Smoothed Particle Hydrodynamics (SPH): corrections and … · 2013-11-19 · For further reading...

Preview:

Citation preview

/centre for analysis, scientific computing and applications

Smoothed Particle Hydrodynamics (SPH):corrections and applications

Q.Z. Hou

April 7 2010

/centre for analysis, scientific computing and applications

Outline

1 Motivation

2 SPH and corrective schemesSmoothed particle hydrodynamicsCorrective schemes

3 Test problemsWater hammerHeat conductionStress wave propagationShock tube

4 Summary and future workBibliography

/centre for analysis, scientific computing and applications

Outline

1 Motivation

2 SPH and corrective schemesSmoothed particle hydrodynamicsCorrective schemes

3 Test problemsWater hammerHeat conductionStress wave propagationShock tube

4 Summary and future workBibliography

/centre for analysis, scientific computing and applications

Physical problem

Laboratory set-up

Bozkus and Wiggert (1997)

/centre for analysis, scientific computing and applications

Practical problem

Industrial pipelines

/centre for analysis, scientific computing and applications

Oil pipeline

/centre for analysis, scientific computing and applications

Steam pipeline

/centre for analysis, scientific computing and applications

Difficulties in traditional CFD methods

Two fundamental frameworks:Eulerian description: mesh fixed in space, objects moveacross it; calculation of mass, momentum and energydistributionLagrangian description: mesh attached to the material,cells deform; no mass flux, flux of momentum and energyat cell boundaries

Difficulties:Eulerian grid: irregular/complex geometry, locations ofinhomogeneities, free surfaces, moving interfaces,deformable boundariesLagrangian grid: mesh generation, large deformationboth: explosion, high-velocity impact, discrete particles

/centre for analysis, scientific computing and applications

Outline

1 Motivation

2 SPH and corrective schemesSmoothed particle hydrodynamicsCorrective schemes

3 Test problemsWater hammerHeat conductionStress wave propagationShock tube

4 Summary and future workBibliography

/centre for analysis, scientific computing and applications

What is SPH?

SPH: smoothed particle hydrodynamics

1. movies3D simulation of dam breakpouring water in a glass (like commercial advertisement forbeer).Hollywood, Pixar Animation Studio, Industry Light andMagic, fluid simulations in Ice Age and 2012

2. key ideafollow particles’ movement, information attached to them isused as weighted average in numerical approximation

/centre for analysis, scientific computing and applications

Fundamentals of SPH

Integral representation

f (x) =

∫Ω

f (ξ)δ(x − ξ,h) dV

Function reconstruction

f (x) ≈∫

Ωf (ξ)W (x− ξ,h) dV

Representation of the derivative

∇f (x) ≈ −∫

Ωf (ξ)∇W (x − ξ,h)dV

with smoothing function W (x − ξ,h)and smoothing length h

/centre for analysis, scientific computing and applications

Smoothing function

Requirements:

unity:∫

Ω W (x − ξ,h) dξ = 1;compact support: W (x − ξ,h) = 0, |x − ξ| > κh;positivity: W (x − ξ,h) ≥ 0, |x − ξ| ≤ κh;symmetry: W (x − ξ,h) = W (ξ − x ,h);δ function consistency:

limh→0

W (x − ξ,h) = δ(x − ξ);

monotonically decreasing with respect to |x − ξ|smooth

/centre for analysis, scientific computing and applications

Cubic spline – smoothing function W

Cubic spline and its derivatives

/centre for analysis, scientific computing and applications

Discretisation (particle approximation)

Ω is divided into N particles, with volume ∆Vj , (j = 1, · · · ,N)

f (x) ≈∫

Ωf (ξ)W (x − ξ,h) dV

⇒ f (x i) ≈∑

f (ξj)W (x i − ξj ,h)∆Vj

movement of particlesgive each particle a fixed mass mj and replace the volume ∆Vjby

∆Vj =mj

ρj

position evolutionxn+1

i = xni + ∆t vn

i

/centre for analysis, scientific computing and applications

Discretisation for derivative

∇f (x) ≈ −∫

Ωf (ξ)∇W (x − ξ,h)dV

⇒ ∇f (x i) ≈ −∑

f (ξj)∇W (x i − ξj ,h)∆Vj

introduce the following notation

Wij := W (x i − ξj ,h),

Wij,α :=(∂Wij∂xα

)i, where α indicates a spatial direction

/centre for analysis, scientific computing and applications

Particle distribution

/centre for analysis, scientific computing and applications

1D example: f (x) = x , x ∈ [0, 1],N = 5, κh = 12

f1 ≈∑5

j=1 fjW1j∆Vj = f1W11∆V4 + f2W12∆V2 = 16 f2 = 1

24 6= 0

function reconstruction derivative approximation

/centre for analysis, scientific computing and applications

Corrective schemes (1D)

Expanding Taylor series for f (ξ) around point x yields

f (ξ) ≈ f (x) +∂f (x)

∂x(ξ − x)

Multiplying with W and integrating over Ω∫Ω

f (ξ)WdV ≈ f (x)

∫Ω

WdV +∂f (x)

∂x

∫Ω

(ξ − x)WdV

∫Ω

f (ξ)WdV ≈ f (x)

∫Ω

WdV

/centre for analysis, scientific computing and applications

Corrective scheme for derivative

Taylor series expansion

f (ξ) ≈ f (x) +∂f (x)

∂x(ξ − x)

Multiplying with ∂W∂ξ and integrating over Ω∫

Ωf (ξ)

∂W∂ξ

dV ≈ f (x)

∫Ω

∂W∂ξ

dV +∂f (x)

∂x

∫Ω

(ξ − x)∂W∂ξ

dV

CSPM: corrective smoothed particle method

/centre for analysis, scientific computing and applications

Corrective scheme in discretized form

function approximation∫Ω

f (ξ)WdV ≈ f (x)

∫Ω

WdV

⇒ f (x) ≈∫

Ω f (ξ)WdV∫Ω WdV

⇒ f (xi) ≈∑

f (xj)Wij∆Vj∑Wij∆Vj

derivative approximation∫Ω

f (ξ)∂W∂ξ

dV ≈ f (x)

∫Ω

∂W∂ξ

dV +∂f (x)

∂x

∫Ω

(ξ − x)∂W∂ξ

dV

⇒(∂f (x)

∂x

)i≈∑

[f (xj)− f (xi)]Wij,x ∆Vj∑(xj − xi)Wij,x ∆Vj

/centre for analysis, scientific computing and applications

Same 1D example

function reconstruction derivative approximation

/centre for analysis, scientific computing and applications

Same 1D example: f = (f1, f2, f3, f4, f5)T

f ≈ Af , ∂f/∂x ≈ Bf

Interpolation matrix A2/3 1/61/6 2/3 1/6

1/6 2/3 1/61/6 2/3 1/6

1/6 2/3

,

4/5 1/51/6 2/3 1/6

1/6 2/3 1/61/6 2/3 1/6

1/5 4/5

Difference matrix B

1∆x

0 1

2−1

2 0 12

−12 0 1

2−1

2 0 12

−12 0

,1

∆x

−1 1−1

2 0 12

−12 0 1

2−1

2 0 12

−1 1

/centre for analysis, scientific computing and applications

2D: f (x , y) = sin πx sin πy , (x , y) ∈ [0,1]× [0,1]

Reconstruction of function f and particle approximation of thefirst derivative fx

/centre for analysis, scientific computing and applications

Outline

1 Motivation

2 SPH and corrective schemesSmoothed particle hydrodynamicsCorrective schemes

3 Test problemsWater hammerHeat conductionStress wave propagationShock tube

4 Summary and future workBibliography

/centre for analysis, scientific computing and applications

Water hammer

Typical reservoir-pipe-valve (RPV) system

Governing equations

∂v∂t

=1ρ

∂p∂x

+∂φ

∂x, (momentum equation)

∂p∂t

= K∂v∂x

(continuity equation)

linear artificial viscosity φ =

Cv,x , v,x < 00, v,x > 0

/centre for analysis, scientific computing and applications

Solving two first order equations by CSPM

CSPM:(∂f∂x

)i≈

∑[f (xj )−f (xi )]Wij,x ∆Vj∑

(xj−xi )Wij,x ∆Vjfor f = v and f = p

IC: v0i = v0; p0

i = p0

Algorithm for single time stepstep 1: apply BC vn+1

N = 0step 2: calculate pressure

pn+1i = pn

i + ∆t[K(∂v∂x

)i

]step 3: apply BC pn+1

1 = p0step 4: calculate velocity

vn+1i = vn

i + ∆t[

(∂p∂x

)i

+

(∂φ

∂x

)i

]

/centre for analysis, scientific computing and applications

Results

instantaneous valve closing linear valve closing

Pressure history at valve

/centre for analysis, scientific computing and applications

Heat conduction in a bar

Temperature distribution at different times, left - energy balance(two first order eqs.), right - heat conduction (one second order

eq.)

/centre for analysis, scientific computing and applications

Wave propagation in an elastic bar

Snapshot at t = 10 µs

/centre for analysis, scientific computing and applications

Shock tube (moving particles)

Shock tube problem

/centre for analysis, scientific computing and applications

Outline

1 Motivation

2 SPH and corrective schemesSmoothed particle hydrodynamicsCorrective schemes

3 Test problemsWater hammerHeat conductionStress wave propagationShock tube

4 Summary and future workBibliography

/centre for analysis, scientific computing and applications

Summary

difficulties in traditional methodsfundamentals of SPHcorrections to SPH1D test problems of CSPM

/centre for analysis, scientific computing and applications

Future work

new corrective schemesirregular particlesmoving particles2D / 3D problems

/centre for analysis, scientific computing and applications

For further reading

Bozkus Z and Wiggert D.C (1997)Liquid slug motion in a voided line. Journal of Fluids andStructures, 11, 947-963

Liu G.R and Liu M.B (2003)Smoothed Particle Hydrodynamics: A Meshfree ParticleMethod.

Chen J.K., Beraun J.E., Carney T.C (1999)A corrective smoothed particle method for boundary valueproblems in heat conduction. International Journal forNumerical Methods in Engineering, 46: 231-252

/centre for analysis, scientific computing and applications

For further reading

Bozkus Z and Wiggert D.C (1997)Liquid slug motion in a voided line. Journal of Fluids andStructures, 11, 947-963

Liu G.R and Liu M.B (2003)Smoothed Particle Hydrodynamics: A Meshfree ParticleMethod.

Chen J.K., Beraun J.E., Carney T.C (1999)A corrective smoothed particle method for boundary valueproblems in heat conduction. International Journal forNumerical Methods in Engineering, 46: 231-252

/centre for analysis, scientific computing and applications

For further reading

Bozkus Z and Wiggert D.C (1997)Liquid slug motion in a voided line. Journal of Fluids andStructures, 11, 947-963

Liu G.R and Liu M.B (2003)Smoothed Particle Hydrodynamics: A Meshfree ParticleMethod.

Chen J.K., Beraun J.E., Carney T.C (1999)A corrective smoothed particle method for boundary valueproblems in heat conduction. International Journal forNumerical Methods in Engineering, 46: 231-252

Recommended