Lagrangian Particle Tracking

14. 5 Release

Multiphase Flow Modeling

in ANSYS CFX
© 2013 ANSYS, Inc. 7-1 Release 14.5
 Overview
• General description
• Particle tracking in ANSYS CFX
• Particle transport equations
• Particle integration algorithm
• General setup of particle tracking in ANSYS CFX
• Particle injection
• Turbulent dispersion
• Wall Treatment
• Particle integration controls
• Particle variables as vertex variables
• Particle-particle collisions
© 2013 ANSYS, Inc. 7-2 Release 14.5
 Introduction

• Method for modeling particle laden flows

• A representative sample of particles is tracked through the continuous fluid

• Ordinary differential equations (ODE) are integrated for position and

velocity of each particle

• The overall mass flow rate of the particle phase is shared amongst the
representative particles, so each particle has its own mass flow rate, and its
own number rate

© 2013 ANSYS, Inc. 7-3 Release 14.5

 Lagrangian or Eulerian?

• Lagrangian particle tracking is an alternative to Eulerian-Eulerian

multiphase modeling. Care should be taken to choose the better
method for the application!

• The same physics is essentially modeled

• Although particles have an associated diameter, they are modeled as

moving points, so they take up no volume of the continuous fluid.
Particle-particle interactions are also neglected and thus the model is
only applicable at low volume fractions

© 2013 ANSYS, Inc. 7-4 Release 14.5

 Simplifying Assumptions
• Moving particle is treated as a moving mass point
– Abstraction from particle shape and volume

• Details of the flow around the particle are neglected, e.g.

– Near particle flow field
– Vortex shedding
– Flow separation
– Boundary layers

• Local properties of the dispersed phase are predicted from spatial averaging
over particle trajectories which are crossing a certain (control) volume

• Not possible to predict the tracks of all physical particles

– Use concept of representative particle trajectories

© 2013 ANSYS, Inc. 7-5 Release 14.5

 Representative Trajectories
• Tracking all particles is impractical. Consider:
Fluid mass flow :
 F  1 kg / s
m
Particle mass flow (10% loading):
m P  0.1 kg / s
Required particles per second:
dP=1cm  ~ 2  105
dP=1mm  ~ 2  108
dP=100m  ~ 2  1011

• Particle equations are solved for representative particle trajectories resulting in

a particle number rate along each predicted particle:

m P 4
n P  with mP  r 3 
mP 3
• The particle number rate is determined based on the chosen diameter
distribution and the specified particle mass flow

© 2013 ANSYS, Inc. 7-6 Release 14.5

 Lagrangian vs. Eulerian

• Eulerian • Lagrangian
– Interpenetrating continua – Continuous phase as in Eulerian
– Definition of phase-weighted approach
averages – Disperse phase as moving mass
points
– PDEs for dependent variables of
– ODEs for particle motion
all phases, including interphase
transfer terms: – Phase interaction terms:
• Mass
• Mass
• Momentum
• Momentum
• Energy
• Energy

© 2013 ANSYS, Inc. 7-7 Release 14.5

 Lagrangian Pros and Cons

• Pro: •Con:
– Complete information about – The model is only valid at low
behavior and residence times volume fractions
of individual particles
– Can be very expensive if it is
necessary to track a large number
– Relatively cheaper than of particles
Eulerian modeling for a wide
range of particle sizes
– Difficult to get smooth
information about local values of
– Better detail for drag, heat and volume fractions, velocities,
mass transfer if particles are forces on walls etc.
changing size

© 2013 ANSYS, Inc. 7-8 Release 14.5

 Implementation in ANSYS CFX

• Fully embedded in CFX-Pre GUI

• Tracking in physical space
• Runs in parallel
• Steady-state or transient
• Full support of multiple frames of reference and multiple zones (all GGI types)
• Arbitrary number of particle types possible
• Combination of one-way and fully coupled particles possible
• All element types supported
• Turbulent tracking (random dispersion) implemented
• Spray drier model
• Coal and oil combustion
• Particle breakup models
• Erosion models
• Wall film models
© 2013 ANSYS, Inc. 7-9 Release 14.5
 Implementation in ANSYS CFX

• Particle momentum
• Particle heat and mass
• Injection on boundary condition patches
– Inlets, openings and walls
– Volume injection supported with particle injection regions
• Tracks information written to the results file

© 2013 ANSYS, Inc. 7-10 Release 14.5

 Force Balance on a Moving Particle
• Particle translation :

d rP  
 UP 
dt UF F 
• Particle momentum: mp UP
 
dU P  rP
mp F z y
dt
x
with
      
F  FD  FB  FR  FV  FP  FH

• Derivation of particles equations of motion from the balance of aerodynamic

and body forces at the location of the particle in the flow.

© 2013 ANSYS, Inc. 7-11 Release 14.5

 Force Balance

 
    
Drag FD  CD AF U P  U F U P  U F
 
Buoyancy FB   P  C g
System      
rotation FR  2 U P  P      rP  P

 
  
Added mass
FV  12 M F dU F dt  dU P dt
Pressure  
gradient FP  M F dU F dt  M F  F p
Basset FH  History of accelerati on

© 2013 ANSYS, Inc. 7-12 Release 14.5

 Particle Drag Law
• Default drag law in ANSYS CFX is Schiller-Naumann
(Grace, Ishi-Zuber also available)
 1  
FD   S CD US US 0  Re p  O(103 )
• with 2
 d2
– Projected particle surface area S
4
– Drag correlation CD 
24
Re

1  0.15 Re 0P.687 

dP U S
– Particle Reynolds number Re P 
F
  
– Slip velocity US  U F U P

© 2013 ANSYS, Inc. 7-13 Release 14.5

 Particle Drag Law
• Non-spherical particles can be obtained by setting the cross-
sectional factor (CCL) to a number different from 1:
 d P2
S  CCSF
4

• The drag coefficient can be multiplied by a factor which can be

specified with the expert parameter

– PT DRAG COEFFICIENT MULTIPLIER

© 2013 ANSYS, Inc. 7-14 Release 14.5

 Particle Injection
• Particles can be injected at boundary patches:
– Supported types: wall, inlet, outlet, opening
• The following information must be set:
– Velocity of the particles
• Can be same as the fluid velocity (via CEL)
– Injection Location
– Number of particle positions
• Not required for injection at integration points or element face centers
– Mass flow rate for all injected particles
• Required for fully coupled solutions. Only used for post-processing if one-way
coupled
– Diameter distribution
• Can also be set in the MORPHOLOGY section if the same distribution is used at all
boundaries

© 2013 ANSYS, Inc. 7-15 Release 14.5

 Turbulent Dispersion

• Instantaneous fluid velocity is decomposed into mean and fluctuating components.

U f  U f  uf
– Mean velocity component affects average particle trajectory.
– Fluctuating velocity component causes dispersion of particles in a
turbulent flow.

• Issues:
– Turbulent flows consist of cascade of turbulent structures.
– Eddies have a spectrum of characteristic sizes, life times and energy levels.
– Particles traveling through a turbulent flow are exposed to entire spectrum of eddies,
interacting with each for some characteristic time or distance.

© 2013 ANSYS, Inc. 7-16 Release 14.5

 Turbulent Dispersion
• Assumptions:
– Single characteristic eddy may be defined.
– Each eddy has a characteristic fluctuating velocity u’f, lifetime te and
length le.
– When a particle enters the eddy, the fluctuating velocity for that eddy is
added to the local mean fluid velocity to obtain the instantaneous fluid velocity.
– Turbulent fluid velocity prevails as long as:
• Particle/eddy interaction time is less than the eddy lifetime.
• Displacement of the particle relative to the eddy is less than the eddy length.
• A GGI boundary is reached.
• If either of these conditions is exceeded, the particle is assumed to be entering
a new eddy with new characteristic u’f, te and le.

© 2013 ANSYS, Inc. 7-17 Release 14.5

 Turbulent Dispersion Example

• Particles with
turbulent dispersion:

• Particles without
turbulent dispersion:

© 2013 ANSYS, Inc. 7-18 Release 14.5

 Wall Treatment
• Coefficient of Restitution
– Perpendicular to wall C R , perp

– Parallel to wall C R , par

– Particles come to rest with CR , perp  0

C R , perp 1 1 <1 0
C R , par 1 <1 1 1

© 2013 ANSYS, Inc. 7-19 Release 14.5

 Symmetry and Periodic Boundaries
• Particles perfectly bounce of symmetry walls

• Particles track through periodic boundaries

© 2013 ANSYS, Inc. 7-20 Release 14.5

 Integration Control
• There are default settings for the particle integration routine which may not
be suitable for all cases
• If you see messages in the particle fate summary that particles have
exceeded the maximum time or distance, then these defaults should be
revisited
• These settings can be accessed from CFX-Pre/Solver Control/Particle Control
under Particle Integration and Particle Integration Control
• For one-way and fully coupled particles:
– Number of Integration Steps per Element
(10)
– Maximum Tracking Time (10 s)
– Maximum Tracking Distance (10 m)
– Maximum Number of Integration Steps
(10000)

© 2013 ANSYS, Inc. 7-21 Release 14.5

 Full Coupling Control
•Additional controls for fully coupled particles:
– First Iteration for Particle Calculation (10)
• Eulerian flow can settle down!
– Iteration Frequency (5)
– Particle Source Target (0.01)
• Based on the rate of change of
sources
– Under Relaxation Factors (0.75)
• Velocity Under Relaxation Factor
• Energy Under Relaxation Factor
• Mass Under Relaxation Factor

© 2013 ANSYS, Inc. 7-22 Release 14.5

 Vertex Variables
• With a few exceptions, Particle variables are available on Particle tracks, but not
on the fluid mesh
– There are occasions when you would like to do post-processing of particle variables on
the fluid mesh (to calculate particle mass flows through a plane, for example)
– Under Output Control, you can specify a number of Particle variables to be copied to the
CFD mesh vertices. If you specify at least one such variable, you will also get the particle
mass flow rates

© 2013 ANSYS, Inc. 7-23 Release 14.5

 Smoothing Vertex Variables
• You can also specify smoothing of particle vertex variables from
Solver Control/Particle Control

© 2013 ANSYS, Inc. 7-24 Release 14.5

 Particle Collision Model - Motivation
• Standard Lagrange simulations are constrained to dilute gas-particle flows
– Particle-particle interactions do not play an important role
– Flow is dominated by momentum exchange between particles and fluid phase
 Two-way coupling

• For flows, where the inter-particle spacing

L 
3
dP 6 P
is less than approx. 10, momentum transfer due to particle-particle collisions
becomes increasingly important and needs to be accounted
for  Four-way coupling

© 2013 ANSYS, Inc. 7-25 Release 14.5

 Particle Collision Model - Conception

• Statistical collision model (Sommerfeld)

– Computational effort of simultaneously tracing all particles is not required
– Instead an iterative approach is used:
• Sequential calculation of particle trajectory
• Compute statistical particle properties (mean and standard deviation of
droplet diameter and velocities)
• Creation of a virtual collision partner according to local statistical mean
particle properties
• Random process decides whether or not a collision takes place

© 2013 ANSYS, Inc. 7-26 Release 14.5

 Particle-Particle Collision - Setup
• Particle-Particle Collision selected on Domain
Fluid Pair Models tab

• Tick “Particle Collison” to activate and get list of

available Collision Models
– Sommerfeld
– User Defined

• Input parameters for Sommerfeld Model:

– Collision Coefficient of Restitution
– 0: Inelastic collision (no coalescence!)
– 1: fully elastic collision

– Collision Static Friction Coefficient

– Collision Kinetic Friction Coefficient
– Steel - Steel: 0.15; Glass – Glass: 0.4; Coal –
Coal: 0.4

© 2013 ANSYS, Inc. 7-27 Release 14.5

 Description of Validation Experiment
Validation by experiment of Fohanno & Oesterlé
• Enforced crossing of trajectories
• Flow induced by gravitation
• Glass particles, dP = 3 mm
• P = 2500 kg/m3
• Collision effects dominate

© 2013 ANSYS, Inc. 7-28 Release 14.5

 Particle Trajectories With Collision Model

without collision model with collision model

© 2013 ANSYS, Inc. 7-29 Release 14.5

 Notes on Particle-Particle Collision

• Validation report available

“A stochastic particle-particle collision model for dense gas-particle flows
implemented in the Lagrangian solver of ANSYS CFS and its validation”, 6th
International Conference on Multiphase Flow, ICMF 2007, Leipzig, Germany, July 9-
13, 2007, Paper No. 148, pp 1-16

• This is an expensive model!

Particle integration time step may become very small compared to non-collision
simulation (up to several (~2 - 4) orders of magnitude)

© 2013 ANSYS, Inc. 7-30 Release 14.5