Documente Academic
Documente Profesional
Documente Cultură
2008
11 March 2008
Kari Myöhänen
kari.myohanen@lut.fi
Presentation Outline
• Introduction
• Modeling options and limitations in Fluent
• Model theory
• Solution strategies
• Example calculation
Introduction
• The discrete phase model (DPM) in Fluent follows the Euler-Lagrange approach.
• The fluid phase (gas or liquid, “continuous phase”) is treated as a continuum by
solving the time-averaged Navier-Stokes equations (Eulerian reference frame).
• The dispersed phase is solved by tracking a number of particles through the
calculated flow field of continuous phase (Lagrangian reference frame).
• The particles may be taken to represent solid particles in gas or liquid, liquid droplets
in gas or bubbles in liquid.
• The dispersed phase can exchange momentum, mass and energy with the fluid
phase.
Discrete Phase Modeling Options in Fluent
For smooth spherical particles, Fluent uses equation by Morsi and Alexander (1972):
For nonspherical particles, the equation by Haider and Levenspiel (1989) is used:
Shape factor
Surface area of sphere with same volume
Actual surface area
Comparison of Drag Coefficient Equations
Coupling
• The discrete phase and the continuous phase can be coupled in a number of ways.
In Fluent, the one-way or two-way coupling are possible to model.
• One-way coupling
– The continuous phase affects the discrete phase, but there is no reverse effect.
– In Fluent, this is referred as “uncoupled approach”.
– The discrete phase is solved once after the continuous phase flow has been
solved.
• Two-way coupling
– Both phases affect each other (exchange of momentum, mass and energy).
– In Fluent, this is referred as “coupled approach”.
– The continuous phase flow field is impacted by the discrete phase and the
calculations of the continuous phase and dispersed phase equations are
alternated until the solution is converged (hopefully).
• Three-way coupling
– Particle disturbance of the fluid locally affects
another particle’s motion, e.g. drafting of a trailing particle.
• Four-way coupling
– Particle collisions affect motion of individual particles.
Two-Way Coupling in Fluent
Momentum exchange
Heat exchange
(without chemical reactions)
Vaporization
Sensible heat
and pyrolysis
Mass exchange
Particle Types and Laws in Fluent
Temperature
Different energy and mass transfer equations are applied during different laws.
Tbp Law 6:
Law 3:
Boiling Volatile fraction
consumed
Tvap
Law 2:
Tinjection Vaporization
Law 1:
Inert heating
before vaporization
Particle time
Mass and Energy Transfer of Drying Droplet
Heat transfer
Convection Radiation
Law 2: Vaporization
Heat transfer
Evaporation
Law 3: Boiling
Different particle boundary conditions can be defined for walls, inlets and outlets:
volatile fraction
flashes to vapor
Normal component:
Tangential component:
Turbulent Dispersion of Particles
In Fluent, the dispersion of particles due to continuous phase turbulence can be modeled by
• a stochastic tracking model (random walk model, eddy interaction model), or
• a particle cloud model.
In the random walk model, the instantaneous continuous phase velocity is formed
of mean velocity and fluctuating component:
The cloud model uses statistical methods to trace the turbulent dispersion of particles
about a mean trajectory
• Mean trajectory is calculated from the ensemble average of the equations of motion
for the particles represented in the cloud.
• Distribution of particles inside the cloud is represented by a Gaussian PDF.
Eddy Interaction Model
The stochastic tracking model in Fluent is based on eddy interaction model.
The discrete particle is assumed to interact with a succession of eddies.
Each eddy is characterized by
• a Gaussian distributed random velocity fluctuation u’i
• a time scale (life time of eddy) e
• a length scale (size of eddy) Le
During interaction, the fluctuating velocity is kept constant.
The interaction lasts until time exceeds the eddy lifetime or the eddy crossing time.
Literature presents several theories for determining the above values (see Graham and James (1996)).
The following presents the equations used in Fluent with k- turbulence model.
Fluid Lagrangian integral time Coefficient CL defined by user. Default value CL = 0.15.
The mass flow of a particle track is determined based on particle mass and
mass flow at the particle injection and particle mass at current location.
The particle mass can change due to evaporation and other phase changes.
The discrete phase concentration inside a cell can be determined from the
residence time and mass flow.
Inside a cell, the particle stream is tracked with n particle time steps. The
residence time of one particle track is the sum of these time steps.
• The particle tracks are calculated in steps. The ”step length factor” determines
approximately the number of steps per fluid cell. The default value is 5, but it should
preferably be higher: 10 – 20.
• Increasing the step length factor (i.e. decreasing the step length) can improve stability of
heat and mass exchange (e.g. when calculating vaporization).
• The ”max. number of steps” limits the number of calculated time steps. This should be
large enough so that the particles can travel from entrance to exit.
• If particles remain suspended in the model (tracking incomplete), then steady state
solution is questionable and transient tracking should be used instead. The transient
calculations in Fluent can be performed in a number of ways and combinations. This
presentation is focused on steady state calculation.
Solution Strategies: Two-Way Coupling
• The solution of the continuous field without coupling is usually the starting point.
In most cases, the continuous flow does not have to be fully converged before
the coupling is started, because the particle tracks will have a large effect on the
continuous flow.
• In a coupled calculation, additional source terms appear in discretized flow Calculate
equations of continuous phase. During particle tracking, each particle is seeing continuous
a ”fresh” cell and makes no notice of particles already visited and marked the phase
cell with their source terms. This leads to overprediction of the source terms and
bad convergence behaviour with evaporation, combustion and radiation.
– Use solution limits to limit the temperature in the domain. Calculate
particle
• Increasing the number of trajectories (especially with random walk model) will
tracks
smooth the particle source terms, which should help convergence.
• The discrete phase source terms can be under-relaxed (e.g. 0.5). The flow
equations may need to be under-relaxed as well (energy and species).
Update
• The number of continuous phase calculations between the trajectory source
calculations can either be small (< 3) or high (>15). In the first choice, the terms
dispersed and continuous flow are closer coupled and the solution of both
should slowly convergence. In the second choice, the flows are decoupled and
the solution of continuous field remains ”better converged” and the calculation is
more stable. In the latter case, the continuous phase may appear to be
converged, but the discrete phase is not.
• If the dispersed phase is not dilute, then convergence is very difficult to achieve
in coupled calculations.
Modeling Example
• The DPM model in Fluent can be used for studying one-way or two-way
coupled dilute dispersed flows, including effects of turbulence.
• The basic model is easy to use and physics are clear and simple.
• The limitations of the DPM model should be carefully considered when
analyzing the results.
• The model neglects particle-particle interaction, thus it is valid for dilute
dispersed phase only.
• The one-way coupling is valid for very dilute flow only. The two-way coupled
solution can be much different from the one-way coupled solution.
• The average flow can be dilute, but it can contain regions, in which the
dispersed phase is dense. In these regions, the model results are false.
Moreover, the convergence is poor, if the dispersed phase is dense and the
momentum, mass and energy exchange to continuous phase is strong.
• Despite the limitations, the DPM model can be (and is) successfully used for
modeling various applications.
References