Indian Institute of Technology Delhi

Two-Fluid vs DEM Modeling of Fluidized

Beds
CLL110 Transport Phenomena Term paper

Author: Naitik Chaudhary Author: Bhoomi Author: Seenu

Entry No.: 2024CH10822 Entry No.: 2024CH10827 Entry No.: 2024CH10619
Department: Chemical Engg. Department: Chemical Engg. Department: Chemical Engg.
E-mail: ch1240822@[Link] E-mail: ch1240827@[Link] E-mail: ch1240619@[Link]

November 14, 2025

 Two-Fluid vs DEM Modeling of Fluidized Beds

Abstract
Fluidized beds display highly nonlinear gas–solid interactions driven by bubble forma-
tion, interface instabilities, and particle-scale dynamics. Accurately modelling such sys-
tems requires numerical methods that capture both global hydrodynamic behavior and local
particle-level effects. In this study, we present a comparative analysis of two major modelling
frameworks: the Two-Fluid Model (TFM), which treats gas and solid phases as interpenetrat-
ing continua, and the Discrete Element Method (DEM), which explicitly resolves individual
particle trajectories and collisions. The mathematical formulation is built on mass and mo-
mentum conservation, incorporating interphase drag laws and granular stress closures.
A custom C++ program was implemented to compute bubble rise velocity, Reynolds
number, and drag coefficient, providing a numerical verification of analytical expressions and
literature correlations. The computed results showed strong agreement with classical models,
thereby validating the theoretical foundation of both TFM and DEM.
To enhance physical interpretation, we also developed a Fluent machining animation using
ANSYS, illustrating material removal, stress distribution, and tool–workpiece interaction.
Although separate from fluidization, this animation served as a visual demonstration of how
numerical models translate into real physical processes, reinforcing the connection between
theory, simulation, and observable behavior.
Overall, our findings show that TFM is highly efficient and reliable for predicting global
parameters such as pressure drop and minimum fluidization velocity, while DEM excels at
capturing microscale flow structures, bubble boundaries, and particle clustering. The study
emphasizes the complementary nature of these models and highlights the potential of hybrid
approaches that combine the accuracy of DEM with the scalability of TFM.

Two-Fluid vs DEM Modeling of Fluidized Beds 1

 Contents

Contents
1 Introduction 1
1.1 Fluidized beds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Two-fluid model(TFM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Eulerian–Lagrangian Discrete Element Method (DEM) . . . . . . . . . . . . . . . 2
1.4 Why Comparing TFM and DEM Matters . . . . . . . . . . . . . . . . . . . . . . . 2

2 Prior works and results 3

2.1 Early Experimental Theoretical Work . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Emergence of Two-Fluid Models (TFM) . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Development of DEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Problem Formulation 4
3.1 Fundamentals of Fluidization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Governing Equations for Gas and Solid Phases . . . . . . . . . . . . . . . . . . . . 5
3.2.1 Two-Fluid Model (TFM, Euler–Euler Approach) . . . . . . . . . . . . . . . 5
3.3 Discrete Element Method (DEM, Euler–Lagrange Approach) . . . . . . . . . . . . 6
3.3.1 Assumptions for this work . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4 Numerical analysis 7
4.1 General algebraic force balance (spherical body) . . . . . . . . . . . . . . . . . . . 8
4.2 Stokes (creeping flow) limit — derivation . . . . . . . . . . . . . . . . . . . . . . . 8
4.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.4 Computational Implementation(C++ Code) . . . . . . . . . . . . . . . . . . . . . 10
4.5 DEM Particle Terminal Update (Per Particle, Per DEM Time Step) . . . . . . . . 10
4.6 TFM Solver (Per Fluid Time Step) . . . . . . . . . . . . . . . . . . . . . . . . . . 11

5 Results and discussions 11

5.1 Minimum Fluidization Velocity Umf . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.2 Pressure Drop at Fluidization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.3 Bubble rise velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.4 Particle terminal velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.5 Relative magnitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

6 Extended results 14
6.1 Minimum Fluidization VelocityUmf . . . . . . . . . . . . . . . . . . . . . . . . . . 14
6.2 Pressure Drop (P) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
6.3 Bubble Rise Velocity (Ub ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
6.4 Particle Terminal Velocity (Upt erm) . . . . . . . . . . . . . . . . . . . . . . . . . . 15
6.5 Using ANSYS Fluent - Fluidised bed TFM Simulation . . . . . . . . . . . . . . . 16

7 Path forward 17

8 Conclusion 18

A Self-assessment 21

Two-Fluid vs DEM Modeling of Fluidized Beds 2

 1 Introduction

Figure 1: Fluidized beds-an overview

1 Introduction
1.1 Fluidized beds
A fluidized bed is a state of a two-phase mixture of particulate solid material and fluid, which
is widely used in many modern technologies for efficient implementation of various physical and
chemical processes. Fluidized beds have been used in technological processes such as: crack-
ing and reforming of hydrocarbons (oil), carbonization and gasification of coal, ore roasting,
Fischer-Tropsch synthesis, polyethylene manufacturing, limestone calcining, aluminum anhydride
production, granulation, vinil-chloride production, combustion of waste, nuclear fuel prepara-
tion, combustion of solid, liquid and gaseous fuels, drying, adsorption, cooling, heating, freezing,
conveying, storing and thermal treating of various particulate solid materials.
The term "fluidized bed" is unavoidably connected to the term "particulate solid material".
Particulate materials are mechanical mixtures of multitude of solid particles. Natural particulate
materials originate from many long-term natural processes: heating, cooling, thermal dilatation,
coliding, crushing, chopping up, atmospheric changes, river erosion and erosion caused by sea
waves. Many technological processes also produce particlate solid material: grinding, chopping
up, milling, evaporation, crystalization, spraying and drying. Particulate materials can also be of
organic (plant) origin: fruits and seeds.
Particulate materials most commonly consist of solid particles with a range of shape and size.
The majority of inorganic particulate solid materials found in nature have an extremely wide
range of particle sizes. Such materials are called polydisperse materials. By certain technological
processes, it is possible to produce particles with practically the same shape and size. Organic
particulate materials found in nature (fruits and seeds) consist of particles of similar shape and
size. Such materials are called monodisperse materials.
1.2 Two-fluid model(TFM)
The Two-Fluid Model (TFM), also known as the Euler-Euler approach, treats both the gas phase
and the solid particle phase as continuous, interpenetrating fluids. Each phase is assigned its
own set of governing equations for mass and momentum conservation, capturing their interaction
through exchange terms such as drag forces. This model simplifies the complex fluid-particle
system by averaging particle behavior over volume, which makes it computationally efficient for
simulating large-scale fluidized beds.
To visualize this, imagine a fluidized bed like a system of two overlapping fields—one repre-
senting the gas flow and the other representing the suspended particles—each with its own velocity
and pressure fields. These continua continuously exchange momentum and energy through their
interfaces. The Euler-Euler model calculates how the particles move collectively within the gas
flow, predicting phenomena like bubble formation and bed expansion without tracking individual
particles explicitly. This approach contrasts with the particle-based Discrete Element Method
(DEM), which follows each particle’s trajectory.

Two-Fluid vs DEM Modeling of Fluidized Beds 1

 1 Introduction

Figure 2: Two-fluid model (Euler-Euler) Figure 3: DEM Model

1.3 Eulerian–Lagrangian Discrete Element Method (DEM)

The Discrete Element Method (DEM) refers to a computer approach for simulating and studying
the behavior of systems that are comprised of discrete particles or solid elements. It can very
much simulate particulate materials such as granular flows, powders, soils, and sands as well as
pneumatic transport systems and fluidized beds. It can allow users to closely observe motion,
contact, and collision between particular particles in a material.
In contrast to continuum approaches such as the Two-Fluid Model (Euler-Euler), where the
two phases are treated as continua, the DEM simulates the interaction and motion of each particle
on an individual basis. The characteristic particle approach thus captures the finer particulars of
granular behavior such as intricate particle-particle collision behavior, interparticle friction, and
breakage of particles.
1.4 Why Comparing TFM and DEM Matters
The Two-Fluid Model (TFM) and Discrete Element Method (DEM) complement each other
by addressing different scales and aspects of fluidized bed behavior. TFM approximates gas and
solid phases as two continuous interacting fluids, which enables simulation of large-scale industrial
systems with relative computational ease. This makes TFM especially suitable for reactor design,
process scale-up, and operational optimization where capturing overall flow structures and average
bubble dynamics is sufficient.
In contrast, DEM tracks individual particles, capturing detailed particle-scale phenomena such
as collisions, clustering, and mixing patterns. This high-resolution approach is critical for under-
standing microscale interactions that influence performance, safety, and reliability, particularly
in sensitive applications like catalyst beds or powder processing. However, DEM simulations re-
quire significantly more computational resources, limiting their practical use to smaller systems
or detailed studies.
Comparing TFM and DEM allows researchers and engineers to select appropriate tools de-
pending on their specific goals and constraints. While TFM provides a cost-effective means for
comprehensive system studies, DEM offers unparalleled insights into particle-level processes. To-
gether, they enable a holistic understanding of fluidized beds, guiding safer and more efficient
designs.

Scope and Structure of the Paper

In summary, this paper compares the Two-Fluid Model (TFM) and the Discrete Element

Two-Fluid vs DEM Modeling of Fluidized Beds 2

 2 Prior works and results

Method (DEM) to evaluate their effectiveness in predicting bubble dynamics and gas–solid in-
teractions in fluidized beds. The study links theoretical formulation with a C++ computational
model to validate parameters such as bubble rise velocity, minimum fluidization velocity, and
drag coefficient.
The remainder of this report is organized as follows: Section 2 reviews prior research on
fluidized beds and numerical modeling; Section 3 presents the problem formulation and governing
equations; Section 4 explains the numerical methodology; Section 5 discusses the results and
observations; Section 7 outlines future scope; and Section 8 concludes the work.

2 Prior works and results

A review of previous research is essential to understand how fluidized beds have been modeled over
time and why modern approaches like TFM and DEM were developed. This section summarizes
the key experimental findings, evolution of numerical models, and their relevance to our study.
2.1 Early Experimental Theoretical Work
Early investigations into fluidized beds were entirely experimental. Researchers observed that
when gas flows upward through a packed bed of particles, a critical velocity is reached at which
particles begin to lift and behave like a fluid. This point was defined as the minimum fluidization
velocity (Umf ).
One of the most influential contributions was by Davidson and Harrison (1963), who studied
bubble formation and motion in gas–solid beds. They proposed that gas bubbles in a fluidized
bed rise similarly to bubbles in a liquid. Based on their experiments, they developed an empirical
relation for bubble rise velocity (Ub ):
p
Ub ≈ U0 + 0.711 gdb
where U0 is the superficial gas velocity, db is bubble diameter, and g is gravitational accelera-
tion. This equation became a benchmark for validating numerical models.
In addition, early researchers established simple pressure drop relations by balancing particle
weight and drag force:

∆P = (1 − εmf )(ρs − ρg )gH

These experimental findings laid the foundation for later numerical approaches. They provided
reference values—such as Umf , bubble velocity, and pressure drop—that are still used to check
whether TFM and DEM simulations are realistic.
2.2 Emergence of Two-Fluid Models (TFM)
The Euler–Euler Two-Fluid Model (TFM) was introduced to treat both gas and solids as in-
terpenetrating continua. Early TFM works (Syamlal, Gidaspow, and others in the 1980s–90s)
developed closure relations for drag and granular stresses based on kinetic theory of granular flow
(KTGF).
The core equation for the drag force per unit volume was formulated as:

Mgs = Kdrag (ug − us ) (1)

with Kdrag derived empirically. A widely used form is the Wen–Yu correlation (for dilute beds):

Kdrag = 3/4Cd (αg αs ρg |ug − us |)/dp (2)

where:
ρg dp |ug −us |
Cd = Re24
p
, (1 + 0.15Re0.687
p ) and Rep = µg
.

Two-Fluid vs DEM Modeling of Fluidized Beds 3

 3 Problem Formulation

Figure 4: Bubble rise and dem simulation

2.3 Development of DEM

The Euler–Lagrange DEM approach models each particle explicitly. This was enabled by com-
putational advances in the 1990s–2000s. DEM solves for each particle’s trajectory:
X
mi dvi /dt = mi g + Fc,ij + Fd,i (3)
with contact forces often modeled by the Hertz–Mindlin spring–dashpot law:
Fc,ij = kn δn,ij nij − γn vn,ij nij (4)
where:

kn = normal stiffness, γn = damping coefficient

δn,ij = normal overlap displacement
vn,ij = relative normal velocity
nij = unit vector along particle contact

3 Problem Formulation
To compare TFM and DEM in a meaningful way, it is necessary to clearly define the physical
system, governing equations, and assumptions used in both models. This section formulates the
basic problem of gas–solid fluidization by outlining the fundamental principles of fluidization,
followed by the mathematical models used in TFM and DEM. These formulations form the
foundation for the numerical analysis and simulation discussed later in this report.
3.1 Fundamentals of Fluidization
A fluidized bed occurs when a fluid, e.g., a gas or liquid flows up a bed of solid particles by
a velocity adequate to counteract the gravitational force on particles. Minimum fluidisation
condition arises when the drag force provided by the fluid equalises the effective weight per unit
bed cross-sectional area of particles. The force balance at minimum fluidization is:
∆P A = (1 − ϵmf )ρs gAH (5)
where:
• ∆ P=presssure drop across the bed(pa)
• A=cross-sectional area of the bed(m2)
• ϵmf =void fraction at minimum fluidization
• ρs =solid particle density(kg/m3)
• g=gravitational acceleration
• H=bed height at minimum fluidization(m)
The fluid velocity here is defined as the minimum fluidisation velocity, asUmf (m/s).
At U0 > Umf , bubbles rise through the bed and create intense gas–solid interaction. These
intense gas–solid interactions are the focal points of transport phenomena (momentum, mass, and
thermal energy transport).

Two-Fluid vs DEM Modeling of Fluidized Beds 4

 3 Problem Formulation

3.2 Governing Equations for Gas and Solid Phases

3.2.1 Two-Fluid Model (TFM, Euler–Euler Approach)
In TFM, both phases are treated as interpenetrating continua, with gas volume fraction αg and
solid volume fraction αs such that:
αg + αs = 1 (6)
Continuity equations (mass conservation):
• Gas phase:

∂(αg ρg )
+ ∇ · (αg ρg ug ) = 0
∂t
(7)

• solid phase

∂(αs ρs )
+ ∇ · (αs ρg us ) = 0
∂t
(8)

Momentum equations (Newton’s second law for each phase):

• Gas phase:

∂(αg ρg ug )
+ ∇ · (αg ρg ug ug ) = −αg ∇p + ∇ · (αg τg )
∂t
+ αg ρg g − Kgs (ug − us )
(9)
• Solid phase:

∂(αs ρs us )
+ ∇ · (αs ρs us us ) = −αs ∇p + ∇ · (αs τs )
∂t
+ αs ρs g + Kgs (ug − us )
(10)
where:
• ug , us = gas and solid velocities (m/s)
• p = pressure (Pa)
• τ g = µg [∇ug + (∇ug )T ] = viscous stress tensor of gas (Pa)
• τ s = solid-phase stress tensor (Pa), modeled using kinetic theory of granular flow
• Kgs = interphase momentum transfer coefficient (N·s/m ), also called drag coefficient
Closure for drag coefficient: For laminar flow and dense regimes, Gidaspow’s correlation is widely
used:
150µg (1 − αg )2 1.75ρg (1 − αg )
Kgs = + |ug − us |
αg d2p dp
(11)

Two-Fluid vs DEM Modeling of Fluidized Beds 5

 3 Problem Formulation

Figure 5: Experimental setup

where dp = particle diameter (m).

This formulation allows TFM to represent bubble formation and rise as regions of low αs (high
gas voidage).
3.3 Discrete Element Method (DEM, Euler–Lagrange Approach)
In DEM, the gas phase is treated as a continuum, while each particle is tracked individually.
Fluid equations (gas):
The gas still satisfies Navier–Stokes (laminar):

∇ · ug = 0 (12)

  PNp
∂ug
ρg ∂t
+ ug · ∇ug = −∇p + µg ∇2 ug + ρg g − 1
Vc i=1 Fd,i (13)

where:
• Vc = volume of computational cell (m 3 3 )
• Np = number of particles in that cell
• Fd,i = drag force from particle i (N)
Particle equations: Each particle i of mass mi is solved using Newton’s law of motion:
c N
dvi X
mi = mi g + Fd,i + Fc,ij
dt j=1

(14)

where:
• mi = π6 ρs d3p = mass of particle i (kg)
• vi = particle velocity (m/s)
• Fd,i = 3πµg dp , (ug − vi ) = laminar Stokes drag (N)
• Fc,ij = collision force from particle j (N), modeled by contact laws
Collision force model (linear spring–dashpot):

Fc,ij = kn δn,ij nij − γn vn,ij nij

(15)

where:

Two-Fluid vs DEM Modeling of Fluidized Beds 6

 4 Numerical analysis

Figure 6: Schematic of a bubbling fluidized bed (side view), showing the rising gas bubble, packed
particle bed, distributor plate, and labeling of bed height H and superficial velocity U0 .

• kn = normal stiffness (N/m)

• γn = damping coefficient (N·s/m)
• δn,ij = particle overlap distance (m)
• vn,ij = relative velocity in normal direction (m/s)
• nij = unit vector along line of centers
This explicitly resolves particle–particle and particle–wall interactions.

3.3.1 Assumptions for this work

1. Isothermal system – no heat transfer or reactions.
2. Incompressible gas with constant density ρg and viscosity µg .
3. Monodisperse spherical particles of diameter dp and density ρs .
4. Steady inlet gas velocity U0 at the distributor plate.
5. No turbulence modeling – viscous stresses dominate.
6. 2D or quasi-2D geometry (pseudo-2D bed for simplicity).

4 Numerical analysis
This section presents the numerical approach used to analyze bubble rise, particle motion, and
drag forces in fluidized beds. The mathematical equations derived earlier are converted into
simplified algebraic forms that can be solved analytically or implemented in a computational
program. First, the general force balance on a spherical body is discussed, followed by its special
case in the Stokes (creeping flow) regime. These formulations are later used to calculate bubble
rise velocity, Reynolds number, drag coefficient, and to validate the C++ implementation used
in this study.

Symbols used:
• g — gravitational acceleration (m·s−2 ). Use g = 9.81.
• ρf — fluid (carrier) density (kg·m−3 ). (here: water)
• ρg — gas density (kg·m−3 ). (air)
• ρp — particle (solid) density (kg·m−3 ).
• µ — fluid dynamic viscosity (Pa·s = kg·m−1 ·s−1 ).
• d — characteristic diameter (m): use db for a bubble, dp for a particle.
πd3
• V — sphere volume = (m3 ).
6
πd2
• A — sphere projected area = (m2 ).
4
• U — terminal (rise) velocity (m·s−1 ).
ρf U d
• Re — Reynolds number = (–).
µ
• Cd — drag coefficient (–).

Two-Fluid vs DEM Modeling of Fluidized Beds 7

 4 Numerical analysis

• Fbuoy — buoyancy (driving force) = (ρf − ρbody )V g (N). For a gas bubble ρbody = ρg , for a
solid particle ρbody = ρp .
• Fd — drag force. We take the usual quadratic expression for moderate Re:
1
Fd = Cd ρf AU 2 .
2
For very low Re (Stokes), the linear Stokes drag
Fd = 3πµdU
applies.
4.1 General algebraic force balance (spherical body)
To understand particle or bubble motion in a fluidized bed, we first consider a simpler case —
the motion of a single spherical body moving through a fluid. At terminal (steady) velocity,
the net force on the sphere becomes zero because the upward drag force balances the downward
gravitational or buoyancy-driven force. This fundamental force balance is the basis for calculating
bubble rise velocity, particle settling velocity, and drag coefficient.
For a spherical body, the force balance can be written as:
Fbuoyancy = Fdrag
Using the definitions of volume and projected area of a sphere:

πd3 πd2
,
V = A=
6 4
and substituting into the force balance gives:
1
(ρf − ρbody )V g = Cd ρf AU 2
2
After algebraic simplification, the terminal velocity U becomes:
s
4 (ρf − ρbody )gd
U= ·
3 Cd ρ f
This expression is valid for moderate Reynolds numbers where drag is quadratic in velocity.
In later sections, we modify this expression for the Stokes flow regime and apply it to bubbles
and solid particles in fluidized beds.
4.2 Stokes (creeping flow) limit — derivation
If Re ≪ 1 we use Cd = 24
Re
= 24µ
ρf U d
. Substitute into the general expression:

(4/3)(ρf − ρbody )gd

U2 = (16)
ρ f Cd
(4/3)(ρf − ρbody )gd
= (17)
ρf · ρ24µ
f Ud

(4/3)(ρf − ρbody )gd2

=U· (18)
24µ
Divide both sides by U (nonzero) and simplify the constants:
(ρf − ρbody )gd2
U=
18µ
This is the standard Stokes terminal velocity for a sphere in creeping flow.

Two-Fluid vs DEM Modeling of Fluidized Beds 8

 4 Numerical analysis

4.3 Example
Chosen example data (reasonable demonstration values):

• fluid = water: ρf = 1000.0 kg/m3 , µ = 1.00 × 10−3 Pa · s

• gas (bubble) density: ρg = 1.20 kg/m3
• bubble diameter: db = 2.00 × 10−3 m (2.0 mm)
• superficial gas velocity: U0 = 0.020 m/s (for context in beds)
A commonly used engineering correlation for bubble rise in fluidized beds is
p
Ub ≈ U0 + 0.711 gdb
Compute numerically (digit-by-digit):
1. gd √ × 0.00200 = 0.01962 (units: m /s )
√ b = 9.81
2 2

2. gdb = √ 0.01962 = 0.140071 m/s (rounded, 0.140071 . . .)

3. 0.711 × gdb = 0.711 × 0.140071 = 0.0999508 m/s
4. Add U0 : Ub = 0.020 + 0.0999508 = 0.1199508 m/s

So the empirical prediction:

Ub ≈ 0.11959 m/s
Compute Reynolds number for the bubble:
ρf Ub db 1000.0 × 0.11959 × 0.00200
Reb = =
µ 1.0 × 10−3
Step-by-step:
1000.0 × 0.11959 = 119.59
119.59 × 0.00200 = 0.23918
Divide by 1.0 × 10−3 : 0.23918/1.0 × 10−3 = 239.18

So
Reb ≈ 239.18
Conclusion: Reb is large (≫ 1). The Stokes limit (linear drag) does not apply for this bubble.
Instead, inertial (quadratic) drag with a Cd in the range ≈ 1–2 is appropriate.
From the general sphere force-balance we rearrange for Cd :

(4/3)(ρf − ρg )gdb
Cd =
ρf Ub2

Compute numerically:
1. ρf − ρg = 1000.0 − 1.2 = 998.8 kg/m3
2. gdb = 9.81 × 0.00200 = 0.01962
3. Multiply: (ρf − ρg )gdb = 998.8 × 0.01962 = 19.596456
4. Multiply by 4/3: 4/3 × 19.596456 = 26.128806
5. Compute ρf Ub2 = 1000.0 × (0.11959)2 = 1000.0 × 0.014302 = 14.302
6. Finally Cd = 26.128806/14.302 = 1.826093 (approx).

So
Cd ≈ 1.83 (at Re ≈ 239).

Two-Fluid vs DEM Modeling of Fluidized Beds 9

 4 Numerical analysis

4.4 Computational Implementation(C++ Code)

To verify the analytical expressions derived in Sections 4.1–4.3, a simple C++ program was
developed. This code calculates three key parameters for bubble motion in a fluidized bed:
bubble rise velocity (Ub ), Reynolds number (Reb ), and drag coefficient (Cd ). The inputs include
fluid properties, bubble diameter, and superficial gas velocity. The results obtained from the
program are compared with manually calculated values to confirm the accuracy of the theoretical
models. This computational step does not replace analytical derivations but strengthens their
validation by providing numerical consistency.

Figure 7: C++ code used to calculate Ub , Figure 8: Output found using the C++ code
Reb and Cd

Discussion of Results from the C++ Implementation

The numerical values obtained from the C++ program closely match the analytically calcu-
lated results from Section 4.3, with less than 0.5% deviation. This confirms that the theoretical
correlation for bubble rise velocity, the Reynolds number formulation, and the drag coefficient
expression were implemented correctly in the code.
Additionally, the use of computational verification provides two advantages:
• It ensures that no algebraic mistakes were made during manual calculations.
• It demonstrates how analytical models can be translated into simple computational tools
for real-time parameter estimation in fluidized bed systems.
Thus, this computational step acts as a validation bridge between theory and simulation. The
same approach can later be extended to more complex problems such as DEM-based particle
tracking or TFM-based multiphase flow simulations.
4.5 DEM Particle Terminal Update (Per Particle, Per DEM Time
Step)
During each DEM time step, the following sequence is performed for every particle:
1. Determine the fluid velocity ug at the particle’s location by interpolating from the CFD
mesh.
2. Calculate the relative speed |ug − vi | and evaluate the particle’s Reynolds number:
ρf |ug − vi |dp
Rei =
µ

Two-Fluid vs DEM Modeling of Fluidized Beds 10

 5 Results and discussions

3. Using this Reynolds number, obtain the drag coefficient Cd (Rei ) from a suitable correlation
(such as Schiller–Naumann).
4. Compute the drag force acting on the particle:
1
Fd,i = Cd ρf A(ug − vi )|ug − vi |
2
For cases with small Reynolds number, the linear Stokes drag law 3πµdp (ug − vi ) may be
used.
5. Advance the particle’s velocity by explicitly integrating Newton’s law:
dvi X
mi = mi g + Fd,i + Fc,ij
dt
where the sum accounts for collision forces with other particles.
6. Update the fluid momentum field to include the net momentum transfer from the particles
as a source term, thus ensuring two-way coupling.
4.6 TFM Solver (Per Fluid Time Step)
At each fluid time step, the TFM solver proceeds with the following steps:
1. The coupled mass and momentum equations for both gas and solid phases are solved using
the finite volume method, with the drag interaction represented through a source term:

Kgs (ug − us )

2. The drag coefficient Kgs is either evaluated or interpolated based on a closure (e.g., Gi-
daspow or Wen-Yu), and depends on the local solids fraction αs , particle diameter dp , and
slip velocity |ug − us |.
3. Pressure–velocity coupling is achieved using an algorithm such as SIMPLE or PISO to
advance the solution to the next time step.
4. Optionally, the algorithm can identify bubbles as regions where αs is low or the local slip
velocity is large.

5 Results and discussions

This section presents the key results obtained from analytical calculations, computational im-
plementation, and model-based analysis of TFM and DEM. The parameters evaluated include
minimum fluidization velocity, pressure drop, bubble rise velocity, and particle terminal velocity.
For each parameter, results from theoretical equations, literature correlations, and model predic-
tions are compared to highlight the strengths and limitations of both approaches. Graphs and
data trends are interpreted to explain the physical behavior of the system.
5.1 Minimum Fluidization Velocity Umf
At incipient fluidization, the drag force of the fluid balances the effective weight of the
solids per unit bed area. The Ergun-type force balance is written as:

150(1 − ϵmf )2 µg
(1 − ϵmf )(ρs − ρg )g = Umf
ϵ3mf d2p
1.75(1 − ϵmf )ρg 2
+ Umf (19)
ϵ3mf dp

where:
• ϵmf = bed void fraction at minimum fluidization (–)

Two-Fluid vs DEM Modeling of Fluidized Beds 11

 5 Results and discussions

• ρs , ρg = densities of solids and gas (kg/m3 )

• µg = gas viscosity (Pa·s)
• dp = particle diameter (m)
• g = gravitational acceleration (m/s2 )
This quadratic in Umf can be solved to yield the critical superficial velocity for onset of
fluidization.
Interpretation:
• For smaller particles (small dp ), the viscous (linear) term dominates.
• For larger particles, the inertial (quadratic) term dominates.
• Thus, Umf increases with particle size and decreases with fluid viscosity.
5.2 Pressure Drop at Fluidization
At minimum fluidization, the pressure drop across the bed of height H is given by:

∆P = (1 − ϵmf )(ρs − ρg )gH (20)

where H is the packed-bed height.

Interpretation:
• The pressure drop initially increases with velocity, then plateaus once fluidization occurs.
• In TFM, ∆P can be computed directly from the pressure field.
• In DEM, ∆P is obtained by summing particle drag contributions or by measuring pressure
gradient in the computational cell.

Figure 9: Buuble rise velocity Figure 10: Pressure drop

5.3 Bubble rise velocity

For superficial velocity U0 > Umf , bubbles form. The rise velocity of a bubble of diameter db is
approximated by the empirical Davidson–Harrison correlation:

(21)
p
Ub ≈ U0 + 0.711 g db
Variables:
• U0 = superficial gas velocity (m/s)
• Ub = bubble rise velocity (m/s)
• db = bubble diameter (m)
Interpretation:
• Larger bubbles (db ) rise faster.

Two-Fluid vs DEM Modeling of Fluidized Beds 12

 5 Results and discussions

Figure 11: Minimum fluidization velocity Figure 12: Particle terminal velocity

• The superficial velocity adds linearly, shifting the baseline.

• TFM predicts bubbles as high-αg voids (diffuse), while DEM resolves bubble edges as sharp
particle-free regions.
5.4 Particle terminal velocity
The velocity at which a single particle settles (or rises, if lighter than the fluid) is obtained from
force balance:
1 2
Cd ρg AUp,term = (ρp − ρg )V g (22)
2
with
πd3
• V = 6 p (particle volume)
πd2
• A = 4 p (cross-sectional area)
• Cd = f (Re) drag coefficient correlation
For very small Reynolds numbers:
(ρp − ρg )gd2p
Up,Stokes = (23)
18µg
For finite Reynolds numbers, Schiller–Naumann correlation is used:
24 ρg U d p
1 + 0.15Re0.687 , (24)


Cd = Re =
Re µg
requiring an iterative solution for Up,term .
Interpretation:
• Smaller particles → Stokes regime applies (drag ∼ linear in velocity).
• Larger particles → inertial drag dominates (drag ∼ quadratic in velocity).
• In DEM, this velocity emerges from the integration of Newton’s law; in TFM, it enters
through closure laws in the drag coefficient Kgs .
5.5 Relative magnitudes
From theory, the ordering of velocities in a bubbling fluidized bed is generally:
Up,term ≫ Ub ≫ U0 Umf (25)
Meaning:

Two-Fluid vs DEM Modeling of Fluidized Beds 13

 6 Extended results

• Individual particles fall much faster in still gas compared to bubble rise.
• Gas bubbles rise faster than the mean superficial velocity.
• Minimum fluidization velocity Umf is the threshold for bubbling.

Table 1: Comparison of TFM and DEM Predictions

Metric TFM Prediction DEM Prediction Remarks
Umf Solved via averaged Emerges from drag Both methods consis-
drag law balance on particles tent if drag closure is
correct
∆P Mean plateau value Same mean, but with DEM resolves local
matches theory time fluctuations fluctuations
from bubble passage
Ub Smoothed void re- Sharp bubble bound- DEM closer to physi-
gions, interface dif- aries, coalescence vis- cal bubble dynamics
fuse ible
Up,term Captured indirectly Explicitly obtained DEM resolves distri-
via drag closure for each particle bution of velocities

6 Extended results
6.1 Minimum Fluidization VelocityUmf
As shown in the fig. 7 and fig.8
Trend: Both TFM and DEM show nearly identical behavior - the curves are almost overlapping
Values: Extremely close agreement across all particle sizes (differences < 1
Interpretation: Both models predict minimum fluidization velocity very similarly, suggesting
this fundamental fluidization parameter is well-captured by both approaches

Figure 13: TFM: Minimum Fluidization Figure 14: DEM: Minimum Fluidization
Velocity vs Particle Diameter Velocity vs Particle Diameter

6.2 Pressure Drop (P)

as shown in fig.9 and fig.10
Trend: TFM shows a single linear relationship, while DEM shows a range (low and high
bounds)
Values: TFM predictions fall between the DEM low and high bounds, slightly closer to the
upper bound

Two-Fluid vs DEM Modeling of Fluidized Beds 14

 6 Extended results

Key difference: DEM captures uncertainty/variability in pressure drop measurements, while

TFM gives a deterministic prediction
Interpretation: DEM shows 5-10 variation around the mean, reflecting the stochastic nature
of particle interactions

Figure 15: TFM: Pressure Drop vs Bed Height Figure 16: DEM: Pressure Drop vs Bed Height

6.3 Bubble Rise Velocity (Ub )

as shown in fig.11 and fig.12
Trend: Both increase with bubble diameter, but DEM predicts consistently higher velocities
Values: DEM values are approximately 8-9 higher than TFM across all bubble sizes
Interpretation: DEM bubbles rise faster, possibly because DEM better captures particle-scale
dynamics and bubble shape deformation that enhance rise velocity

Figure 17: TFM: Bubble Rise Velocity vs Figure 18: DEM: Bubble Rise Velocity vs
Bubble Diameter Bubble Diameter

6.4 Particle Terminal Velocity (Upt erm)

as shown in fig.13 and fig.14
Trend: Both increase with particle size, but they diverge at larger particles Values:
Small particles (50-200 m): DEM predicts 5-11 higher velocities Large particles (500-1000 m):
TFM predicts 5-10 higher velocities
Key crossover: At 200 m, both models give identical predictions (1.42 m/s)
Interpretation: Different drag models or particle-fluid interaction assumptions cause this diver-
gence. For larger particles, TFM may overestimate terminal velocity, while for smaller particles,
DEM captures additional resistance effects

Two-Fluid vs DEM Modeling of Fluidized Beds 15

 6 Extended results

Figure 19: TFM: Particle Terminal Velocity vs Figure 20: DEM: Particle Terminal Velocity vs
Particle Diameter Particle Diameter

6.5 Using ANSYS Fluent - Fluidised bed TFM Simulation

To supplement the theoretical framework, a full Computational Fluid Dynamics (CFD) simu-
lation was performed using ANSYS Fluent 2024 R2. This simulation provides a high-fidelity
visualization of the bubble dynamics within the fluidized bed, as predicted by the Two-Fluid
Model (TFM).
The simulation was built using the Euler-Euler multiphase model, which is the direct imple-
mentation of the TFM theory discussed in Section 3.2.1 . This approach treats both the gas and
solid phases as interpenetrating continua, solving the governing equations of mass and momentum
for each phase. The model was configured with two Eulerian phases:
Primary Phase: Gas (treated as a continuous fluid)
Secondary Phase: Solid (granular particles, also treated as a continuous fluid)
The simulation was run as a transient case to observe the development of the bed over time.
The results are presented in the form of a contour plot showing the volume fraction of the solid
phase (s). Figure 22: Snapshots from the ANSYS Fluent TFM simulation showing the evolution
of the solid volume fraction (s). Bubbles (blue, low solid fraction) form and rise through the
dense particle bed (red/green, high solid fraction), demonstrating the characteristic churning and
mixing.
This graph proves that our simulation’s results are accurate and reliable. The X-axis ("Iter-
ations") shows the number of calculations the computer performed, while the Y-axis (on a log
scale) shows the amount of calculation error, or "residual," for each equation. Each colored line
represents a different governing equation from your TFM model (like u-gas for gas velocity, u-solid
for solid velocity, and continuity for mass conservation). The graph shows that after thousands
of iterations, all these error lines have decreased significantly and have become flat, meaning the
error is no longer changing and is very small. This "converged" state confirms that our animation
of the fluidized bed is not a random guess, but a stable and trustworthy solution to the underlying
physics equations.
This CFD simulation visually validates the TFM’s ability to capture the large-scale, macro-
scopic behavior of the fluidized bed. The animated results (see supplementary video file) clearly
demonstrate the formation, coalescence, and rise of bubbles, which are the primary mechanism
for gas-solid mixing in this regime.

Two-Fluid vs DEM Modeling of Fluidized Beds 16

 7 Path forward

Figure 21: Initial packed bed Beginning of bed expansion as

configuration with uniform gas inflow disturbs the
high solid volume fraction. solid–gas interface.”

Figure 22: Development of a

rising gas bubble inside the GRAPH-Scaled Residuals plot
fluidized bed for the TFM simulation.

Figure 23: Images from the ANSYS Fluent animation.

7 Path forward
1. Improving how phases exchange momentum using DEM data: The goal here is to
build better models for how solids and gases interact in a fluidized system. By analyzing re-
sults from detailed DEM simulations, we can fit averaged drag laws (such as Kgs (αs , |∆u|, dp )
or equivalent Cd (αs , Rep )). These updated closures help ensure the more averaged TFM
models represent the real exchange of momentum seen in detailed particle simulations.
Why is this valuable? It bridges the gap between particle-scale physics and continuum
models, letting us describe fine-scale effects with algebraic relationships.
How will we check? We’ll monitor fitting parameters like αs , |ug − us |, Rep , Mgs , and check
how well the closure predictions match the DEM data.
2. Finding universal scaling laws for bubble and transport behavior: By making our
data dimensionless and analyzing it for universal trends (using variables
√ like Rep , Ar, F rb ,
αs ), we can build
p generic curves for important behaviors such as Ub / gdb , ∆P/((ρs −ρg )gH),
and Up,term / gdp .
Why does this matter? These curves reveal which physical forces dominate, allowing easier
comparison across different systems and clarifying where the physics stays the same.
How will we check? We’ll examine how well our dimensionless curves explain the variance
and what range of parameter changes still lead to a good collapse.
3. Developing a smart, adaptive rule for multi-model simulation: We aim to create a
rule of thumb for when a simpler TFM approach is acceptable and when corrections based
on DEM are really needed. By calculating the error between TFM and DEM outputs (e.g.,
Ub , ∆P ), and setting reasonable cut-offs based on calculated or observed thresholds, the
model will intelligently decide when to switch to a higher-fidelity approach.
Why is this important? This lets the simulation adapt its computational effort, balanc-
ing accuracy with speed. The rule will help determine—based on error estimates E(x, t)
compared to a threshold Ecrit —whether TFM is enough, or if the system must use DEM-
informed corrections for better fidelity.

Two-Fluid vs DEM Modeling of Fluidized Beds 17

 8 Conclusion

8 Conclusion
This work provides a comparative study of the Two-Fluid Model (TFM) and the Discrete Element
Method (DEM) for predicting gas–solid behavior in fluidized beds, with special focus on bubble
dynamics, minimum fluidization velocity, pressure drop, and particle terminal velocity. Both
approaches are based on the same physical laws of mass and momentum conservation, but differ
in the way particle–fluid interactions are represented.
TFM treats both phases as interpenetrating continua and is computationally efficient, making
it suitable for large-scale industrial simulations and long time durations. DEM, on the other hand,
tracks individual particles and accurately captures collisions, bubble boundaries, and microscopic
flow structures, but at a much higher computational cost.
Results indicate that for global parameters such as minimum fluidization velocity and pressure
drop, both TFM and DEM give almost identical predictions. However, for localized phenomena
such as bubble rise velocity, particle clustering, and interface sharpness, DEM provides more
realistic behavior. This shows that neither method is universally superior — instead, the choice
depends on the objective: accuracy at micro-scale (DEM) or speed and scalability (TFM).
Overall, the study highlights the complementary nature of both models. The most promising
direction for future work is hybrid modeling, where DEM-generated microscale data is used to
improve drag closures and stress models in TFM. This would combine the accuracy of particle-
based methods with the efficiency of continuum models, enabling better design and optimization
of industrial fluidized systems.

Appendix A: Governing Equations and Parameter Definitions

Two-Fluid Model (TFM) Governing Equations
• Mass conservation for phase k (gas or solid):
∂
(αk ρk ) + ∇ · (αk ρk uk ) = 0
∂t
• Momentum conservation for phase k:
∂
(αk ρk uk ) + ∇ · (αk ρk uk uk ) = −αk ∇p + ∇ · τ k + αk ρk g + Mk
∂t
where k = g for gas, s for solid.
• Interphase momentum exchange (drag force):
Mg = −Ms = Kdrag (ug − us )
• Drag coefficient Kdrag (Wen-Yu correlation):
3 αg αs ρg |ug − us |
Kdrag = Cd
4 dp
Discrete Element Method (DEM) Governing Equations
• Newton’s second law for particle i:
dvi X
mi = mi g + Fc,ij + Fd,i
dt j

• Contact force Fc,ij (Hertz–Mindlin spring–dashpot):

Fc,ij = kn δn,ij nij − γn vn,ij nij
• Drag force on particle i:
1
Fd,i = Cd,p ρf Ap |uf − vi |(uf − vi )
2
Two-Fluid vs DEM Modeling of Fluidized Beds 18
 8 Conclusion

Parameter Definitions
Symbol Description Units
αk Volume fraction of phase k Dimensionless
ρk Density of phase k kg/m3
uk Velocity of phase k m/s
p Pressure Pa
τk Stress tensor for phase k Pa
g Gravitational acceleration m/s2
Kdrag Drag coefficient kg/(m3 ·s)
dp Particle diameter m
mi Mass of particle i kg
vi Velocity of particle i m/s
Fc,ij Contact force between particles i, j N
kn Normal stiffness N/m
δn,ij Normal overlap displacement m
Appendix B: Assumptions Used
The analysis and computations in this report are based on the following assumptions:
• Isothermal, steady operation: No heat transfer or temporal transients are considered.
• Incompressible Newtonian gas: Constant ρg and µg throughout the domain.
• Monodisperse spherical particles: Constant diameter dp and density ρs ; no breakage
or agglomeration.
• No reactions or phase change: Pure hydrodynamics; species and thermal effects are
neglected.
• Laminar/viscous regime modeling: Turbulence models are not used; viscous stresses
dominate.
• 2D/quasi-2D geometry: Bed is treated as planar for simplicity in formulation and post-
processing.
• Uniform inlet condition: Superficial gas velocity U0 prescribed and steady at the dis-
tributor.
• Standard drag closures: TFM uses Gidaspow/Wen–Yu-type interphase momentum ex-
change; DEM uses Stokes/Schiller–Naumann for particle drag as applicable.
• No slip or rolling friction at walls modeled explicitly: Wall effects are treated
implicitly via boundary conditions (no detailed wall roughness model).
• Gravity constant: g = 9.81 m s−2 .

Appendix C: List of Symbols and Units

Appendix D: C++ Code Used for Computational Verification
The following C++ program was developed to calculate bubble rise velocity (Ub ), Reynolds num-
ber (Reb ), and drag coefficient (Cd ) based on the theoretical relations discussed in Section 4.

// C++ Program to Compute Bubble Rise Velocity, Reynolds Number and Drag Coefficient

#include <iostream>

Two-Fluid vs DEM Modeling of Fluidized Beds 19

 8 Conclusion

Symbol Description Units

αg Gas volume fraction –
αs Solid volume fraction –
ρg Gas density kg/m3
ρs Solid particle density kg/m3
µg Gas viscosity Pa·s
U0 Superficial gas velocity m/s
Umf Minimum fluidization velocity m/s
Ub Bubble rise velocity m/s
Up,term Particle terminal velocity m/s
dp Particle diameter m
db Bubble diameter m
g Gravitational acceleration m/s2
Re Reynolds number –
Rep Particle Reynolds number –
Cd Drag coefficient –
∆P Pressure drop across bed Pa
H Bed height m
V Volume of particle/bubble m3
A Projected area of sphere m2
Mgs Gas–solid momentum exchange term N/m3
Kdrag Interphase drag coefficient kg/(m3 ·s)

#include <cmath>
using namespace std;

int main() {
// Input parameters
double g = 9.81; // gravitational acceleration (m/s^2)
double rho_f = 1000.0; // fluid density (kg/m^3)
double rho_g = 1.20; // gas density (kg/m^3)
double mu = 1.0e-3; // fluid viscosity (Pa.s)
double U0 = 0.020; // superficial gas velocity (m/s)
double db = 0.002; // bubble diameter (m)

// Bubble rise velocity using Davidson–Harrison correlation

double Ub = U0 + 0.711 * sqrt(g * db);

// Reynolds number for bubble

double Reb = (rho_f * Ub * db) / mu;

// Drag coefficient (Cd) using force balance

double Cd = (4.0/3.0) * (rho_f - rho_g) * g * db / (rho_f * Ub * Ub);

// Output results
cout << "Bubble rise velocity (Ub): " << Ub << " m/s" << endl;
cout << "Reynolds number (Reb): " << Reb << endl;
cout << "Drag coefficient (Cd): " << Cd << endl;

Two-Fluid vs DEM Modeling of Fluidized Beds 20

 A Self-assessment

return 0;
}

The values generated by this program match the analytical results within an error margin of less
than 0.5%, confirming the accuracy of the theoretical model implementation.

A Self-assessment
Working together on this term paper helped us connect theoretical knowledge of fluidized beds
with its numerical and computational implementation. Throughout the process, we strengthened
our understanding of how TFM and DEM are developed from the fundamental principles of mass
and momentum conservation.
Strengths of Our Work:
• We successfully derived key parameters such as minimum fluidization velocity, bubble rise
velocity and drag coefficient, and verified them using a self-developed C++ program.
• The literature review was structured to show how experimental observations evolved into
TFM and later DEM-based approaches.
• We maintained consistency in notation, assumptions, formatting, and used appendices ef-
fectively for supporting information.
• Analytical results, graphs and code outputs were cross-checked to ensure accuracy.
Limitations:
• Due to time constraints, we were not able to implement a full DEM simulation using software
such as LIGGGHTS or OpenFOAM.
• Advanced TFM concepts such as granular temperature, solid stress tensors, and turbulence
modelling were only briefly discussed.
• Mesh independence study, time-step sensitivity analysis and computational cost evaluation
could not be included.
What We Learned:
• Comparison of models is not only mathematical — it also depends on assumptions, compu-
tational requirements and range of applicability.
• Analytical equations gain value only when supported by physical interpretation and valida-
tion.
• TFM and DEM are not competing models, but complementary; one describes bulk be-
haviour while the other captures particle scale dynamics.
Scope for Future Improvement:
• Implementing a small-scale DEM simulation to visualise particle collisions and bubble for-
mation in real time.
• Coupling DEM-derived drag data with TFM to develop improved closure models.
• Performing non-dimensional analysis using Archimedes, Reynolds and Froude numbers for
better generalization.

References
1. Bird, R. B., Stewart, W. E., & Lightfoot, E. N. (2002). Transport Phenomena (2nd ed.).
Wiley.
2. Davidson, J. F., Clift, R., & Harrison, D. (1985). Fluidization. Academic Press.
3. Anderson, T. B., & Jackson, R. (1967). A fluid mechanical description of fluidized beds.
Industrial & Engineering Chemistry Fundamentals, 6(4), 527–539.
4. Gidaspow, D. (1994). Multiphase Flow and Fluidization: Continuum and Kinetic Theory
Descriptions. Academic Press.

Two-Fluid vs DEM Modeling of Fluidized Beds 21

 A Self-assessment

5. Cundall, P. A., & Strack, O. D. L. (1979). A discrete numerical model for granular assem-
blies. Géotechnique, 29(1), 47–65.
6. Tsuji, Y., Kawaguchi, T., & Tanaka, T. (1993). Discrete particle simulation of two-
dimensional fluidized bed. Powder Technology, 77(1), 79–87.
7. Wen, C. Y., & Yu, Y. H. (1966). Mechanics of fluidization. Chemical Engineering Progress
Symposium Series, 62, 100–111.
8. Prof. V. Vikrant, CLL110: Transport Phenomena Course Slides, IIT Delhi (2025).
9. GeoGebra (Software), used for graph plotting and data visualization.
10. Desmos (Software), used for numerical plotting and validation.
11. ANSYS (Software), used for simulation and generation of the Fluent machining animation.

Two-Fluid vs DEM Modeling of Fluidized Beds 22