MINIMIZATION OF EXIT FLOW NON-UNIFORMITIES IN SUPERSONIC NOZZLES

Abstract:
Supersonic nozzles, commonly known as convergent-divergent nozzles, are used to
accelerate pressurized fluids to speeds beyond that of sound in order to achieve thrust. They
are used in aircrafts, jets, rockets, turbines and even in some unconventional machining
processes such as Laser Beam Machining (LBM) and Abrasive Jet Machining (AJM). The flow
field entering the nozzle usually consists of a high pressure, high temperature fluid that
expands and cools as it accelerates along the nozzle. However, this expansion is not uniform
largely because of the friction that exists between the fluid and the nozzle walls, and the
nature of the inlet flow field with regards to the thermodynamic properties. This non-uniform
expansion leads to strong radial gradients of pressure, temperature, density and velocity that
are undesirable in many cases. Recent researches have shown that the nozzle geometry has
a significant impact on exit flow non-uniformities. The objective of this research is to minimize
the non-uniformities in flow at the exit section of supersonic nozzles by controlling the nozzle
geometry. By varying the nozzle geometry using a suitable set of parameters and simulating
the flow inside the nozzle for each case using ANSYS FLUENT, the exit flow properties can be
estimated. The parameter values and the estimated flow properties are used as input for
numerical optimization using Gravitational Search Algorithm (GSA) and Tabu Search
Algorithm (TSA) in MATLAB to further improve the solution. The best profiles obtained from
the numerical optimization are simulated using FLUENT for validation. The proposed research
will thus establish a family of non-dominating solutions, which will guide designers to choose
the right kind of nozzle with the right profile for the requisite application.

Background:
A fluid flow is said to be non-uniform if the flow properties like pressure, density,
velocity and temperature change with respect to space at a given instant of time. In other
words, the flow gradients are non-zero. In practice, almost all fluid flows exhibit certain
degrees of non-uniformity in flow properties. These non-uniformities are consequences of
inherent fluid characteristics such as friction and viscosity [7].
It is to be noted here that in the context of nozzles, the flow is axial and the term ‘non-
uniformity’ very specifically refers to the variation in flow properties along the radial
direction, i.e. normal to the direction of flow. The non-uniformities in the axial direction are
essential to achieve thrust and any efforts to minimize them would defeat the very purpose
of using a nozzle.
For flows involving pressure close to atmospheric pressure and temperature under
500K, the effects of such non-uniformities are minimal and are generally neglected [5].
However, under the high temperature, high pressure conditions commonly encountered in
supersonic nozzles, such non-uniformities in flow properties can have some adverse and
undesirable effects including

 Additional stresses on nozzle walls leading to premature failure [10]

 Damage to rotor blades on steam turbines [10]
 Lowered average exit thrust, leading to increased fuel/energy consumption and
reduced efficiency [2]
 Formation of swirls and vortices just outside the nozzle exit, thereby increasing the
exit pressure [5]
 Possibility of shock wave formation leading to premature expansion [5]
 Increased transitions between laminar and turbulent flows leading to energy losses
[9]
Non-uniformities in flows inside supersonic nozzles primarily occur due to the
following causes [10]:
1. Non-uniformities in the upstream region of the nozzle (i.e. from the exit of the
combustion chamber) due to inhomogeneous combustion or extremely rapid
combustion.
2. Pre-swirl existing in the combustion chamber
3. Insufficient mixing of fuel and oxidant
4. Wall friction, which directly depends on the nozzle profile

This research is targeted at the final cause. By adjusting the profile of the nozzle in a
subtle manner, wall friction can be minimized, which in turn will minimize the variation in
radial gradients of the flow properties [2].

Literature Review:
The effect of nozzle profile on flow non-uniformities was not discussed until the late
1980s. In 1989, a research team led by Doty was the first to infer the existence of a
relationship between the extent of non-uniformities upstream of a convergent-divergent
nozzle and the average exit thrust from the nozzle. The team compared the results from a
uniform inlet profile to those from a non-uniform inlet profile and found drastic differences,
highlighting the losses caused by the non-uniformities [2].
Later, in the same year, Doty’s team studied the effect of non-uniformities on exit
thrust in different types of nozzles including variants of bell nozzles and double bell nozzles.
They deduced that nozzles with smoother profiles suffered less non-uniformities [3].
Two years later, another research team led by Snelling identified the same effect on
scramjet engines. Scramjet engines contain a diffuser upstream of a nozzle. It was observed
that the pitching moment and the exit thrust improved significantly when both the diffuser
and the nozzle profiles were smoothened [4].
 The next major breakthrough was achieved in the new millennium, when Jackson and
Mullen, after considerable experimentation, recommended the use of flow straighteners
between combustion chambers and nozzles for relatively low velocity applications (Mach
Number between 0.8 and 1.3). The flow straighteners consist of vanes designed to allow only
axial flow through them. This arrangement could be especially useful in steam turbines where
potential damage to the rotor blades can be avoided. However, accelerated life test data
showed the limited reliability of flow straighteners and their need for frequent replacement.
This result shifted the emphasis back to nozzle geometry [8].
In 2012, a research team led by Yu carried out Reynolds-Averaged Navier-Stokes
(RANS) simulations to investigate the effect of nozzle geometry on the turbulence
characteristics of fluid flow through nozzles. They considered four different nozzle types for
their simulations: a simple baseline conical nozzle and three modified versions namely,
extended, grooved and ringed nozzles. Their simulations strongly correlated with actual
results from a physical model [11].
In the same year, Karla Quintao published a master thesis focusing on optimizing
nozzle profiles to minimize non-uniformities in density, Mach number and static temperature
at the exit section of the nozzle. 40 initial nozzle designs were generated randomly using a
FORTRAN code, and modelled and simulated in FLUENT. The results of simulation along with
the data were imported into the optimization package modeFRONTIER and optimized using
the Genetic Algorithm (GA). The fittest solutions offered by the GA were simulated in FLUENT
for validation. This thesis represents one of the most comprehensive researches in this field
[10].
In 2015, Jean-Baptiste Mulumba Mbuyamba published a dissertation regarding nozzle
design for a cold gas dynamic spray. While not directly related to rocket nozzles, the
dissertation considers de Laval nozzles for design. It also describes several theoretical
elements regarding compressible gas flow in a convergent-divergent nozzle as well as
methods to simulate and calculate specific parameters [12].

Research Gap:
From the literature survey, it was inferred that most researches targeted specific
nozzle applications and that there is no generic model that can be rendered suitable to a
variety of applications. Also, the models that have been specified did not take into account
complex phenomena such as swirls, vortices and transitional flows, which would necessitate
the use of 3D models and higher order turbulence equations. Additionally, no efforts have
been made to model and simulate different types of nozzles comparatively. This research is
aimed that bridging this very specific gap.

Research Objective:
To minimize the exit flow non-uniformities in different types of supersonic nozzles by
controlling the nozzle geometry and thereby develop a set of generic, non-dominant Pareto
solutions that will guide the selection of nozzle type and geometry depending upon the
intended requirement.
Methodology:
The methodology for experimentation and analysis are briefly detailed as follows:
1. Four broad categories of nozzles namely, conical, bell-shaped, double bell-shaped and
a hypothetical 6th order polynomial curve-shaped, are considered for this study. Also,
four variants in each category namely, baseline, extended, grooved and ringed are
considered, making a total of 16 different nozzle types.
2. The design parameters are defined for each type of nozzle. These parameters include
inlet and exit radius, inlet and exit wall angle, throat radius, length of
convergent/divergent portions, groove length, ring depth and intermediate bell
radius [10][11]. It is to be noted that not all parameters are applicable for all nozzle
types. To ensure wider suitability of the model, the throat radius is standardized as
unity and the other dimensions are scaled relative to the throat radius. Bounds are
defined for every parameter to ensure convergence of solutions [10].
3. 100 initial designs are generated for each nozzle type using SOBOL [1] sequence
generation technique to ensure a uniform distribution of data points over the
respective parametric intervals.
4. These designs are modelled in 3D using SOLIDWORKS package and exported to IGS
format.
5. These designs are imported into ANSYS FLUENT, meshed and simulated for 3 different
inlet conditions corresponding to different pressure and temperature ranges. Two
different turbulence models are used namely, RANS and Transition-SST-K-Ω models.
These models are selected despite their higher computational requirements as they
better represent the actual system, accounting for transitions between laminar and
turbulent flows [7].
6. The results of simulation are exported to MS Excel and the standard deviation of the
flow properties including pressure, density, temperature and velocity are determined
at the exit section of the nozzle [6].
7. These results are now imported into MATLAB and they act as the initial population
for the numerical optimization using GSA and TSA. 500 virtual designs are generated
in addition to the 100 initial designs for each nozzle type and the optimization is
performed until the terminating condition, i.e. Max. Norm. Std. Dev. < 0.5 is satisfied
[6].
8. The set of non-dominating virtual design solutions are identified. These designs are
modelled using SOLIDWORKS and simulated in FLUENT under their respective set of
inlet conditions to validate the results. Preliminary study with a limited number of
models revealed that the maximum deviation between the results of numerical
optimization and those of CFD simulation was 13.1%. Anticipating a higher degree of
accuracy with the increased number of models, a confidence level of 90% has been
set for accepting the results of the numerical optimization.
9. The validated set of non-dominating solutions for each type of nozzle and inlet
condition is established as the Pareto solution set, which will assist in the selection of
the appropriate nozzle type depending upon the intended application
References:

1. Sobol, I. M.: “Uniformly distributed sequences with an additional uniform property”,

USSR Computational Mathematics and Mathematical Physics (16) 236-242, 1976.

2. Doty, John H., “Performance Prediction and Design of Maximum Thrust Planar
Supersonic Nozzles Using a Flux-Difference-Splitting Technique”, PHD Dissertation.
Purdue University, Aug 1989.

3. Doty, John H., Thompson, H. Doyle, and Hoffman, Joe D., “Optimum Thrust Two-
Dimensional NASP Nozzle Study”, NASP CR1069, Nov 1989.

4. Snelling, Sandra L., “Effect of Non-Uniform Entrance Flow Profile on Hypersonic Nozzle
Pitching Moment”, Master Thesis. The Air Force Institute of Technology, Air University,
Dec 1991.

5. Gai, S.L., “Free piston shock tunnels: Developments and capabilities”, Progress in the
Aerospace Sciences 29, 1–41 (1992).

6. Deb, K. 2002, “Multi-Objective Optimization Using Evolutionary Algorithms”, John

Wiley & Sons, New York.

7. Cebeci, T., Shao, J.P., Kafyeke, F., and Laurendeau, E. 2005, “Computational Fluid
Dynamics for Engineers”, Springer Publications, NY

8. Jackson, D.R., and Mullen, O.D., “Comparison of Nozzles and Flow Straighteners for
Tank Waste Sluicing Applications”, U.S. Department of Energy, 2005.

9. Ahmadi, G. 2010, “Introduction to Turbulence Modeling”, Department of Mechanical

and Aeronautical Engineering, Clarkson University, Potsdam, NY 13699-5725.

10. Quintao, Karla K., " Design Optimization of Nozzle Shapes for Maximum Uniformity of
Exit Flow" [2012], FIU Electronic Theses and Dissertations, Paper 779.

11. Yu, Y., Shademan, M., Barron, R. M., and Balachandar R., “CFD Study of Effects of
Geometry Variations on Flow in a Nozzle”, Taylor and Francis Group, 2012.

12. Mbuyamba, Jean-Baptiste Mulumba, “Calculation and Design of Supersonic Nozzles

for Cold Gas Dynamic Spraying using MATLAB and ANSYS Fluent” [2013].