Large Scale Simulation of Cloud Cavitation Collapse Ursula - - PowerPoint PPT Presentation

▶
large scale simulation of cloud cavitation collapse
SMART_READER_LITE
LIVE PREVIEW

Large Scale Simulation of Cloud Cavitation Collapse Ursula - - PowerPoint PPT Presentation

May 22, 2023 282 likes •566 views

ICCS 2017 Zurich, June 11 - 14, 2017 Large Scale Simulation of Cloud Cavitation Collapse Ursula Rasthofer with: Fabian Wermelinger, Panagiotis Hadjidoukas, Petros Koumoutsakos CSE lab Computational Science & Engineering Laboratory

slide-1
SLIDE 1

CSElab

Computational Science & Engineering Laboratory http://www.cse-lab.ethz.ch

Large Scale Simulation of Cloud Cavitation Collapse

Ursula Rasthofer

ICCS 2017 Zurich, June 11 - 14, 2017

with: Fabian Wermelinger, Panagiotis Hadjidoukas, Petros Koumoutsakos

slide-2
SLIDE 2

Power of Cavitation

2

engineering

https://de.wikipedia.org/wiki/Kavitation extracted from: Brennen, Hydrodynamics of pumps, Oxford University Press, 1994 http://www.forddoctorsdts.com http://www.forddoctorsdts.com extracted from: Bazan-Peregrino et al., Cavitation-enhanced delivery of a replicating oncolytic adenovirus to tumors using focused ultrasound , Journal of Controlled Release Volume 169, Issues 1–2, 2013, 40-47

biomedical applications

https://en.wikipedia.org/wiki/ Extracorporeal_shock_wave_lithotripsy

nature

https://de.wikipedia.org/wiki/ Knallkrebse herve.cochard.free.fr

slide-3
SLIDE 3

Cloud Cavitation Collapse

3

esource.alstrbology.com www.jotformeu.com www.amc.edu.au

  • thousands of bubbles
  • cloud dynamics dominated by

bubble-bubble interactions

  • experiments
  • challenging high frequencies and

microscopic length scales

  • risk of damage
  • theory
  • based on Rayleigh-Plesset-like equations
  • spherical bubbles usually assumed
  • simulations
  • Lagrangian approaches for bubbles
  • fully resolved bubble clouds usually

restricted to small clouds

slide-4
SLIDE 4

Outline

  • Computational Approach
  • Simulations
  • Conclusions

4

slide-5
SLIDE 5

Outline

  • Computational Approach
  • Simulations
  • Conclusions

5

slide-6
SLIDE 6

Modeling Strategies

6

complexity/fidelity computational efficiency

two-fluid models

Lagrangian methods as well as Eulerian methods, e.g., based

  • n level-set description

single-fluid models

e.g., diffuse interface methods

mixture approches

e.g., thermodynamic equilibrium cavitation model

slide-7
SLIDE 7

Diffuse Interface Method

where and

7

[Kapila et al. 2001, Murrone & Guillard 2005, Saurel et al. 2009, Tiwari et al. 2013, …]

∂α1ρ1 ∂t + r · (α1ρ1u) = 0 ∂α2ρ2 ∂t + r · (α2ρ2u) = 0 ∂ (ρu) ∂t + r · (ρu ⌦ u + pI) = 0 ∂E ∂t + r · ((E + p) u) = 0 ∂α2 ∂t + u · rα2 = K r · u

K = α1α2(ρ1c2

1 − ρ2c2 2)

α1ρ2c2

2 + α2ρ1c2 1

α1 + α2 = 1

O(h)

FLUID 1 FLUID 2 I N T E R F A C E

slide-8
SLIDE 8

Godunov-Type Finite Volume Method

  • reformulation of gas-volume-

fraction equation

[Johnsen & Colonius 2006]

  • high-order reconstruction of

face values of primitive variables using WENO3/5

[Liu et al. 1994, Jiang & Shu 1998]

  • approximate HLLC Riemann

solver for flux reconstruction

[Toro et al. 1994]

  • low-storage 3rd-order Runge-

Kutta scheme for time discretization

[Gottlieb et al. 2001]

8

∂α2 ∂t + r · (α2u) = α2ρ1c2

1

α1ρ2c2

2 + α2ρ1c2 1

r · u

ci ci-1 ci+1 xi-1/2 xi+1/2 Ui-1 Ui Ui+1 Fi+1/2 Fi-1/2 h WL WR

s r c x y U- U1 U2 U+ UL* UR*

slide-9
SLIDE 9

Cubism-MPCF

  • compressible multicomponent flow solver

tailored to HPC systems [Rossinelli et al. 2013, P

. E. Hadjidoukas et al. 2015]

  • 3D Cartesian-grid finite volume solver
  • wavelet-based compression of simulation data

9

PFLOPS (% Peak) 14.4 PFLOPS (72%) 0.1 - 3% (TUM) 1.3 - 6.4% (Stanford) SIZE 1.3 E13 - 15K bubbles 1.2 E08 - 0.15K bubbles (TUM) 0.4 E13 - Turbulence (Stanford) I/O COMPRESSION 10-100X

  • TIME TO SOLUTION (no I/O)

Tw = 1.8 Tw= 29.7 (TUM) Tw= 16.3 - 39.0 (Stanford)

slide-10
SLIDE 10

Cubism-MPCF: Software Layout

  • cluster (MPI)
  • node (OpenMP)
  • core (SIMD): WENO, HLLC

10

Core Node Cluster

grid cell computational domain subdomain block

www.cscs.ch www.alcf.anl.gov www.fz-juelich.de

slide-11
SLIDE 11

Node and Cluster Layer

  • OpenMP parallelization
  • dynamic work scheduling
  • 1 thread exclusively processes 1 block
  • MPI parallelization
  • non-blocking P2P communication for ghost blocks
  • communication time ~ O(time for processing 1-2 blocks)

11

node i node i+1 node i-1

slide-12
SLIDE 12

Core Layer

  • block-based memory layout
  • increases spatial locality
  • instruction and data level parallelism
  • explicit vectorization
  • code fusion techniques
  • temporal locality
  • ring buffers for active data slices, e.g., in

WENO, HLLC

12

slide-13
SLIDE 13

Outline

  • Computational Approach
  • Simulations
  • Conclusions

13

slide-14
SLIDE 14
  • gas-volume fraction
  • cloud interaction parameter
  • cloud generation
  • locations: random
  • radii: constant or random
  • collapse driven by increase
  • f ambient pressure

Cloud Setup

14

RB rB x y z RC RB dG

β = α ✓ Req Ravg ◆2

α = 1 R3

C nB

X

i=1

R3

B,i

slide-15
SLIDE 15

50K Bubbles, 64 Billion Cells

15

25K time steps, 72h x 32K cores

slide-16
SLIDE 16

−12 −8 −4 4 8 12 −12 −8 −4 4 8 12 x y

Micro-Jet Formation

  • due to pressure gradient

along interface

  • directed toward cloud

center

16

slide-17
SLIDE 17

Bubble Reconstruction

17

α2 < 0.5 α2 > 0.5

Γint(t) = {x ∈ Ω | α2(x, t) = 0.5}

sphericity porosity

φ = VB V h

B

Ψ = ⇣ 6π

1 2 VB

⌘ 2

3

SB

Ψ = 1.00 φ = 1.00 Ψ = 0.96 φ = 0.98 Ψ = 0.60 φ = 0.67

slide-18
SLIDE 18

Sphericity and Porosity

18

0.99 1

φ

0.98 0.99 1

φ

0.2 0.4 0.6 0.8 1 0.80 0.90 1

t/tC φ

0.99 1

Ψ

0.96 0.98 1

Ψ

0.2 0.4 0.6 0.8 1 0.60 0.80 1

t/tC Ψ

increasing cloud interaction parameter

slide-19
SLIDE 19

Keller-Miksis Equation for Clouds

  • spherical bubble collapses assumed
  • weakly compressible liquid flow
  • extended form including bubble translation

19

[Keller & Miksis 1997, Mettin et al. 1997, Doinikov 2004]

1 − ˙ RBi c1 ! RBi ¨ RBi + 3 2 − ˙ RBi 2c1 ! ˙ R2

Bi = 1

ρ1 1 + ˙ RBi c1 ! (pBi − p1) + RBi ρ1c1 d dt (pBi − p1) + 1 4 ˙ x2

Bi

−

nB

X

j=1;j6=i

" 1 dij ⇣ R2

Bj ¨

RBj + 2RBj ˙ R2

Bj

⌘ + R2

Bj

2d3

ij

  • xBi − xBj
  • ·

⇣ RBj ¨ xBj + ˙ RBj ˙ xBi + 5 ˙ RBj ˙ xBj ⌘ − R3

Bj

4d3

ij

˙ xBj · ˙ xBi + 2 ˙ xBj

  • + 3

d2

ij

˙ xBj ·

  • xBj − xBi

xBi − xBj

  • ·

˙ xBi + 2 ˙ xBj !# 1 3RBi ¨ xBi + ˙ RBi ˙ xBi =

N

X

j=1;j6=i

" 1 d3

ij

  • xBi − xBj

⇣ RBiR2

Bj ¨

RBj + 2RBiRBj ˙ R2

Bj + ˙

RBi ˙ RBjR2

Bj

⌘ − R2

Bj

2d3

ij

⇣ RBiRBj ¨ xBj + ⇣ ˙ RBiRBj + 5RBi ˙ RBj ⌘ ˙ xBj ⌘ + 3R2

Bj

2d5

ij

  • xBi − xBj
  • xBi − xBj
  • ·

✓ RBiRBj ¨ xBj + ⇣ ˙ RBiRBj + 5RBi ˙ RBj ⌘ ˙ xBj ◆!# pBi = pBi(0) ✓RBi(0) RBi ◆3γ2

radius position

slide-20
SLIDE 20

Bubble Radius

  • increasing deviations with increasing cloud interaction

parameter, in particular, if bubble translation is neglected

20

increasing cloud interaction parameter

  • A: center-most
  • B: closest to RC/2
  • C: outer-most
  • solid: 3D simulations
  • dashed: without bubble translation (KM)
  • dash-dot-dot: with bubble translation (DK)

10 20 30 40 50 0.5 0.6 0.7 0.8 C B A t [µs] RB [mm] 20 40 60 0.5 0.6 0.7 0.8 C B A t [µs] RB [mm] 20 40 60 80 0.5 0.6 0.7 0.8 C B A t [µs] RB [mm]

slide-21
SLIDE 21

Large-Scale Cloud

21

  • 12500 air bubbles in water
  • pC = 1 bar
  • p∞ = 10 bar
  • cloud radius: 45 mm
  • average bubble radius: 0.69 mm
  • gas-volume faction: 5%
  • cloud interaction parameter: 28
slide-22
SLIDE 22

Cloud Collapse

22

100 200 300 400 0.2 0.4 0.6 0.8 1 t [µs]

pS pS,max V2 V2(0)

slide-23
SLIDE 23

Collapse Wave

  • Morch model: collapse wave in cloud of vapor bubbles

23

R!! R + ( 3

2 − 1 2 (1−γ )(1−αC)) !

R2 = − p∞ − pvapor ρliquidαC

[Morch 1989]

50 100 150 200 250 300 0.2 0.4 0.6 t [µs] uW [km/s]

simulation Morch model

slide-24
SLIDE 24

Erosive Potential

24

1 2 3

1 2 3 4 103 104 105 shell 1 shell 2 shell 3 dP [mm] cpr [

1 cm2s]

simulation fit

1 2 3 4 2 4 6 8 ·104 dP [mm] pcr [ 1

cm s]

shell 1 shell 2 shell 3

probability density function of coverage rate cumulative impact rate

  • pits: plastic deformations in

form of small indents

  • due to impulsive load

generated by bubble collapses


slide-25
SLIDE 25

Outline

  • Computational Approach
  • Simulations
  • Conclusions

25

slide-26
SLIDE 26

Summary & Outlook

  • diffuse interface method and Cubism-MPCF

✓ important to account for gas expansion and compression in interface ✓ compressible multicomponent flow solver capable of processing up to 7x1011 cells per second

  • HPC for collapsing clouds with up to 50K bubbles

✓ comparison with bubble-particle approaches ✓ insights into induced pressure and velocity fields, wave dynamics, bubble shape evolution as well as erosive potential

➡ integration of mass transfer and application to turbulent cavitating flow

26

slide-27
SLIDE 27

27

Thank you for your attention!

  • 1. U. Rasthofer, F. Wermelinger, P

. Hadjidoukas, P . Koumoutsakos, Large-scale simulation of cloud cavitation collapse, Procedia Computer Science 108C (2017) 1763–1772.

  • 2. U. Rasthofer, F. Wermelinger, P

. Koumoutsakos, A comparative study on cloud cavitation collapse: Rayleigh-Plesset-like equations versus full three-dimensional simulations, 2017, in preparation.

  • 3. U. Rasthofer, F. Wermelinger, P

. Karnakov, J. Sukys, P . Hadjidoukas, P . Koumoutsakos, Numerical investigation of collapsing clouds with up to O(104) gas bubbles, 2017, in preparation.

  • 4. J. Sukys, U. Rasthofer, F. Wermelinger, P

. Hadjidoukas, P . Koumoutsakos, Optimal fidelity multi-level Monte Carlo for quantification of uncertainty in simulations of cloud cavitation collapse, 2017, submitted for publication.

  • 5. P

. Hadjidoukas, D. Rossinelli, F. Wermelinger, U. Rasthofer, P . Koumoutsakos, A high-performance compression framework for extreme-scale CFD data, 2017, in preparation.