mus sa -
is y, a leas , he ollowing cons ain s:
1. I mus be no malized and end o δwhen h ends
o ze o, i.e:
lim
h→0W( − 0, h) = δ( − 0); ZV
WdV 0= 1.
(3)
2. I mus be posi i e and dec ease con inuously
wi h ( − 0).
3. I mus be symme ic wi h espec o ( − 0).
Quan i a i ely, he modelling o a con inuous
medium using a pa icles sys em en ails ans o ming
he in eg al, equa ion (2), by a sum whe e he mass o
each pa icle depic s a olume elemen whose mass is
ρdV , i.e:
h ( )i=ZΩ
( 0)
ρ( 0)W( − 0, h)ρ( 0)d 0
≈X
j∈N(i)
mj
j
ρj
W( j− i, h),(4)
whe e hi e e s o he app oxima ed alue o unc ion
since he second o de e m has no been conside ed,
N(i)depic s he neighbou pa icles jo each pa icle
iand ρjis he mass densi y alloca ed o he pa icle j.
F om he symme y p ope y (i em 3), he g adi-
en o any scala magni ude ( )can be simpli ied as:
h∇ ( )i=∂
∂ Z ( 0)
ρ( 0)W( − 0, h)ρ( 0)d 0
≈X
j∈N(i)
mj
j
ρj∇W( j− i, h).(5)
Gene alizing he equa ion (5), i is possible o o -
mula e a alid equa ion o any di e en ial o de . This
gene alized equa ion is:
h∇l ( i)i=X
j∈N(i)
mj
j
ρj∇lW( j− i, h),(6)
whe e l e e s o he di e en ial o de equa ion.
I is no ed ha he equa ion (6) does no ega d he
second o de e m [21]. Also, i iola es he conse a-
ion o angula momen um [13]. As consequence, al-
hough he equa ion (6) sa is ies he SPH p emises, he
accu acy o he ob ained esul s can be comp omised.
To p e en his issue, his equa ion mus be modi ied
o be applied o luid low simula ion, acco ding is de-
sc ibed by [13]. Ne e heless, his symme ical o -
mula ion can inc ease he ins abili y [2, 7, 21].
WSEAS TRANSACTIONS on FLUID MECHANICS
Juan J. Pe ea, Juan M. Co de o
E-ISSN: 2224-347X
51
Volume 13, 2018
3 Dynamic Equa ions
The dynamic o an incomp essible and in iscid luid,
whose low e ol es adiaba ically wi h high Reynolds
numbe is desc ibed by he below PDEs sys em:
∇· = 0,(7)
D
D =−1
ρ(∇p) + F, (8)
whe e D/D =∂/∂ + ·∇ e e s o he con ec i e
de i a i e, ρis he mass densi y, is he eloci y, pis
he p essu e and Fdepic s he body o ce ec o pe
uni olume.
To close his sys em o equa ions, i is necessa y
o ega d a ela ionship be ween he p essu e and he
mass densi y. This ela ionship is se by he equa ion
o s a e (EOS). Al hough he e a e se e al o mula-
ions o he EOS, in SPH is usually used Tai ’s equa-
ion exp essed as:
p= (γ−1)ρ, (9)
whe e γis he adiaba ic index.
The modelling o he equa ions (7) and (8) by
SPH shows p oblems o s abili y and accu acy. To
sol e hese p oblems, se e al p oposals ha e been de-
eloped [15, 21, 23, 30]. The mos ou s anding a e he
esea ch de eloped by P ice [21], appoin ed as MPM,
and he echnique desc ibed by Inu suka [7] known as
GPSH.
Rega ding he a ia ional app oach, P ice [21]
models he equa ion (8) as:
d i
d =−X
j∈N(i)
mj(χi∇Wij(hi) + χj∇Wij(hj)) +
X
j∈N(i)
αmi
ρij
s ij ·ˆ
Rij∇˜
Wij +F(10)
whe e he subsc ip iand j e e s o pa icle iand j
espec i ely, χ= (Ω−1p/ρ2),Ωis a co ec ion e m
ela ed o he smoo hing leng h, ˆ
Rij is he uni ec o
de ined by he line joining he pa icles iand j,˜
Wij
is he a e aged ke nel and s=ci+cj− ij ·ˆ
Rij.
In P ice [21] is shown a de ailed explana ion o he
de i a ion o his equa ion.
On he o he hand, he echnique GSPH is seeking
o es o e consis ency o he disc e e densi y es ima e
by he ke nel unc ion con olu ions using a Riemann
sol e . Acco ding o he GSPH echnique, he equa-
ion (8) is o mula ed as:
d i
d =−X
j∈N(i)
mjp∗
ij hV2
ij(hi)∇Wij(hi√2)+
V2
ij(hj)∇Wij(hj√2)i
≈ − X
j∈N(i)
mjp∗
ij 1
ρ2
i∇˜
Wij
1
ρ2
j∇˜
Wij!+F(11)
whe e p∗is he in e media e p essu e a ising om he
solu ion o a Riemman sol e pa icula ized o he pa -
icles pai i, j and Vij a e he in e media e eloci ies
ela ed o he solu ion o Riemann sol e pa icula -
ized o wo pa icles [24]. The pape s [7, 8] show a
de ailed explana ion o he de i a ion o he equa ion
(11).
4 P oposed Model
In his sec ion, i will be desc ibed he ea u es o
ou p oposal o imp o e he s abili y and accu acy o
SPH simula ions. Fi s ly, he undamen al hypo he-
sis o he model will be se . F om his hypo hesis
will be o mula ed ou p oposal o ob ain he p es-
su e o ce. Nex , i will be desc ibed he analy ic ea-
u es ha should be sa is ied by he p essu e o ce o
gua an ee s abili y. These ea u es a e se om he
mos ou s anding s udies in s able simula ions scope
by SPH. F om hese cons ain s and he undamen al
hypo hesis, i will be o mula ed he p essu e o ce.
I s o mula ion will allow en o cing he incomp ess-
ibili y. Once he incomp essibili y is en o ced, i will
be explained he p ocess o calcula e he densi y us-
ing bo h he SPH o mula ion and he p essu e o ce.
This joined o mula ion allows ul illing he incom-
p essibili y cons ain .
Analy ically, when ex e nal o ces ac ing on a
luid, he dis ances be ween adjacen egions change.
Then, in o he luid appea s a o ce ha ends o e-
s o e he balance condi ions [34]. This in e nal o ce
is quan i ied by he p essu e g adien . Ex apola ing
his c i e ion o he modelled luid by a pa icle sys-
em, we can o mula e ou undamen al hypo hesis,
namely: he p essu e o ce is due o change o he dis-
ance be ween pa icles.
Taking in o accoun his p emise, we a oid he
p oblems o s abili y ela ed o modelling o he p es-
su e g adien by SPH, especially when he numbe
o neighbo ing pa icles is low [3, 22]. Fu he mo e,
we can en o ce he incomp essibili y cons ain on he
p essu e o ce and he densi y. On he o he hand, as
he o ce only depends on dis ance, we can o mula e
i om a conse a i e po en ial [34]. Ne e heless,
WSEAS TRANSACTIONS on FLUID MECHANICS
Juan J. Pe ea, Juan M. Co de o
E-ISSN: 2224-347X
52
Volume 13, 2018
any conse a i e po en ial is no alid o ob ain s able
luid simula ions. To se a sui able o mula ion mus
be aking in o accoun he cons ain s ha gua an ee
he s abili y. These cons ain s i will be desc ibed be-
low.
4.1 P essu e Fo ce
Se e al esea ch ha e been ca ied ou o se he ela-
ionship be ween s abili y–accu acy and he p essu e
g adien . The mos highligh ing s udies asse ha he
p essu e o ce:
1. I should ensu e s abili y o a wide ange o
he numbe o neighbou pa icles. This ea u e
can be deduced om he s udies de eloped by
[3, 12, 21, 28]. These esea ch asse ha i he
numbe o pa icles is oo low, he simula ions
show an e a ic sca e ing o pa icles, mainly in
he bounda y o luid. On he o he hand, i he
numbe o pa icles is high, a clus e ing o pa i-
cles can appea .
2. I should be conca e in espec o he abscissa
axis. This cons ain is deduced om he e-
sea ch de eloped by [1, 3, 24], whe e is de-
sc ibed ha a ke nel unc ion whose g adien ,
ha appea in p essu e o ce, shows an in lec ion
poin and conca i y, imp o e he s abili y.
3. I should show an asymme ic p o ile wi h ega d
o he minimum alue o he o ce. Thus, wo
equi emen s a e sa is ied: i s ly, he epulsi e
o ce is inc eased when he dis ance is dec eased,
as i can be deduced om [9] and las ly, he in-
e ac ion can be weighed by he dis ance, as is
sugges ed by [3, 21]. A ha monic o ce can only
sa is y he i s condi ion bu i would no ensu e
he las one [21].
4. I should allow us o en o ce he incomp essibil-
i y [12, 28], namely, i s o mula ion should limi
he minimum dis ance among pa icles. Thus,
he o mula ed o ce should show an asymp o ic
beha io wi h dis ance. I s aim would be o a oid
ha he dis ance be ween pa icles will be incom-
pa ible wi h incomp essibili y cons ain .
F om hese ea u es, we o mula e he conse a-
i e po en ial, om which, we will ob ain he p essu e
o ce. The p oposed po en ial is:
Vi( i) = 21
24πhiX
j∈N(i) 2
(Rij −0.9ε)2−
7π2
2ζ(Rij −(0.85ε)) + (3π2σ)!,(12)
whe e Rij = j− i,εis named as asymp o ic ac o ,
ha allows us o con ol he incomp essibili y a e, ζ
is appoin ed as dep h ac o , by his magni ude, we
can a oid he pa icles clus e ing and σis named as
ail coe icien , by means o his coe icien , i is pos-
sible o con ol he po en ial a he bounda y o he
suppo ed domain. Empi ically, we ha e e alua ed he
alue ange, o hese pa ame e s, ha o e he bes e-
sul s, hese alues a e: 0.3h≤ε≤h,0.5≤ζ≤2.0
and 0.8≤σ≤2.25. The Figu e 1 shows he p o ile
o he o mula ed po en ial.
Figu e 1: G aph o he p oposed po en ial. The pa-
ame e alues a e h= 3,ε= 0.6,ζ= 1.25 and
σ= 1.0.
Analy ically he po en ial, equa ion (12), only de-
pends on posi ion. As consequence, he o ce ha is
de i ed can be w i en as ~
Fpi=−∇V( i), i.e:
~
Fpi=−5.25
6πhiX
j∈N(i) 3.5π2
ζ(Rij −(0.85ε))2−
4
(Rij −0.9ε)3!(h−Rij)~
Rij
Rij
,(13)
F om he Figu e 1, we can desc ibe he ea u es o
ou p oposed po en ial. Fi s ly, i shows a conca i y in
espec o he abscissa, as i was es ablished in i em 2.
Secondly, i shows an asymme ic p o ile wi h espec
o minimum o o ce, acco ding o he i em 3. Las ly,
as a consequence o asymp o ic limi , o low alues
o , we can con ol he incomp essibili y, as i was
equi ed in i em 4. I is no ed, as consequence o he
p essu e o ce is de i ed om his po en ial, his ea-
u es will be inhe i ed. On he o he hand, analyzing
he equa ion (13) we can say ha he o ce is sa u a ed
wi h he dis ance [34]. Fu he mo e, his sa u a ion is
weighed by he dis ance. Thus, he p essu e o ce sa -
is ies he cons ain sugges ed by [3, 13, 17, 32]. Also,
WSEAS TRANSACTIONS on FLUID MECHANICS
Juan J. Pe ea, Juan M. Co de o
E-ISSN: 2224-347X
53
Volume 13, 2018
he o ce o mula ed in equa ion (13) sa is ies he con-
se a ion o momen um.
Then, once he p essu e o ce has been ob ained,
we a e going o desc ibe ou p oposal o con ol he
incomp essibili y, aking ad an age o he asymp o ic
limi shown by he o mula ed p essu e o ce.
Due o he ac ha he p essu e o ce is ob ained
om a po en ial, ou echnique will he eina e be e-
e ed o as P–SPH.
4.2 The En o cemen o he Incomp essibil-
i y
Quali a i ely, he incomp essibili y is ela ed o he
maximum o ce ha can be bo ne by he luid, wi hou
changing i s local olume [34]. Acco ding o his ac ,
he incomp essibili y in a luid modelled by a pa icle
sys em can be explained as he minimum dis ance ha
he pa icles can app oach. In his con ex , he o ce
calcula ed by he equa ion (13), allows us o model
he incomp essibili y by he asymp o ic alue when R
is low. Ne e heless, i he incomp essibili y coe i-
cien is lowe han 2%, we mus in oduce a con ol
p ocess o educe he nume ical e o s accumula ed in
each ime s ep.
Ou p oposal o con ol he incomp essibili y
when i s coe icien lowe han 2% can be di ided in o
wo s ages. In he i s s age, he p essu e o ce is
calcula ed by he equa ion (13). In he las s age, we
check he pa icles whose p essu e o ce iola es he
incomp essibili y cons ain . I he p essu e o ce o
any pa icle iola es he incomp essibili y cons ain ,
hen he o ce excess is quan i ied and bo h he dy-
namic and he posi ion o hese pa icles a e modi ied
o sa is y he incomp essibili y cons ain .
To implemen his p ocess, i s ly we calcula e he
maximum comp ession o ce, namely, he alue om
which he incomp essibili y cons ain is iola ed. To
his end, we pa icula ize he equa ion (13) wi h a dis-
ance, whose alue is in line wi h he es densi y luid.
Acco ding o his desc ip ion, he ob ained equa ion
is:
~
FC
j|i=−5.25
6πh 3.5π2
ζ(ξ−(0.85ε))2−4
(ξ−0.9ε)3!~
R
R,
(14)
whe e ~
FC
j|iis he maximum o comp ession o ce ha
he pa icle jcan acqui e in espec o pa icle iand ξ
is he minimum dis ance ha he pa icle jcan ap-
p oxima e o pa icle iwi hou iola es he incom-
p essibili y. The equa ion ha sa is ies his c i e ia is:
ξ=3
s3m
4π(1 −η)ρ0
,(15)
whe e ηis he comp essibili y coe icien and ρ0is he
es densi y.
I |~
FC
j|i|>|~
Fpj|i|, hen, we ha e o con ol he
incomp essibili y. Thus, we change he o ce o he
pa icle jand i s posi ion, o he maximum ole a ed
alues, i.e.:
~
Fpj|i=~
FC
pj|iy ~ 0
j=ξ~
R
R,(16)
whe e ~
Fpj|iis he p essu e o ce o pa icle jin espec
o pa icle iand ~ 0
jis i s new posi ion.This change only
is alid i he di e ence be ween |FC
pj|i|and |FC
pj|i|is
lowe han 5%. Al hough his limi a ion seems es ic-
i e, as he p essu e o ce con ol he incomp essibil-
i y, i allows us ha he di e ence be ween wo o ces
is lowe han his alue.
5 Resul s
In his sec ion, we will ca y ou a se o es s o show
he imp o emen s ha o e ou p oposal. To ha
end, we will simula e a luid whose dynamic is de-
sc ibed by he equa ions (7) and (8). To sol e his sys-
em o equa ions, i will be ega ded he echniques
MPM [21], GSPH [7] and P–SPH. The aim is o com-
pa e he ob ained esul s om each echnique o show
he imp o emen ha can be ob ained by he p o-
posed echnique. I will be implemen ed h ee es s
commonly used o quan i y he s abili y and accu acy
[24, 26, 27], hese a e: se ling o a andom pa icles
dis ibu ion,Sod’s shock– ube and blas –wa e. In
each es , we will ca y ou an analysis o he ange
o he numbe o neighbou pa icles whe e each ech-
nique is s able.
We ha e ega ded he mos o he alues o sim-
ula ion pa ame e s sugges ed by [24]. Each es has
di e en alues ha be speci ied in each simula ion.
Ne e heless, in all hem, he alue o he adiaba ic
pa ame e γ, equa ion (9), will be γ= 1,46. Rega d-
ing ou p oposed p essu e o ce, equa ion (13), he se-
lec ed alues o εand ζa e ε= 0.45 and ζ= 1.0.
Likewise, all simula ions will be implemen ed using
a Wendland C4as ke nel unc ion. We ha e selec ed
his unc ion due o i s analy ic ea u es ha a ou he
s abili y and accu acy, acco ding is highligh ed by [3].
5.1 Se ling o a Random Pa icles Dis ibu-
ion
Se e al esea ch show a consensous which ega ds he
se ling o a andom pa icle dis ibu ion as a sui able
es o quan i y he s abili y in SPH [3, 21, 29]. De-
sc ip i ely, in his es , a pa icles sys em e ol e om
WSEAS TRANSACTIONS on FLUID MECHANICS
Juan J. Pe ea, Juan M. Co de o
E-ISSN: 2224-347X
54
Volume 13, 2018

P-SPH
Figu e 2: E olu ion om a andom pa icles dis ibu ion o a se led pa icles dis ibu ion.
a andom dis ibu ion o a s a e ha shows a le el o
egula i y. The achie ed egula i y allows quan i ying
he s abili y deg ee o any SPH simula ion [21, 24].
To implemen his es , we ha e used 980 pa icles
whose mass is m= 0.00213. The simula ion limi s
a e [50 ×30] and he ime s ep used ∆ = 1.5 10−4s
and h= 0.023. F om his alues, we ha e ob ained
he simula ion shown in Figu e 2.
To analyse he ange o he numbe o neighbou
pa icle ha allow ob aining s able and accu acy sim-
ula ion, we ha e changed om he lowes numbe o
highes . In his es , he used alues a e:
MPM GSPH P–SPH
Ni120–375 95–375 15–375
Ni180 165 25
Table 1: Range o he numbe o neighbou pa icles.
Ni e e s o ange ha allows s abili y and Niis he
alue ega ded o ob ain he simula ions shown in Fig-
u e 2.
F om he Figu e 2, we can say ha he bes esul
is ob ained om P–SPH. This one allows ob aining
he esul s o he highes egula i y. To quan i y his
la ice deg ee, we ha e o mula ed an o de pa ame e
ς ha sa is ies:
ς=Ps
k=1(|dl−dk
ij|)
s,(17)
whe e s= (m−1)(n−1), and mand na e he pa -
icles numbe in each space di ec ion, dlis he la ice
leng h, dk
ij is he dis ance be ween each pai o neigh-
bou pa icles. F om he o mula ion o he pa ame e
ς, i can be said ha he lowe alue, he mo e la ice
deg ee is achie ed and, consequen ly, he mo e s abil-
i y is ob ained.
The ob ained alues o his pa ame e a e shown
in Table 2.
MPM GSPH P–SPH
ς21.7 16.4 4.7
Table 2: Ob ained alues o he o de pa ame e om
Figu e 2.
5.2 Sod’s Tube–Shock Tes
Sod’s shock– ube es [27] allow quan i ying he accu-
acy and s abili y o SPH [10, 14, 26]. Desc ip i ely,
he shock– ube es is buil by a ube whe e a non–
po ous memb ane is in oduced. This memb ane de-
ines wo sec ions ha a e illed wi h gas. In a sec ion,
ha is named as high sec ion, he gas is a ound 10
imes dense and mo e con ined han in he o he sec-
ion which is named low sec ion. F om hese ini ial
cons ain s, he memb ane is ins an ly aken away, a
= 0. A his momen , due o di e ences o p essu e
and densi y be ween high–low sec ions, a shock wa e
is gene a ed. This one e ol es in o he low sec ion.
A he same ime, a a e ac ion wa e appea s in o he
high sec ion. Be ween each wa e, a con ac discon i-
nui y is a isen. Quan i e i ely, om he ob ained ac-
cu acy in he simula ion o hese h ee egions, i is
possible o quan i y he sui abili y o he used ech-
nique.
To ca y ou his es we ha e used a domain
[-0.5,0.5]. The low sec ion is limi ed be ween
[-0.5,0.0)and he high sec ion is limi ed by
WSEAS TRANSACTIONS on FLUID MECHANICS
Juan J. Pe ea, Juan M. Co de o
E-ISSN: 2224-347X
55
Volume 13, 2018
(a) (b) (c)
(d) (e) ( )
Figu e 3: Compa a i e g aphs o he exac solu ion and nume ical ob ained alues om MPM, GSPH and P-SPH
o Sod’s shock– ube es .
(0.0,0.5]. The emaining used pa eme e s a e he
sugges ed by [24].
F om hese ini ial alues, we ha e ob ained he e-
sul s shown in Figu e 3 a ime =0.15. I is ega ded
he densi y and he eloci y o e alua e he ob ain ac-
cu acy, acco ding o he sugges ion gi en by [24, 26].
The Figu e 3 shows a compa ison be ween he
exac solu ion, solid line, and he nume ical solu ion
shows as do s.
In each g aphs, i is possible o iden i y h ee e-
gion: he con ac discon inui y, he a e ac ion wa e
and he shock wa e. The con ac discon inui y appea s
a ound xcd = 1.004, he a e ac ion wa e appea om
x<xcd and he shock wa e om x>xcd. Al hough
he posi ion o he con ac discon inui y is he same in
all g aphs, bo h he spa ial wid h o he con ac dis-
con inui y and i s wa ing, shows signi ican disc ep-
ancies.
To quan i y he ob ained accu acy, we ha e se-
lec ed he nume ic alues o each pa o he low,
i.e., a e ac ion wa e, con ac discon inui y and shock
wa e, speci ically he wid h ha a e appoin ed as: W
he wid h o he a e ac ion ail, Wcd he wid h o
con ac discon inui y and Wsw he wid h o he shock
wa e. The ob ained alues a e shown in Tables 3 and
4.
5.3 Blas –Wa e Tes
The blas –wa e es is a mo e ex eme e sion o
Sod’s shock– ube es and is o mula ed o a e y
MPM GSPH P–SPH
20.9651 0.9708 0.9891
W 0.105 0.073 0.0205
Wcd 0.078 0.061 0.026
Wsw 0.128 0.104 0.051
Table 3: The mos ele an alues associa ed wi h he
densi y g aphs o Sod’s shock– ube es , Figs. 3a, 3b
and 3c.
MPM GSPH P–SPH
20.9651 0.9688 0.9891
W 0.126 0.099 0.039
Wcd 0.101 0.075 0.043
Wsw 0.089 0.065 0.037
Table 4: The mos ele an alues associa ed wi h he
eloci y g aphs o Sod’s shock– ube es , Figs. 3d, 3e
and 3 .
high Mach numbe , a ound M= 200. I was de-
eloped by Woodwa d and Colella [31] o highligh
he beha iou o he con ac discon inui y, mainly o
he eloci y, a high densi y and p essu e alues. In
blas –wa e es , he low e olu ion shows he same
ea u es ha Sod’s shock– ube es , namely, a con ac
discon inui y will appea inse ed be ween he shock
wa e and he a e ac ion wa e.
WSEAS TRANSACTIONS on FLUID MECHANICS
Juan J. Pe ea, Juan M. Co de o
E-ISSN: 2224-347X
56
Volume 13, 2018
(a) (b) (c)
(d) (e) ( )
Figu e 4: Compa a i e g aphs o he exac solu ion and nume ical ob ained alues om MPM, GSPH and P-SPH
o blas –wa e es .
As a esul o he simila i ies wi h Sod’s shock–
ube es , we ha e only changed he ini ial alues
o some simula ion pa ame e s. The new alues
o mass densi y, p essu e and he mal ene gy a e:
(ρh, ph, uh) = (1.0,1 103,0.0) and (ρl, pl, ul) =
(1.0,0.01,0.0), acco ding o he sugges ions gi en by
[26, 31]. The new ime s ep is ∆ = 8 10−5. Us-
ing hese alues, we ha e ob ained he esul s ha a e
shown in Figu e 4 a ime = 7.5 10−4, whe e he
exac solu ions and nume ic solu ions a e depic ed by
solid line and do s, espec i ely.
Simila ly o Sod’s shock– ube es , he h ee e-
gions ha a ise when he memb ane is aken away
can be loca ed in he g aphs o Figu e 4: he con ac
discon inui y posi ion xcd = 0.1081, he a e ac ion
wa e x<xcd and he shock wa e x>xcd. Likewise,
in each g aph o 4 he wid h o con ac discon inu-
i y changes o echnique ob aining he lowes wid h
when ou p oposal is implemen ed.
Analogously as he p e ious sec ion 5.2 o shown
he ob ained accu acy, we ha e selec ed he mos ele-
an alues associa ed o he g aphs o 4. These alues
a e shown in Tables 5 and 6.
6 Conclusions
In his pape , we ha e e iewed some o he mos
ele an ea u es ha in luence on he accu acy and
s abili y o low simula ion by SPH. F om hese ea-
MPM GSPH P–SPH
20.9618 0.9791 0.9903
W 0.094 0.081 0.046
Wcd 0.035 0.018 0.009
Wsw 0.094 0.079 0.038
Table 5: The mos ele an alues associa ed wi h he
densi y g aphs o blas –wa e shock es , Figs. 4a, 4b
and 4c.
MPM GSPH P–SPH
20.9621 0.9788 0.9901
W 0.0945 0.0813 0.0585
Wcd 0.0826 0.0582 0.0348
Wsw 0.087 0.026 0.009
Table 6: The mos ele an alues associa ed wi h he
eloci y g aphs o blas –wa e es , Figs. 4d, 4e and
4 .
u es, we ha e p oposed a hyb id po en ial–SPH based
me hod. To show he imp o emen s ob ained by he
p oposed echnique, we ha e implemen ed a se o
es s using an incomp essible and in iscid luid. In
each implemen ed es we ha e compa ed ou p o-
posal wi h he MPM and GSPH echniques which a e
o mula ed o a oid he ins abili ies ela ed o he p es-
su e g adien . Pa icula ly, analysing he es s’ esul s,
WSEAS TRANSACTIONS on FLUID MECHANICS
Juan J. Pe ea, Juan M. Co de o
E-ISSN: 2224-347X
57
Volume 13, 2018
we can conclude:
1. The P-SPH echnique allows us o ob ain a
high s abili y. This conclusion can be deduced
om he ob ained esul s in se ling es , Figu e
2, whe e he ob ained esul s by P-SPH show
g ea e la ice deg ee han he ob ained ones by
he o he wo echniques. Quan i a i ely, his im-
p o emen is shown by he esul s o he pa am-
e e ς, Table 2, whe e he ob ained alues by P-
SPH a e he lowes ones.
2. The P-SPH allow us o ob ain highly accu a e e-
sul s in simula ions whe e he shock wa e and
con ac discon inui y appea . This a i ma ion is
based on he ob ained esul s in Sod’s shock– ube
es . The ob ained g aphs o he densi y and e-
loci y, Figu e 3, a e mo e accu a ely simula ed
by P-SPH han by MPM o GSPH. This accu-
acy is mo e appa en in he a e ac ion ail, he
con ac discon inui y and he shock wa e. To e-
in o ce his ac , we ha e selec ed he alues as-
socia ed o hese egions ha a e shown in Tables
3 and 4. All ob ained alues e idence ha he
bes accu acy is o e ed by P-SPH.
3. The imp o emen s ob ained in Sod’s shock– ube
es is also accomplished in a mo e ex eme e -
sion o his es . This conclusion can be ob ained
om he esul s shown in he blas –wa e es .
Quali a i ely, he ob ained esul s a e in ag ee-
men wi h he exac solu ions. Ne e heless, he
mos accu a e esul s a e ob ained by he P-SPH,
as can be seen in he g aphs o Figu e 4. Quan i-
a i ely, he imp o emen is shown in he alues
o Tables 5 and 6.
4. The P-SPH allow us o ob ain s able simula ions
o a wide ange o he numbe o neighbou pa -
icles. Also, he s abili y is achie ed wi h he
lowes numbe , as consequence. The P–SPH no
only a ou he s abili y bu also he e iciency,
om he compu a ional cos poin o iew. This
conclusion can be deduced om Table 1.
F om hese esul s, we can conclude ha he P-
SPH imp o es bo h he s abili y and he accu acy o
luid low simula ions by SPH
Acknowledgemen s: This esea ch has been sup-
po ed by he Pololas p ojec (TIN2016-76953-C3-2-
R) o he Spanish Minis y o Economy and Compe -
i i eness.
Re e ences:
[1] R. M. Cabez´
on, D. Ga c´
ıa-Senz, and A. Re-
lano. A one–pa ame e amily o in e pola -
ing ke nels o smoo hed pa icle hyd odynam-
ics s udies. Jou nal o Compu a ional Physics,
227(19):8523–8540, Oc 2000.
[2] S. J. Cummins and R. M. An imp o ed sph
me hod o modeling liquid sloshing dynamics.
Jou nal o Compu a ional Physics, 152:584–
607, Jul 1999.
[3] W. Dehnen and H. Aly. Imp o ing con e gence
in smoo hed pa icle hyd odynamics simula ions
wi hou pai ing ins abili y. Mon hly No ices o
he Royal As onomical Socie y, 425(2):1068–
1082, Ap 2012.
[4] C. T. Dyka, P. W. Randles, and R. P. Ingel. S ess
poin s o ension ins abili y in sph. In e na-
ional Jou nal o Nume ical Me hods in Engi-
nee ing, 40(13):2325–2341, Jul 1997.
[5] R. Gingold and J. Monaghan. Smoo hed pa -
icle hid odynamics: Theo y and applica ion o
non–sphe ical s a s. Royal As onomical Soci-
e y, 181:375–389, 1977.
[6] M. Ihmsen, J. O hmann, B. Solen hale ,
A. Kolb, and M. Teschne . Sph luids in com-
pu e g aphics. In Eu og aphics 2014 - S a e o
he A Repo s. The Eu og aphics Associa ion,
2014.
[7] S. Inu suka. Re o mula ion o smoo hed pa i-
cle hyd odynamics wi h iemann sol e. Jou nal
o Compu a ional Physics, 179(1):238–267, Jun
2012.
[8] K. Iwasaki and S. Inu suka. Smoo hed pa icle
magne ohyd odynamics wi h a iemann sol e
and he me hod o cha ac e is ics. Mon hly
No ices o he Royal As onomical Socie y,
418(3):1668–1688, Dec 2011.
[9] G. Liu and M. Liu. Smoo hed Pa icle Hyd o-
dynamics:A Mesh ee Pa icle Me hod. Wo ld
Scien i ic, 2003.
[10] M. Liu, G. Liu, and K. Lam. Cons uc ing smoo-
hing unc ions in smoo hed pa icle hyd ody-
namics wi h applica ions. Jou nal o Compu-
a ional and Applied Ma hema ics, 155(2):263–
284, Jun 2003.
[11] L. B. Lucy. A nume ical app oach o he es ing
o he ission hypo hesis. As onomical Jou nal,
82(12):1013–1024, Dec 1977.
WSEAS TRANSACTIONS on FLUID MECHANICS
Juan J. Pe ea, Juan M. Co de o
E-ISSN: 2224-347X
58
Volume 13, 2018