Professional Documents
Culture Documents
2020-3763
August 24-28, 2020, VIRTUAL EVENT
AIAA Propulsion and Energy 2020 Forum
B. Dequick∗ , M. Lefebvre†
Royal Military Academy, Avenue de la Renaissance 30, Brussels, 1000, Belgium
P. Hendrick‡
Université Libre de Bruxelles, Avenue F.D. Roosevelt 50, Brussels, 1050, Belgium
Hybrid rocket engines (HREs) are rocket engines that combine a solid fuel and a liquid
(or gaseous) oxidizer, or vice versa. A setup with a solid fuel is most commonly used. For
some years now, researchers at Université Libre de Bruxelles (ULB) are running tests with a
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
test bench HRE. This engine has a target thrust of 1 kN and uses a paraffin fuel with liquid
N2 O as oxidizer. A first computational fluid dynamics (CFD) simulation of this engine is
presented here. It is a steady-state RANS simulation, using the standard 𝑘-𝜖 turbulence model
together with the Eddy Dissipation Model (EDM) for the turbulence chemistry interaction
(TCI). Oxidizer and fuel enter the domain in a gaseous phase. An ambient area is included in
the computational domain as well. This avoids imposing a boundary condition at the nozzle
exit section, and therefore a more correct pressure profile can be formed. Furthermore, it
allows to compare and investigate the exhaust plume structure. The results are compared with
test firing data and show an average deviation of less than 10%. Also, a parametric study is
done regarding the oxidizer and fuel inlet temperatures and the oxidizer to fuel mass ratio (O/F
ratio). Using the CFD model, future work will aim on improving the engine’s performance by
focusing on the design influence of the post combustion chamber and nozzle.
I. Introduction
Copyright © 2020 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
A recent and detailed overview of the state of the art and current challenges with respect to HREs in which liquefying
fuels such as paraffin-based fuels are used, is provided by [4].
applies for the study presented in the current paper. Details are provided in section III.
A recent and detailed overview of the state of the art of CFD techniques for HREs with classical and liquefying fuels
such as paraffin-based fuels is provided by [6]. It is clear that improved CFD modeling, including model validation
using experimental data, is necessary.
1. Latest design
The ULB-HRE was first developed in 2010 and has meanwhile undergone several modifications. The engine is
designed to have a theoretical thrust of 1 kN and a burning time of 10 s. The latest design is shown in Fig. 1 [7], on
which four typical parts of a HRE can be recognized from left to right: the pre-chamber with the oxidizer injector and
ignition cartridge, the combustion chamber with a single port fuel grain, the post-chamber, and the nozzle.
2
The dimensions were obtained based on classical internal ballistics theory for hybrid propulsion. Details are
presented in [7]. The total internal axial length measured from the injector plate to the nozzle exit is ≈360 mm. At
the pre-chamber and post-chamber, the internal diameter is at its maximum of ≈130 mm. The fuel grain consists
out of pure paraffin and has a single port circular configuration. Its total length is a relatively short 110 mm and the
initial port diameter varies between tests, ranging from 20 to 50 mm. The nozzle throat and exit diameter are 22 and
50 mm, respectively. The oxidizer is liquid N2 O, which is fed through the injector plate after which it atomizes in the
pre-chamber.
B. Main purpose
The research conducted with the ULB-HRE has mainly been experimental so far. The development of a numerical
model of the entire engine is therefore desirable, such that it adds a predictive capacity to the research if it proves to be
sufficiently reliable.
The main purpose of the current paper is to establish a first CFD model of the ULB-HRE. The available experimental
results are used for comparison purposes. This allows to evaluate the accuracy of the model.
Future work will focus on the post-combustion chamber and nozzle design influence, in order to propose potential
engine performance improvements.
A. Computational domain
The computational domain is 2D axisymmetric with respect to the main flow direction and includes the interior of
the engine and the exhaust plume region. The dimensions of the latter are based on the general size of the observed
plume during firing tests. The simulations are steady-state and so the fuel port diameter is fixed at 70 mm, which
represents a typical intermediate value during engine operation. Ignition features are not included, as this is not of
interest in the current work. The geometry of the circular plate and bolts which keep the nozzle in place are ignored as
well, since their impact on the exhaust plume structure is assumed to be negligible. Figure 2 shows an overview.
B. Mesh
A series of different meshes were created in order to study mesh convergence and the influence of inflation layer
cells near the boundaries. All meshes are structured-like and have quadrilateral cells. Triangular cells do appear in
some areas, but they represent less than 0.02% in number.
3
Fig. 2 Computational domain
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
𝜌𝐶 𝜇0.25 𝑘 0.5
𝑃 𝑦𝑃
𝑦★ ≡ (1)
𝜇
Here, 𝜌 and 𝜇 are the fluid density and dynamic viscosity, 𝑘 𝑃 is the turbulence kinetic energy in P, and 𝑦 𝑃 is the
actual distance from P to the wall. 𝐶 𝜇 is a constant equal to 0.09.
As said before, NWTs generally require 𝑦★ to be in a certain range. Depending on the NWT, 𝑦★ should typically be
less than 1 or greater than 30. However, the NWT that will be chosen here (see III.C.2) is claimed to be insensitive to
the values of 𝑦★. If this is correct, results should not depend on the thickness (within reasonable limits) of the cell
layers near the walls. In the current paper, these results are represented by the obtained global operating values such
as chamber pressure and nozzle exhaust speed. They depend less on what is happening in the near-wall region than
parameters such as drag coefficients. Therefore it is predicted, especially with the chosen insensitive NWT, that the
operating parameters are not strongly affected by the near-wall mesh resolution.
This will now be checked by comparing the results obtained with 3 meshes, all with equal cell sizes of ≈ 1 mm.
The only difference between these meshes are the values of 𝑦★ and the number of cell layers in the near-wall regions.
Figure 3 shows a top-left corner of each mesh to visualize the inflation layers.
Fig. 3 Top-left corner of three meshes with different levels of near-wall refinement.
The 3 simulations are done with boundary conditions that fall within the range of those that will be used for the
actual simulations of the experiments which are discussed in section IV.
The results are compared in two ways. First, by looking at the resulting global operating values. They are summarized
together with some mesh properties in Table 1. Next, by looking at the obtained field values at two sections of the engine:
a section in the post-combustion chamber and the nozzle exit section. The obtained field values are shown in Fig. 4.
4
Table 1 Comparison of global operating parameters obtained with 3 meshes with different levels of near-wall
refinement.
Fig. 4 Profile of axial velocity, temperature and static pressure along a section in the post-chamber and along
the nozzle exit section, for 3 levels of near-wall mesh refinement. The legend shown in (a) is valid for all charts.
5
From Table 1 it can be concluded that overall differences for typical operating parameters between the three
meshes are small, especially between mesh 2 and mesh 3, where differences between 𝑃post-ch , 𝑇 post-ch and 𝑣 nozzle exit are
respectively 0.14%, 1.3%, and 0.29%. This is confirmed in Fig. 4, from which it is clear that mesh 2 and mesh 3 result
in very similar field value profiles.
Therefore, in the context of the current study and modeling approach presented later in section III.C.2, a moderate
near-wall mesh refinement is considered to be sufficient.
2. Mesh convergence
Different series of meshes were created during the initial mesh design process. The differences between these series
are essentially related to the near-wall mesh refinement (discussed in section III.B.1). Within each series, meshes only
differ in global cell size.
In order to obtain mesh-independent results, mesh convergence must be reached. This means refining the mesh cell
sizes until there is no significant change in the solution when all other parameters are kept unchanged. Within each
series, consecutive mesh refinements were done and the resulting engine operating values were monitored.
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
Figure 5 shows the evolution of the chamber pressure for 5 series of meshes, all with a single representative set
of boundary conditions. Initially, a first order differencing scheme was used for the turbulence field values (𝑘 and
𝜖). For these mesh series (shown in gray), a high overall refinement was necessary to observe some convergence just
above 22 bar. On the other hand, when using second order differencing for the turbulence field values (series shown in
black), the chamber pressure always falls within the range of 22 to 23 bar, with a general trend towards just above 22 bar
for finer meshes. Any fluctuations are solely related to the mesh, since all simulations have converged perfectly, as
explained in section III.C.4.
Based on these observations, a global mesh size of 1 mm × 1 mm is considered to be sufficient when using second
order differencing for all field values. This mesh resolution provides results that differ only 1% from those of the
0.4 mm × 0.4 mm mesh (for series with near-wall mesh refinement). The final mesh used for the CFD simulations in
section IV is shown in Fig. 6. It has 120602 cells.
D
20 E
19
Fig. 5 Mesh convergence for chamber pressure for 5 series of meshes, each with increasing global refinement
from right to left. The impact of the differencing scheme for the turbulence field values becomes clear.
6
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
C. Flow modeling
7
RANS is the most common way by which the turbulent flows inside HREs are modeled today [4, 6]. In the associated
equations, any fluctuations caused by turbulence are time averaged and they appear as an extra stress term which needs to
be modeled by some turbulence model. This is why RANS simulations provide a steady-state solution which represents
an average flow (when boundary conditions are constant in time). Running a series of steady RANS simulations can
represent snapshots of an evolving flow, such as in a HRE with increasing port diameter during operation. Some
examples of such a quasi-steady approach include [6, 18, 20] and many more. In this paper however, only one fuel port
diameter is considered to establish the first CFD model of the ULB-HRE. Many researches have obtained representative
results by using RANS to model paraffin-fueled HREs. Recent examples include [6, 21–24]. The computational cost is
low compared to LES and DNS, and a 3D mesh is not required. Instead, equations can be solved in a 2D axisymmetric
domain. With this in mind, it seems that the RANS approach is a good choice to establish a first CFD model of the
ULB-HRE.
As said before, the RANS equations need to be closed by some turbulence model. Commonly used models are the
𝑘-𝜖 [25], 𝑘-𝜔 [26], 𝑘-𝜔 SST [27] and Spalart-Allmaras [28] models. For now, the model used for this study is the 𝑘-𝜖
model, as the use of other models do not yet provide realistic results. In some cases, no diffusion flame is formed at all
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
in the (relatively short) combustion chamber. Along with a turbulence model, a proper wall treatment must be chosen
to take into account the viscous sublayer in which turbulence effects are blocked because of the presence of the wall.
The Menter-Lechner near-wall treatment is used here, as it is presented as a 𝑦★-independent model by the software
developer [13].
Next, a combustion model has to be chosen. Two common approaches are based on which factor is considered to
limit the chemical reaction rate: the reaction kinetics or the turbulent mixing. If the mixing of the reactants is considered
fast, then the reaction progress is limited by some finite-rate reaction mechanism, described by Arrhenius laws. In this
case, reaction progress depends on local averaged field values and does not take into account any turbulence-chemistry
interaction (TCI). On the other hand, if the chemical reactions are considered fast compared to the transport phenomena,
then the reaction progress is limited by the turbulent mixing. One way to implement this TCI is the Eddy Dissipation
Model (EDM) [29], in which the different species source terms are calculated from the turbulent kinetic energy, the eddy
dissipation rate and the species mass fractions. This implies that transport equations for each species must be solved,
but it avoids including complex chemical kinetics. The EDM is typically suited for a one-equation global reaction, as
the reaction rates for multiple reactions would all be based on the same turbulence field values, which would lead to
incorrect results [13]. The model, which is of the type mixed-is-burnt, is one of the popular choices for simulating the
turbulent non-premixed combustion in HREs. Recent examples include [21, 23, 30]. The EDM will be applied in the
current study as well. An alternative way to model combustion with TCI, thus also assuming fast chemistry, would be
the mixture fraction approach [31]. In this case, no transport equations for the different species are solved. Instead,
transport equations for the averaged mixture fraction 𝑓 = 𝑓 (𝑂/𝐹) and it’s variance 𝑓 02 are solved. By assuming a
Probability Density Function (PDF) coupled with chemical equilibrium, a look-up table can be constructed from which
the temperature and species mass fractions are obtained. Examples include [6, 11, 18, 32]. Like the EDM, it is also a
mixed-is-burnt model.
To model the fuel entering the combustion chamber, a fuel mass source is added in a very thin layer of cells adjacent
to the axial grain wall, all at a fixed temperature of 740 K, which is the average of Tboiling and Tcritical of paraffin. The
grain sides are assumed non-reactive. Fuel mass flows are obtained from experimental data, and thus the regression
rate is not solved and there is no gas-surface interaction (GSI). The low carbon gasses resulting from the pyrolysis of
paraffin fuel consist mainly out of ethylene (C2 H4 ) and a small fraction of hydrogen [33] of about ≈ 0.5 m% when
C28 H58 is considered as the paraffin wax hydrocarbon [11]. As a first approximation, the fuel mass source is set to 100%
of gaseous ethylene. Future work may include the liquid fuel droplet entrainment such as presented in [11]. For a DNS
of the liquid film instabilities the reader is referred to [34].
The inlet oxidizer spray is simulated as gaseous N2 O entering the domain at a specified mass flow rate and
temperature. It is thus assumed that the liquid oxidizer atomizes and vaporizes upon entering the flow domain. This is
a common approach, recent examples include [6, 21, 22, 24, 32]. Because of the 2D axisymmetric domain, all real
inlet nozzle geometries are reduced to a disk or annulus shaped area, although the injection velocity vectors can differ
from the normal direction. In the current study, simulations where the oxidizer is entirely injected as ∅ 30 𝜇m liquid
droplets of N2 O at 280 K, suggest that they evaporate completely in the pre-chamber. These simulations however are not
presented here, as they are preliminary. An example of a CFD model considering such a multiphase oxidizer injection is
presented in [23].
8
The ambient area is bounded by a wall on the left (next to the nozzle exit), a 1 atm pressure inlet (298 K, air) at the
top, and a 1 atm pressure outlet on the right. The pressure inlet is due to the exhaust plume which draws in air from the
sides. With this setup, no reversed flows are present in the converged solution. Furthermore, the presence of an ambient
area allows the formation of a more correct nozzle exit pressure profile.
Heat losses through the motor housing are ignored, and thus, at this stage, all walls are set to be adiabatic except for
the grain surface wall, which is kept at a fixed temperature as mentioned before. Furthermore, a no slip condition is valid
at every wall. Radiation effects are ignored for the sake of simplicity. It is assumed that their contribution is negligible.
Figure 7 shows a summary of this section.
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
3. Equations
This section provides an overview of the fluid equations that apply for the modeling strategy as discussed in section
III.C.2.
Starting from an arbitrary, stationary, and non-deformable control volume, one can derive a general transport
equation for any extensive property 𝜙 (some property per unit mass),
𝜕
(𝜌𝜙) + ∇ · (𝜌®𝑣 𝜙) = ∇ · (Γ∇𝜙) + 𝑆 (2)
𝜕𝑡
which is known as the convection-diffusion equation. From left to right there is a non-steady term, a convective
term, a diffusive term, and a source term. Γ is known as the diffusivity. If the transported property is mass, then 𝜙
equals 1 (mass per unit mass), Γ equals zero, 𝑆 equals zero and Eq. 2 becomes
𝜕𝜌
+ ∇ · (𝜌®𝑣 ) = 0 (3)
𝜕𝑡
which is known as the continuity equation (conservation of mass). If the transported property is momentum, then 𝜙
equals 𝑣® (momentum per unit mass), and Eq. 2 becomes
𝜕
(𝜌®𝑣 ) + ∇ · (𝜌®𝑣 𝑣®) = ∇ · (Γ∇®𝑣 ) + 𝑆 (4)
𝜕𝑡
As surface forces on a fluid element, represented by a stress tensor 𝜏, ¯ relate to diffusion of momentum, and as body
® provide a source of momentum, it can be shown that Eq. 4 becomes
forces (𝑏)
𝜕
(𝜌®𝑣 ) + ∇ · (𝜌®𝑣 𝑣®) = ∇ · 𝜏¯ + 𝜌 𝑏® (5)
𝜕𝑡
which is known as Navier’s equation of equilibrium, valid for all fluids. For Newtonian, isotropic, homogeneous,
and Stokian fluids, 𝜏¯ becomes
2
𝜏¯ = −𝑃𝐼 + 𝜇 (∇®𝑣 + ∇®𝑣𝑇 ) − ∇ · 𝑣® 𝐼 (6)
3
9
and in this case Eq. 5 is known as the Navier-Stokes equations, as one equation for each component of 𝑣® can be
written. As explained in section III.C.2, to model turbulence, the Navier-Stokes equation is averaged in time which
leads to an extra stress term. This term is modeled using the Boussinesq Hypotheses. In this case Eq. 5 becomes
𝜕 𝑇 2 2
(𝜌®𝑣 ) + ∇ · (𝜌®𝑣 𝑣®) = −∇𝑃 + ∇ · (𝜇 + 𝜇𝑡 ) (∇®𝑣 + ∇®𝑣 − ∇ · 𝑣® 𝐼) + ∇ · (−𝜌𝑘 𝐼) (7)
𝜕𝑡 3 3
in which the turbulent viscosity 𝜇𝑡 and the turbulent kinetic energy 𝑘 are modeled via the standard 𝑘-𝜖 turbulence
model. Therefore, two additional transport equations are solved: one for 𝑘 and one for 𝜖 (turbulent dissipation rate).
𝜕 𝜇𝑡
(𝜌𝑘) + ∇ · (𝜌®𝑣 𝑘) = ∇ · (𝜇 + )∇𝑘 + 𝐺 𝑘 + 𝜌𝜖 (8)
𝜕𝑡 𝜎𝑘
𝜖2
𝜕 𝜇𝑡 𝜖
(𝜌𝜖) + ∇ · (𝜌®𝑣 𝜖) = ∇ · (𝜇 + )∇𝜖 + 𝐶1𝜖 𝐺 𝑘 + 𝐶2𝜖 𝜌 (9)
𝜕𝑡 𝜎𝜖 𝑘 𝑘
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
in which 𝐺 𝑘 represents the generation of turbulence kinetic energy due to the mean velocity gradients, and 𝜎𝑘 and 𝜎𝜖
are the turbulent Prandtl numbers for 𝑘 and 𝜖, respectively. 𝐶1𝜖 and 𝐶2𝜖 are constants: 𝐶1𝜖 = 1.44 and 𝐶2𝜖 = 1.92.
The turbulent viscosity 𝜇𝑡 is calculated as
𝑘2
𝜇𝑡 = 𝜌𝐶 𝜇 (10)
𝜖
in which 𝐶 𝜇 is a constant equal to 0.09.
Furthermore, the transport equation is also solved for energy, from which the temperature field can be obtained.
Equation 2 then becomes
𝜕
(𝜌ℎ) + ∇ · (𝜌®𝑣 ℎ) = ∇ · (𝜅∇𝑇) + 𝑆 ℎ (11)
𝜕𝑡
in which ℎ is the sensible enthalpy, 𝜅 the thermal conductivity, and 𝑆 ℎ the heat source (for example, due to chemical
reactions).
To model the chemical reactions and the TCI, the EDM has been chosen. As a first approximation at this stage of the
CFD study, a simple global chemical reaction model has been considered:
𝜕 ∇𝑇
(𝜌𝑌𝑖 ) + ∇ · (𝜌®𝑣𝑌𝑖 ) = ∇ · 𝜌𝐷 𝑖,𝑚 ∇𝑌𝑖 + 𝐷 𝑇 ,𝑖 + 𝑅𝑖 (13)
𝜕𝑡 𝑇
Here, 𝐷 𝑖,𝑚 is the mass diffusion coefficient and 𝐷 𝑇 ,𝑖 is the thermal diffusion coefficient for species 𝑖 in the mixture.
𝑅𝑖 is the net rate of production of species 𝑖 by the chemical reaction. For the EDM, 𝑅𝑖 is the smallest of the two
following equations.
00 0 𝜖 𝑌𝑅
𝑅𝑖 = (𝜈𝑖 − 𝜈𝑖 )M𝑖 𝐴𝜌 min 0 (14)
𝑘 𝑅 𝜈𝑖 M 𝑅
Í !
00 0 𝜖 𝑃 𝑌𝑃
𝑅𝑖 = (𝜈𝑖 − 𝜈𝑖 )M𝑖 𝐴𝐵𝜌 Í 00 (15)
𝑘 𝑗 𝜈𝑗 M𝑗
Here, 𝜈𝑖0 and 𝜈𝑖00 are the reaction coefficients for reactant 𝑖 and product 𝑖, respectively. M is the molecular weight, 𝑌𝑅
is the mass fraction of any reactant and 𝑌𝑃 is the mass fraction of any product. 𝐴 and 𝐵 are model constants equal to 4
and 0.5 respectively.
10
From basic rocket propulsion theory, the equations for thrust (𝐹) and specific impulse (𝐼sp ) are also of interest.
𝐹 = ( 𝑚𝑣)
¤ exit + (𝑃exit − 𝑃ambient ) 𝐴exit (16)
𝐹
𝐼sp = (17)
𝑚¤ exit 𝑔0
in which the subscript exit relates to the nozzle exit section.
In this way, highly converged solutions are obtained, with all residuals (scaled) below 10−6 and a series of monitored
field values leveling out ’completely’. The deviation between mass inflow and mass outflow is always less than 0.001%.
IV. Results
Fig. 8 Overview of the 4 types of showerhead injectors used during the experiments.
(a) SH1: 11× ∅ 1.4 mm (b) SH2: 11× ∅ 1.9 mm (c) SH3: 21× ∅ 1.4 mm (d) SH4: 71× ∅ 0.8 mm for SH4
.
The space-time averaged results of the SH experiments are summarized in Table 2. The reader is referred to [35] for
details about how these values were obtained.
11
Table 2 Summary of experimental results with 4 different showerhead injectors, used for comparison with the
CFD results [35].
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
B. Numerical results
12
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
(a) (b)
(c)
Fig. 9 Comparison between experimental and numerical chamber pressure (a), thrust (b),
and specific impulse (c).
13
(a)
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
(b)
Fig. 10 Contours of temperature (a) and Mach number (b) for the simulation of test SH1-01.
14
Chamber pressure Chamber pressure
23.0 23.0
22.9 22.9
22.8 22.8
Chamber pressure [bar]
Fig. 12 Charts showing how the chamber pressure is affected by the inlet temperatures.
2. Parametric study
This section shows how the model responds to the oxidizer and fuel inlet temperature (Fig. 12), and mass ratio (O/F)
(Fig. 13). The monitored field value is the chamber pressure. The following observations are made.
Figure 12a. Within the range of simulated oxidizer inlet temperatures (220 − 600 K), the maximum and minimum
chamber pressure only differ 0.25 bar. A minimum is observed in the region of 400 K. In reality, the oxidizer inlet
temperature can be controlled in the sense that the storage tanks are usually at ambient temperature. Small temperature
variations in the feeding lines are assumed to be negligible. The simulated chamber pressure difference between, for
example, −20 ◦ C (253 K) and 40 ◦ C (313 K) is only 0.1 bar or 0.48%.
Figure 12b. As the fuel inlet temperature is increased, the chamber pressure increases as well. Assuming that in
reality, the temperature of the expelled gasses from the paraffin wax pyrolysis process lies somewhere within the boiling
and critical temperature region (700 − 780 K [11]), the model introduces a chamber pressure uncertainty of 0.051 bar or
0.23%. Note that the fuel gasses inlet temperature is of course not directly controllable during experiments.
Figure 13 shows how the modeled chamber pressure is affected by the oxidizer to fuel mass flow ratio. The total
mass flow rate is kept constant at 0.7 kg/s, a typical value for the SH4-injector test runs. So, for example, when the O/F
ratio is 1, both oxidizer and fuel mass flow rates are set to 0.35 kg/s. According to Eq. 12, the stoichiometric O/F ratio
is 6.57. This is consistent with Fig. 13, on which the modeled chamber pressure reaches a maximum for an O/F ratio
between 5 and 7. The experimental O/F ratios as shown in Table 2 are lower than this, ranging from 2.4 to 3.6. This
is possibly because in reality, an important part of the fuel can leave the engine as partially unburnt droplets due to
entrainment effects, contributing far less to the gaseous fuel phase in the combustion chamber. If this is true, keeping in
mind the range of the model’s optimal O/F ratio (5 − 7) and the range of the experimental O/F ratio (2.4 − 3.6), the mass
percentage of entrained fuel during the experiments can be estimated to range from 28% to 66%. These estimations can
be realistic values according to [3], taking into account the experimental port mass fluxes considered in this study of
min. 98 𝑚𝑘𝑔2 𝑠 at the upstream port section, and max. 184 𝑚𝑘𝑔2 𝑠 at the downstream port section.
15
24
23
22
Fig. 13 Chart showing how the chamber pressure is affected by the O/F mass ratio.
Furthermore, a parametric study has been carried out to monitor the model’s sensitivity to the oxidizer and fuel
inlet temperatures. Within a broad, but physically realistic range of inlet temperatures for both fuel and oxidizer, the
differences in operating chamber pressures stay below 0.1 bar or 0.5%. Therefore it can be concluded that the impact
of these parameters is relatively limited. Also, a parametric study of the inlet oxidizer to fuel mass flow ratio (O/F
ratio) has been done in order to determine which ratio leads to a maximum chamber pressure in the model. Simulations
show a maximum chamber pressure for an O/F ratio between 5 and 7, which is consistent with the stoichiometric O/F
ratio of 6.57. Looking at the lower experimental O/F ratios which are between 2.4 and 3.6, the experimental fuel mass
entrainment is estimated to range from 28% to 66%.
Elements which can be explored in the future model include the following.
To model entrained fuel droplets, a spray of liquid paraffin originating from the grain surface can be added. The
droplet temperature could be set to 515 K, which is the average of Tmelting and Tboiling of paraffin. Both phases in the
flow are coupled as the droplet size cannot be ignored. The two-phase flow is then solved in an Eulerian-Lagrangian
framework. This strategy has also been adopted by [11], although in this case, the gaseous fuel entered the domain
through a velocity inlet rather than as a mass source.
To improve the oxidizer inlet modeling, other inlet geometries can be considered, as well as a two-phase flow such
as proposed above for the fuel entrainment. Switching from an axisymmetric 2D to a full 3D computational domain
would allow to model the injector’s geometry exactly.
The chemical reaction can be altered to a more precise version, or, the entire combustion model can be altered by,
for example, switching to a mixture fraction approach instead of using the Eddy Dissipation Model (EDM).
The CFD model will be used for future parametric studies, which will definitely include modifying the dimensions
of the post combustion chamber and nozzle. As this will influence the exhaust plume structure, the use of a thermal
camera during the experiments can be interesting for comparison purposes with the CFD model’s temperature contours.
16
References
[1] Sutton, G. P., and Biblarz, O., Rocket Propulsion Elements, 9th ed., 2017, pp. 5–9.
[2] Altman, D., and Holzman, A., Fundamentals of Hybrid Rocket Combustion and Propulsion, 2007, Progress in Astronautics and
Aeronautics, Vol. 218, Chap. 1.
[3] Karabeyoglu, M. A., Cantwell, B. J., and Altman, D., “Development and Testing of Paraffin-based Hybrid Rocket Fuels,” 37th
AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, Salt Lake City, UT, July, 2001.
[4] Leccese, G., Cavallini, E., and Pizzarelli, M., “State of Art and Current Challenges of the Paraffin-based Hybrid Rocket
Technology,” AIAA Propulsion and Energy 2019 Forum, Indianapolis, IN, August, 2019.
[5] Ferziger, J. H., and Perić, Computational Methods for Fluid Dynamics, 1999, Chap. 2.
[6] Di Martino, G. D., Carmicino, C., Mungiguerra, S., and Savino, R., “The Application of Computational Thermo-Fluid-Dynamics
to the Simulation of Hybrid Rocket Internal Ballistics with Classical or Liquefying Fuels: A Review,” Aerospace, Vol. 6,
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
[7] Bouziane, M., Bertoldi, A. E. M., Milova, P., Hendrick, P., and Lefebvre, M., “Development and Testing of a Lab-scale
Test-bench for Hybrid Rocket Motors,” 2018 SpaceOps Conference, Marseille, France, May, 2018.
[8] Bouziane, M., Bertoldi, A. E. M., Dahae, L., Milova, P., Hendrick, P., and Lefebvre, M., “Design and Experimental Evaluation
of Liquid Oxidizer Injection System for Hybrid Rocket Motors,” 7th European Conference for Aeronautics and Space Sciences
(EUCASS), Milan, Italy, May, 2017.
[9] Bouziane, M., Bertoldi, A. E. M., Dahae, L., Milova, P., Hendrick, P., and Lefebvre, M., “Experimental Investigation of
Injectors Design and Their Effects on 1kN Performance Hybrid Rocket Motor,” 69th International Astronautical Congress
(IAC), Bremen, Germany, October, 2018.
[10] Bouziane, M., Bertoldi, A. E. D., Milova, P., Hendrick, P., and Lefebrve, M. H., “Performance Comparison of Oxidizer Injectors
in a 1-kN Paraffin-Fueled Hybrid Rocket Motor,” Aerospace Science and Technology, Vol. 89, 2019, pp. 392–406.
[11] Milova, P., Blanchard, R., and Galfetti, L., “A Parametric Study of the Effect of Liquid Fuel Entrainment on the Combustion
Characteristics of a Paraffin-based Hybrid Rocket Motor,” 6th European Conference for Aeronautics and Space Sciences
(EUCASS), Kraków, Poland, July, 2015.
[12] Bredberg, J., “On the Wall Boundary Condition for Turbulence Models,” Chalmers University of Technology, Department of
Thermo and Fluid Dynamics, Göteborg, Sweden, 2000.
[13] ANSYS Fluent Theory Guide (2019 R3), ANSYS, Inc., 2019, Chaps. 4, 7.
[14] Marxman, G. A., Wooldridge, C. E., and Muzzy, R. J., “Fundamentals of Hybrid Boundary Layer Combustion,” AIAA
Heterogeneous Combustion Conference, Palm Beach, FL, December, 1963.
[15] Muzzy, R. J., “Schlieren and Shadowgraph Studies of Hybrid Boundary-Layer Combustion,” AIAA Journal, Vol. 1, Number 9,
1963.
[16] Chiaverini, M. J., and Kuo, K. K., Fundamentals of Hybrid Rocket Combustion and Propulsion, 2007, Progress in Astronautics
and Aeronautics, Vol. 218, pp. 603–606.
[17] Waxman, B. S., Cantwell, B. J., and Zilliac, G., “Effects of Injector Design and Impingement Techniques on the Atomization of
Self-Pressurizing Oxidizers,” 48th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, Atlanta, GA, July, 2012.
[18] Di Martino, G. D., Malgieri, P., Carmicino, C., and Savino, R., “A Simplified Computational Fluid-Dynamic Approach to the
Oxidizer Injector Design In Hybrid Rockets,” Acta Astronautica, Vol. 129, 2016.
[19] Echekki, T., and Mastorakos, E. (eds.), Turbulent Combustion Modeling, 2011, Fluid Mechanics and its Applications, Vol. 95,
pp. 28–34.
[20] Bianchi, D., Nasuti, F., and Carmicino, C., “Hybrid Rockets with Axial Injector: Port Diameter Effect on Fuel Regression Rate,”
Journal of Propulsion and Power, Vol. 32, Number 4, July, 2016.
[21] Di Martino, G. D., Mungiguerra, S., Carmicino, C., and Savino, R., “Two-Hundred-Newton Laboratory-Scale Hybrid Rocket
Testing for Paraffin Fuel-Performance Characterization,” Journal of Propulsion and Power, Vol. 35, Number 1, 2019.
17
[22] Leccese, G., Bianchi, D., and Nasuti, F., “Modeling and Simulation of Paraffin-Based Hybrid Rocket Internal Ballistics,” AIAA
Propulsion and Energy 2018 Forum, Cincinnati, OH, July, 2018.
[23] Paccagnella, E., Gelain, R., Barato, F., Pavarin, D., van den Berg, P., and Barreiro, F., “CFD Simulations of Self-Pressurized
Nitrous Oxide Hybrid Rocket Motors,” AIAA Propulsion and Energy 2018 Forum, Cincinnati, OH, July, 2018.
[24] Bianchi, D., Nasuti, F., and Delfini, D., “Modeling of Gas-Surface Interface for Paraffin-Based Hybrid Rocket Fuels in
Computational Fluid Dynamics Simulations,” Progress in Propulsion Physics, Vol. 11, 2019.
[25] Jones, W. P., and Launder, B. E., “The Prediction of Laminarization with a Two-Equation Model of Turbulence,” Int. Journal
for Heat and Mass Transfer, Vol. 15, 1972.
[26] Wilcox, D. C., “Reassessment of the Scale-Determinig Equation for Advanced Turbulence Models,” AIAA Journal, Vol. 26,
Number 11, 1988.
[27] Menter, F. R., “Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications,” AIAA Journal, Vol. 32,
Number 8, 1994.
Downloaded by CARLETON UNIVERSITY on August 23, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-3763
[28] Spalart, P. R., and R, A. S., “A One-Equation Turbulence Model for Aerodynamic Flows,” AIAA 30th Aerospace Sciences
Meeting and Exhibit, Reno, NV, January, 1992.
[29] Magnussen, B. F., and Hjertager, B. H., “On Mathematical Modeling of Turbulent Combustion With Special Emphasis on Soot
Formation and Combustion,” 1977.
[31] Sivathanu, Y. R., and Faeth, G. M., “Generalized State Relationships for Scalar Properties in Nonpremixed Hydrocarbon/Air
Flames,” Combustion and Flame, Vol. 82, 1990.
[32] Di Martino, G. D., Mungiguerra, S., Carmicino, C., and Savino, R., “Computational Fluid-dynamic Simulations of the Internal
Ballistics of Hybrid Rocket Burning Paraffin-based Fuel,” AIAA Propulsion and Energy 2018 Forum, Cincinnati, OH, July,
2018.
[33] Karabeyoglu, M. A., Cantwell, B. J., and Stevens, J., “Evaluation of the Homologous Series of Normal Alkanes as Hybrid
Rocket,” 41st AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, Tucson, AZ, July, 2005.
[34] Adachi, M., and Shimada, T., “Liquid Films Instability Analysis of Liquefying Hybrid Rocket Fuels Under Supercritical
Conditions,” AIAA Journal, Vol. 53, Number 6, 2015.
[35] Bouziane, M., “Influence of the Oxidizer Injector Design on the Performance of a 1-kN Paraffin-Fueled Hybrid Rocket Motor,”
Ph.D. thesis, Université Libre de Bruxelles, Brussels, Belgium, 2019.
18