Coupling an Actuator Line Model with

Partial Pressure Fields Towards Future

Thrust-Drag Accounting in the Navier-Stokes
Equations
Logan J. Szajnecki 1
The Pennsylvania State University, State College, PA, 16801, United States

A variation of the actuator line model (“Numerical Modeling of Wind Turbine Wakes”,
ASME. J. Fluids Eng. June 2002; 124(2): 393-399) is presented and synthesized with a partial
pressure field approach ("Drag Decomposition Using Partial-Pressure Fields in the
Compressible Navier-Stokes Equations", AIAA Journal, Vol.57, No.5, 2019, pp. 2030-2038). The
purpose of the present study is to develop a two-dimensional simulation wherein the effect of
an actuator line model (ALM) on a lifting surface can be quantified using partial pressure
fields (PPFs). The resulting flowfield variations are discussed; the implementation of both the
ALM and the PPFs acting on the RAE-2822 airfoil is visualized. The ALM spreads a body
force along a user-defined vertical line to simulate a propeller. The ALM’s length and thrust
magnitude is adequately representative of a Cessna 172 because of its sufficiently low cruise
speed and its well published performance metrics. A PPF to quantify the ALM’s influence on
the incompressible Navier-Stokes equation is analyzed. All PPFs presented in this study give
insight into the sources of the airfoil’s net aerodynamic forces. This two-dimensional study on
full aircraft coefficients serves as a base upon which further research can be performed in
three dimensions. The incompressible, subsonic (M=0.18) flow domain in question is resolved
using a steady Reynolds Averaged Navier-Stokes (RANS) solver and a one-equation Spalart-
Allmaras eddy-viscosity turbulence model. Actuator line and partial pressure field histories
are discussed and a brief on grid selection and verification methods is performed. Results show
that a propulsive body force represented by an ALM three meters below the RAE-2822 airfoil
reduces its lift coefficient by 48.6%. The total drag increases by 6.44 drag counts in this
scenario. Repositioning the ALM to 1.5 meters below the airfoil gives way to a 77.6%
reduction in lift coefficient. In this scenario, the total drag decreases by 1.5 counts.

Nomenclature
A = area
c = chord [m]
𝐶𝐶𝑑𝑑,𝑝𝑝𝜇𝜇 = dissipative pressure field drag coefficient
𝐶𝐶𝑑𝑑,𝑝𝑝𝐸𝐸 = inviscid Euler pressure field drag coefficient
𝐶𝐶𝑑𝑑,𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 = propeller pressure field drag coefficient
𝐶𝐶𝑑𝑑,𝑝𝑝 = static pressure field drag coefficient, 𝐶𝐶𝑑𝑑,𝑝𝑝𝜇𝜇 + 𝐶𝐶𝑑𝑑,𝑝𝑝𝐸𝐸 + 𝐶𝐶𝑑𝑑,𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝
𝐶𝐶𝑑𝑑,𝜏𝜏 = viscous drag coefficient
𝐶𝐶𝑑𝑑 = total drag coefficient, 𝐶𝐶𝑑𝑑,𝑝𝑝 + 𝐶𝐶𝑑𝑑,𝜏𝜏
𝑑𝑑𝑒𝑒𝑒𝑒 = equivalent diameter
𝐹𝐹⃗ = body force source term, 𝜂𝜂 ∙ 𝑓𝑓 [N/m^3]
𝑓𝑓 = forcing function to be normally distributed, constant [N]

1
Undergraduate Student Researcher, Penn State Department of Aerospace Engineering

1
𝑔𝑔⃗ = gravitational acceleration vector
M = Mach number
𝑚𝑚̇ = mass flow rate, [kg/s]
𝑃𝑃𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 = operating pressure, 𝑃𝑃 = 𝜌𝜌𝜌𝜌𝜌𝜌 [Pa]
𝑝𝑝𝜇𝜇 = viscous dissipative pressure, [Pa]
𝑝𝑝𝐸𝐸 = inviscid Euler pressure, [Pa]
𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 = propeller pressure, [Pa]
p = static pressure, [Pa]
𝑅𝑅𝑅𝑅 = Reynolds number, 𝑅𝑅𝑅𝑅 = 𝜌𝜌𝜌𝜌𝜌𝜌/𝜇𝜇
𝑅𝑅 = ideal gas constant, 287 [J/𝑘𝑘𝑘𝑘 ∙ 𝐾𝐾]
𝑇𝑇
𝑟𝑟𝑝𝑝
���⃗ = grid cell coordinates, ���⃗
𝑟𝑟𝑝𝑝 = �𝑥𝑥𝑝𝑝 , 𝑦𝑦𝑝𝑝 � , [m]
���⃗
𝑟𝑟𝑁𝑁 = Gaussian coordinates, ���⃗ 𝑟𝑟𝑁𝑁 = (𝑥𝑥𝑁𝑁 , 𝑦𝑦𝑁𝑁 )𝑇𝑇 , [m]
𝑟𝑟⃗ = net distance vector, |𝑟𝑟⃗| = �����⃗ 𝑟𝑟𝑁𝑁 �, [m]
𝑟𝑟𝑝𝑝 − ���⃗
𝑟𝑟𝑒𝑒𝑒𝑒 = equivalent radius, [m]
𝑆𝑆𝜑𝜑 = transport equation source term
t = time, [s]
�⃗
𝑇𝑇 = thrust vector, [N]
𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 = reference temperature = 300 [K]
𝑉𝑉𝑐𝑐 = forward (climb) velocity, [m/s]
𝑣𝑣𝑖𝑖 = induced velocity, [m/s]
𝑣𝑣⃗ = vector velocity field, (𝑢𝑢, 𝑣𝑣)𝑇𝑇 [m/s]
w = rotor disk downwash parameter
𝑦𝑦0 = Gaussian initialization parameter, [m]
𝛼𝛼 = angle of attack, [deg]
𝛾𝛾 = ratio of specific heats, = 𝐶𝐶𝑝𝑝 /𝐶𝐶𝑣𝑣
Δ = spreading length scale, [m]
𝜀𝜀 = planar Gaussian projection radius, [m]
𝜂𝜂𝑁𝑁 = Gaussian normalization function
𝜇𝜇 = fluid dynamic viscosity, [kg/𝑚𝑚 ∙ 𝑠𝑠]
Ξ = diffusion coefficient
ρ = density [kg/m^3]
𝜏𝜏 = viscous stress tensor, [Pa]
𝜑𝜑 = transport equation scalar

2
 Table of Contents

I. INTRODUCTION ...........................................................................................................................................4

II. STATIC PRESSURE DECOMPOSITION OF THE NAVIER-STOKES EQUATION WITH A BODY

FORCE .........................................................................................................................................................................4

III. ACTUATOR LINE DEVELOPMENT AND IMPLEMENTATION .........................................................5

A. EFFECTIVE PROPELLER THRUST AND AREA FORMULATION ...............................................................................5

B. JET VELOCITY/THRUST FORMULATION ..............................................................................................................6
C. FORMULATION OF GAUSSIAN DISTRIBUTION ......................................................................................................8

IV. COMPUTATIONAL/NUMERICAL METHODS........................................................................................8

A. POISSON EQUATIONS IN FLUENT ........................................................................................................................9

B. AIRFOIL/GRID SELECTION AND MODIFICATIONS ................................................................................................9
C. VERIFICATION METHODS ................................................................................................................................. 10

V. RESULTS ....................................................................................................................................................... 12

A. BASE-CASE RESULTS ........................................................................................................................................ 13

B. ACTUATOR LINE MODEL 3M BELOW RAE-2822 AIRFOIL ................................................................................ 14
C. ACTUATOR LINE MODEL 1.5M BELOW RAE-2822 AIRFOIL ............................................................................. 16
D. COMPREHENSIVE DISCUSSION .......................................................................................................................... 18

VI. CONCLUSION .............................................................................................................................................. 19

VII. APPENDICES ................................................................................................................................................ 20

A. FLOW FIELD DENSITY AND OPERATION PRESSURE CALCULATION................................................................... 20

B. C++ SOURCE CODE .................................................................................... ERROR! BOOKMARK NOT DEFINED.
C. MATLAB R2020B POST PROCESSING SCRIPT ................................................................................................... 20

ACKNOWLEDGMENTS ......................................................................................................................................... 22

REFERENCES .......................................................................................................................................................... 22

3
 I. Introduction

T he use of computational tools to describe aerodynamic phenomena is of utmost interest for the field of
aerospace engineering. Specifically, it is crucially important to characterize aerodynamic drag when
considering various aircraft designs. The longevity of the modern aircraft is heavily dependent on its ability to
minimize drag and fuel consumption. Furthermore, to appease the continuous quarrel between aerodynamicists and
propulsion engineers, comprehensive aerodynamic drag summaries must be developed. To better understand the
aerodynamic forces acting on an aircraft, aerospace engineers utilize computational fluid dynamics (CFD); CFD is an
effective way to simulate aerodynamic performance without the need for expensive and impractical experimentation.
There are many types of CFD methods ranging from highly accurate Direct Numerical Simulation (DNS) to lower-
fidelity Reynolds Averaged Navier-Stokes (RANS) simulations; each of which has their pros and cons. A given CFD
method demonstrates some of its advantages and disadvantages through its turbulence modeling capabilities [1]. For
instance, DNS simulations can resolve all turbulent scales of motion, down to the Kolmogorov scale, see Schaefer et.
al [2] for a detailed discussion; however, this ability comes at a computational cost. On the other hand, steady Reynolds
Averaged Navier-Stokes (RANS) simulations depend less heavily on resolving the smallest of turbulence scales and
can resolve flows with significantly higher Reynolds number. For decades now, steady RANS simulations have aided
in reducing simulation time, cost, and effort [1]. The present study makes use of a steady RANS solver to decompose
the aerodynamic drag acting on the RAE-2822 airfoil [3]. The goal of this study is to question the complex
aerodynamic interaction between a propulsive system and a lift generating surface. Specifically, the effects of near
field flow variations introduced by a propulsive system are to be analyzed via computational simulation.
Understanding this complicated phenomenon will provide a deeper insight into the influence of propulsive systems
on airfoil aerodynamics, while simultaneously bridging the gap between aerodynamicists, propulsion engineers, and
aerospace design specialists. To achieve this, it is required to implement a propulsive system that injects a body force
to a given flow field and to analyze the resulting interaction. This novel approach builds upon the longstanding method
introduced by Shen & Sørensen [4], which will be used to develop a variation of the actuator line model (ALM). The
ALM is attractive in the wind-energy community due to less complicated mesh requirements and lower computational
cost. The ALM is a powerful computational method with an extensive research history; see Ref. [5,6] for a detailed
discussion on ALM applications. The ALM developed in this thesis is a novel computational model for a body force
that will enhance the stability of the simulation. The development of this ALM is covered in section III and its effects
are discussed in section V.
To decompose the aerodynamic drag acting on the RAE-2822 airfoil, the methods introduced by Schmitz & Hart
[7,8] will be expanded upon. Partial pressure field (PPF) theory leverages the linearity of the static pressure field and
allows for a comprehensive summary of forces acting on a lifting surface. The PPF method is widely applicable to a
broad range of CFD cases, ranging from incompressible subsonic flow to compressible transonic flow [8]. The
successful implementation and combination of the ALM with PPFs will give insight into complicated airfoil-body
force interactions in CFD. The decomposition of the static pressure field in the Navier-Stokes equation with a body
force term is discussed in section II. A discussion of the ALM and its implications is made in section III. Section IV
follows with a discussion on the computational methods and assumptions in place. Results and conclusions are
discussed in sections V and VI, respectively. Finally, an appendix is included to detail specific calculations, MATLAB
post processing scripts, and a C++ source code.

II. Static Pressure Decomposition of the Navier-Stokes Equation with a Body Force
The static pressure decomposition of the Navier-Stokes equation in Ref. [7,8] does not consider body forces. The
present study seeks to quantify the interaction between a propulsive system and the aerodynamics of the RAE-2822
airfoil [3]. As a result, it is required to analyze the implications of a body force on the Navier-Stokes equation. The
momentum equation can be easily modified to include an external body force represented by:

𝜕𝜕 (1)
(𝜌𝜌𝑣𝑣⃗) + ∇ ∙ (𝜌𝜌𝑣𝑣⃗𝑣𝑣⃗) = −∇𝑝𝑝 + ∇ ∙ (𝜏𝜏) + 𝜌𝜌𝑔𝑔⃗ + 𝐹𝐹⃗
𝜕𝜕𝜕𝜕

Here the body force term is depicted on the right-hand side of Eq. (1) by the vector 𝐹𝐹⃗ . The development of 𝐹𝐹⃗ is rooted
in actuator line theory and is covered later, see section (III.B). Furthermore, the present study is to be conducted in a
steady flow environment, such that time variations in the flow are neglected. Finally, the effect of the gravitational
acceleration vector 𝑔𝑔⃗ is to be neglected as well. With these assumptions in place, Eq. (1) can be written as:

4
 ∇ ∙ (𝜌𝜌𝑣𝑣⃗𝑣𝑣⃗) = −∇𝑝𝑝 + ∇ ∙ (𝜏𝜏) + 𝐹𝐹⃗ (2)

The viscous stress tensor 𝜏𝜏̅ associated with Eq. 2 is particularly difficult to define, yet the linearity of the scalar pressure
field permits an analytical static pressure decomposition (i.e., ∇𝑝𝑝𝜇𝜇 = ∇𝑝𝑝 − ∇𝑝𝑝𝐸𝐸 − ∇𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 ) where 𝑝𝑝𝐸𝐸 is the Euler partial
pressure, 𝑝𝑝𝜇𝜇 is the dissipative partial pressure, and 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 is the propeller partial pressure. With this static pressure
decomposition in mind, the momentum equation can be rewritten as follows:

∇𝑝𝑝𝐸𝐸 + 𝜌𝜌(𝑣𝑣⃗ ∙ ∇)𝑣𝑣⃗ = �−∇𝑝𝑝𝜇𝜇 + ∇ ∙ (𝜏𝜏)� − �∇𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 − 𝐹𝐹⃗ � (3)

Here each PPF is grouped with its associated physical terms. Recognizing that the pressure gradient in the far-field of
the flow is equal to zero [7,8] one can develop the following three pressure gradients:

∇𝑝𝑝𝐸𝐸 = −𝜌𝜌(𝑣𝑣⃗ ∙ ∇)𝑣𝑣⃗ (4)

∇𝑝𝑝𝜇𝜇 = ∇ ∙ (𝜏𝜏) (5)

∇𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 = 𝐹𝐹⃗ (6)

Equations 4-6 represent the static pressure decomposition of the momentum equation with the inclusion of a body
force term. To prepare these equations for implementation into a computational domain, it is required to make use of
the gradient operator, i.e.:

∇2 𝑝𝑝𝐸𝐸 = ∇ ∙ (−𝜌𝜌(𝑣𝑣⃗ ∙ ∇)𝑣𝑣⃗) (7)

∇2 𝑝𝑝𝜇𝜇 = ∇ ∙ (∇ ∙ 𝜏𝜏̅) (8)

∇2 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 = ∇ ∙ 𝐹𝐹⃗ (9)

Equations 7-9 are three Poisson equations that can be numerically solved using computational methods, see section
IV.A. It is necessary to define boundary conditions such that Eqs. 7-9 can be solved. In the farfield (i.e., inlets and
sides), there are no viscous effects; Euler pressure is enforced to be equal to the static pressure, and the propulsor
pressure 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 is constrained to be zero. In other words, the following Dirichlet boundary conditions are in place at
the inlets and sides: 𝑝𝑝𝜇𝜇 = 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 = 0 and 𝑝𝑝𝐸𝐸 = 𝑝𝑝. The dissipative pressure normal to the outlet is defined by ∇𝑝𝑝𝜇𝜇 ∙ 𝑛𝑛� =
0 as a von Neumann boundary condition. In addition, a no-slip boundary condition is enforced at the walls. Thus, at
the walls, ∇𝑝𝑝𝜇𝜇 = ∇ ∙ 𝜏𝜏̅, ∇𝑝𝑝𝐸𝐸 = 0, and ∇𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 ∙ 𝑛𝑛� = 0. With these boundary conditions in place, the Poisson equations
above can be solved for each PPF, and their drag contributions can be determined by integrating each respective PPF
in the nearfield.

III. Actuator Line Development and Implementation

The following section concerns the development and implementation of the body force source term 𝐹𝐹⃗ . The source
term in question serves to appropriately express 𝐹𝐹⃗ such that the amount of injected force is representative of a Cessna
172 [9]. A Cessna 172 was selected for analysis because its propulsion system is well researched, and its cruise speed
(M=0.18) avoids any transonic effects. To simplify the simulation, the propulsive system of the Cessna 172 must be
modified from a front-mounted single engine propeller to an effective system of two wing-mounted propellers. The
resulting thrust produced from two smaller wing-mounted engines is to be equal to the thrust produced by one front-
mounted engine. To achieve proper actuator line implementation, it is required to establish the effective propeller size
(III.A), an estimation of the exit velocity introduced by the propeller (III.B), and a Gaussian regularization function
that spreads 𝐹𝐹⃗ over a given range of grid cells (III.C).

A. Effective Propeller Thrust and Area Formulation

5
 The present study is not concerned with complicated jet wake-airframe interactions, thus is it required to reposition
the propeller such that only the interaction between a propeller and lifting surface (airfoil) is captured. Simulating the
propeller as a fictitious wing-mounted configuration instead of a front-mounted propeller configuration allows for the
omission of the airframe entirely. It can easily be shown that the total 2-D area of a front-mounted propeller is:

𝜋𝜋𝑑𝑑 2 (10)
𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 =
4

where d is the diameter of the front-mounted propeller [9]. Thus, the equivalent area of each wing-mounted propeller
can be defined as half of the total propeller area (i.e., Aeq=Atotal / 2). The required radius of the equivalent wing-
mounted propeller is then:

𝐴𝐴𝑒𝑒𝑒𝑒 𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡
𝑟𝑟𝑒𝑒𝑒𝑒 = � = � (11)
𝜋𝜋 2𝜋𝜋

The results of Eq. 11 are later used to develop the length of the propeller model to be implemented into ANSYS Fluent,
see section IV. C. With the dimensions of the effective propulsion system defined, a control volume analysis rooted
in momentum theory can be performed on the equivalent propeller, from which an expression for the estimated jet
velocity can be developed.

B. Jet Velocity/Thrust Formulation

The analysis shown in section III.A determines the equivalent area and radius of a theoretical wing-mounted
propeller whose propulsive capability resembles a Cessna 172. To formulate an estimate for the jet velocity introduced
by this wing-mounted propeller requires a control-volume analysis to be performed on a rotor disk in climb. The
control-volume analysis performed by Leishman [10] and Schmitz [11] is shown below. The following simplifying
assumptions are made to idealize the flow through the control volume:

1) The flow is strictly one-dimensional (1-D)

2) The flow is quasi-steady (i.e., neglect time variations in the flow)
3) The flow is inviscid & incompressible

With these assumptions in place, a control volume of a rotor disk in climb can be modeled as follows:

Fig. 1 Flow model for momentum theory analysis of a rotor in axial climbing flight [10,11].

6
It is worthy to note that the flow model in Fig. 1 is for a rotor disk in vertical flight; however, the analysis is identical
for a rotor in horizontal flight. In either case, the same flow assumptions are made, and the only discrepancy is a
difference in rotor orientation and flow direction. If the velocity is continuous across the disk and if the control volume
is impermeable (i.e., zero mass flux across boundaries), the conservation of mass can be applied as follows:

𝜕𝜕 (12)
�⃗ ∙ 𝑛𝑛��𝑑𝑑𝑑𝑑 = 0
� 𝜌𝜌𝜌𝜌𝜌𝜌 + � 𝜌𝜌�𝑉𝑉
𝜕𝜕𝜕𝜕 𝑉𝑉 𝑆𝑆

Here 𝑛𝑛� is a unit vector normal to the control surface S. Additionally, since the flow is quasi-steady, the time derivative
in Eq. 12 is equal to zero. Furthermore, for a rotor disk in climb:

(13)
�⃗ ∙ 𝑛𝑛��𝑑𝑑𝑑𝑑 = �
𝑚𝑚̇ = � 𝜌𝜌�𝑉𝑉 �⃗ ∙ 𝑛𝑛��𝑑𝑑𝑑𝑑
𝜌𝜌�𝑉𝑉
𝑆𝑆 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅

Since the flow is 1-D and incompressible, the mass flow rate 𝑚𝑚̇ is given by:

𝑚𝑚̇ = 𝜌𝜌𝐴𝐴∞ (𝑉𝑉𝑐𝑐 + 𝑤𝑤) = 𝜌𝜌𝜌𝜌(𝑉𝑉𝑐𝑐 + 𝑣𝑣𝑖𝑖 ) = 𝜌𝜌𝐴𝐴0 𝑉𝑉𝑐𝑐 ; 𝐴𝐴1 = 𝐴𝐴2 = 𝐴𝐴 (14)

Equation 14 is a statement of the conservation of mass (and resulting mass flow rate) inside the control volume. The
preceding mass conservation analysis can be coupled with the conservation of momentum to develop a relationship
between thrust and mass flow rate.
To begin a momentum analysis on the control volume in Fig. 1, it is required that 𝑝𝑝 = 𝑝𝑝∞ on most of S except for
the actuator disk. In this case, the momentum principle can be applied which gives:

𝐹𝐹⃗ = 𝑇𝑇
�⃗ = � 𝑝𝑝𝑛𝑛� 𝑑𝑑𝑑𝑑 + � �𝜌𝜌𝑉𝑉
�⃗�𝑉𝑉
�⃗ ∙ 𝑛𝑛� 𝑑𝑑𝑑𝑑 ; 𝑝𝑝 = 𝑝𝑝∞ (15)
𝑆𝑆 𝑆𝑆

The assumption that 𝑝𝑝 = 𝑝𝑝∞ on most of S allows Eq. 15 to be simplified such that the pressure flux term is zero. As a
result, the net change in momentum inside of 𝑑𝑑𝑆𝑆⃗ is:

𝐹𝐹⃗ = 𝑇𝑇
�⃗ = � �𝜌𝜌𝑉𝑉
�⃗�𝑉𝑉
�⃗ ∙ 𝑛𝑛� 𝑑𝑑𝑑𝑑 − � �𝜌𝜌𝑉𝑉
�⃗�𝑉𝑉
�⃗ ∙ 𝑛𝑛� 𝑑𝑑𝑑𝑑 (16)
∞ 0

Thus,

𝑇𝑇 = � 𝜌𝜌(𝑉𝑉𝑐𝑐 + 𝑤𝑤)(𝑉𝑉𝑐𝑐 + 𝑤𝑤) 𝑑𝑑𝑑𝑑 − � 𝜌𝜌𝑉𝑉𝑐𝑐 𝑉𝑉𝑐𝑐 𝑑𝑑𝑑𝑑 = 𝑚𝑚̇(𝑉𝑉𝑐𝑐 + 𝑤𝑤) − 𝑚𝑚̇𝑉𝑉𝑐𝑐 = 𝑚𝑚̇𝑤𝑤 (17)
∞ 0

where T is the rotor disk’s thrust and w is the downwash resulting from the induced velocity at the rotor disk. To obtain
an expression for the downwash of a given rotor disk, an analysis based on the conservation of energy is required. The
analysis performed by Leishman [10] can be summarized by the following well-known relation:

𝑤𝑤 = 2𝑣𝑣𝑖𝑖 (18)

Substituting Eq. 14 and Eq. 18 into the findings of Eq. 17 gives the following:

𝑇𝑇𝑒𝑒𝑒𝑒 = 2𝜌𝜌𝐴𝐴𝑒𝑒𝑒𝑒 (𝑉𝑉𝑐𝑐 + 𝑣𝑣𝑖𝑖 )𝑣𝑣𝑖𝑖 (19)

According to McIver [12], the single front-mounted propeller of the Cessna 172 can produce 1041 lbf (4631 N) of
static thrust (𝑇𝑇𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 ) under idealized conditions. Furthermore, since 𝑇𝑇𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 of the doubly mounted wing propeller system
must equal that of the singly mounted front propeller system, the equivalent thrust of each propeller must be 𝑇𝑇𝑒𝑒𝑒𝑒 =
𝑇𝑇𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 / 2.

7
 Therefore, upon inspection, the only unknown quantity in Eq. 19 is the induced velocity 𝑣𝑣𝑖𝑖 . Note, the climb velocity
𝑉𝑉𝑐𝑐 is taken to be equal to the cruise speed of the Cessna 172 (i.e., 𝑉𝑉𝑐𝑐 = 62.5856 m/s) [9]. Additionally, the fluid density
𝜌𝜌 depends on a given Reynolds number, flight speed, and dynamic viscosity; see Appendix A. for a detailed derivation.
From Eq. 19, the following quadratic equation can be derived:

𝑇𝑇𝑒𝑒𝑒𝑒 (20)
𝑣𝑣𝑖𝑖2 + 𝑉𝑉𝑐𝑐 𝑣𝑣𝑖𝑖 − =0
2𝜌𝜌𝐴𝐴𝑒𝑒𝑒𝑒

When solved, the only valid solution to Eq. 20 is such that 𝑣𝑣𝑖𝑖 > 0. Thus, the induced velocity is determined and an
estimate for the downwash w can be made. Referencing Fig. 1, the jet velocity introduced by the propeller as a result
of the induced velocity is 𝑉𝑉𝑐𝑐 + 𝑤𝑤. As a result of the preceding analysis, the expected jet velocity from the effective
propulsion system is 71 m/s. This value will be used to validate the accuracy of the rotor disk model to be implemented
into ANSYS Fluent, see section IV.C.

C. Formulation of Gaussian Distribution

Abruptly injecting a large body force into a given computational domain is both inefficient and inaccurate. A large
gradient in momentum across two adjacent cells yields stability issues for a given case, and it is commonplace in the
wind-energy community to “spread” a force over a range of cells, see Schmitz & Jha [13]. The ALM developed by
Shen & Sørensen [4] achieves this goal and applying a variation of it to the study in question yields both accurate and
time-efficient results. According to Ref. [4], the required body force introduced to a computational domain takes the
form:

𝐹𝐹(𝑥𝑥, 𝑦𝑦) = − � 𝑓𝑓(𝑥𝑥, 𝑦𝑦)𝜂𝜂𝑁𝑁 (21)

𝑁𝑁

Note, f is a function of spatial coordinates and η is a 2-D Gaussian projection function. The subscript N assigns a
positive integer value (i.e., {𝑁𝑁: 𝑁𝑁 ∈ ℤ+}) to a given Gaussian distribution function. Further, a forcing function f (that
remains constant throughout the domain) is introduced, which is then multiplied into the 2-D Gaussian projection
function. This product is then integrated as a volume sum integral throughout the entire domain of interest to validate
its implementation, see section IV.C. Schmitz [11] defines the 2-D Gaussian distribution function as follows:

1 |𝑟𝑟⃗|
2
(22)
𝜂𝜂 = 2 exp �− � � �
𝜋𝜋𝜀𝜀 Δ 𝜀𝜀

Here 𝜀𝜀 is defined as a planar Gaussian projection radius, frequently accepted as twice the local grid spacing [6]. The
magnitude of 𝜀𝜀 determines the strength of the normalization function [4]. In addition, |𝑟𝑟⃗| = �����⃗ 𝑟𝑟𝑁𝑁 � is the distance
𝑟𝑟𝑝𝑝 − ���⃗
𝑇𝑇
between a given grid cell, 𝑟𝑟���⃗
𝑝𝑝 = �𝑥𝑥𝑝𝑝 , 𝑦𝑦𝑝𝑝 � , and a Gaussian location, ���⃗
𝑟𝑟𝑁𝑁 = (𝑥𝑥𝑁𝑁 , 𝑦𝑦𝑁𝑁 )𝑇𝑇 . The parameter ∆ is defined as a
length scale in the direction of propeller thrust (i.e., −𝚤𝚤̂). This parameter ∆ is included so that the units of 𝜂𝜂 are [1/m3]
[11], such that when multiplied with f [N], the resulting body force 𝐹𝐹 ���⃗𝑝𝑝 has units [N/m3]; Fluent requires that x-
momentum source terms are in units [N/m ] [14]. It can be verified through a domain sum volume integral that this
3

implementation is correct, see Fig. 6. To address the forcing function f, some simplifying assumptions need to be
made. In particular, the forcing function is assumed to be constant, time invariant, and precisely equal to the thrust
produced by one propeller, see section IV.B. Using Eq. 21 with the definition of f and 𝜂𝜂 in place gives way to a source
term that is interpreted by Fluent and appropriately modifies the x-momentum flow equation.

IV. Computational/Numerical Methods

The preceding sections focus on developing the tools required to computationally implement the ALM and PPFs.
Specifically, the work of Shen & Sorensen [4], Schmitz [7], and Schmitz & Hart [7] has been extensively used thus
far. The Poisson equations outlined by Eqs. 7-9 are of particular interest for quantifying aerodynamic drag.
Determining a solution to these equations permits the calculation of the respective drag forces associated with each
PPF. The following sections outline the computational methodologies used to properly synthesize the ALM with the
PPFs. Computations are to be performed using ANSYS Fluent 2021 R1 [15], a computational tool with advanced
physics modeling capabilities. All solution data are post processed in MATLAB R2020b. Further, it is required to

8
make various assumptions about the fluid flow to achieve a realistic solution. Notably, the flow must be incompressible
(constant 𝜌𝜌), time invariant, and must avoid transonic effects (i.e., low Mach number). With these assumptions in
mind, a flow field can be developed that represents a Cessna 172 in a steady, level flight condition.
As in Schmitz & Hart [8], the one-equation Spalart-Allmaras eddy-viscosity turbulence model was used, second
order upwinding was used to calculate numerical solutions, and gradients were determined using the Green-Gauss
cell-based method. To support the solution data, it is required to discuss how Fluent interprets the Poisson equations
of interest (IV.A), the grid modifications required to utilize the ALM and PPF decomposition (IV.B), and the
verification steps taken to ensure realistic ALM implementation (IV.C).

A. Poisson Equations in Fluent

The general form of a user-defined scalar transport equation is given in the Ansys Fluent UDF Manual [14] as:

𝜕𝜕(𝜌𝜌𝜌𝜌) (23)
+ 𝜌𝜌(𝑣𝑣⃗ ∙ ∇)𝜑𝜑 = ∇ ∙ (Ξ∇φ) + 𝑆𝑆𝜑𝜑
𝜕𝜕𝜕𝜕

Here 𝜑𝜑 is a given scalar quantity, 𝑆𝑆𝜑𝜑 is a source term, and Ξ is defined as a diffusion coefficient. The convective term
above accounts for the transport of 𝜑𝜑 due to the presence of the velocity field 𝑣𝑣⃗ and the diffusive term above accounts
for the transport of 𝜑𝜑 due to its gradients.
Equation 23 can be simplified because of the assumptions made about the flow field. Specifically, the steady nature
of the flow eliminates the transient term and, by using the appropriate settings in Fluent, one can disable the convective
transport of 𝜑𝜑. Finally, the diffusive coefficient Ξ is set to be one; the result is a Poisson equation that Fluent solves
numerically. The choice of 𝑆𝑆𝜑𝜑 in the present study is the right-hand-side of Eqs. 7-9. The boundary conditions outlined
in section II were implemented using the Fluent graphical user interface (GUI). The result of this method is three user-
defined Poisson equations, one for each PPF in the flowfield.

B. Airfoil/Grid Selection and Modifications

Various decisions needed to be made regarding the grid selection, spacing, and refinement. The airfoil in question
for analysis is the RAE 2822 transonic airfoil, shown below:

Fig. 2 Schematic of RAE 2822 airfoil [3].

The RAE 2822 airfoil has a maximum thickness of 12.1% at 37.9% chord and a max camber of 1.3% at 75.7% chord
[3]. Although this study does not consider the transonic regime, this airfoil was selected due to its extensive use in
CFD validation cases, see Doig et. al [16]. The RAE-2822 CFD grid used is a structured C-grid with 145,014
quadrilateral cells, 291,410 faces, and 146,396 nodes.

9
 It is worthy to note that both PPF decomposition and ALM implementation are subject to sources of error unless
the computational grid is reasonably fine. Therefore, it is required to use tools in Fluent such as Adapt & Refine Mesh
[15]. The grid is refined in a given region of interest for where the ALM is implemented (i.e., sufficiently close to the
RAE 2822 airfoil). For reference, consider the following figure:

Fig 3. Example of mesh refinement in region of interest

The region of interest in this example is below and sufficiently close to the airfoil such that a velocity/pressure
interaction can be analyzed, see section IV.C and V.B-C. Furthermore, note that the mesh gradually increases in cell
size with increasing x-distance from the airfoil, so as to not introduce abrupt changes in cell size to the grid.
Ultimately, the location of mesh refinement region will be varied throughout this study and is chosen such that the
ALM and its wake is properly captured.

C. Verification Methods
This section outlines the visual and computational methods used to verify the implementation of the ALM. As
mentioned in section III., the actuator line model should introduce a body force into the flow domain that is normally
distributed across an arbitrary column of cells. In particular, the findings of Eq. 11 (i.e., 𝑟𝑟𝑒𝑒𝑒𝑒 ) explicitly constrain the
length of cells (in the y direction) that the ALM should act. According to the Cessna 172 Pilot’s Operating Handbook
(POH) [9], the diameter (𝑑𝑑𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 ) of the Cessna 172’s propeller is 75 in (1.905 m). Following the process outlined in
section III. A, one finds that the required diameter of the effective propulsion system is 𝑑𝑑𝑒𝑒𝑒𝑒 = 2𝑟𝑟𝑒𝑒𝑒𝑒 = 1.347𝑚𝑚.
Ultimately, the length of the ALM in Fluent is governed by the number of Gaussian distributions (𝑁𝑁) and it is
required to develop an expression for 𝑦𝑦𝑁𝑁 , see section III. C. Note, it is unnecessary to define an expression for 𝑥𝑥𝑁𝑁 ,
since the only variations in the ALM are in the y-direction; 𝑥𝑥𝑁𝑁 is considered a constant and is arbitrarily chosen to be
slightly behind the leading edge of the RAE-2822 airfoil. Consider the following linear expression for a given
Gaussian location 𝑦𝑦𝑁𝑁 :

𝑦𝑦𝑁𝑁 = 𝑦𝑦0 − (𝑚𝑚 ∙ 𝑖𝑖) (24)

Here 𝑦𝑦0 is the y-location of the top of the ALM, i is a positive integer representing the indices of a C++ for loop, see
Appendix B., and m is a slope defined by:

𝑑𝑑𝑒𝑒𝑒𝑒 (25)
𝑚𝑚 = ; 𝑁𝑁 = 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 𝑜𝑜𝑜𝑜 𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
𝑁𝑁

10
The purpose of defining this linear function is to express 𝑦𝑦𝑁𝑁 such that when 𝑖𝑖 = 0, the location of the first Gaussian
distribution is 𝑦𝑦𝑁𝑁 = 𝑦𝑦0 [𝑚𝑚] and when 𝑖𝑖 = 𝑁𝑁, 𝑦𝑦𝑁𝑁 = 𝑦𝑦0 − 𝑑𝑑𝑒𝑒𝑒𝑒 ; thus giving a range of 𝑦𝑦𝑁𝑁 that exactly satisfies the length
requirement of the ALM (𝑑𝑑𝑒𝑒𝑒𝑒 ). Suppose the desired location of the top of the ALM is 𝑦𝑦0 = −3 𝑚𝑚, and the desired
number of Gaussian distributions is 𝑁𝑁 = 20. Using Eqs. 24-25 and remaining consistent with C++ coding syntax, the
implementation of the ALM is coding into Fluent using an interpreted source code, see appendix B. The resulting
implementation is shown by the following figure:

Fig 4. Implemented actuator line model of length 𝒅𝒅𝒆𝒆𝒆𝒆 (𝑦𝑦0 = −3 𝑚𝑚 , 𝑥𝑥𝑁𝑁 = 0.0125 𝑚𝑚 , 𝜀𝜀 = 0.1, 𝑁𝑁 = 20)

Figure 4 verifies the findings of Eqs. 24-25 and demonstrates an ALM whose strength is spread normally across
multiple columns of cells. The following figure allows for easier visualization of each Gaussian location:

Fig. 5 Post-processing of Gaussian distributions, MATLAB R2020b

Note that Fig. 5 represents N=20 Gaussian distributions that are evenly distributed across the desired y-values
[𝑦𝑦𝑜𝑜 , 𝑦𝑦0 − 𝑑𝑑𝑒𝑒𝑒𝑒 ]. It is equally important to verify that the ALM is introducing the proper forcing magnitude to the
flowfield. In particular, the volume integral sum of 𝐹𝐹(𝑥𝑥, 𝑦𝑦) = 𝜂𝜂 ∙ 𝑓𝑓 across every cell within the domain should be

11
equal to Teq = 2315.5 N, as per section III. B. Making use of the built-in volume integral tool in Fluent gives the
following results:

Fig 6. Volume integral sum of 𝑭𝑭(𝒙𝒙, 𝒚𝒚) throughout flow domain (UDS-5)

Although the magnitude of the volume integral sum is slightly higher than the expected value of 𝑇𝑇𝑒𝑒𝑒𝑒 , the implemented
ALM has regions of overlapping Gaussian distributions, which is a leading source of error. Decreasing the number of
Gaussian distributions N will slightly mitigate this issue; however, this discrepancy in added force is small and can be
considered insignificant when interpreting the results in section V.
Finally, since the ALM is injecting thrust into the flow field, it is expected that the jet velocity from the ALM is
consistent with the findings in section III. B. Consider the following velocity contour for example:

Fig 7. Flow field velocity magnitude with ALM and airfoil interaction, 𝒚𝒚𝟎𝟎 = −𝟑𝟑𝟑𝟑

Notably, Fig. 7 depicts a jet velocity of 77 m/s. While this is slightly higher than the predicted value from the
previous section, the difference can be attributed to the slightly higher forcing value seen in Fig. 6. Furthermore, by
nature of the 2-D simulation, tip vortices that would induce the flow to contract the wake cannot be resolved.
Nevertheless, Fig. 7 illustrates an interesting velocity (and hence pressure) interaction between the ALM and airfoil
that is to be quantified in the following section.

V. Results
The following section is concerned with quantifying the sources of aerodynamic forces acting on the airfoil via
ANSYS Fluent 2021R1 [15]. Note, all such cases were run with Re = 5 million, 𝛼𝛼 = 0 degree, and M = 0.18. The
assumptions outlined in the preface of the previous section are in place. Further, convergence was determined through
sufficiently small magnitudes of solution residuals. Finally, to develop intelligible results, the location of the ALM is

12
to be varied in the y-direction such that its influence can be analyzed. In particular, the following section is broken up
as: the base-case (V.A), 𝑦𝑦0 = −3 𝑚𝑚 (V.B), and 𝑦𝑦0 = −1.5 𝑚𝑚 (V.C).

A. Base-case Results
It is required to establish a “base-case” so that later simulations (with the inclusion of the ALM) will reflect the
changes imposed by the ALM. The work performed in Ref. [7,8] is directly applicable to the base-case due to their
omission of a body force term. Essentially, the propeller source term and resulting Poisson equation shown by Eq. 9
can be dropped from analysis for the base-case. The result is an undisturbed flow domain wherein the sole subject of
analysis is the airfoil boundary condition.
The distribution of static, Euler, and dissipative pressure coefficients, respectively, can be visualized as follows:

Fig 8. RAE-2822 airfoil – Static, inviscid, and viscous 𝑪𝑪𝒑𝒑 vs x/c (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18, 𝐶𝐶𝑙𝑙 = 0.194)

As expected, it is supported by Fig. 8 that the inviscid Euler PPF is consistent with the static pressure acting on the
RAE-2822 airfoil. It is worthy to note that error is introduced at the trailing edge of the RAE-2822. In particular, the
viscous dissipative pressure approaches a vertical asymptote at x/c = 1. The contours of the static pressure coefficient
within the vicinity of the airfoil surface are shown below:

Fig. 9 RAE-2822 airfoil – static pressure magnitude, No ALM (𝑪𝑪𝒑𝒑 ) (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18)

The preceding pressure information can be used to quantify the aerodynamic drag acting on the airfoil surface.
Specifically, multiplying the PPFs obtained from the now solved Poisson equations by the x-face area of a given cell
and summing across the entire domain gives a numerical value for the aerodynamic drag [N] imposed by each PPF.
Note, this is only possible since the cases in question were run at 𝛼𝛼 = 0𝑜𝑜 . If angle-of-attack studies are to be
conducted, then it is required to consider the y-face area of a given cell. Nondimensionalizing and tabulating the PPF
results gives the following base-case scenario:

13
 Table 1. RAE-2822 airfoil – PPF decomposition, No ALM (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18, 𝛼𝛼 = 0𝑜𝑜 )
Lift Coefficient:
Fluent Force Report [15]
𝑪𝑪𝒍𝒍 0.1938906
Total Drag (× 𝟏𝟏𝟏𝟏−𝟒𝟒 ):
Fluent Force Report [15]
𝑪𝑪𝒅𝒅 85.03065
Euler Pressure Drag (× 𝟏𝟏𝟏𝟏−𝟒𝟒 ):
PPF Poisson Eq. 7
𝑪𝑪𝒅𝒅,𝒑𝒑𝑬𝑬 4.578591
−𝟒𝟒
Dissipative Pressure Drag (× 𝟏𝟏𝟏𝟏 ):
PPF Poisson Eq. 8
𝑪𝑪𝒅𝒅,𝒑𝒑𝝁𝝁 8.924545
−𝟒𝟒
Pressure Drag (× 𝟏𝟏𝟏𝟏 ):
Fluent Force Report [15]
𝑪𝑪𝒅𝒅,𝒑𝒑 13.50314
Viscous Drag (× 𝟏𝟏𝟏𝟏−𝟒𝟒 ):
Fluent Force Report [15]
𝑪𝑪𝒅𝒅,𝝉𝝉 71.52752

Interestingly, despite the magnitude of the viscous dissipative pressure coefficient being lower than that of the inviscid
Euler pressure coefficient, see Fig. 8, the dissipative pressure drag dominates the Euler pressure drag. In other words,
the viscous dissipative pressure force is small, yet it contributes significantly to the total drag acting on the RAE-2822
airfoil.

B. Actuator Line Model 3m Below RAE-2822 Airfoil

With a base-case in place, the effects introduced by an ALM can be interpreted. With the implementation of an
ALM, the PPF methodology differs from section V.A. In particular, Eq. 9 is now considered in analysis. Appropriately
implementing the ALM using the aforementioned methodology gives a flowfield where the airfoil-ALM interaction
can be quantified via PPFs. The pressure interaction shown by the following figure is to be analyzed:

Fig. 10 RAE-2822 airfoil – ALM/airfoil static pressure interaction (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18, 𝑦𝑦0 = −3𝑚𝑚)

Based on Fig. 7 and Fig. 10, it is apparent that a flowfield interaction is occurring which can be attributed to the ALM.
To quantify this interaction, the propeller Poisson equation is solved in Fluent, and its corresponding drag is calculated.
At 𝛼𝛼 = 0𝑜𝑜 , chordwise pressure coefficient distributions are plotted as follows:

14
 Fig 11. RAE-2822 airfoil – Static, inviscid, viscous, and propeller 𝑪𝑪𝒑𝒑 (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18, 𝑦𝑦0 = −3𝑚𝑚)

When compared with Fig. 8, it is clear that the ALM is influential on the inviscid Euler pressure distribution,
particularly with the pressure differential amongst the upper and lower surfaces of the RAE-2822. This “compression”
of the Euler pressure coefficient distribution has prominent effects on the airfoil’s lifting capabilities. The lift and drag
values are tabulated as:

Table 2. RAE-2822 airfoil – PPF decomposition, ALM with 𝒚𝒚𝟎𝟎 = −𝟑𝟑𝟑𝟑 (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18, 𝛼𝛼 = 0𝑜𝑜 )
Lift Coefficient:
Fluent Force Report [15]
𝑪𝑪𝒍𝒍 0.0997072
−𝟒𝟒
Total Drag (× 𝟏𝟏𝟏𝟏 ):
Fluent Force Report [15]
𝑪𝑪𝒅𝒅 91.47096
−𝟒𝟒
Euler Pressure Drag (× 𝟏𝟏𝟏𝟏 ):
PPF Poisson Eq. 7
𝑪𝑪𝒅𝒅,𝒑𝒑𝑬𝑬 20.04357
Dissipative Pressure Drag (× 𝟏𝟏𝟏𝟏−𝟒𝟒 ):
PPF Poisson Eq. 8
𝑪𝑪𝒅𝒅,𝒑𝒑𝝁𝝁 9.021762
−𝟒𝟒
Propeller Drag (× 𝟏𝟏𝟏𝟏 ):
PPF Poisson Eq. 9
𝑪𝑪𝒅𝒅,𝒑𝒑𝒑𝒑𝒑𝒑𝒑𝒑 -8.684941
−𝟒𝟒
Viscous Drag (× 𝟏𝟏𝟏𝟏 ):
Fluent Force Report [15]
𝑪𝑪𝒅𝒅,𝝉𝝉 71.09057
−𝟒𝟒
Pressure Drag (× 𝟏𝟏𝟏𝟏 ):
Fluent Force Report [15]
𝑪𝑪𝒅𝒅,𝒑𝒑 20.37628

By Table 2 and Fig. 11, it is concluded that the ALM reduces the lifting capability of the RAE-2822 airfoil by
48.6%. The large velocity introduced by the ALM into the flowfield below the RAE-2822 ultimately decreases the
pressure differential between the upper and lower surfaces of the airfoil. This effect deserves further examination and
is likely attributed to the limitations of a 2-D simulation (i.e., inability to resolve tip vortices). Furthermore, the total

15
drag increased by 6.44 drag counts and the Euler pressure drag increased drastically by 15.5 drag counts. Interestingly,
the propeller PPF nearly cancels the dissipative PPF. In other words, despite this large increase in Euler pressure drag,
the neutralization of the dissipative pressure drag on behalf of the ALM results in a still reasonable total drag increase.

C. Actuator Line Model 1.5m Below RAE-2822 Airfoil

The following section is concerned with further closing the distance between the RAE-2822 airfoil and the ALM.
Based on the preceding section, the drag associated with the Euler PPF is expected to further increase due to the
implications of the ALM on the fluid’s inertial properties. The computational methodology is consistent with the
previous sections; it is required to make a simple change in the source code (i.e., 𝑦𝑦0 = −1.5𝑚𝑚) and to reconverge the
flow with the newly refined mesh. The ALM-airfoil interaction becomes even more pronounced in this condition and
is visualized by the following:

Fig. 12 RAE-2822 airfoil – ALM/airfoil pressure interaction (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18, 𝑦𝑦0 = −1.5𝑚𝑚)

Fig. 13 RAE-2822 airfoil – ALM/airfoil velocity interaction (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18, 𝑦𝑦0 = −1.5𝑚𝑚)

The interaction above is again quantified through PPFs, and the following pressure coefficient distributions on the
airfoil surface can be developed:

16
 Fig 14. RAE-2822 airfoil – Static, inviscid, viscous, and propeller 𝑪𝑪𝒑𝒑 (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18, 𝑦𝑦0 = −1.5𝑚𝑚)

The visualization of each PPF gives insight into the effects of the ALM on the airfoil surface. Specifically, the
distribution of 𝐶𝐶𝑝𝑝,𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 shown in Fig. 14 indicates that the propeller PPF has an appreciable impact on the airfoil’s
aerodynamics. This effect is quantified by the further reduction in lift coefficient from both section V.A and V.B.
Consider the following table:

Table 3. RAE-2822 airfoil – PPF decomposition, ALM with 𝒚𝒚𝟎𝟎 = −𝟏𝟏. 𝟓𝟓𝟓𝟓 (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18, 𝛼𝛼 = 0𝑜𝑜 )
Lift Coefficient:
Fluent Force Report [15]
𝑪𝑪𝒍𝒍 0.0433861
−𝟒𝟒
Total Drag (× 𝟏𝟏𝟏𝟏 ):
Fluent Force Report [15]
𝑪𝑪𝒅𝒅 90.59697
−𝟒𝟒
Euler Pressure Drag (× 𝟏𝟏𝟏𝟏 ):
PPF Poisson Eq. 7
𝑪𝑪𝒅𝒅,𝒑𝒑𝑬𝑬 22.10267
Dissipative Pressure Drag (× 𝟏𝟏𝟏𝟏−𝟒𝟒 ):
PPF Poisson Eq. 8
𝑪𝑪𝒅𝒅,𝒑𝒑𝝁𝝁 9.597234
−𝟒𝟒
Propeller Drag (× 𝟏𝟏𝟏𝟏 ):
PPF Poisson Eq. 9
𝑪𝑪𝒅𝒅,𝒑𝒑𝒑𝒑𝒑𝒑𝒑𝒑 -23.96220
−𝟒𝟒
Viscous Drag (× 𝟏𝟏𝟏𝟏 ):
Fluent Force Report [15]
𝑪𝑪𝒅𝒅,𝝉𝝉 82.348362
−𝟒𝟒
Pressure Drag (× 𝟏𝟏𝟏𝟏 ):
Fluent Force Report [15]
𝑪𝑪𝒅𝒅,𝒑𝒑 8.248611

The total drag coefficient 𝐶𝐶𝑑𝑑 generated by a Fluent force report [user manual] is within 0.51 drag counts of the Poisson
equation’s total drag coefficient (90.086 drag counts). As shown by Tables 2 & 3, relocating the ALM from 𝑦𝑦0 =
−3𝑚𝑚 to 𝑦𝑦0 = −1.5𝑚𝑚 causes a significant reduction to the lift coefficient of the RAE-2822 airfoil. Specifically, 𝐶𝐶𝑙𝑙 in

17
this case decreases by 56.4% from section V.B and 77.6% from section V.A. Furthermore, when compared with the
base-case, this near field ALM configuration causes a significant increase in the Euler drag, consistent with the
findings in section V.B. The Euler pressure drag continues to reflect notable influences resulting from the ALM, likely
a result of the Euler PPFs nonlinear dependence on the velocity field, see Eq. 7. Consistent with Fig. 14, Table 3
demonstrates the heightened effect of the propeller PPF drag on the airfoil’s aerodynamics. In contrast with section
V.B, the propeller PPF counteracts the sharp increase in Euler pressure drag. Interestingly, the total drag counts 𝐶𝐶𝑑𝑑
seen in case C. is lower than the total drag counts seen in case B. In essence, the ALM negatively impacts the lift
coefficient of the RAE-2822 airfoil, while the total drag counts seem to be within control.

D. Comprehensive Discussion
Combining the results from section V.A-C into one comprehensive section will allow for easier extraction of
conclusions. Consider the following figures that superimpose each PPF, allowing for easier visualization:

Fig. 15 RAE-2822 airfoil – Superimposed static pressure coefficients, 𝑪𝑪𝒑𝒑 (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18)

Fig. 16 RAE-2822 airfoil – Superimposed Euler PPF coefficients, 𝑪𝑪𝒑𝒑,𝑬𝑬 (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18)

The Euler PPF is particularly subject to the effects of the ALM, and the impact on lift is visualized in Figures 15-16.
Of course, the lower surface of the RAE-2822 demonstrates a significant change in both 𝐶𝐶𝑝𝑝 and 𝐶𝐶𝑝𝑝,𝐸𝐸 since the ALM
is introducing a large positive velocity below the airfoil. The following figures outline the changes in the dissipative
PPF, 𝐶𝐶𝑝𝑝,𝜇𝜇 , and the propeller PPF, 𝐶𝐶𝑝𝑝,𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 , respectively.

18
 Fig. 17 RAE-2822 airfoil – Superimposed dissipative PPF coefficients, 𝑪𝑪𝒑𝒑,𝝁𝝁 (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18)

Fig. 18 RAE-2822 airfoil – Superimposed propeller PPF coefficients, 𝑪𝑪𝒑𝒑,𝒑𝒑𝒑𝒑𝒑𝒑𝒑𝒑 (𝑅𝑅𝑅𝑅 = 5 × 106 , 𝑀𝑀∞ = 0.18)

The static and Euler PPF coefficients shown in Fig. 15-16 are consistent with one another. Furthermore, both figures
represent the impact on the upper and lower surface pressure differential from the ALM. The net effect from closing
the distance between the ALM and the RAE-2822 negatively impacts the lift coefficient of the airfoil. Furthermore,
from Fig. 18, the propeller PPF is notably influential when brought closer to the RAE-2822. The trailing edge
instabilities are consistent throughout each case and become relevant at about 98% x/c.

VI. Conclusion
The actuator line method [4] was used in conjunction with the partial pressure field [7,8] approach to decompose
the static pressure field acting on the RAE-2822 airfoil [3]. The interaction between an effective propeller
representative of a Cessna 172 [9] and a lifting surface was captured, and its implications were discussed. The steady
and incompressible flowfield was simulated using ANSYS Fluent 2021 R1 [14,15]. The RANS equations were
discretized and solved using a second-order Green Gauss cell-based method, and a single equation Spalart-Allmaras
Eddy-viscosity turbulence model was used. The development of the ALM used was verified using momentum theory
and ANSYS Fluent. An analysis of the PPF decomposition obtained declares that the propeller PPF is heavily
influential on the Euler drag coefficient 𝐶𝐶𝑑𝑑,𝐸𝐸 when within 𝑦𝑦0 = −1.5𝑚𝑚 of the RAE-2822 airfoil. In other words, the
fluid’s inertial effects are emphasized by the ALM, and the implications are apparent to 𝐶𝐶𝑑𝑑,𝐸𝐸 because of its nonlinear
dependence on 𝑣𝑣⃗.
When considering beyond the range of 𝑦𝑦0 = −1.5𝑚𝑚 (i.e., 𝑦𝑦0 = −3𝑚𝑚), 𝐶𝐶𝑑𝑑,𝜇𝜇 is subject to a small increase of 1 drag
count and 𝐶𝐶𝑑𝑑,𝐸𝐸 is subject to a significant increase of 15.5 drag counts. However, despite a large increase in 𝐶𝐶𝑑𝑑,𝐸𝐸 , 𝐶𝐶𝑑𝑑,𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝
takes away from 𝐶𝐶𝑑𝑑 and the net result is a total drag increase of 6.4 drag counts. It is worthy to note the limitations

19
introduced by a 2-D simulation. By nature, it is impossible to resolve the tip vortices that inevitably close the wake of
the ALM. As a result, both the magnitude of the injected body force and the velocity below the RAE-2822 airfoil is
slightly increased. This effect propagates throughout the data by means of the lift coefficient and PPFs. Still, the
present work is a canonical study where these inaccuracies are not of significant importance. It would be interesting
to scale the methods in this study to a three-dimensional simulation, where the actuator line model and its effects can
be further explored. Ultimately, the present study is a summary of the methods used to synthesize two computational
methods that both provide insight into complicated thrust drag aerodynamic interactions.

VII. Appendices

A. Flow Field Density and Operation Pressure Calculation

The density and operating pressure used in ANSYS Fluent is calculated below. The flow is incompressible; thus,
both the density and operating pressure are constant throughout the domain. Consider the following statements of
Mach number, Reynolds number, and the Ideal-Gas Law:

𝑈𝑈 (26)
𝑀𝑀 =
�𝛾𝛾𝛾𝛾𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟

𝜌𝜌𝜌𝜌𝜌𝜌 (27)
𝑅𝑅𝑅𝑅 =
𝜇𝜇

𝑃𝑃𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 = 𝜌𝜌𝜌𝜌𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 (28)

The parameter c is a reference chord length and is equal to one in this case. Furthermore, the velocity U is the horizontal
cruise velocity, 𝑉𝑉𝑐𝑐 m/s. The objective is to determine an expression for 𝑃𝑃 that depends on known quantities (i.e.,
𝑅𝑅𝑅𝑅, 𝜇𝜇, 𝑀𝑀, 𝛾𝛾, 𝑅𝑅, 𝑇𝑇, 𝑐𝑐). Simple algebraic simplification of Eq. 26-28 gives the following equations for density 𝜌𝜌 and P:

𝜇𝜇𝜇𝜇𝜇𝜇 (29)
𝜌𝜌 =
𝑀𝑀�𝛾𝛾𝛾𝛾𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟

(30)
𝜇𝜇𝜇𝜇𝜇𝜇 𝑅𝑅𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟
𝑃𝑃𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 = �
𝑀𝑀 𝛾𝛾

Equations 29-30 give a direct analytical solution for the density and operating pressure to be coded into ANSYS
Fluent. This analysis is possible because of the incompressible nature of the fluid and the assumption of an ideal gas.

B. MATLAB R2020b Post Processing Script

This appendix serves to outline the post-processing steps taken in this study. Specifically, the following figure
outline the post processing of the solution data from ANSYS Fluent.

20
21
 Fig 19. Script for post-processing ANSYS Fluent [14,15] solution data

Acknowledgments
I would like to thank Dr. Sven Schmitz for providing me the opportunity to develop this undergraduate thesis.
Furthermore, I would like to thank PhD candidate Pierce Hart for his immense help in familiarizing me with PPFs,
ALM, and CFD as a whole. Without these two, this work would not be possible, and I would not have uncovered my
deep-rooted interest in computational aerodynamic research.

References

[1] Menter, F.R., “Turbulence Modeling for Engineering Flows,” ANSYS, Inc., Canonsburg, PA, 2011

[2] Schaefer, P., Gampert, M., Goebbert, J.H., Wang, L., Peters, N., “Testing of Model Equations for the Mean Dissipation using
Kolmogorov Flows,” Flow Turbulence Combust, Vol. 85, 2010, pp. 225–243. https://doi.org/10.1007/s10494-010-9273-4.

[3] Selig, M.S., “RAE-2822 transonic airfoil,” UIUC Airfoil Coordinates Database, UIUC Applied Aerodynamics Group.,
Department of Aerospace Engineering at the University of Illinois at Urbana-Champaign., Urbana, IL, Mar. 2022.

[4] Sørensen, J. N., Shen, W. Z., "Numerical Modeling of Wind Turbine Wakes," Journal of Fluids Engineering, Vol. 124, No. 2,
June 2002, pp. 393-399. https://doi.org/10.1115/1.1471361.

22
[5] Chruchfield, M.J., Schreck, S.J., Martinez, L.A., Meneveau, C., Spalart, P.R., “An Advanced Actuator Line Method for Wind
Energy Applications and Beyond,” 35th Wind Energy Symposium, January 2017. https://doi.org/10.2514/6.2017-1998.

[6] Troldborg, N., "Actuator Line Modeling of Wind Turbine Wakes," Ph.D. Dissertation, DTU Department of Mechanical
Engineering, Technical University of Denmark, June 2008.

[7] Schmitz, S., "Drag Decomposition Using Partial Pressure Fields in the Compressible Navier-Stokes Equations," AIAA Journal,
Vol. 57, No. 5, 2019, pp. 2030-2038. https://doi.org/10.2514/1.J057701.

[8] Hart, P.L., Schmitz, S., "Application of Partial-Pressure Field Drag Decomposition to the ONERA M6 Wing," AIAA Aviation
2021 Forum, August 2021. https://doi.org/10.2514/6.2021-2555.

[9] Cessna Aircraft., “Pilot’s Operating Handbook and FAA Approved Airplane Flight Manual,” Cessna 172SPHUSR05, July 2004.

[10] Leishman, J.G., “Fundamentals of Rotor Aerodynamics,” Principles of Helicopter Aerodynamics Computational, 2nd ed.,
Cambridge Aerospace Series 12, Cambridge, MA, 2020, pp. 55-110.

[11] Schmitz, S., “Advanced Computational Methods,” Aerodynamics of Wind Turbines: A Physical Basis for Analysis and Design, 1st
ed., Wiley, New York, 2020, pp. 218–220.

[12] McIver, J., “Cessna Skyhawk II/100 Performance Assessment,” Temporal Images., Ormond, VIC, Australia, 2010.

[13] Schmitz, S., Jha, P., “Modeling the Wakes of Wind Turbines and Rotorcraft Using the Actuator-Line Method in an OpenFOAM -
LES Solver,” 69th AHS Forum. May 2013.

[14] ANSYS Fluent UDF Manual 2021R1, ANSYS, 2021.

[15] ANSYS Fluent User’s Guide 2021R1, ANSYS, 2021.

[16] Doig, G., Barber, T., Neely, A., Myre, D., “Aerodynamics of an aerofoil in transonic ground effect: Numerical study at full-scale
Reynolds numbers,” The Aeronautical Journal, Vol. 116, No. 1178, 2012, pp. 407-430.
https://doi.org/10.1017/S0001924000005297

23