Upload
others
View
19
Download
0
Embed Size (px)
Citation preview
M. B. Liu (刘谋斌)
November 30, 2016
@NTU
Recent Progress on meshless methods for Flow Simulations—SPH, MPS, DPD
Smoothed Particle Hydrodynamics - Methodology and Applications
MES, Peking University—History
Peiyuan Zhou
• 1952—Established by Peiyuan Zhou at Peking University, 1st Dept. Mech. in PRC
• 1958—China’s first low-speed wind tunnel
• 2000—State Key Laboratory for Turbulence and Complex Systems
• 2006—Joined in COE
• 2012— Ranked as NO.1 in PRC
• 450 faculty, 3200 B.S., 820 M.S., 280 Ph.D.
60th Anniversary Celebration
The MES serves the world by providing a high-quality learning
environment, outstanding research laboratories, and challenging
experiences for our students in science, industry and technology.
2/68
09:55
MES, Peking University—Divisions
Fluid Mechanics •Turbulence
•Computational fluid dynamics methods
•Aerodynamics
•Propulsion, heat transfer and advanced engines
Solid Mechanics •Micro/nano-mechanics, mechanics and physics of
complex materials
•Mechanics of smart materials and structures
•Elasticity and plasticity
•Experimental mechanics
Engineering Mechanics •Computational mechanics,
•Fluid-structure interaction
•Mesh and meshfree methods
•Structural dynamics
•Mesh generation
Biomechanics and Medical Engineering •Cell mechanics •Bio-solid mechanics •Bio-fluid mechanics •Mechanobiology •Bio-imaging and bio-instrumentation
Dynamics and Control •Cooperative control of multi-agent systems
•Robust tracking control
•Redundant control and dynamic control allocation
•Nonlinear dynamics and control
•Dynamics and control of systems with group symmetries
3/68
09:55
Research team – Computational FSI
Impact and penetration
LNG船液舱
液面变形、破碎
结构变形、破坏
强流固耦合作用
Liquid sloshing
网格类,FVM…
粒子类,SPH…
耦合类,SPH/FEM
Numerical methods Key issues
CPU,GPU,CPU+GPU
高性能计算技术 复杂载荷作用下的流体运动规律、流固耦合效应、结构失稳与破坏及防护
Computational Mechanics, FSI, Multi-scale modeling
底板
焊板
Explosive welding
Powder Transport Gas-particle two-
phase flow
4/68
09:55
Outline
Background
SPH methodology
SPH for hydrodynamics
SPH for environmental flows
SPH for explosion and impact
Prospects and future directions
1
2
3
4
5
6
5/68
09:55
1. Background
Explosion and impact Water discharge and dam collapse with free surface
Tsunami with complex interface dynamics
1.1 Continuum scale complex problems Examples: tsunami, dam collapse, penetration…
Difficulties:
Free surfaces, deformable boundaries, moving interfaces (for FDM)
Large deformations (for FEM)
Complex mesh generation and mesh adaptivity (for both FDM and FEM)
6/68
09:55
Lagrangian grid-based methods
Eulerian grid-based method
Grid moving with object
Easy to get time history of movement and deformation
Convenient to track moving features
Difficult to treat large deformations
Grid fixed on space
Easy to treat large deformations
Difficult to get time history
Difficult to track moving features
Particle methods such as SPH combine
advantages of FDM & FEM
1. Background Difficulties for FDM/FEM
7/68
09:55
1. Background
It is a great challenge to generate or adapt mesh: structured/unstructured, body-fitting, mesh adaptivity, moving mesh, mesh rezoning…, Cost 2/3 work load
Mesh adaptivity
Moving mesh Mesh rezoning
Meshing problems
8/68
09:55
1. Background
Landslide and mud flow
Discontinuous problems Granular flows in environmental, chemical, geophysical, bio-engineering…
9/68
09:55
1. Background 1.2 Micro/Nano scale problems Traditional numerical methods based on continuum scale
constitutive equations may not be valid, especially when the dimension diminishes…
1.3 Problems with multi-scale physics Coupling MD with FDM/FEM is complicated and not natural…
Evolution of a polymer drop break-up
Cross section of a bilayer of lipid in water molecules
Water droplets in oil
10/68
09:55
1. Background
http://www.worldscientific.com/worldscibooks/10.1142/9017
http://www.worldscientific.com/worldscibooks/10.1142/5430
11/68
09:55
2. SPH methodology – Basic concepts
2.1 History Originally invented for solving astrophysical problems in open space
Recently applied to general fluid dynamic problems
2.2 Numerical approximation Weight function (or smoothing function ) , W, centered on particles
and describe continuous or discrete field function,
Kernel approximation:
Particle approximation:
( ) ( )i if f W d x x x
1
N
i j ij j j
j
f f W m
, ,( )i if f W d x x
, ,
1
N
i j i j j
j
f f W m
i
j
rij Rcut
SW
SPH approximations in a two-dimensional space
12/68
09:55
Liu & Liu, WSPC, 2003; 2015
2.3 SPH equations of motion
N
j
ijji Wm
1
N
j
ij
ji
Wm
Dt
D
1
)(
i
jix
v-v
N
j i
ij
ji
jjii
j
i
ijN
j ji
ji
j
Wm
Wppm
Dt
D
11
xx
v i
i
ijN
j j
jj
i
iij
i
ijN
j j
j
i
ij
Wm
Wppm
Dt
D
xx
v i
122
122
)()(
)3
2(
111
ijiji
N
j j
jij
ji
N
j j
jij
ji
N
j j
j
i WmWmWm
vx
vx
v
ii
ii
i
iN
j
ij
ji
ji
ji
Wppm
Dt
De
2)(
2
1
1
i
jix
v-v
ii
i
iN
j
ij
j
j
i
ij
iWpp
mDt
De
2))((
2
1
122
i
jix
v-v
Density
Momentum
Energy
Strain rate
2. SPH methodology – Basic concepts 13/68
09:55
2. SPH methodology – Basic concepts
2.4 Features and advantages
Features
• Particles are used to represent the state of a system
• Governing equations are discretized and approximated on particles
• Particles move with the object
Merits/demerits
• Mass rigorously conserved, no loss/gain
• Easy to treat large deformations
• Easy to obtain time history of movement and deformations
• Method under development
• Heavier computational cost
Particle method
Meshfree method
Lagrangian method
SPH combines advantages of FDM &
FEM
14/68
09:55
2.5 Constructing smoothing function
2. SPH methodology – Improvements
Original SPH empirically specifies that the weight function W satisfies some special requirements like (a) normalization , (b) Delta function property , and (c) symmetric (even) property.
Weight function constructing conditions
( , ) 1W h d
x x x
0lim ( , ) ( )h
W h
x x x x
( )
1
0
( )( ) ( , ) ( )
!( )
jnj
n
j
fW h df r
j
x
x x x x x x xx
2
0
1
2
( , ) 1
( ) ( , ) 0
( ) ( , ) 0
( ) ( , ) 0n
n
W h d
W h d
W h d
W
M
M
h d
M
M
x x x
x x x x x
x x x x x
x x x x x
( )
1
0
( 1) ( )( ) ( ) ( )
!
n j jj
n
j
ff r
j
xx x x x x
Taylor series analysis ( ) () , )( Wf h df
x x xxx
Example: a quartic weight function
20 1 2( , ) ,n
nW h a a R a R a R Rh
x xx x
15/68
09:55
Liu et al., JCAM, 2003
2.6 Improving accuracy - FPM
2. SPH methodology – Improvements
The original SPH method does not have C0 consistency for particle approximation, so the numerical accuracy of the original SPH method is quite low especially for disordered particles or boundary particles.
Finite particle method (FPM), retaining the advantages of SPH with higher order accuracy .
FPM has C0 and C1 consistency for particle approximation.
FPM is not influenced by disordered particle distribution and the selection of smoothing function.
FPM is more computational expensive than the original SPH.
1
1 1 1
,
, , ,
1 1 1
( )
( )
j i
j i
N N N
ij j ij j j ij j
j j ji
N N Ni
ij j ij j j ij j
j j j
W v W v f W vf
fW v W v f W v
x x
x x
Numerical scheme of FPM
Original SPH particle approximation 1
N
i j ij j j
j
f f W m
, ,
1
N
i j i j j
j
f f W m
16/68
09:55
Liu et al., AMM, 2005; ANM, 2006.
0 0.5 1 1.5 2 2.5 3 3.5
x 10-5
0
0.2
0.4
0.6
0.8
1x 10
-3
velocity (m/s)
z (
m)
Analytical solutions
SPH results
0.01 s 0.1 s
1 s
Accuracy and stability of SPH and FPM
0 0.5 1 1.5 2 2.5
x 10-5
0
0.2
0.4
0.6
0.8
1x 10
-3
velocity (m/s)
z (m
)
FPM results
Analytical solutions
0.01 s
0.1 s 1 s
Conventional SPH
Improved scheme (FPM) numerical
oscillation removed
1
N
i j ij j j
j
f f W m
, ,
1
N
i j i j j
j
f f W m
1
1 1 1
,
, , ,
1 1 1
( )
( )
j i
j i
N N N
ij j ij j j ij j
j j ji
N N Ni
ij j ij j j ij j
j j j
W v W v f W vf
fW v W v f W v
x x
x x
2.6 Improving accuracy - FPM
2. SPH methodology – Improvements 17/68
09:55
2.7 Adapting SPH (ASPH)
2. SPH methodology – Improvements
W
i
j
rij
h
S W
i
j
S
rij h
1h
2
SPH: isotropic weight function
ASPH: anisotropic weight function different axes to match the particle spacing as it varies in time, space and direction Capable of catching anisotropic deformation suitable for problems with large dimension ratio
18/68
09:55
Liu et al., JSW, 2006
SPH: flow in a micro channel with 800 particles
SPH: a metal bar impacting on a solid wall
2.7 Adapting SPH (ASPH)
ASPH: flow in a micro channel with 200 particles
ASPH: a metal bar impacting on a solid wall
2. SPH methodology – Improvements 19/68
09:55
Liu et al., JSW, 2006
2.8 Solid boundary treatment (SBT)
2. SPH methodology – Improvements
As a meshfree, Lagrangian particle method, SPH is difficult in exactly implementing solid boundary conditions (SBC).
Virtual particles are usually used to implicitly implement SBC, and the distribution of virtual particles and the calculation of their properties influence the accuracy of SPH simulation.
A new SBT algorithm—coupled dynamic SBT:
1、Two types of virtual particles:Repulsive and ghost particles
2、New repulsive force and new approximation scheme
/(0.75 )ij ijr h
2
2
2 / 3 0 2 / 3
(2 1.5 ) 2 / 3 1( )
0.5(2 ) 1 2
0
f
otherwise
2
20.01 ( )
ij
ij
ij
c fr
x
F
1 0 1.51.5
ij
ij
rr d
d
20/68
09:55
Liu et al., JHD, 2013
2.8 Solid boundary treatment (SBT)
2. SPH methodology – Improvements
Conventional SBT algorithms in SPH lead to pressure oscillation, and large errors in calculating pressure load on structures
The new SBT algorithm produces smooth and accurate pressure field and is more attractive for FSI problems.
A typical example: dam collapse
21/68
09:55
3. SPH for hydrodynamics – Free surface flows
Liu M. B. et al., ArCME, 2010
With fixed, moving or deformable boundaries • Dam collapse and water discharge
• Tsunami and flooding
• Wave dynamics
• Liquid drop dynamics
• …
22/68
09:55
3. SPH for hydrodynamics – Free surface flows
Dam collapse
Shao, Liu et al., IJCM, 2010
Water discharge
3.1 Dam collapse and water discharge
23/68
09:55
3. SPH for hydrodynamics – Free surface flows
3.2 Water impact – dam break against an obstacle
Liu & Shao, SCPMA, 2012
24/68
09:55
3. SPH for hydrodynamics – Free surface flows
3.2 Wave impact – water impact on building
Without mitigation
Mitigation with a dike
Top-bottom view
Front-back view
Initial particle distribution
25/68
09:55
3. SPH for hydrodynamics – Free surface flows
3.3 Injection flow
Water injection into an empty container
Water injection into a filled container
26/68
09:55
3. SPH for hydrodynamics – Free surface flows
3.3 Water injection
Experiment SPH simulation
Water injection
27/68
09:55
3. SPH for hydrodynamics – Free surface flows
3.4 Multi-phase flows – liquid drop impact on thin liquid film
28/68
09:55
Ma, Liu et al., APS, 2012
3. SPH for hydrodynamics – Free surface flows
3.4 Multi-phase flows – Rayleigh-Taylor instability
Heavier fluid: 1800 kg/m3 Lighter fluid: 1000 kg/m3
Present HU et al. 2007
29/68
09:55
Chen, Liu* et al., JCP, 2013
3. SPH for hydrodynamics – Free surface flows
3.4 Multi-phase flows – Air bubble rising
30/68
09:55
Chen, Liu* et al., JCP, 2013
3. SPH for hydrodynamics – FSI problems
3.5 Water entry – cylinder
Liu and Shao, ICCEES, 2011
Penetration depth
31/68
09:55
3. SPH for hydrodynamics – FSI problems
3.5 Water entry – wedge
Falling velocity
32/68
09:55
Liu et al., IJNMF, 2014
3. SPH for hydrodynamics – FSI problems
3.6 Water exit – cylinder
33/68
09:55
Liu et al., IJNMF, 2014
3. SPH for hydrodynamics – FSI problems
3.6 Water exit – Projectile launch
Projectile launch directly from water
Projectile launch from a canister
34/68
09:55
3. SPH for hydrodynamics – FSI problems
3.6 Water exit – Projectile launch
The Principle of Independence of the Cavity Sections Expansion
Vlandimir et al., artificial supercavitation, 2002.
35/68
09:55
3. SPH for hydrodynamics – FSI problems
3.7 Liquid sloshing
Water sloshing in a reservoir
LNG船液舱
Liquid fuel sloshing in aerospace and aeronautical vehicles。
LNG sloshing in a LNG ship
36/68
09:55
3. SPH for hydrodynamics – FSI problems
3.7 Liquid sloshing – in a rectangular container
Shao, Liu, and Liu, CAS, 2012
Fluent (up) vs SPH (bottom) Water level at an observation point
cos(2 / )S A t T
37/68
09:55
3. SPH for hydrodynamics – FSI problems
3.7 Liquid sloshing – in a rotating rectangular container
Shao, Liu, and Liu, CAS, 2012 Shao, Liu, and Liu, CAS, 2012 Yang, Peng and Liu, IJCM 2011
1.0 rad s
8.2 rad s 4.1rad s
2.0 rad s 1.0 rad s
4.1rad s
38/68
09:55
3. SPH for hydrodynamics – FSI problems
3.7 Liquid sloshing – Ballast water
Shao, Liu, and Liu, CAS, 2012 Shao, Liu, and Liu, CAS, 2012 Yang, Peng and Liu, IJCM 2011 Yang, Peng and Liu, IJCM 2011
39/68
09:55
3. SPH for hydrodynamics – FSI problems
3.7 Liquid sloshing – in other shapes of containers
Shao, Liu, and Liu, CAS, 2012 Shao, Liu, and Liu, CAS, 2012 Yang, Peng and Liu, IJCM 2011 Yang, Peng and Liu, IJCM 2011
40/68
09:55
Shao, Liu et al., EC, 2015.
Vertical baffle
Horizontal … Porous …
Multiple compartments
M>T>I>-
3.7 Liquid sloshing – sloshing mitigation
3. SPH for hydrodynamics – FSI problems 41/68
09:55
3. SPH for hydrodynamics – FSI problems
3.8 Boom and oil spill — Rigid boom
•Free surfaces
•Multiphase
•Fluid-solid interaction
•Wave-current
Yang and Liu, SCPMA 2012
42/68
09:55
Evolution of flow pattern
Frontal area and two ends
Drag force
Flow pattern
Criterion for stability
0.41
( ) arctan( )
x
x
L
L
Lu C u
L
2 VD U Urigid:V = 0
flexible:V = -2/3
3.8 Boom and oil spill — Flexible boom (SPH+EBG)
3. SPH for hydrodynamics – FSI problems 43/68
09:55
Yang & Liu, PRE, 2015; JHD, 2016.
3. SPH for hydrodynamics – FSI problems
3.9 Hydro-elasticity – dam collapse with elastic plate
Displacement of plate free end
Water level
44/68
09:55
Shao & Liu, JHD, 2013.
4. SPH for environmental flows 4.2 Landslide
• Fast landslide – flow like
• Fluid model – non-Newtonian
• Surface topography – GIS data
Tangjiashan Landslide: Pre- and post-earthquake topographies.
45/68
09:55
Hu, Liu et al., Environmental Earth Sciences, 2015.
4. SPH for environmental flows 4.2 Landslide
3D profile SPH model of Tangjiashan landslide
Pre- and post-earthquake topographies
46/68
09:55
Hu, Liu et al., Environmental Earth Sciences, 2015.
4. SPH for environmental flows 4.4 Subsurface flows – porous media
SPH simulation Comparison: SPH (top) and VOF (bottom)
Multiphase flow in a fractured porous media
47/68
09:55
Liu et al., Water Resources Research, 2007.
4. SPH for environmental flows 4.4 Subsurface flows – fracture network
Multiphase flow in a fracture network
DPD模拟 VS VOF模拟
SPH VS VOF、Experiment
48/68
09:55
49
5. SPH for explosion and impact 5.1 1D TNT slab detonation
Liu et al., Shock Waves, 2003
TNT 0.1 m
Ignition point
Solid boundary
Detonation direction
TNT slab detonation
x(m)
P(G
Pa)
0 0.02 0.04 0.06 0.08 0.1
5
10
15
20
25 Von Neumann Spike
C-J State Pressure
Ideal explosive
Non-ideal explosive
Liu et al., Shock Waves, 2013
49/68
09:55
5. SPH for explosion and impact 5.2 2D confined UNDEX
Liu et al., C&F, 2003
Pressure field Bubble evolution
50/68
09:55
5. SPH for explosion and impact 5.3 Water mitigation (WM)
Liu et al., Shock Waves, 2002
Water held in plastic bags as a bomb shelter to mitigate blast effects
0 0.25 0.5 0.75 1 1.25 1.5 1.75 20.4
0.5
0.6
0.7
0.8
0.9
1
Normalized water shield thickness
Nor
mal
ized
pea
k sh
ock
pres
sure
Point 4Point 3Point 2Point 1
Peak shock pressure
0.5
0.0
25
HE
Water
Air
P1 P2 P3
P4
Contact WM
51/68
09:55
5. SPH for explosion and impact 5.4 Middle & high velocity impact
Liu et al., Shock Waves, 2006
2D Taylor bar impact with adaptive SPH
3D Taylor bar impact with SPH
52/68
09:55
5. SPH for explosion and impact 5.5 Penetration
Liu et al., Shock Waves, 2006
Aluminum
6180 m/s
10 c
m
0.4 cm
D=1 cm
Al disk impacting/penetrating al plate at 6180 m/s
Other examples
53/68
09:55
5. SPH for explosion and impact 5.6 Contact explosion
Feng & Liu, Shock Waves, 2013
钢板
TNT炸药
D1 D2
The shape of the crate is closely related to type, size and shape of the explosive charge
54/68
09:55
5. SPH for explosion and impact 5.7 Shape charge with jet formation
Feng & Liu, C&F, 2013
高能炸药引信
外壳
药型罩 装药直径
装药头长 空腔长度
装药长度
外壳厚度
锥角
Metal jets
Simulation
Experiment
Metal jets without & with metal case
55/68
09:55
5. SPH for explosion and impact 5.7 Shape charge with jet formation & penetration
T (s)
Ene
rgy
(MJ)
0.0E+00 5.0E-05 1.0E-04 1.5E-04
Kinetic EnergyInternal EnergyTotal Energy
10
15
20
25
30
5
Feng & Liu, C&F, 2013
56/68
09:55
5. SPH for explosion and impact 5.8 Explosive-driven welding
v
Welding from direct impact Deribas (1967), Hydrodynamic model
57/68
09:55
6. Prospects and future directions
Better theories Relationship between particle-particle interactions,
Theoretical foundation for multi-scale modeling,
Fluid-solid interaction model.
Applications:
Protective technology,
Coastal engineering and ocean hydrodynamics,
Civil and environmental engineering,
Bio/nano engineering,
Movie and animation making…
59/68
09:55
60
6. Prospects and future directions
Target:3D SPH code for FSI with wave and currents
1 Features • 2 and 3D • Lagrangian, particle method • Nonlinear • Explicit algorithm • Multi-materials, Multi-scales,
Multi-physics
2 Typical applications • Free surface flows(breaking waves、
dam collapse, flood…)
• FSI(slamming、sloshing、water-on-deck、 high speed water entry/exit)
• Extreme load(explosion, impact, penetreation)
2 Advantages • Meshfree: easy in treating large
deformation • Lagrangian particle: easy in
treating free surfaces and moving interfaces.
• FSI: simutaneously coupling
Code • Released the first SPH open source code; • Developing a 3D SPH code for hydrodynamics and
environmental flows
60/68
09:55
5. Prospects and future directions
Coupling approaches: 1. Coupling SPH/DEM for problems with solid particles (landslide, e.g.),
landslide mud flow
61/68
09:55
5. Prospects and future directions
Coupling approaches: 2. Coupling particle methods with FEM/FDM for boundary enhancement
and fluid-solid interaction,
SPH+FEM modeling 911
62/68
09:55
5. Prospects and future directions
Coupling approaches: 3. Coupling MD/DPD/SPH for multi-scale simulation,
Multi-scale modeling:from nano to micro and to macro scale
DPD particles
MD atoms SPH particles
63/68
09:55
5. Prospects and future directions
Coupling approaches: 4. Coupling multiple physics (viscous, elastic, heat, chemical,
electronic…).
Electrochemical Phenomena: Induced Polarization
64/68
09:55
Collaborators
G. R. Liu, Univ. Cincinnati
W. K. Liu, NWU
Paul Meakin, H. Huang, INL
S. K. Tan, NTU
Z. Shang, Daresbury Lab, UK
Team members
X. Wen, Z. L. Zhang, P. Y. Yang, PKU
J. R. Shao, X. F. Yang, …,Imech
Z. Chen, G. X. Zhu, DLUT
L. Cheng, SCU
J. Z. Chang, H. T. Liu, …, NUC
Chinese Academy of Sciences
NSFC, China
LMFS,LNM, LHMRE
Funding organizations
National University of Singapore (NUS)
A-Star, Singapore
DSTA, Singapore
Idaho National Laboratory (INL)
DoE, US
Lee Kuan Yew (LKY) Foundation
Acknowledgement 65/68
09:55