You are on page 1of 112

Aerodynamic Design of the NASA

Rotor 67 for Non Uniform Inflow Due


to Boundary Layer Ingestion

Master Thesis Report

Faculty of Aerospace Engineering

Department of Energy Technology

Done By:
Tan Yi Yun Raynold

Thesis Supervisors:
Dr Matteo Pini
Dr. Arvind Gangoli Rao
Dr. Bjorn Laumert
Ir. Salvatore Vitale
Master Thesis Report

Executive Summary

As the demand to improve the fuel efficiency of current commercial aircraft increases, new commer-
cial airliner concepts such as the Blended Wing Body has been researched on and studied in various
aspects over the years as an efficient alternative to the conventional transport configuration. One
particular aspect of the Blended Wing Body is the use of the propulsive fuselage concept. In this
concept, the fuselage boundary layer is ingested by the engine and this is aimed at producing benefits
such as improved fuel efficiency, reduced ram drag as well as lower structural weight of the engine.
During the ingestion process, the low momentum boundary layer is re-energized by the propulsion
system before exiting into the atmosphere. In this way, the ingested flow does not contribute to the
wake deficit and hence, the overall drag of the aircraft is reduced. Since thrust equal drag in steady
and level flight, and power is equal to thrust multiplied by velocity, the reduction in drag implies a
reduction in the power required to drive the vehicle.

In essence, the ingestion of the boundary layer which leads to a lower inlet stagnation pressure rep-
resents a direct thermodynamic penalty. However, the momentum deficit captured by the engine
represents a drag reduction to the aircraft. In this way, the propulsion system performance suffers a
decrease in engine efficiency while the aircraft drag is reduced in proportion to the amount of boundary
layer flow that is ingested. Therefore, a trade-off exists between the increase in aircraft drag reduction
and the decrease in engine performance as more boundary layer is consumed. Another important
concern is the significant flow distortion which can lead to increased vibration and fatigue of the fan
and compressor blades in particular. This flow distortion is characterised by the distortion coefficient,
a standard widely used in the aircraft engine industry. While it was found that the ingestion of the
boundary layer can provide a decrease in fuel burn of several percentages, the benefits of boundary
layer ingestion have shown to be very sensitive to the magnitude of the fan and duct losses. Hence,
it is crucial that fan designers are able to design new rotor blades that are able to withstand theses
flow distortion while ensuring that engine performance degradation is kept to a minimum in order to
maximise the overall gain in fuel efficiency.

The main aim of this research is therefore to understand and analyse the rotor performances un-
der both uniform and non uniform inflow condition. This will then provide insights into the main
fluid mechanism affecting rotor performances under such conditions. As such, the early phase of this
research was focused on the development of an in-house blade modeller which was then later used in
the parametrisation and reconstruction of the NASA Rotor 67. Other than the development of the
Blade Modeller, this research was also focused on the coupling of an open-source meshing software,
SALOME to the Blade modeller which will then allow the user to achieve automated meshing needed
for the design optimization process.

A literature study was carried out in order to understand the effect of boundary layer ingestion
on inlet performance and subsequently its influences on rotor performance. Extensive literature was
found which characterized and analyse the influence of boundary layer ingestion on the engine perfor-

I
Master Thesis Report

mance as a whole but very few literatures were found which pry into the detailed fluid mechanism of
boundary layer ingestion influencing the rotor performance. Detailed analysis of the blade to blade
domain under non uniform inflow condition were not found. Therefore, the main highlight of this
thesis will be on the detailed analysis of the blade to blade domain as well as the overall rotor perfor-
mance under non uniform inflow condition.

The framework of this report is as follows; The first part of the report will focus on blade parametri-
sation as well as the process of blade reconstruction. This is followed by the procedures and approach
taken to carry out the simulations using the open source Computational Fluid Dynamics (CFD) Tools,
SU2. The CFD simulations and analyses conducted in this thesis are divided into two main stages.
The first stage will deal with the computational analysis of the 2D fitted profiles in order to validate
the profile fitting process. The second stage will deal with the comparison of the 3D simulation results
under both uniform and non-uniform inflow.

From the 2D results, it was found that the 2D fitted and Actual point coordinate sources of profile
1265 and 1865 have the closest fit in terms of blade loading, Mach/Pressure Contour as well as their
stagnation pressure loss results. However, in the case of profile 2265 and 1565, it was found that the
2d fitted and surface fitted profile have the closest fit in terms of their blade loading, Mach/Pressure
Contour and stagnation pressure loss results. Therefore, this result suggests that it is difficult to
pinpoint a particular source of error for the blade fitting procedures. However, it can be seen that
minor deviation in the leading/trailing edge position and curvature of the blade profile can bring
about a significant shift in the shockwave position. It can also be concluded that the BSpline surface
generation algorithm gives a better fit to the 2d fitted profiles closer to the tip radius than the hub
radius. It should be pointed out that the current blade geometry was generated based on a previous
version of the Blade Modeller using a different technique for the generation of BSpline surfaces as well
as the flaring angle. The conclusion made from this results are based on the older version of the Blade
Modeller.
The 3D simulation results composed of two parts. The first part of this result will looks into the
comparison of results between the case of a uniform and non uniform inflow boundary condition. It
was found that the effect of the non uniform inflow is an increase in entropy production and a decrease
in isentropic efficiency. The decrease in efficiency is mainly a result of the change in inlet incidence
angle as the inlet total pressure varies. Consequently, this has an effect on the blade loading, shock
position as well as the trailing edge wake pattern. For the comparison of the results between the case
of a rotor with casing and without casing under uniform inflow, it was found that the case of the rotor
without casing exhibits a reduction in isentropic efficiency. This is mainly because of the increased tip
leakage losses as compared to the case of a rotor with casing. However, it is noted that the pressure
ratio over the unducted rotor is slightly higher as compared to the ducted rotor. In this simulation,
the total duct length is relatively short and therefore the pressure induced by the shroud does not
significantly add to the total pressure ratio.

As described above, the variation in the inlet total pressure along the span direction leads to an
increase/decrease in the inlet incidence angle. In order to improve the efficiency of the compressor/fan
blade, adjustment to the blade profile can be made in term of the inlet blade metal or stagger angle.
However, the change in the inlet blade metal angle influences the enthalpy/pressure change across
the blade profile and consequently the thermodynamic properties of the engine cycle. Therefore, it
is not recommended to change this angle if possible. Implementing changes to the stagger angle is a
possibility. Follow up studies on the numerical optimization of the optimal stagger angle at each span
position is necessary.

II
Master Thesis Report

Acknowledgements

The thesis work presented in this report would not have been possible without the help of a number
of people. Firstly, I would like to extend my greatest gratitude to my daily supervisor, Ir. Salvatore
Vitale for his endless support in helping me to get started with this thesis work and also providing me
with the technical support and advice throughout my thesis work. In addition, he was instrumental
in motivating me and providing valuable feedback on my work.

Secondly, I would like to thank Professor Dr. Arvind G. Rao and Dr. Matteo Pini for their help
in the formulation of this thesis topic and providing me the opportunity to carry out this work at
TU Delft. The opportunity to perform my thesis work at the TU Delft has enabled me to learn more
about the techniques used in Computational Fluid Dynamics (CFD) to obtain a series of results. In
particular, I was exposed to the opportunity to learn more about SU2, SALOME and python during
the course of this work.

Thirdly, I would like to thank my fellow friend and MSc student, Roel de Koning for introducing
me to the Blade Modeller which he has started developing and also providing me the opportunity
to work on both the developmental aspect and usage of the Blade Modeller. He was very helpful in
sharing his knowledge and I am truly grateful to him for his relentless support and time invested on
my thesis work.

Last but not least, I would like to thank all my fellow master students, friends and family for their
patience, motivation and encouragement which have kept me going despite all the difficulties faced in
recent times.

Raynold Tan Yi Yun


Delft, December 2015

III
Master Thesis Report Contents

Contents

Executive Summary I

Acknowledgements III

Nomenclature XI

1 Introduction 2

2 Literature Review 4
2.1 Propulsion Airframe Integration (PAI) . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Cycle Analysis of the Influence of BLI . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Blade Parameterization, Construction and Performance Characterization . . . . . . . . 15
2.4 Capabilities and Function of SU 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3 Research Goals and Objectives 26


3.1 Research Question and Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

4 Parameterisation and Reconstruction of the NASA Rotor 67 28


4.1 Step 1: Extraction of the blade geometry from the CGNS mesh . . . . . . . . . . . . . 28
4.2 Step 2: Fitting the camberline of the 2D Profiles . . . . . . . . . . . . . . . . . . . . . 29
4.3 Step 3: Fitting the Suction/Pressure side curves of the 2D Profile . . . . . . . . . . . . 32
4.4 Step 4: Stacking of Blade Profiles and Obtaining 3D Blade Shape . . . . . . . . . . . . 33
4.5 Comparison of Actual and Reconstructed Rotor Geometry . . . . . . . . . . . . . . . . 34
4.6 Design Trend of the NASA Rotor 67 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

5 Verification of 2D Fitting Results 38


5.1 2D Inlet Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2 2D Outlet Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.3 2D Mesh Topology and CFD Solver settings . . . . . . . . . . . . . . . . . . . . . . . . 43
5.4 2D Pressure & Mach Contour, Blade loading and Geometry Comparison Results . . . 47
5.5 Comparison of 2D Stagnation Pressure Loss and Blade Channel Area Distribution . . 61

6 Development of Salome 2D & 3D Meshing template and Grid Generator 63


6.1 Development of 2D Meshing Template . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2 Comparison of 2D CFD Results between SALOME and UMG 2 Mesh . . . . . . . . . 65
6.3 Development of 3D Meshing Template . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.4 3D CFD Results of Salome Generated Mesh . . . . . . . . . . . . . . . . . . . . . . . . 68
6.5 Development of SALOME-SU 2 Grid Generator . . . . . . . . . . . . . . . . . . . . . . 68

IV
Master Thesis Report Contents

7 3D Flow Simulation of the NASA Rotor 67 71


7.1 3D Uniform Inflow Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.2 3D Non Uniform Inflow Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . 72
7.3 ANSYS CFX Solver Settings and Mesh Convergence Studies . . . . . . . . . . . . . . 79
7.4 3D Uniform and Non Uniform Inflow Simulation Results . . . . . . . . . . . . . . . . . 81
7.5 3D Uniform Inflow Simulation Results With/Without Casing . . . . . . . . . . . . . . 90

8 Conclusion and Recommendation 96

Bibliography II

V
Master Thesis Report List of Figures

List of Figures

2.1 BWB with podded engines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4


2.2 BWB with embedded engines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Closer look on the Embedded Engine Concept [1] . . . . . . . . . . . . . . . . . . . . . 5
2.4 Podded versus Embedded Engine Concept [1] . . . . . . . . . . . . . . . . . . . . . . . 6
2.5 LHS terms of Power Balance Equation [2] . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.6 RHS terms of Power Balance Equation [2] . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.7 Control Volume of aircraft with and without BLI configuration [3] . . . . . . . . . . . 8
2.8 Control Volume Analysis of a Engine with Boundary Layer Ingestion [4] . . . . . . . . 9
2.9 Enthalpy - Entropy Diagram of a Non Ideal Brayton Cycle . . . . . . . . . . . . . . . 11
2.10 Plot of total pressure rise efficiency versus flow quantity coefficient [5] . . . . . . . . . 13
2.11 Comparison of the Influence of distortion pattern amplitude on the stability margin
between experimental and computational results [6] . . . . . . . . . . . . . . . . . . . . 14
2.12 Definitions used in characterising 2D Blade Profile . . . . . . . . . . . . . . . . . . . . 15
2.13 Camberline Parameterization [7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.14 Effect of w on camberline shape [7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.15 Use of two control points to adjust camberline [7] . . . . . . . . . . . . . . . . . . . . . 17
2.16 Blade Thickness Definition [8] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.17 Influence of arctangent stretching parameter on points distribution along ’u’ parameter
[7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.18 Construction of Blade Profile Suction/Pressure Side [7] . . . . . . . . . . . . . . . . . . 18
2.19 Stacking of a 3D Axial Turbine Blade [8] . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.20 End wall losses [9] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.21 Secondary flow losses [9] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.22 Tip leakage losses [9] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.23 Type of Inlet Distortion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.24 Definitions for Pav and Pav,low . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

4.1 Original NASA Rotor 67 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28


4.2 Sliced NASA Rotor 67 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3 Sliced 2D Profiles of the NASA Rotor 67 . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4 Plot of Original Point Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.5 Plot of Re-ordered Points Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.6 Plot of Re-ordered Points Coordinates with sep. suction and pressure side indexes . . 31
4.7 Plot of Re-ordered Points with camberline . . . . . . . . . . . . . . . . . . . . . . . . 31
4.8 Illustration of projection algorithm method . . . . . . . . . . . . . . . . . . . . . . . . 31
4.9 Illustration of geometry fitting algorithm method . . . . . . . . . . . . . . . . . . . . . 32
4.10 Illustration of profile before optimization . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.11 Illustration of profile after optimization . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.12 Illustration of stacked profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

VI
Master Thesis Report List of Figures

4.13 Illustration of full profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34


4.14 Original NASA Rotor 67 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.15 Reconstructed NASA Rotor 67 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.16 Angle Distribution of NASA Rotor 67 . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.17 Thickness Distribution of Suction Side NASA Rotor 67 . . . . . . . . . . . . . . . . . . 36
4.18 Thickness Distribution of Pressure Side NASA Rotor 67 . . . . . . . . . . . . . . . . . 36
4.19 Thickness Distribution of Pressure Side NASA Rotor 67 . . . . . . . . . . . . . . . . . 37

5.1 Concept of Radial Equilibrium [10] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41


5.2 Calculation procedure for profile 1865 outlet boundary condition . . . . . . . . . . . . 42
5.3 Calculation procedure for profile 1265, 1565 and 2265 outlet boundary condition . . . 43
5.4 Illustration of Mesh Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.5 Overall mesh topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.6 Mesh region near leading edge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.7 Mesh region near trailing edge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.8 Plot of Density Residuals versus Number of Iterations . . . . . . . . . . . . . . . . . . 47
5.9 Comparison of Pressure contour at different domain length . . . . . . . . . . . . . . . 48
5.10 Comparison of Mach contour at different domain length . . . . . . . . . . . . . . . . . 48
5.11 Comparison of 1865 Mach Contour between 2D fitted and Actual Point Profile . . . . 49
5.12 Comparison of 1865 Pressure Contour between 2D fitted and Actual Point Profile . . . 49
5.13 Comparison of 1865 Mach Contour between 2D fitted and Surface fitted Profile . . . 50
5.14 Comparison of 1865 Pressure Contour between 2D fitted and Surface fitted Profile . . 50
5.15 Comparison of 1865 Mach Contour between Surface fitted Profile and Actual point profile 50
5.16 Comparison of 1865 Press. Contour between Surf. fitted Profile and Actual point profile 50
5.17 Comparison of 1865 Blade loading between 2D fitted Profile and Actual point profile . 51
5.18 Comparison of 1865 Blade loading between 2D fitted Profile and Surface fitted profile 51
5.19 Comparison of 1865 Blade loading between Surface fitted Profile and Actual point profile 51
5.20 Comparison of 1865 Blade Profiles from 3 Coordinate sources . . . . . . . . . . . . . . 51
5.21 1865 Blade Profile leading edge zoom . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.22 1865 Blade Profile Trailing edge zoom . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.23 Comparison of 1265 Mach Contour between 2D fitted and Actual Point Profile . . . . 52
5.24 Comparison of 1265 Pressure Contour between 2D fitted and Actual Point Profile . . 52
5.25 Comparison of 1265 Mach Contour between 2D fitted and Surface fitted Profile . . . . 53
5.26 Comparison of 1265 Pressure Contour between 2D fitted and Surface fitted Profile . . 53
5.27 Comparison of 1265 Mach Contour between Actual point and Surface fitted Profile . . 53
5.28 Comparison of 1265 Pressure Contour between Actual point and Surface fitted Profile 53
5.29 Comparison of 1265 Blade loading between 2D fitted and Actual point profile . . . . . 54
5.30 Comparison of 1265 Blade loading between 2D fitted and Surface fitted profile . . . . 54
5.31 Comparison of 1265 Blade loading between Surface fitted and Actual point profile . . 54
5.32 Comparison of 1265 Blade Profiles from 3 Coordinate sources . . . . . . . . . . . . . . 54
5.33 Comparison of 1565 Mach Contour between 2D fitted and Actual Point Profile . . . . 55
5.34 Comparison of 1565 Pressure Contour between 2D fitted and Actual Point Profile . . 55
5.35 Comparison of 1565 Mach Contour between 2D fitted and Surface fitted Profile . . . . 55
5.36 Comparison of 1565 Pressure Contour between 2D fitted and Surface fitted Profile . . 55
5.37 Comparison of 1565 Mach Contour between Actual point and Surface fitted Profile . . 56
5.38 Comparison of 1565 Pressure Contour between Actual point and Surface fitted Profile 56
5.39 Comparison of 1565 Blade loading between 2D fitted and Actual point profile . . . . . 56
5.40 Comparison of 1565 Blade loading between 2D fitted and Surface Fitted profile . . . . 56
5.41 Comparison of 1565 Blade loading between Surface fitted and Actual point profile . . 57
5.42 Comparison of 1565 Blade Profiles from 3 Coordinate sources . . . . . . . . . . . . . . 57

VII
Master Thesis Report List of Figures

5.43 1565 Blade Profile leading edge zoom . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57


5.44 1565 Blade Profile trailing edge zoom . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.45 Comparison of 2265 Mach Contour between 2D fitted and Actual Point Profile . . . . 58
5.46 Comparison of 2265 Pressure Contour between 2D fitted and Actual Point Profile . . 58
5.47 Comparison of 2265 Mach Contour between 2D fitted and Surface fitted Profile . . . . 58
5.48 Comparison of 2265 Pressure Contour between 2D fitted and Surface fitted Profile . . 58
5.49 Comparison of 2265 Mach Contour between Actual point and Surface fitted Profile . . 59
5.50 Comparison of 2265 Pressure Contour between Actual point and Surface fitted Profile 59
5.51 Comparison of 2265 Blade loading between 2D fitted and Actual point profile . . . . . 59
5.52 Comparison of 2265 Blade loading between 2D fitted and Surface Fitted profile . . . . 59
5.53 Comparison of 2265 Blade loading between Surface fitted and Actual point profile . . 60
5.54 Comparison of 2265 Blade Profiles from 3 Coordinate sources . . . . . . . . . . . . . . 60
5.55 2265 Blade Profile leading edge zoom . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.56 2265 Blade Profile trailing edge zoom . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.57 Enthalpy-Entropy diagram of a Stator Blade . . . . . . . . . . . . . . . . . . . . . . . 61
5.58 Area Distribution for various blade profiles . . . . . . . . . . . . . . . . . . . . . . . . 62

6.1 Features of a good quality 2D mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63


6.2 2D Mesh Generated by Salome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.3 Leading edge zoom on Salome 2D Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.4 Trailing edge zoom on Salome 2D Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.5 Boundary Layer zoom on Salome 2D Mesh . . . . . . . . . . . . . . . . . . . . . . . . 64
6.6 Comparison of Profile 1265 Salome and UMG 2 Mesh Mach Contour . . . . . . . . . . 65
6.7 Comparison of Profile 1265 Salome and UMG 2 Mesh Pressure Contour . . . . . . . . 65
6.8 Comparison of Profile 1265 Salome and UMG 2 Mesh Blade Loading . . . . . . . . . . 65
6.9 Features of a good quality 3D mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.10 Fine Discretized Mesh across the Leading/Trailing edge and the Top/Bottom Blade
Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.11 Clipped Vertical View of NASA Rotor 67 Internal Mesh Domain . . . . . . . . . . . . 67
6.12 Clipped Horizontal View of NASA Rotor 67 Internal Mesh Domain . . . . . . . . . . . 67
6.13 3D mesh of Centrifugal Rotor with tip clearance . . . . . . . . . . . . . . . . . . . . . 67
6.14 Clipped Vertical View of Centrifugal Rotor Internal Mesh Domain . . . . . . . . . . . 68
6.15 Clipped Cross-Section View of Centrifugal Rotor Internal Mesh Domain . . . . . . . . 68
6.16 Mach Streamlines of Wing Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.17 Top view of Mach contour of Wing Domain . . . . . . . . . . . . . . . . . . . . . . . . 68
6.18 Pressure Coefficient Plot of Salome Meshed Wing Slice . . . . . . . . . . . . . . . . . . 69
6.19 SU 2 Mesh Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

7.1 Spanwise Total Pressure at Inlet/Outlet Boundary . . . . . . . . . . . . . . . . . . . . 72


7.2 Spanwise Total Temperature at Inlet/Outlet Boundary . . . . . . . . . . . . . . . . . . 72
7.3 Spanwise Static Pressure at Inlet/Outlet Boundary . . . . . . . . . . . . . . . . . . . . 72
7.4 Measurement Points at Inlet and Outlet Boundary . . . . . . . . . . . . . . . . . . . . 72
7.5 Total Pressure Contour at the Inlet and Outlet Plane of the Inlet Configuration . . . . 73
7.6 Velocity Contour at the Inlet and Outlet Plane of the Inlet Configuration . . . . . . . 74
7.7 Sampled Points along the Outlet Plane of inlet configuration . . . . . . . . . . . . . . 76
7.8 Plot of total pressure versus angular position for varying radial position . . . . . . . . 77
7.9 Plot of velocity versus angular position for varying radial position . . . . . . . . . . . . 77
7.10 Calculation of distortion intensity at mid radius . . . . . . . . . . . . . . . . . . . . . . 78
7.11 Plot of Circumferential Averaged Total Pressure at various Spanwise Radial Position . 79
7.12 Isometric View of 3D Blade Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

VIII
Master Thesis Report List of Figures

7.13 Meridional View of Single Blade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81


7.14 20% Span blade loading Uniform Inflow . . . . . . . . . . . . . . . . . . . . . . . . . . 82
7.15 20% Span blade loading Non Uniform Inflow . . . . . . . . . . . . . . . . . . . . . . . 82
7.16 50% Span blade loading Uniform Inflow . . . . . . . . . . . . . . . . . . . . . . . . . . 82
7.17 50% Span blade loading Non Uniform Inflow . . . . . . . . . . . . . . . . . . . . . . . 82
7.18 80% Span blade loading Uniform Inflow . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7.19 80% Span blade loading Non Uniform Inflow . . . . . . . . . . . . . . . . . . . . . . . 83
7.20 20% Span Uniform Inflow Blade to Blade Domain Velocity Vector . . . . . . . . . . . 83
7.21 20% Span Non Uniform Inflow Blade to Blade Domain Velocity Vector . . . . . . . . . 83
7.22 50% Span Uniform Inflow Blade to Blade Domain Velocity Vector . . . . . . . . . . . 84
7.23 50% Span Non Uniform Inflow Blade to Blade Domain Velocity Vector . . . . . . . . . 84
7.24 80% Span Uniform Inflow Blade to Blade Domain Velocity Vector . . . . . . . . . . . 84
7.25 80% Span Non Uniform Inflow Blade to Blade Domain Velocity Vector . . . . . . . . . 84
7.26 20% Span Uniform Inflow Blade to Blade Domain Entropy Change . . . . . . . . . . . 85
7.27 20% Span Non Uniform Inflow Blade to Blade Domain Entropy Change . . . . . . . . 85
7.28 50% Span Uniform Inflow Blade to Blade Domain Entropy Change . . . . . . . . . . . 85
7.29 50% Span Non Uniform Inflow Blade to Blade Domain Entropy Change . . . . . . . . 85
7.30 80% Span Uniform Inflow Blade to Blade Domain Entropy Change . . . . . . . . . . . 86
7.31 80% Span Non Uniform Inflow Blade to Blade Domain Entropy Change . . . . . . . . 86
7.32 20% Span Uniform Inflow Blade to Blade Domain Static Pressure . . . . . . . . . . . . 86
7.33 20% Span Non Uniform Inflow Blade to Blade Domain Static Pressure . . . . . . . . . 86
7.34 50% Span Uniform Inflow Blade to Blade Domain Static Pressure . . . . . . . . . . . . 87
7.35 50% Span Non Uniform Inflow Blade to Blade Domain Static Pressure . . . . . . . . . 87
7.36 80% Span Uniform Inflow Blade to Blade Domain Static Pressure . . . . . . . . . . . . 87
7.37 80% Span Non Uniform Inflow Blade to Blade Domain Static Pressure . . . . . . . . . 87
7.38 Visualisation of Uniform Inflow Streamlines . . . . . . . . . . . . . . . . . . . . . . . . 88
7.39 Visualisation of Non Uniform Inflow Streamlines . . . . . . . . . . . . . . . . . . . . . 89
7.40 20% Span Blade Loading Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.41 50% Span Blade Loading Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.42 80% Span Blade Loading Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.43 95% Span Blade Loading Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.44 99% Span Blade Loading Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
7.45 Circumferential Average Total Pressure Versus Normalized Span Position at Inlet . . . 93
7.46 Circumferential Average Total Pressure Versus Normalized Span Position at Outlet . . 93
7.47 Visualisation of Rotor Streamlines (with casing and Uniform Inflow) . . . . . . . . . . 93
7.48 Visualisation of Rotor Streamlines (without casing and Uniform Inflow) . . . . . . . . 94

IX
Master Thesis Report List of Tables

List of Tables

5.1 Design and Geometry Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38


5.2 Atmospheric Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.3 Calculated BC Data at tip radius . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.4 Calculated Inlet BC Data at other radial position . . . . . . . . . . . . . . . . . . . . . 40
5.5 Inlet and Outlet BC Data at R = 0.1265, 0.1565, 0.1865 & 0.2265 . . . . . . . . . . . . 43
5.6 Common Parameters for Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . 46
5.7 Flow Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.8 Turbulent Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.9 Comparison of Stagnation Pressure Loss Coefficient between different profile types . . 61

7.1 BWB Flight Operating Condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75


7.2 Mass flow Averaged Total Condition Across Inlet Configuration . . . . . . . . . . . . . 75
7.3 Comparison of Outlet Plane Total Condition vs. Aft Section Total Condition without
BLI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
7.4 Comparison of Inlet Plane Total Condition vs. Aft Section Total Condition without BLI 76
7.5 Performance of the Rotor under Non Uniform Inflow Boundary Condition . . . . . . . 90
7.6 Performance of the Rotor under Uniform Inflow Boundary Condition . . . . . . . . . . 90
7.7 Performance of the Rotor (with casing) under Uniform Inflow Boundary Condition . . 95
7.8 Performance of the Rotor (without casing) under Uniform Inflow Boundary Condition 95

X
Master Thesis Report List of Tables

Nomenclature

( ∆p c
p )i Circumferential distortion intensity Φf uselage Jet Dissipation
Φjet Jet Dissipation
Q
( )comp Compressor Pressure Ratio

(θi )¯ Circumferential distortion extend Φshock Shock Dissipation


Φtotal Total viscous dissipation rate
(Pav )i Circumferential Averaged Static Pressure
for a given ring Φvortex Vortex Dissipation

βin Inlet Blade Metal Angle Φwake Wake Dissipation

βout Outlet Blade Metal Angle ρ Density


τs Temperature Ratio
δloss stagnation pressure loss coefficient
τw Wall shear stress
Ėa Wake Streamwise Kinetic Energy Deposi-
tion Rate alocal Local Speed of Sound

Ėp Wake Pressure Defect Work Rate Cθ Absolute Tangential Velocity

Ėv Wake Transverse Kinetic Energy Deposi- Cax Axial Chord Length
tion Rate DA Airframe Drag
Ėw Wave Pressure Work and Kinetic Energy Fengine Propulsive force provided by the engine
Outflow Rate
P01,rel Relative Stagnation Pressure
ηcomp Adiabatic Compressor Efficiency
Pk Net Propulsor Mechanical Energy Flow
ηp Propulsive Efficiency Rate
Ps2 Outlet Static Pressure
ηth Thermal Efficiency
Ps Net Propulsor Shaft Power
ηtt Total to Total Isentropic Efficiency
PT2 Outlet Total Pressure
ηturb Adiabatic Turbine Efficiency
Pv Net Pressure Volume Power
Q1
nD3
Flow Quantity coefficient
Q1 Volumetric Flow Rate
γ Stagger Angle
T01,rel Relative Stagnation Temperature
λf reevortex Free vortex constant U∞ Free Stream Velocity
µ Dynamic Viscosity of Air Uj Exit Velocity
Φduct Inlet Duct Dissipation Uw Wake Velocity

XI
Master Thesis Report List of Tables

V0 Inlet Velocity BLI Boundary Layer Ingestion

V2 Exit Velocity BWB Blended Wing Body

W ḣ Potential Energy Rate CFD Computational Fluid Dynamics

Wθ Relative Tangential Velocity FPR Fan Pressure Ratio

xle Leading edge axial position FW-H Ffowcs Williams - Hawking

xte Trailing edge axial position H Blade Height

yle Leading edge tangential position PAI Propulsion Airframe Integration

yte Trailing edge tangential position RANS Reynold Average Navier Stokes

ALE Arbitrary Lagrangian Eulerian SST Menter Shear Stress Transport

1
Master Thesis Report Introduction

1|Introduction

In Europe, the Advisory Council for Aeronautical Research has established ambitious targets to sub-
stantially reduce fuel consumption, C02, NOx and noise emission by 2020 under the Clean Sky project.
These requirements demand a novel configuration for air vehicles, which suggest a highly integrated
components for optimum overall performance. Improved propulsion system architecture are of pri-
marily importance in achieving these goals [11].

According to recent estimates, a 1% reduction on fuel burn can result in savings of 1 million per
engine year. Other than fuel consumption, noise also plays an important role in the commercial
engine business with growing tighter regulations. Increased noise emission can result in curfew at air-
ports, which increases the overall operational costs due to reduced airport capacity in a given period of
time. With regards to engine design, fuel consumption, emissions, noise, drag and weight plays a vital
role in the overall aircraft performance. Propulsion airframe integration (PAI) is also an important
element in meeting future aviation goals.

Hybrid wing body has become increasingly popular as a result of increased aerodynamic efficiency
as compared to conventional airframe design. Such a configuration blurs the division between fuselage
and wings. The Blended Wing Body (BWB) is one such configuration, where the propulsion system
is mounted on the upper aft surface of the fuselage. PAI on a BWB can be achieved by either pod-
ding the engines or embedding them onto the aft fuselage section. While podded engines offer the
advantage of ingesting uniform flow into the engine, this concept is however associated with increased
structural mass, drag and consequently increased fuel consumption. Embedding the engines into the
fuselage implies the ingestion of the boundary layer. This concept has shown to provide reduction in
the overall fuel burn percentages as a result of reduced momentum deficit.

The momentum wake trailed by an aircraft has both lift induced and viscous contributors. While
the lift induced component is essentially fixed, the viscous portion can be reduced through boundary
layer ingestion. The low momentum fuselage boundary layer is re-energized by the propulsion system
before exiting into the atmosphere. As such, the entire momentum wake deficit of the aircraft and
consequently the total drag of the aircraft is reduced. As a result of BLI, the inlet suffers from a lower
mass averaged stagnation pressure. This results in an overall loss to the thermodynamic cycle of the
engine as well as increased viscous losses. Another associated concept with boundary layer ingestion
in BWB is the use of a distributed propulsion system over the aft fuselage. It has been shown by [12]
that a 3 engine configuration demonstrated a 5% reduction in fuel consumption for a 2% increase in
pressure loss. It was also found that the potential of BLI is better for larger engines than for smaller
engines as the latter incur high losses in efficiency due to scale effect which overrides the benefits of
BLI. Therefore a careful selection of the number of engines as well as the proper design of the fan
blade are required in order to maximise engine performance while taking advantage of the reduced
momentum deficit.

2
Master Thesis Report Introduction

Chapter 2 of this report will present a short summary of the literature study. This will include a
study on the comparison between embedded and podded engine configurations, a cycle analysis on the
effect of boundary layer ingestion on the thermodynamic and propulsive efficiency, a study on blade
parametrisation, construction, loss mechanism and performance characterisation of the rotor blade
and finally a literature review on the capabilities and function of SU2 Computational Fluid Dynamics
(CFD) Code. After the literature review, chapter 3 of the report will specifically present on the overall
research objective, goal and the scope of this thesis work. Chapter 4 of the report will then focus on
the actual methodology used to parametrise and reconstruct the NASA Rotor 67. This is followed by
a presentation on the design trend of the rotor blade. In order to validate the blade reconstruction
method, Chapter 5 of the report will touch on the results of the 2D blade profile simulation. Since
a part of the research work was focused on coupling Salome, an open source meshing software onto
the blade modeller, a part of the 2D CFD simulation results will present on the comparison of per-
formances results of the same profile under the use of different meshing software. Chapter 6 of the
report will then touch on the development of meshing template for both 2D and 3D blade profiles in
SALOME as well as the coupling of SALOME to the blade modeller. Chapter 7 of the report will
then present on the 3D boundary conditions and simulation results for both uniform and non uniform
inflow condition. Finally, Chapter 8 of the report will present on the conclusion and recommendations
from this research.

3
Master Thesis Report Literature Review

2|Literature Review

In this chapter, a review of the most relevant literature is summarised and presented. The aim of this
chapter will be to provide the reader a brief introduction of concepts that will be used throughout
this entire report. The first sub-section of this chapter will touch on propulsion airframe integration.
The second sub-section will present a cycle analysis on the effect of boundary layer ingestion on the
thermodynamic and propulsive efficiency. The third sub-section of this report will focus on blade
parametrisation , construction, loss mechanism as well as the performance characterisation of the
rotor blade. Finally, the last sub-section of this chapter will touch on the capabilities and function of
SU2 (Stanford University Unstructured), a CFD tool that was used throughout this research work.

2.1 Propulsion Airframe Integration (PAI)


Propulsion airframe integration (PAI) is an important element to consider when meeting modern
aviation requirement of reduced emissions, noise and fuel consumption. In the case of the Blended
Wing Body concept, PAI can be achieved by either podding the engines onto the aft fuselage or
embedding the engine into the fuselage.

Figure 2.2: BWB with embedded engines


Figure 2.1: BWB with podded engines

An illustration of both concept is shown in figure 2.1 and 2.2 above. Fundamentally, the podded
engine concept allows a clean undisturbed airflow into the engine which provides a high total pressure
recovery and low inlet distortion. Secondly, the podded engine concept is a well understood and proven
installation technique that can be implemented easily. The main disadvantage of such a concept is
the increased wetted area which means an increase in the profile drag. The use of pylon also implies
additional structural weight which can reduce the overall fuel efficiency of the aircraft. The presence of
unshielded inlet design also leads to larger noise pollution comparatively. Lastly, the higher thrust line
of the podded engine concept implies a larger horizontal control surface needed for trimming the flight.

4
Master Thesis Report Literature Review

A closer look at the embedded engine concept is shown in figure 2.3. The main advantage of this
concept is the lower propulsion system thrust line which implies lower balance requirement and there-
fore smaller trim drag. The absence of the pylons also leads to a reduction in both the structural weight
as well as the wetted surface area. Finally, embedded engine configuration features shielded intakes
which leads to lower noise production. Embedded configurations also give rise to size limitation which
impose constraint on the selection of an optimum thermodynamic cycle for a given flight profile. Most
importantly, The main drawback of the embedded engine concept is the significant flow distortion into
the engine, resulting in both engine stability issues and performance loss. The elevated distortion level
also leads to increased maintenance to address possible high cycle fatigue and operability with the
compression system. Taken these limitations together, they have a significant impact on the engine
subsystem performance. Lastly, the use of an embedded engine design leads to a more complex and
integrated design for the aft fuselage section.

Figure 2.3: Closer look on the Embedded Engine Concept [1]

A mathematical analysis of an embedded engine configuration with BLI versus podded engine by
A.Plas [1] has shown that an embedded engine configuration requires lesser power as compared to the
podded engine case with no BLI. This analysis is made on the assumption that the total weight of
either configuration is the same. This is mainly because the inlet velocity into the propulsion system
for the embedded engine concept is lesser than the podded engine concept.
In the podded engine concept, the flow entering the engine has a freestream velocity of U∞ . The
engine then accelerates the flow to a velocity of Uj . For an ideally expanded nozzle, the momentum
excess created by the jet balances the momentum deficit created by the drag of the airframe (DA ).

Fengine = ṁ(Uj − U∞ ) = ṁ(U∞ − Uw ) = DA (2.1)


The rate of (kinetic) energy added to the flow by the engine is given as follows:

ṁ 2 2 ṁ Fengine
PnoBLI = (Uj − U∞ ) = (Uj − U∞ )(Uj + U∞ ) = (Uj + U∞ ) (2.2)
2 2 2
In the embedded engine concept, all the fuselage boundary layer is assumed to be ingested by the
engine and the engine increases the wake velocity to the far upstream value. The wake velocity, Uw is

5
Master Thesis Report Literature Review

Figure 2.4: Podded versus Embedded Engine Concept [1]

then used as the inlet velocity into the engine. The propulsive force provided by the engine balances
the momentum deficit caused by the airframe drag:

Fengine = ṁ(Uj − Uw ) = ṁ(U∞ − Uw ) = DA (2.3)


The mechanical power required to produce this force in the presence of boundary layer ingestion is
then formulated as such:

ṁ 2 ṁ Fengine
PBLI = (Uj − Uw2 ) = (Uj − Uw )(Uj + Uw ) = (Uw + U∞ ) (2.4)
2 2 2
Comparing equation 2.2 and 2.4, it can be seen that the embedded engine concept with BLI requires
lesser mechanical power since Uw < Uj .

Power Balance Method

Another way of investigating the benefit of boundary layer ingestion is to use the power balance
method developed by Drela [2]. This method is based on the conservation of total power. The rate of
energy entering into the system equates the rate of energy leaving the system. Specifically, the sum
of the total mechanical power supply is equal to the sum of the total viscous dissipation rate, Φtotal
and the mechanical energy flow rate out of the control volume (CV), ε̇. The total mechanical power
supply is made up of three terms; the net propulsor shaft power Ps , the net pressure volume power,
Pv power and the net propulsor mechanical energy flow rate into the CV, Pk . Mathematically, the
power balance equation can be expressed as follows:

Ps + Pv + Pk = ε̇ + Φtotal (2.5)
The first term on the R.H.S represents the energy flow out of the control volume while the second
term represents the energy loss within the control volume. The mechanical energy outflow rate, ε̇ can
be decomposed into 5 main terms; the potential energy rate, W ḣ, the wake streamwise kinetic energy
deposition rate, Ėa , the wake transverse kinetic energy deposition rate, Ėv , the wake pressure defect
work rate, Ėp and the wave pressure work and kinetic outflow rate, Ėw . Taken all these together, the
power balance equation can be expressed as follows:

Ps + Pv + Pk = W ḣ + Ėa + Ėv + Ėp + Ėw + Φtotal (2.6)

6
Master Thesis Report Literature Review

An illustration of the control volume representing all the different terms in equation 2.6 is shown in
figure 2.5 and 2.6. The definition of each terms given in equation 2.6 is presented below:

Figure 2.5: LHS terms of Power Balance Equation Figure 2.6: RHS terms of Power Balance Equation [2]
[2]

• Ps The net propulsor shaft power is defined as the integrated force multiplied by velocity on all
moving surface

• Pv The net pressure volume power is a volumetric mechanical power provided by the fluid
expanding against atmospheric pressure.

• Pk The net propulsor mechanical flow rate is the net pressure work and kinetic energy flow rate
across the SB . This is to account for power sources whose moving blade is not covered by SB .

• W ḣ The potential energy rate is the power consumption needed to increase the aircraft’s potential
energy and become a power source during descent.

• Ėa The wake streamwise kinetic energy deposition rate term is the rate of streamwise kinetic
energy being deposited in the flow out of the control volume through the treffez plane (see figure
2.6), also known as the outlet plane of the CV.

• Ėv The wake transverse kinetic energy deposition rate term is the rate of transverse kinetic
energy being deposited in the flow out of the control volume through the treffez plane (see figure
2.6).

• Ėp The wake pressure defect work rate is the rate of pressure work done of the fluid crossing the
treffez plane at some pressure P different from the ambient P∞ .

• Ėw The wave pressure work and kinetic energy outflow rate is the rate of pressure work and
kinetic energy deposition crossing the side cylinder.

• Φ The viscous dissipation rate is the rate at which the kinetic energy of the flow is converted
into heat inside the control volume

For a simplified aircraft configuration, the total dissipation comes from the following sources; the
jet dissipation Φjet , the wake dissipation Φwake , the fuselage dissipation Φf uselage and the vortex
dissipation Φvortex . This is illustrated in figure 2.7. In other more complex cases, the shock dissipation
Φshock as well as the inlet duct dissipation Φduct may be included.

7
Master Thesis Report Literature Review

Figure 2.7: Control Volume of aircraft with and without BLI configuration [3]

In actual situation, all form of kinetic energy outflow are converted into dissipations downstream of
the aircraft. Using this fact, the dissipation term in the power balance equation can be separated into
various sources as follows:

Ps + Pk =
W ḣ + (Ėp + Φvortex )+
+ (Ėw + Φwave ) + (Φf use − Pv,f use ) (2.7)
+ (Ėa,wake + Ėp,wake + Φwake − Pv,wake )
+ (Ėa,jet + Ėp,jet + Φjet − Pv,jet )

Defining the mechanical energy loss as follows:

Φ∗ =
X X X
Ė + Φ− PV (2.8)

The power balance equation can be simplified as follows:

Ps + Pk = W ḣ + (Φ∗ )vortex + (Φ∗ )wave + (Φ∗ )f use + (Φ∗ )wake + (Φ∗ )jet (2.9)

Considering that the flight configuration of both system in figure 2.7 is level (therefore, the flight
path angle is equal to zero); W ḣ = W V∞ sinγ = 0 and that no components are moving relative to the
control volume, Ps = 0. The power balance equation of a non BLI configuration can then be expressed
as shown in equation 2.10:

Pk = (Φ∗ )vortex + (Φ∗ )wave + (Φ∗ )f use + (Φ∗ )wake + (Φ∗ )jet (2.10)

As for the BLI configuration, the net propulsive power is the sum of the mechanical flow power without
the jet dissipation, (PK −(Φ∗ )jet ). The (Φ∗ )wake is reduced by a factor, fBLI as a result of the ingested
wake by the engine. The remaining dissipation term from the fuselage, wave pressure as well as the
vortex source are approximately the same for both the BLI and non BLI concept. The power balance
equation of an BLI configuration can be expressed as shown in equation 2.11:

Pk − (Φ∗ )jet = (1 − fBLI )(Φ∗ )wake + (Φ∗ )f use + (Φ∗ )vortex + (Φ∗ )wave (2.11)

8
Master Thesis Report Literature Review

2.2 Cycle Analysis of the Influence of BLI


In this section, the influence of boundary layer ingestion on the size of the propulsion system, ther-
modynamic as well as propulsive efficiency will be discussed. It should be pointed out that the exact
inflow condition is dependent on the aircraft geometry, placement of the propulsor as well as the flight
condition. The concept of inlet recovery ratio will be introduced as it is an indicator of the amount
of boundary layer ingested.

Figure 2.8: Control Volume Analysis of a Engine with Boundary Layer Ingestion [4]

Y PT1
= (2.12)
loss
PT0
A greater ingestion of the boundary layer implies a lower inlet recovery ratio. In order to understand
the influence of boundary layer ingestion on engine design and cycle performance, the following math-
ematical relations between inlet recovery ratio and related design parameters were derived below. The
thrust produced by an engine is given by the following expression:

T = ṁ(Vexit − Vinlet ) = ṁ(V2 − V0 ) (2.13)


Assuming incompressible flow, the bernouli equation can be used to calculate the exit velocity as
follows:

ρV22
PT2 = Ps,2 + (2.14)
2
s
2
V2 = PT − Ps,2 (2.15)
ρ 2
Defining the fan pressure ratio as:

PT2
FPR = (2.16)
PT1
The exit velocity can be expressed as:
s
2
V2 = F P R ∗ PT1 − Ps,2 (2.17)
ρ

9
Master Thesis Report Literature Review

Substituting the expression for the inlet recovery ratio into the exit velocity:
s
2 Y
V2 = FPR ∗ ∗PT0 − Ps,2 (2.18)
ρ loss

The total thrust produced can then be expressed as follows:


s
2 Y
T = ṁ ∗ ( FPR ∗ ∗PT0 − Ps − V0 ) (2.19)
ρ loss

Defining the mass flow as:

ρπDp2 V1
ṁ = ρAp V1 = (2.20)
4
The total thrust produced can finally be expressed as follows:

ρπDp2 V1
s
2 Y
T = ∗( FPR ∗ ∗PT0 − Ps,2 − V0 ) (2.21)
4 ρ loss

Considering that the total thrust produced has to be constant in steady and level flight, the following
relation can be observed:

• As the inlet recovery ratio decreases, the fan diameter has to increase in order to produce the
same thrust. This implies an increase in the structural mass of the propulsion system.

• As the inlet recovery ratio decreases, the exit velocity also reduces which implies a reduction in
the specific thrust.

From this relation, it can be said that an embedded propulsion system with lower pressure recovery
tends to suffer from a lower specific thrust and therefore require a larger fan diameter to achieve the
same level of thrust. As such, more horsepower is required in order to drive the same fan pressure
ratio with greater airflow. This additional power has to come from a larger core airflow and hence
increased fuel flow for the same combustor exit temperature. Therefore, the result of a reduced inlet
recovery ratio is a larger core to power the propulsion system.

Boundary layer ingestion also has an influence on both the thermal and propulsive efficiency. The
propulsive efficiency is related to the level of wasted kinetic energy trailed by the wake of the propulsion
system. It is a measure of how effective the net work output is converted to thrust power.

P ropulsiveP ower T V∞
ηp = = 1 2 2
(2.22)
Rateof P rod.of P ropulsiveKE 2 ṁ(Vj − V∞ )

Substituting equation 2.13 into equation 2.22,

ṁ(Vj − V∞ )V∞ 2
ηp = 1 = V
(2.23)
2 ṁ(Vj − V∞ )(Vj + V∞ ) 1 + V∞j
Substituting the expression for the exit velocity, the propulsive efficiency can then be expressed as
follows:

2 2
ηp = Vj
= q (2.24)
1+ 2
Q
V∞ ρ
F P R∗ loss
∗PT0 −Ps,2
1+ V∞

10
Master Thesis Report Literature Review

From equation 2.24, it can be seen that a reduction in the inlet recovery ratio leads to a reduction in
the exit velocity and consequently an increase in the overall propulsive efficiency. Strictly speaking,
equation 2.23 is not valid for engine configurations that incorporate boundary layer ingestion. This
is because the incoming inlet velocity is not equal to the free-stream velocity but to a different lower
incoming velocity. Denoting the inlet velocity into the engine as Vin , the propulsive efficiency of an
engine configuration using BLI can be re-expressed as follows:

ṁ(Vj − Vin )V∞


(ηp )BLI = 1 (2.25)
2 ṁ(Vj − Vin )(Vj + Vin )

2(Vj − Vin )V∞ 2V∞ 2


(ηp )BLI = = = Vj
(2.26)
(Vj − Vin )(Vj + Vin ) (Vj + Vin ) + Vin
V∞ V∞

Substituting the expression for the exit velocity, the propulsive efficiency can then be expressed as
follows:

2
(ηp )BLI = q (2.27)
2
Q
ρ
F P R∗ loss
∗PT0 −Ps,2
Vin
V∞ + V∞

In equation 2.27, a similar conclusion can be reached as in equation 2.24. However, it is noted that
the increase in propulsive efficiency is greater as compared to equation 2.24 as more boundary layer
is ingested since both terms in the denominator of equation 2.27 reduces.

The thermal efficiency is a measure of how effective the thermodynamic cycle converts the ther-
mal energy to net work output. It is a ratio of the work output to the heat input. Considering the
illustration of a non ideal brayton cycle given below, the thermal efficiency of a jet engine can be
expressed as follows:

Figure 2.9: Enthalpy - Entropy Diagram of a Non Ideal Brayton Cycle

W orkOutput W4−5 − W3−0 (T4 − T5 ) − (T3 − T0 ) T5 − T0


ηth = = = =1− (2.28)
HeatInput Q4−3 (T4 − T3 ) T4 − T3

11
Master Thesis Report Literature Review

The adiabatic turbine and compressor efficiency can be expressed as follows:

h4 − h5 T4 − T5
ηturb = = (2.29)
h4 − h5s T4 − T5s

h3s − h0 T3s − T0
ηcomp = = (2.30)
h3 − h0 T3 − T0
Using equation 2.29 and 2.30, T5 and T3 can be derived respectively as follows:

T5s Y k−1
T4 − T5 = ηturb T4 (1 − ) = ηturb T4 (1 − ( )turb
k
) (2.31)
T4
T5s Y k−1
T5 = T4 − ηturb T4 (1 − ) = T4 − ηturb T4 (1 − ( )turb
k
) (2.32)
T4
k−1
T3s − T0 T0 ( TT3s0 − 1) − 1)
k
Q
T0 (( )comp
T3 − T0 = = = (2.33)
ηcomp ηcomp ηcomp
k−1
T3s − T0 T0 ( TT3s0 − 1) − 1)
k
Q
T0 (( )comp
T3 = T0 + = T0 + = T0 + (2.34)
ηcomp ηcomp ηcomp
Substituting expressions for T5 and T3 into equation 2.28 and using the following assumption for the
isentropic pressure ratio and temperature ratio.
Y 1 Y
( )comp = Q = (2.35)
( )turb
T3s Y k−1
τs = = ( )) k (2.36)
T0
As such, the thermal efficiency of a non ideal brayton cycle can be expressed as follows:
k−1
T4 − ηturb T4 (1 − ( )turb − T0
Q k
ηth = 1 − k−1 (2.37)
1
T4 − T0 (1 + ηcomp (( )comp − 1))
Q k

By having a common denominator for all the terms in equation 2.37 and then perform the following
operation:
ηcomp
• Multiply the equation by T0 on both the numerator and denominator terms
1
• Factorise the numerator term and denominator term by (1 − τs ) and ηcomp respectively

Then the thermal efficiency of the non ideal brayton cycle can be re-expressed as follows:
1 T4
(1 − τs )( T0 ηturb ηcomp − τs )
ηth = (2.38)
ηcomp ( TT40 − 1) − τs + 1
The influence of BLI on cycle performance is expressed directly via a reduction in the adiabatic
compressor efficiency, ηc with the following relation:

(ηcomp )BLI = ηBLI ∗ ηcomp (2.39)


With

12
Master Thesis Report Literature Review

1 − τs ( k−1
k )(1 − ( )loss )
Q
ηBLI = (2.40)
τs − 1
Substituting equation 2.39 into equation 2.38 results in an expression which relates the thermal effi-
ciency of a non ideal Brayton cycle to the inlet recovery ratio. It can be seen that the inlet recovery
ratio has a strong negative effect on the overall thermal efficiency. It should be pointed out that the
loss in thermal efficiency is more significant than the gain in propulsive efficiency [4]. As a result, the
overall impact is a reduction in the overall efficiency of the propulsion system.

A related experiment which was conducted by NASA on the influence of wall boundary layer on
the performance of an axial flow fan rotor [5] had shown that fan peak efficiency ηH reduces as much
Q1
as 2.5 % when a 18 inch spoiler was used. At the peak efficiency flow quantity coefficient nD 3 , the
measured displacement boundary layer thickness was found to be 0.069 inch and 0.060 inch for the
inner and outer surface respectively. The flow quantity coefficient is defined as ratio of the volumet-
ric flow rate Q1 to the product of the rotation speed n multiplied by the cube of the diameter D.
Q1
At the low blade loadings,(that is at high value of nD 3 ), the thickest boundary layer reduced the
efficiency approximately 8 %. It was concluded that the overall loss in efficiency may possibly be re-
duced by decreasing the blade pitch angle in the boundary layer to conform to the upstream velocity
profile. Essentially, the increase in the boundary layer thickness leads to a reduction of the axial
velocity, resulting in an increase in the flow incidence angle. Therefore, the blade pitch angle should
be reduce to better match the flow direction. The results of this experiment is illustrated in figure 2.10.

Figure 2.10: Plot of total pressure rise efficiency versus flow quantity coefficient [5]

In another related experiment, Kimzey [6] modelled the influence of inlet total pressure distortion on
the stability of a XC-1 compressor, by varying the distortion in a 60◦ sector, located on the bottom
of the inlet. The results are shown in figure 2.11. In the figure, N is the compressor rotor speed, P is

13
Master Thesis Report Literature Review

the total pressure, W is the circumferential velocity component and δ is the ratio of the compressor
entry total pressure to the standard sea level static pressure. It can be seen that the stability margin
of all the compressors tested decreases as the distortion amplitude increases.

Figure 2.11: Comparison of the Influence of distortion pattern amplitude on the stability margin
between experimental and computational results [6]

Finally, the overall influence of boundary layer ingestion on the engine design and cycle performance
can be summarised as presented below. An increase in ingestion of the fuselage boundary layer into
the engine leads to the following outcome:

• Reduced specific thrust


• Reduced thrust to weight ratio (since more mass flow is required to achieve the same thrust
level)
• Increased fan diameter
• Increased core size
• Reduced thermal efficiency
• Increased propulsive efficiency
• Reduced overall engine efficiency
• Reduced stability margin

14
Master Thesis Report Literature Review

2.3 Blade Parameterization, Construction and Performance Char-


acterization
After introducing the concept of boundary layer ingestion and analysing the influence of BLI on
engine design and cycle performance, this section of the report will present a brief introduction on the
parametrisation and construction of axial blades (used in the Blade Modeller) as well as the parameters
used to characterise the performance of turbomachinery cascade.

2.3.1 2D and 3D Axial Blade Parameterization and Construction


The parameterisation of a complex 3D axial blade such as the NASA Rotor 67 starts from the
parametrisation of a 2D Profile. After the construction of 2D profiles at various span positions,
the profiles are then stacked together to form a 3D blade shape. Depending on the type of stacking
distribution, the blade can be shaped to have lean, sweep or both characteristics. An illustration of
the commonly used definition for blade profile geometry is shown in figure 2.12.

Figure 2.12: Definitions used in characterising 2D Blade Profile

Blade parameterization is an important design tool to be applied in blade optimization. The prudent
selection of relevant geometry parameters is one of the most critical aspects of any shape optimization
procedures [8]. A good choice of the selected parameters and its range should allow the blade shape
to be described accurately. However, the total number of parameters as well as the range of each
individual parameter needs to be limited in order to reduce the search space. Simplifications should
be applied to the optimization problem by omitting parameters that have little or no impact on

15
Master Thesis Report Literature Review

the performance of the geometry. Most design of shapes rely on curves. In particular, Beizer and
BSpline curves are most suited for shape parameterization [8]. These curves have simple mathematical
formulation and their characteristics are strongly coupled to the underlying polygon of control points
[8].

Figure 2.13: Camberline Parameterization [7] Figure 2.14: Effect of w on camberline shape [7]

The parameterization of a 2D Blade profile begins with the definition of the camberline. This is
illustrated in figure 2.13. The camberline can be described by the following parameters; the inlet
blade metal angle βin , the exit blade metal angle βout , the axial chord length Cax and finally the
stagger angle γ. The leading edge position is first specified by the user. Using the axial chord length
and the stagger angle, the trailing edge position can be derived. The intersection of the inlet and
outlet line tangential to the inlet and outlet blade metal angle defines the position of the mid control
point. For this camberline, the points located at the LE, TE as well as the middle position defines
the control points of the beizer curve. Mathematically, the position of the trailing edge point can be
derived as follows:

xte = xle + Cax (2.41)

yte = yle − Cax tan γ (2.42)


By adjusting the weight w of the mid control point, the desired curvature of the beizer curve can be
obtained. This is illustrated in figure 2.14. The other possibility to construct the camberline curve in
the Blade Modeller is to use four control points instead. In this method, the position of the two mid
control points are controlled by the t1 and t2 parameter which moves the control points linearly along
the inner and outer line tangential to the inlet and outlet blade metal angle.
Mathematically, the location of these mid control points can be calculated as shown in equation 2.43
[7].

xp1 = xle + t1 (xmid − xle ) (2.43a)


yp1 = yle + t1 (ymid − yle ) (2.43b)
xp2 = xmid + (1 − t2 )(xte − xmid ) (2.43c)
yp2 = xmid + (1 − t2 )(yte − ymid ) (2.43d)

The parameters t1 and t2 are illustrated in figure 2.15 respectively. Once the camberline is con-
structed, the next step is to specify a thickness distribution for both the suction and pressure side

16
Master Thesis Report Literature Review

Figure 2.16: Blade Thickness Definition [8]

Figure 2.15: Use of two control points to adjust


camberline [7]

curve. The thickness distribution is specified by providing a list of points for which a 2nd degree
BSpline curve can be constructed. It should be pointed out that these points specifically control the
thickness and not the curvature of the profile. The x values of these points represent a normalized
coordinate system ranging from 0 to 1, where 0 and 1 represent the LE and TE position respectively.
The y values would then represent the thickness. Optional parameters such as the number of points
that control the thickness as well as a thickness factor is included in the blade modeller. The thickness
factor is a parameter which scales the thickness value of all the distributed points along the camberline.

The distribution of these points along the camberline can be specified by using an arctangent or
simply an equispaced function. The arctangent method enables more points to be specified near the
leading and trailing edge position. In addition, a stretching factor parameter is available which en-
ables the user to control the actual spacing distribution of the points. Mathematically, the arctangent
relation between a point distributed along the normalized axial coordinate and the ’u’ parameter is
expressed as follows:

u = arctan (bx) (2.44)


Where ’b’ is the stretching parameter. A larger value of the stretching factor results in more points
being distributed near the leading and trailing edge position. The variation of the points distributed
along the ’u’ parameter with different stretching factor is illustrated in figure 2.17. Once the distri-
bution of the points have been determined, the distributed points would then offset perpendicularly
from the camberline with reference to the thickness value specified earlier. This process is illustrated
in figure 2.18.
For both the suction and pressure side curves, a B-Spline curve is used to represent the thickness
distribution. The distributed points along the curve represent the thickness value used to shape the
pressure and suction side curve. Therefore, by controlling the control points of the B-Spine curve, the
user will be able to directly control the thickness distribution of either the suction or pressure side.
Finally, the last aspect to consider is the blade profile’s leading and trailing edge shape. Within the
blade modeller, the edge can be constructed by using either the radius values of the B-Spline curve (see
figure 2.18) or a circular arc. The circular arc is constructed using an additional B-Spline curve based
on three control points. In order to obtain the 2D blade profile, the circular arc is then connected to
both the suction and pressure side curves taking G1 continuity into account.

After all the 2D blade profiles have been constructed, the 3D blade can then be generated through the

17
Master Thesis Report Literature Review

Figure 2.17: Influence of arctangent stretching


Figure 2.18: Construction of Blade Profile Suc-
parameter on points distribution along
tion/Pressure Side [7]
’u’ parameter [7]

construction of a B-Spline surface along the span-wise direction. Within the blade modeller, the blade
can be either stacked through the center of gravity of the mid control profile or through the leading
edge of each control profile. During the generation of the B spline surfaces for the suction/pressure
side as well as the LE/TE circular arc surfaces, the total number of control points along each control
profile as well as the degree of all the spanwise B-Spline curves has to be kept constant. Some other
examples of blade stacking can be found in the paper of Tom Verstraete [8]. In his paper, the stacking
process of an axial turbine is being demonstrated. This is illustrated in figure 2.19

Figure 2.19: Stacking of a 3D Axial Turbine Blade [8]

The rotor blade profiles are stacked via its center of gravity while the vanes are stacked through its
trailing edge points. The straight stacked shape is then altered by applying a lean or sweep distri-
bution to several intermediate profiles between the hub and shroud. The lean and sweep law can be
defined using beizer curves for which the control points are the optimization parameters [8]. In other
cases, the lean and sweep law can also be defined by angle and distance parameters so that a straight
forward relation between the lean/sweep parameters and performance parameters can be obtained.

18
Master Thesis Report Literature Review

After the blade profiles are stacked accordingly, the next step is to generate B-Spline surfaces for
the pressure/suction side as well as the LE/TE edges. The generation of the B-Spline surfaces de-
pends on a number of parameters. They are the degree, multiplicities as well as the knots. They
control the continuity of the span-wise surface as well as the accuracy with which the surface approxi-
mates the individual control profile. A selection of higher degree leads to a higher continuity but lower
accuracy. Intermediate profiles are constructed based on interpolation using the generated span-wise
distribution of the various design parameters (blade inlet/outlet metal angles, stagger angle, leading
edge position, axial chord, thickness distribution, etc).

For any typical turbomachinery blade, the hub and tip control profiles may be flared. Flaring is
defined as the inclination of the blade profile with respect to a certain reference axis. In the case of
the NASA Rotor 67, the control profiles at the hub and tip location are flared with respect to the
axial direction (x-axis). Flaring is controlled in the Blade Modeller using a 2D meridional channel.

2.3.2 Performance characterisation of Turbomachinery Cascade


Firstly, a short description of the various loss mechanisms and flow phenomenon that initiate these
losses in turbomachinery cascade will be explained. Then, the parameters used to characterize the
performance of turbomachinery cascade will be described in detail.

Loss Mechanism

A review of the most common type of losses are presented below:

• Profile Losses: The profile loss consists of a combination of the wall shear stress losses and the
trailing edge mixing losses. Since both losses can hardly be separated in experimental results,
they are often combined under the term known as the profile losses. Wall shear stress losses
originate in the boundary layer. Trailing edge mixing losses occur when the boundary layer
mixes with the free stream. These losses can be minimised by increasing the surface quality
and decreasing the trailing edge radius. However, there are manufacturing limitations to both
solutions.

• Shock Losses: Shock losses can occur especially in fans that operate in transonic flow field. It
contributes greatly to the overall entropy rise and has the same order of magnitude of losses as
that of the profile loss.

• EndWall/Secondary Losses Secondary losses are generated by the build up of the endwall
boundary layer, secondary flow as well as the tip clearance. Secondary flow can be described as
flow that flows at right angle to the intended primary flow. They are mostly caused by a pressure
gradient that is generated between the pressure and suction side of two adjacent blades within a
flow channel. This pressure gradient is a function of the blade loading present on each blade. A
positive pressure gradient gives rise to an increase in the boundary layer thickness. Secondary
flow generates vibration, noise and flutter because of the unsteady pressure field between the
blades and the rotor-stator interaction.

• Tip Leakage Flow Losses: Tip leakage losses are generated as result of fluid flowing from
the pressure side to the suction side of the blade. This occurs because of the gap that is present
between the annulus wall and the blade. This leakage flow induces three dimensionality effects
such as the formation of vortices.

19
Master Thesis Report Literature Review

Other than these major classification of losses in turbomachinery, it is prudent to look into the various
mechanisms that initiate these losses. The two main flow phenomenon concerned are separation and
flow deviation.

• Separation: Separation is characterized by the flow completely detaching from the blade profile.
Prior to separation, the boundary layer grows very thick. At the onset of the separation bubble,
the fluid particles located at the wall has no velocity relative to the blade. In the bubble itself,
the fluid is turned to flow in the opposite direction due to the adverse pressure gradient, leading
to an area of recirculation. Separation can occur in various area of the blade, although it
most commonly occurs on the aft suction side of the blade. Other than the usual separation
at the trailing edge of the airfoil, leading edge separation is another uncommon but localised
phenomenon. It can occur after an excessive high suction peak on the suction side of the airfoil or
an incorrect blade metal inlet angle onto the pressure side of the airfoil. In the case of the suction
side, the flow tends to reattach mainly because of the strong acceleration on the suction side. On
the pressure side, such a separation will lead to an extensive re-circulation zone. Separation in
transonic fan is promoted by the shock boundary layer interaction due to high pressure gradient
across the shockwave. Detailed analysis reveals a significant growth of the boundary layer in the
shock region. If the shock is strong enough, separation can occur.

• Flow Deviation: Flow deviation represents the discrepancies in angle between the inlet/outlet
flow and its respective blade metal angle. However, deviation is not easy to predict by analytical
expression. As the incidence angle increases, the flow over the blade deteriorates until the point
of stall. It is possible to achieve un-disturbed flow over the blade with zero degree incidences
when operating at design rotation speed.

An illustration of some of the loss mechanisms described earlier is shown in figure 2.20, 2.21 and 2.22.

Figure 2.20: End wall losses [9] Figure 2.21: Secondary flow losses [9]

After a discussion on the various loss mechanism, the next part of this subsection will describe the
various parameters that characterise the performance of turbomachinery cascade. Other than that,
the distortion coefficient parameter that is used to characterised flow distortion will be explained as
well.
Characterization of Turbomachinery Performance

An important performance parameter that is used to characterise the loss in turbomachinery rotor
is known as the stagnation pressure loss coefficient. It is defined as shown in equation 2.45. It is

20
Master Thesis Report Literature Review

Figure 2.22: Tip leakage losses [9]

noted that the last terms in the denominator can be either the static pressure at the inlet or outlet
respectively.

P01,rel − P02,rel
δloss = (2.45a)
P01,rel − P2
P01,rel − P02,rel
δloss = (2.45b)
P01,rel − P1

The bar sign that appears above each quantities implies that mass averaged values have been used for
both the inlet and outlet condition. The total to total isentropic efficiency of the fan (NASA Rotor
67) can be defined as follows:

h02ss − h01
(ηtt )f an = (2.46)
h02 − h01
Where ss defines the total condition for which the process is isentropic. h01 represents the total
condition at the inlet of the rotor, h02 represents the total condition at the outlet of the rotor. h02ss
represents the isentropic total condition at the outlet of the rotor. Since the kinetic energy of the flow
after the rotor is being used, the total to total efficiency is used instead of the total to static efficiency.
Another useful measure of the rotor efficiency is to compute the entropy change across the rotor. The
entropy change across the rotor can be derived starting from the second law of thermodynamics:

dp
T ds = dh − (2.47)
ρ
Considering the process of a perfect gas passing through the rotor at constant relative stagnation
enthalpy (h01rel = h02rel ), the second law of thermodynamics can be expressed as follows:

dp
T ds = (2.48)
ρ
p01,rel − p02,rel
T01,rel ∆srotor = (2.49)
ρ01,rel

21
Master Thesis Report Literature Review

Where the negative sign in T ds disappears since the relative stagnation pressure is decreasing through
the rotor. It should be pointed out that since the rotation speed is constant, a constant relative
stagnation enthalpy implies a constant rothalpy across the rotor. Substituting equation 2.45 and
using the ideal gas law relation, the entropy change across the rotor can be expressed as follows:

p01,rel − p02,rel R(p01,rel − p02,rel ) Rδloss (P01,rel − P1 )


∆srotor = = = (2.50)
ρ01,rel T01,rel P01,rel P01,rel

P1
∆srotor = Rδloss (1 − ) (2.51)
P01,rel
Inlet Distortion

Inlet total pressure distortion can be classified as either circumferential or radial with respect to the
machine axis. Circumferential distortions are considered to be more critical than radial distortions as
it produces a disturbance normal to the airfoil motion, resulting in a change in the angle of attack
and stall margin. An illustration of both type of distortion is shown in figure 2.23.

Figure 2.23: Type of Inlet Distortion

In order to assess the circumferential type inlet distortion, the distortion analysis methodology de-
scribed by the Society of Automative Engineers document ARP 142019 was used. The concept of
the circumferential distortion intensity as well as the circumferential distortion extent are being being
introduced. The circumferential distortion extend for a given ring, (θi )¯, is the angular distance in
degrees for which the local total pressure falls below the ring average total pressure, where i refers to
a single ring along the contour of the circle (See figure 2.24).

(θi )¯ = θ2i − θ1i (2.52)

The circumferential distortion intensity, ( ∆p c


p )i , which quantifies the amount of total pressure variation
around the ring is computed as follows:

∆pc Pav − Pav,low


( )i = ( )i (2.53)
p Pav
Where:

22
Master Thesis Report Literature Review

1 n=N
Z 360
1 X
(Pav )i = P (θ)i dθ = P (θ)i dθ (2.54)
360 0 N ∗ dθ n=1
N is the total number of sampling points taken over a 360◦ sector and dθ is the angular size of each
sampling point.
Z θ2,i N dθ
2
1 1 X
(Pav,low )i = P (θ)i dθ = P (θ)i dθ (2.55)
(θi )¯ θ1,i (θi )¯ N dθ
1

N1 is the starting index position where the total pressure falls below the (Pav )i and N2 is the index
position where the total pressure rises above the (Pav )i value. A low total pressure region occurs when
the local total pressure P (θ)i lies below the ring average total pressure, Pav,i . If two low total pressure
regions occur within 25◦ of each other, it is treated as a single once-per-revolution low total pressure
region. If two low total pressure regions are separated by more than 25◦ , the circumferential distortion
intensity is taken to be the maximum of the two. An illustration of the various definitions presented
above is shown in figure 2.24.

Figure 2.24: Definitions for Pav and Pav,low

2.4 Capabilities and Function of SU 2


In this section, a brief introduction on the capabilities and function of the Stanford University Un-
structured (SU 2 ) CFD Code will be described. This is a tool that was used throughout this research
work. SU 2 is a computational analysis and design software that is being developed to solve complex
multi-physics analysis and optimization task using arbitrary unstructured meshes. Fundamentally, it
is able to discretize and solve problems described by Partial Differential Equation (PDE) as well as
PDE constrained optimization problems such as optimal shape design.

Multidisciplinary problem usually involves the interaction of several separate software module that
represent different physical models. In order to avoid integration problems and the loss of efficiency,
having a common set of numerical solutions algorithms and code architecture is beneficial [13]. There-
fore, the ability to integrate multiphysics models, parameterize the system as well as to optimise the
system is critical. For the purpose of optimization, it is necessary to obtain the gradient of the re-
sponse to variation with respect to a large number of design parameters. As such, SU2 is in-built

23
Master Thesis Report Literature Review

with an adjoint solver that computes the gradient of the response. Solutions from the adjoint solver
are then used to compute the functional driven mesh adaptation techniques [13]. In SU 2 , both the
continuous and discrete adjoint equations have been implemented. In the continuous approach, the
adjoint equations are derived from the governing partial differential equation and then subsequently
discretized. Wheras in the discrete approach, the adjoint equations are directly derived from the
discretized governing equation[13].

The SU 2 suite is composed of several analysis modules as described below [13]:

• SU 2 CFD: The main Partial Differential Equation solution module, which is primarily an Euler
or Reynolds Navier Stoke Solver, also includes the adjoint equations for many of the supported
governing equation system.

• SU 2 DDC: The Domain Decomposition Code is used for computation which involves multiple
processors. The specified volumetric grid is partitioned for use with several other core tools in
parallel.

• SU 2 MAC: The Mesh Adaptation Code is used to refine unstructured computational meshes
so as to improve the accuracy of the predictions. Based on the analysis of converged flow, adjoint
or linearized solution, grid adaptation is applied to strategically refine the mesh about the key
flow features.

• SU 2 GPC: The Gradient Projection Code is used in the calculation of sensitivities for the
purpose of optimization and uncertainty quantification. It uses the surface sensitivities computed
using SU2 CFD, the flow solution and the definition of the geometric design variable to evaluate
the derivative of a particular function.

• SU 2 MDC: The Mesh Deformation Code is used to perturb an existing unstructured volume
mesh to a new surface geometry. It computes the geometric deformation of a surface within the
computation mesh and the surrounding volumetric grid.

• SU 2 PBC: The Periodic Boundary Code is used for the solutions of PDE with periodic bound-
ary condition. A new mesh containing the proper communication structure between the periodic
faces is generated.

• SU 2 SMC: The Sliding Mesh Code enables solution on meshes that slide past each other. Ghost
cells in the computation domain are created for performing simulations with sliding surfaces.
A new multi zone mesh containing the proper communication structure between the sliding
interfaces is generated.

SU 2 is composed of several groups of python scripts which enable the coupling of the various modules
for performing complex activities including design optimization and adaptive grid refinement.

• High Fidelity Analysis Script: These scripts are designed to enhance the flexibility of the
SU2 framework. They simplify the execution of parallel tasks, grid adaptation or the interface
with other softwares.

• Optimal Shape Design Script: These scripts are designed to automate the shape design
process

• Automatic Differentiation Script: These scripts are designed to create differentiated versions
of the code section in SU2.

24
Master Thesis Report Literature Review

Brief Description of Modelling Equations in SU 2


Several form of the Reynold Average Navier Stokes (RANS) equation have been implemented in SU 2 .
They include the compressible, incompressible, Arbitrary Lagrangian-Eulerian,etc. As a subset of
the RANS equation, both the laminar Navier Stokes as well as the inviscid Euler equations are also
available by disabling turbulence modelling and removing viscosity respectively. In the case of turbu-
lence modelling, both the Spalart-Allmaras (S-A) model as well as the Menter Shear Stress Transport
(SST) model have been implemented. The Menter SST turbulence model is a two equation model that
blends the traditional k − ω and k −  model. For the modelling of laminar to turbulence transition,
the γ − Reθt transition model which is an adaptation of the model of Langtry and Menter [14] have
been implemented. This model is an correlation based model that augments the Spalart-Allmaras
turbulence model with two equations. For simulations that involve steady rotation of aerodynamic
bodies, it is advantageous to transform the system of flow equation into a reference frame (a.k.a ro-
tating frame) that rotates with the body of interest [13]. As such, a flow field that is unsteady when
viewed from the inertial frame can be solved in a steady manner without the need for grid motion.
This is achieved by applying a transformation of the RANS equation from the inertial reference frame
to the rotating frame. In the case that the rotation of the aerodynamic bodies includes unsteady flow
phenomenon, solutions which includes grid motion is necessary. Therefore, the Arbitrary Lagrangian-
Eulerian (ALE) formulation is also implemented in SU 2 .

For wave related equations, both the Ffowcs Williams - Hawking (FW-H) equation and the linear
elasticity equation was implemented for aeroacoustic and linear elastic small deformation problems
respectively. Other physical models such as the heat equation, Gauss’s law as well as the plasma
equation have also been implemented in SU 2 . The Gauss law is part of the Maxwell equations and is
used to relate the distribution of electric charge to the resulting electric field [13]. The plasma flow
equation is tightly coupled with the gauss law for the calculation of electric field generated by any
local separation of charge[13].

In the context of this research, the compressible/incompressible RANS equation as well as the Menter
SST turbulence model have been mainly used for the 2D and 3D simulations.

25
Master Thesis Report Research Goals and Objectives

3|Research Goals and Objectives

This chapter will look into the research objective, goals, motivation and scope of the thesis.

3.1 Research Question and Objective

The main research question of this thesis is as follows:

• How does Boundary Layer Ingestion influences the performance of an engine fan/compressor
blade such as the NASA Rotor 67.
• Which fluid dynamics phenomena are different as compared to the uniform inlet flow?

The objective of this thesis is therefore "to conduct a thorough analysis of the flow field of a representative
3D Rotor (such as the NASA Rotor 67) under both uniform and non uniform inflow conditions using
Computational Fluid Dynamics (CFD) tool (e.g SU2 and ANSYS CFX).

In order to achieve this objective, a number of preliminary goals has to be accomplished:

• To parameterize and reconstruct the rotor blade


• To model both uniform and non uniform boundary conditions in ANSYS CFX.
• To analyse the rotor performance and flow physics under both BLI and non BLI boundary
conditions

In order to achieve the aforementioned goals, there is a number of sub-goals to be achieved:

• To extract the original blade geometry from the given CGNS mesh file
• To obtain CFD results from the Aerodynamic Interface Plane (AIP) of the provided inlet duct
geometry and compare them with its inlet condition
• To develop a geometry fitting algorithm to fit all the NASA Rotor 67 2D profiles with BSpline
curves.
• To reconstruct the 3D blade using BSpline surfaces
• To construct meshes of both the 2D and 3D geometries using SALOME, UMG2 and UMG3
• To obtain the appropriate 2D and 3D boundary conditions for the simulations
• To perform CFD simulation and analysis on the 2D/3D geometries

In addition to these sub-goals, CFD related questions such as the required size of the domain , optimal
solver settings for the time and spatial integration scheme as well as mesh quality should be addressed.

26
Master Thesis Report Research Goals and Objectives

3.2 Motivation
From the literature study that was conducted, it was found that there exists a large number of de-
velopment in various aspect of BLI for a BWB. Most of the literature that were found focuses on the
system benefit of BLI. A small number of papers looked specifically into the inlet design for a em-
bedded engine configuration. As compared to the podded engine configuration, it was found that the
embedded engine configuration can provide an overall reduction in fuel consumption due to reduced
momentum deficit. However, the benefits of BLI have shown to be very sensitive to the magnitude of
the fan and duct losses. Therefore, the design of ingestion tolerant fan is critical in achieving increased
fuel efficiency.

Secondly, few literature exists that investigate the flow phenomenon on the rotor blade and on the
parametric design of ingestion tolerant fan blades for embedded engines configurations on a BWB.
Therefore, the main goal of this thesis is to investigate the flow physics and performance of the rotor
under non uniform inflow condition and then to propose possible design improvement for the fan blade
so as to minimise entropy loss.

3.3 Scope
The scope of this thesis is to focus on the continued development of an in-house blade modeller which
is subsequently used to parameterize and reconstruct the NASA Rotor 67. After the re-construction of
the NASA Rotor, the other aspect of this thesis is to focus on the coupling and development of meshing
templates for 2D/3D geometries based on an open source meshing software, SALOME. Thereafter,
the next part of this thesis will focus on the development of a mesh writer that prints SU 2 Mesh
file format based on extracted information from SALOME. When the necessary mesh files have been
generated, the final aspect of this thesis will focus on the simulation and analysis of various blade
geometries. Taking into consideration the total time frame of this project, the scope of this thesis will
not include the development of an optimization algorithm for the blade modeller.

27
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

4|Parameterisation and Reconstruction


of the NASA Rotor 67

In this chapter, the actual reconstruction process of the NASA Rotor 67 will be demonstrated. Starting
from the process of geometry extraction to obtaining the final 3D blade shape, the reconstruction
process of the NASA Rotor 67 will be demonstrated. Intermediate steps such as the fitting of the
blade camberline as well as the development of a geometry fitting algorithm for fitting the 2D profiles
will be explained. After the blade is reconstructed, a comparison of the reconstructed and original
NASA Rotor 67 as well as the design trend of the NASA Rotor 67 will be presented

4.1 Step 1: Extraction of the blade geometry from the CGNS mesh
In this section, the extraction of the rotor geometry from the CGNS mesh is illustrated. Using tecplot,
2d profile slices are extracted manually from the mesh file.

Y Y

X
Z Z
X

Figure 4.1: Original NASA Rotor 67 Figure 4.2: Sliced NASA Rotor 67

As seen in figure 4.2, several horizontal slices were made within the mid location of the blade. Near
the hub and tip location, the slices made were flared according the geometrical shape of the rotor. An
illustration of the various slices made along the blade is illustrated in figure 4.3. In total, 13 control
profiles were extracted from the CGNS mesh of the NASA Rotor 67.

28
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

Figure 4.3: Sliced 2D Profiles of the NASA Rotor 67

4.2 Step 2: Fitting the camberline of the 2D Profiles


After extracting the profiles from the CGNS mesh, the next step is to fit the individual 2d profiles.
This process begins with fitting the camberline of each 2d profile. The camberline fitting process is
a mix of automatic and manual work. The goal is to first use the interface of the blade modeller to
fit the 2d profiles and then to later develop a geometry fitting algorithm which extracts information
from the discretized points and refit the camberline based on the extracted information.

Using the ’plane cutting’ feature of tecplot, each of the 2d slices are discretized into many points
along the contour of the profile. Before they can be imported into the blade modeller (BM), an
import and export function/interface for the discretized 2d points was developed for the BM. After
the points were imported, a separate script was developed in such a way that the user is able to
interactively select the point closest to the leading and trailing edge of these points for the formation
of the initial camberline. As previously mentioned, the camberline is defined by the leading edge
position, axial chord length the stagger angle as well as the blade metal angle at the inlet and outlet
of the profile (See figure 2.13). From the selected leading and trailing edge position of the discretized
points, the BM would then recalculate the axial chord length and stagger angle respectively. The
blade inlet/outlet metal angle were then determined by the following approach in the script belonging
to the class of the geometry fitting algorithm:

• Calculate the tangent values of the points closest to the leading and trailing edge for both the
suction and pressure side curve of the 2D profiles

• Calculate the difference in tangent values between adjacent pair of points for both the suction
and pressure side curve

29
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

• Determine the index value of the point at which the change in tangent value between adjacent
point reaches the minimum for both the suction and pressure side curve

• Using the index values of both the pressure and suction side curves, compute the average tangent
values for both the blade metal inlet and outlet angle.

With these information, the BM will be able to calculate the initial camberline. Interactively, the
steps taken to construct the camberline on the blade modeller is as follows:

• Import and plot the original points on the BM

• Reorder the points using the selected first and second index values

• Specify the location of the suction side of the blade and the points closest to the leading and
trailing edge of the blade. Use the ’Extract Main Info’ command to determine the initial cam-
berline

• Re-plot the ordered points with separate numbering for the pressure and suction side curve
respectively.

Once the ’Extract Main Info’ command has been initiated, the computed camberline will then appear
on the BM interface together with the plotted points. Further manual adjustment of the camberline’s
leading edge position, trailing edge position as well as its mid control point are possible on the BM
interface. Within the BM, the ’pos LE’ and ’pos TE’ enables minor adjustment away from the selected
leading and trailing edge marker. An illustration of the plot of original and re-ordered points are shown
in figure 4.4 and 4.5 respectively. The plot of the separately reordered suction and pressure side points
as well as the plot of points with initial camberline are shown in figure 4.6 and 4.7.

−0.028

−0.028
6
−0.029
68
69
70
5 6
145 146
144 147 135 136 66
67
7 130 131 132 133 134 6465
−0.030 5253 54 55 63
143 −0.030
56 57 58 59 60 61 62
51
142 50
49
141 48
47
140 46
−0.031 139 45
138 44
−0.032 43
137 42
41
40
8
39
−0.032 122 38
37
123 36
35
124
125 34
−0.034 33
126 32

−0.033 127
128
129
4
−0.089 −0.088 −0.087 −0.086 −0.085 −0.084 −0.083
−0.090 −0.088 −0.086121 −0.084 −0.082 −0.080
120
119
118
117

Figure 4.4: Plot of Original Point Coordinates


Figure 4.5: Plot of Re-ordered Points Coordinates

Before a discussion is made on the fitting algorithm for the pressure and suction side curve, the
development of the ordering algorithm will be presented. This algorithm is used earlier when the
points are re-ordered for the purpose of reconstructing the camberline and the pressure/suction side
curves. From an initial starting point selected by the user, the ordering algorithm is used to find the
next adjacent point. In order to find the adjacent point, a tangential line is projected from the current
point using the computed discretized axial distance (dx) and its derivative of the previous pair of

30
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

−0.028

18
−0.029
17
16
15
1314
1 2 3 4 12
5 6 10 11
7 8 9
−0.030 0
1

2
3
−0.031 4
5
6
7
−0.032 8
9
10
11
12
−0.033 13
14
15
16
17
−0.089 −0.088 −0.087 −0.086 −0.085 −0.084 −0.083 −0.082 18
19
20
21 22
Figure 4.7: Plot of Re-ordered Points with camberline
Figure 4.6: Plot of Re-ordered Points Coordinates
with sep. suction and pressure side indexes

points. The next point is then selected based on the closest distance from the projected point. This
process is illustrated in figure 4.8.

Figure 4.8: Illustration of projection algorithm method

31
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

4.3 Step 3: Fitting the Suction/Pressure side curves of the 2D Pro-


file
In this section, the fitting algorithm that is used to determine the closest fit of the BSpline curve for the
points distributed along the contour of the blade profile will be explained. As previously mentioned,
the pressure and suction side curve are constructed using a separate list of distributed points where
the curvature is determined by its thickness distribution. This thickness distribution is then directly
controlled by the use of a BSpline curve where the control points are the variables. Using such an
approach, the user will be able to adjust the thickness distribution of the pressure/suction side curve
using a minimum number of variables.

In order to ensure a good fit between the pressure/suction curve with the distributed point along
the contour of the blade profile, the geometry fitting algorithm was developed. This algorithm is
based on the minimisation of error between the distributed points of the actual blade profile and the
pressure/suction side curve formulated by the control points which controls the thickness distribution.
Using the ’x’ coordinate of the suction/pressure side points, the algorithm searches for the closest ’u’
parameter on the curve. The ’u’ parameter is a normalized coordinate that moves along the BSpline
curve. The ’y’ coordinate of the curve is then determined by using the associated ’u’ parameter. The
algorithm then computes the differences in ’y’ coordinate between a particular point of interest on the
distributed points surrounding the blade contour and the segment of the curve belonging to the asso-
ciated ’u’ parameter. An error function is then composed by using the summation of the error along
the entire blade contour. The error function is then minimised by using the gradient based optimizer
in the Scipy library of python. Both the simplex and sequential least square quadratic programming
algorithm is being implemented in the BM. User can decide to use either the mid control points or
all the control points for the optimization. In the case that the ’x’ position of the adjacent point is
lesser than the current point, the algorithm readjust the value of the starting U parameter so that the
associated ’u’ parameter closest to the current ’x’ coordinate can be found.

Figure 4.9: Illustration of geometry fitting algorithm method

The control points belonging to the leading and trailing edge has to be adjusted manually to obtain a
good fit before the optimization algorithm can be applied. An illustration of the blade profile before
and after optimization is presented in figure 4.10 and 4.11.
Comparing figure 4.10 and 4.11, it can be seen that the pressure and suction side curves exhibits a
better fitting at the mid region after the optimization process. The only drawback thus far is that the
implementation of using all control points for geometry fitting is not sufficiently robust.

32
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

Figure 4.10: Illustration of profile before optimization

Figure 4.11: Illustration of profile after optimization

4.4 Step 4: Stacking of Blade Profiles and Obtaining 3D Blade


Shape
After all the 2D profiles have been constructed, the next step is to reassemble all the control profiles
in the BM and then to generate a BSpline surface over the profiles. This process is illustrated in figure
4.12 and 4.13 respectively.
It should be pointed that flaring was applied to both the hub and tip control profiles since the geometry
that was extracted was already flared at an angle of 15◦ and 27.7◦ at the hub and tip respectively. In
order to generate the BSpline surfaces, the following points have to be taken into account. Firstly, the
number of control points used to generate the suction and pressure side of each control profile has to
be kept constant. This also applies to both the surfaces at the leading and trailing edge if a circular
arc is being used. In this case, the profiles are stacked along the leading edge line. After the suction
and pressure side surfaces have been generated, the next step is to form a blade solid by generating

33
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

Figure 4.12: Illustration of stacked profiles Figure 4.13: Illustration of full profiles

surfaces at the hub/tip and then fusing all the surfaces together using the FreeCAD modules.

4.5 Comparison of Actual and Reconstructed Rotor Geometry


In this section, a visual comparison of the reconstructed rotor and the original rotor is made. This is
presented in 5.11 and 5.12.

Y Z

X Y
Z X

Figure 4.14: Original NASA Rotor 67 Figure 4.15: Reconstructed NASA Rotor 67

Comparing figure 5.11 and 5.12, it can be seen that there are minor differences in the tip curvature

34
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

between the reconstructed and actual rotor. This is because only a straight slice of the face closest to
the tip section can be executed during the geometry extraction process. There could also be minor
differences in the curvature of the blade leading and trailing edge although this is difficult to inspect
visually. A comparison of their 3D results will be demostrated in the later part of this thesis.

4.6 Design Trend of the NASA Rotor 67


By using the BM to reconstruct the rotor, the complete design information of the rotor can be extracted
and analysed. A discussion of their general design trend will be presented in this section. In figure
4.16, the angle distribution of the NASA Rotor 67 is being presented.

NASA Rotor 67 Design Data


Beta In Distribution
100 Beta Out Distribution
Stagger Distribution
Beta Out - Beta In

50
Angle Distribution [ ◦ ]

−50

0.0 0.2 0.4 0.6 0.8 1.0


Normalized Span

Figure 4.16: Angle Distribution of NASA Rotor 67

The following trend can be observed. Firstly, it can be seen that the stagger angle increases towards
the tip. This is done mainly to allow for maximum stagger angle at the tip. Maximum stagger angle
at the tip allows for a lower blade loading and tip losses [15]. In the ASME paper written by Yong
Sang Yoon, Seung Jin Song and Hyoun Woo Shin [15], it was found that the pressure rise across
the rotor blade reduces as the stagger angle increases. A reduction in the pressure rise across the
blade implies a reduction in the blade loading. The other general trend is that the angular value
of βout − βin decreases towards the tip while the radial value increases toward the tip. The is done
mainly to keep the enthalpy change and consequently the blade loading across the rotor to be constant.

The other important aspect is the thickness distribution of the rotor. It can be seen that the overall
thickness of the rotor decreases from hub to tip for both the suction and pressure side. Also, the

35
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

leading edge portion of the blade reduces in thickness sharply as the blade profile approaches the tip.
This could be done mainly to reduce profile losses near the tip section due to the much higher relative
speed of the incoming flow.

0.06 Suction Side Thickness Distribution

0.05 Profile 0
Profile 1
Profile 2
0.04
Profile 3
Thickness Distribution

Profile 4
0.03
Profile 5
Profile 6
Profile 7
0.02 Profile 8
Profile 9
Profile 10
0.01 Profile 11
Profile 12

0.00
0.0 0.2 0.4 0.6 0.8 1.0
Non Dimensional Chord Length

Figure 4.17: Thickness Distribution of Suction Side NASA Rotor 67

0.07 Pressure Side Thickness Distribution

0.06
Profile 0
Profile 1
0.05 Profile 2
Profile 3
Thickness Distribution

0.04 Profile 4
Profile 5
Profile 6
0.03 Profile 7
Profile 8
0.02 Profile 9
Profile 10
Profile 11
0.01
Profile 12

0.00
0.0 0.2 0.4 0.6 0.8 1.0
Non Dimensional Chord Length

Figure 4.18: Thickness Distribution of Pressure Side NASA Rotor 67

Other than the reduction in thickness of the blade as the profile approaches the tip, the other geomet-

36
Master Thesis Report Parameterisation and Reconstruction of the NASA Rotor 67

rical trend of the rotor design is the overall reduction in chord length as the blade profile approaches
towards the tip. Other than aerodynamic reasons, a reduction in chord length and thickness of the
blade as the profile approaches the tip implies a smaller centrifugal load. This is advantageous in
terms of the maximum level of vibration as well as the fatigue life of the rotor blade.

0.09 NASA Rotor 67 Design Data


Chord Distribution

0.08
Chord Distribution

0.07

0.06

0.05

0.04
0.0 0.2 0.4 0.6 0.8 1.0
Normalized Span

Figure 4.19: Thickness Distribution of Pressure Side NASA Rotor 67

37
Master Thesis Report Verification of 2D Fitting Results

5|Verification of 2D Fitting Results

In this chapter, the report will focus on the verification of the 2D fitting approach that was used to
reconstruct the rotor. In order to do so, the first part of this chapter will look into the methodology
used to obtain the correct boundary conditions. After that, the second part of this report will focus
on the mesh topology and analysis of the 2D results.

5.1 2D Inlet Boundary Conditions


The 2D boundary conditions were derived based on information provided by the NASA Technical
report 2879 [16]. A summary of the key information regarding the design and geometry data of the
NASA Rotor 67 can be found in table 5.1.
Table 5.1: Design and Geometry Data

Parameters Values Units


Design Rotation Speed 16043 rpm
Inlet Tip Relative Mach No. 1.38 -
Mass Flow 33.25 kg/s
Hub to Tip Ratio 0.375 -
Tip Radius 0.257 m
No. of Rotor Blade 22 -

Other than using the above information, the static pressure, temperature and density were assumed
to be at ambient condition. Also, the air is considered to be an ideal gas.

Table 5.2: Atmospheric Data

Parameters Values Units


Pressure 101325 Pascal
Temperature 288.15 K
Density 1.225 kg/m3
γair 1.4 -

The calculation procedure for the boundary conditions can be split into both inlet and outlet condition
respectively. The inlet boundary condition was first calculated assuming a free vortex design rule. This
imply that the absolute axial velocity along the span will be constant and that the free vortex constant
is equal to the product of the absolute tangential velocity and the radius. Using the information
provided in table 5.1, the hub radius can be computed as follows:

Rhub
Rhub = ∗ Rtip (5.1)
Rtip

38
Master Thesis Report Verification of 2D Fitting Results

The absolute axial velocity can then calculated as follows:

M assF low M assF low


Cax = = (5.2)
Area 2πRmean H
Using the local speed of sound, the inlet tip relative velocity can be calculated.

Vrel,tip = Mrel,tip ∗ alocal (5.3)


The relative stagnation pressure and temperature can then be computed using the isentropic relation
as follows:
γair − 1 2 γ γair−1
P01,rel = P1 ∗ [1 + M1 ] air (5.4)
2
γair − 1 2
T01,rel = T01,rel ∗ [1 + M1 ] (5.5)
2
Using the inlet relative velocity and the constant axial velocity Wax = Cax , the relative tangential
velocity can be computed:
q
Wθ = W 2 − Wax
2 (5.6)
Using the peripheral speed U = Rm,1 Ωdesign , the absolute tangential speed can be calculated as follows:

Cθ,1 = Wθ − U (5.7)
From the computed absolute tangential velocity and tip radius, the free vortex constant and the
absolute tangential velocity at other radial position can be found.

λf reevortex = Rtip Cθ,1 (5.8)

λf reevortex
Cθ = (5.9)
Rother
Using the value of Cθ , axial velocity Cax as well as the peripheral speed U , the value of the absolute
and relative velocity at other radial position can be computed.
q
C= 2 + C2
Cax (5.10)
θ

q q
W = 2 + W2 =
Wax 2 + (C + U )2
Cax (5.11)
θ θ

With the given new value of W , the value of P01,rel , T01,rel and βin at a new radial position can be
computed using equation 5.4, 5.5 and 5.12.

Wax
βin = arccos (5.12)
W
A summary of the calculated boundary values at the tip radius are provided in table 5.3.

39
Master Thesis Report Verification of 2D Fitting Results

Table 5.3: Calculated BC Data at tip radius

Parameters Symbols Values Units


Speed of Sound alocal 340.263 m/s
Relative Stagnation Pressure P01,rel 313517.943 Pascal
Relative Stagnation Temperature T01,rel 397.901 K
Peripheral Speed U 431.765 m/s
Relative Tangential Speed Wθ 444.21 m/s
Absolute Tangential Speed Cθ 12.44 m/s
Hub Radius Rhub 0.096375 m
Absolute axial speed Caxial 152.215 m
Inlet relative flow angle βin 71.085 ◦

In order to ensure that the inlet relative flow angle matches the metal blade inlet angle, the value of
the axial velocity was adjusted to ensure a small or zero incidence angle for the various radial position
considered. The computed inlet boundary conditions for profiles at other radial position is presented
in table 5.4. For the purpose of verifying the 2D fitting approach, 4 different blade profiles located
at various span position distributed across the hub and tip was used. Their radial position is also
presented in table 5.4.

Table 5.4: Calculated Inlet BC Data at other radial position

Parameters Symbols Radius Radius Radius Radius Units


0.1265 0.1565 0.1865 0.2265
Peripheral Speed U 212.52 262.923 313.32 380.524 m/s
Absolute tangential Speed Cθ 25.27 20.43 17.145 14.117 m/s
Absolute Speed C 233.373 232.89 232.63 232.429 m/s
Relative Speed C 332.22 366.215 403.774 457.784 m/s
Relative Stagnation Pressure P01,rel 186627.41 210111.6973 241482.33 298778.283 Pascal
Relative Stagnation Temperature T01,rel 343.09 354.906 369.3015 392.463 K
Inlet Relative Flow Angle βin 45.71 50.69 54.93 59.55 ◦

Inlet Blade Metal Angle M etalin 44.01 49.608 55.4 60.86 ◦

Inlet Incidence Angle - 1.697 1.083 -0.47 -1.31 ◦

Axial Velocity Cax 232 232 232 232 m/s

5.2 2D Outlet Boundary Conditions


In order to compute the outlet boundary conditions, the radial equilibrium condition was used to
compute the stagnation condition at each radial position. Since the static pressure at 1 outlet condi-
tion is required, the outlet pressure at the mean radius was estimated based on the design pressure
ratio and the assumed inlet static pressure. It should be pointed out that a simplification of the
radial equilibrium condition was used when computing the outlet stagnation condition at each radial
position.

Theory of Radial Equilibrium


The theory of radial equilibrium is based on the fact that the flow in axial turbomachines are in
general three dimensional. This implies that the radial velocities through a blade row may become

40
Master Thesis Report Verification of 2D Fitting Results

appreciable and the consequent redistribution of mass flow may affect the outlet velocity profile [10].
The principle of radial equilibrium is based on the fact that the pressure forces should balance the
centrifugal forces. An illustration of this concept is presented in figure 5.1.

Figure 5.1: Concept of Radial Equilibrium [10]

Using figure 5.1, the radial equilibrim equation can be derived as follows:

1 dmCθ2
(p + dp)(r + dr)dθ − prdθ − (p + dp)drdθ = (5.13)
2 r
Considering that the mass element per unit depth may be approximated assuming small value of dr
and dθ:

dm = ρrdθdr (5.14)
Substituting the expression for the mass flow, the radial equilibrium may be expressed as follows:
1
dpdrdθ + rdpdθ = ρdrdθCθ2 (5.15)
2
Neglecting the second order smaller terms, the radial equilibrium equation can be expressed as follows:

1 dp C2
= θ (5.16)
ρ dr r
Integrating on both sides, equation 5.16 can be expressed as follows:

Z2
dr
P2 − Pmid = ρCθ2 (5.17)
r
mid
Equation 5.17 can only be solved if the value of ρ and Cθ which is a function of the radius is known
beforehand. Therefore, the following simplification was applied to the problem considering the lack

41
Master Thesis Report Verification of 2D Fitting Results

of information. The fluid is assumed to be incompressible and that the value of Cθ at the outlet is
assumed to be a constant.
Using this assumption, the radial equilibrium can be expressed as follows:

P2 − Pmid = ρCθ2 [ln R2 − ln Rmid ] (5.18)


Where R2 is the new radial position being considered. For the initial calculation involving the blade
profile at R = 0.1865, the value of Cθ at the outlet is not known beforehand. Therefore, an initial guess
for P2 is to use the outlet static pressure at the mean radius. After the first CFD run, the value of Cθ
is then calculated based on the mass averaged axial ū velocity, transverse velocity v̄ of the CFD results
as well as the peripheral speed. Using this solution, the new value of P2,new is re-computed for use in
the next simulation run as well as comparison with the initial value of P2,old . This iterative process
repeats until the point where the value of P2,new matches the value of P2,old . The entire calculation
procedure is presented schematically in figure 5.2.

Figure 5.2: Calculation procedure for profile 1865 outlet boundary condition

As for the remaining blade profiles (R = 0.1265, 0.1565, 0.2265), the calculation procedure is given
schematically as shown in figure 5.3.

42
Master Thesis Report Verification of 2D Fitting Results

Figure 5.3: Calculation procedure for profile 1265, 1565 and 2265 outlet boundary condition

The value of Cθ,2 is first calculated based on free vortex rule. Thereafter, its value is being iterated
until the point where the chosen value of Cθ,2 matches with the Cθ,2 from the CFD results. The final
value of the chosen Cθ,2 , P2 and inlet boundary condition values for the various profiles are presented
in table 5.5.

Table 5.5: Inlet and Outlet BC Data at R = 0.1265, 0.1565, 0.1865 & 0.2265

Parameters R = 0.1265 R = 0.1565 R = 0.1865 R = 0.2265


- Profile Profile Profile Profile
P01 186627.41 210111.7 241482.33 298778.3
T01 343.09 354.91 369.3 392.46
Cθ,2 199.98 134.0 105.0 83.7363
P2 148806.65 162491.04 165889.71 167293.04

5.3 2D Mesh Topology and CFD Solver settings


Before any CFD results can be deemed valid, it is necessary to first verify the quality of the mesh as
well as its CFD solver setting. Therefore, the mesh topology and the optimal solver setting will be
described first before any discussion on the 2D results is made.

5.3.1 2D Mesh Topology


For the generation of the 2D mesh, an in-house mesh generator, UMG 2 was used to generate the
mesh for the various blade profiles. Since there is no way to extract the mesh quality information

43
Master Thesis Report Verification of 2D Fitting Results

interactively as in CFX, the quality of the mesh was inspected visually based on its aspect ratio, skew-
ness and smoothness. The aspect ratio is a ratio of the longest length to the shortest mesh length.
The skewness is a measure of the angle between the associated sides of the mesh. For the case of a
triangular mesh, a 60◦ angle between the sides is optimal. Finally, the smoothness is a measure of the
transition in cell size of the mesh domain. A large change in cell size across different mesh represents
a bad smoothness. The mesh topology is dependent on parameters such as the y-plus (y + ), total
thickness of the boundary layer and the number of layers, local edge length as well as the fineness. A
description of these meshing parameter and their associated values will be described below.

y-plus (y + )
The y-plus is known as the non dimensional wall distance. It is defined as the physical distance from
the wall to the first layer of the boundary layer. A smaller value of the y + implies a closer node
distance between the first boundary layer and the wall. This leads to a finer mesh and hence a better
ability to capture the flow characteristic accurately. Mathematically, the equation can be expressed
as follows:

y ρτw
y+ = (5.19)
µ
Where µ is the dynamic viscosity of the air, τw is the wall shear stress and ρ is the density of the air.
For the various blade profiles tested, it is considered good practice to keep the y + value less than 5
for all cases, especially in regions with important boundary layer phenomena such as the separation
zone. In UMG 2, this is basically controlled by a parameter which defines the thickness of the first
boundary layer (BL).

Overall thickness of BL and number of layers


The overall thickness of the boundary layer influences how accurately viscous effect due to shear stress
is being modelled. It has an influence on the overall prediction of momentum deficit. The number of
layer affects the discretization within the boundary layer and its ability to accurately predict the flow
characteristics within the wall region. Through a series of extensive testing, it was found that this
parameter has a large influence on the blade loading. An optimum number of 28 layers along with a
thickness of 0.007 was used for the various blade profile. This was decided based on the skewness as
well as the aspect ratio of the mesh within the boundary layer.

Fineness and Local edge control


The fineness parameter describes the growth rate of the mesh. Within UMG 2, the mesh within
the boundary layer is structured while the mesh outside of the boundary layer is unstructured. The
fineness of both set of meshes are controlled within the spacing control file of UMG2.

Figure 5.4: Illustration of Mesh Zone

44
Master Thesis Report Verification of 2D Fitting Results

The mesh of any general 2D blade profile generated through UMG 2 is divided into 5 zone. This
is presented in figure 5.4.The spacing control parameters of UMG 2 allows local control of the mesh
discretization within the 5 zones as shown in figure 5.4. In addition, the size and discretization
around a radial zone encircling the leading and trailing edge is controlled through the 0 radius0 and
0 h0 parameter of the spacing control file. The fineness and local edge control of the mesh is controlled

through the 0 h0min and 0 h0max parameters within the spacing control file. An illustration of the final
mesh topology for the R=0.1265 blade profile is shown in figure 5.5.
y

0 0.02 0.04 0.06 0.08 0.1


x

Figure 5.5: Overall mesh topology

0.005

0
y

-
-0.005

-0.04
-0.01

-0.005 0 0.005 0.01 0.015 0.075 0.08 0.085 0.09 0.095


x x

Figure 5.6: Mesh region near leading edge Figure 5.7: Mesh region near trailing edge

45
Master Thesis Report Verification of 2D Fitting Results

The varying level of mesh discretization within the entire 2D domain is illustrated in figure 5.5, 5.6
and 5.7. It can be seen that the leading and trailing edge radial zone has a much denser mesh as
compared to the zone around the blade wall. Also, the zone around the blade wall exhibits a denser
mesh as compared to the region of the inflow, outflow, periodic 1 and periodic 2 zone.

5.3.2 SU 2 CFD Solver Settings


Once an appropriate mesh has been generated by UMG 2, the next step is to set up the right numer-
ical/spatial integration scheme, turbulence model and boundary conditions for the configuration file
of SU 2 . The menter shear stress transport (SST) turbulence model which is composed of a blend-
ing of the k − ω and k −  model is used for all the simulation run. Secondly, the inlet stagnation
pressure/temperature and the outlet static pressure are based on the values provided in table 5.5.
Thirdly, since a non rotating simulation is applied, the relative flow is used in order to account for
the peripheral speed. This implies that the relative flow angle will be imposed on the inlet boundary
condition and that the flow angle in the outlet plane is also in the relative frame of reference.

For all the mesh generated, a periodic boundary condition is imposed on the top and bottom side of
the blade to blade domain since it is assumed that the flow solution is periodic across the blade row
for every pitch distance. This is a simplification of a real flow simulation since every blade within the
blade row is different due to manufacturing uncertainties and that the blade row may be exposed to
unsteady flow phenomenon.

A brief description of the common numerical settings used for all the 2D simulation is presented
below. The solver settings can be divided into 4 main categories; common parameters for numerical
scheme, flow numerical methods, turbulent numerical methods as well as the convergence parameter.
A list of the commonly considered parameters for the simulation run are presented in table 5.6, 5.7
and 5.8.
Table 5.6: Common Parameters for Numerical Methods

Parameters Type
Numerical Method For Spatial Gradient Weighted Least Square
CFL Number 10
Number of Total Iterations 7500

Table 5.7: Flow Numerical Methods

Parameters Type
Convective Numerical Method Roe
Spatial Numerical Order Integration 2nd Order Limiter
Slope Limiter Venkatakrishnan
Time Discretization Scheme Euler Implicit

It should be pointed out that the total number of iterations required was finally decided based on the
convergence of the density parameter as well as its order of magnitude reduction. Other than the use
of the density parameter, other solutions exist for use as an convergence parameter. A plot of the
density convergence results for one of the tested profile (1865) is presented in figure 5.8.

46
Master Thesis Report Verification of 2D Fitting Results

Table 5.8: Turbulent Numerical Methods

Parameters Type
Convective Numerical Method Scalar Upwind
Spatial Numerical Order Integration 2nd Order Limiter
Slope Limiter Venkatakrishnan
Time Discretization Scheme Euler Implicit

Plot of Density Reisdual versus Number of Iterations

-3.5
Res_Flow[0]

-4
Density Residual (Res_Flow [0])

-4.5

-5

-5.5

-6

-6.5

-7

0 2000 4000 6000 8000


Iteration

Figure 5.8: Plot of Density Residuals versus Number of Iterations

5.4 2D Pressure & Mach Contour, Blade loading and Geometry


Comparison Results

In this section, the 2D results for the various profiles will be presented. A comparison of the same
profile belonging to three different coordinate sources will be made. In particular, their mach/pressure
contour, blade loading and profile geometry will be compared. Due to spurious effect, it is necessary
to first investigate the appropriate size of the simulation domain before any comparison of the dif-
ferent coordinate sources can be made. Therefore, different lengths of the simulation domain were
investigated.

47
Master Thesis Report Verification of 2D Fitting Results

5.4.1 Comparison of Different Simulation Domain Length


Within the BM, the size of the various inflow/outflow domain lengths were generated as a percentage
of the axial chord length. In total, 5 different simulation domain length ranging from 50% to 150% of
the axial chord length were generated and compared. A comparison of the pressure/Mach contour of
the 1865 profile using different simulation domain length is presented in figure 5.9 and 5.10.

Figure 5.9: Comparison of Pressure contour at different domain length

Figure 5.10: Comparison of Mach contour at different domain length

48
Master Thesis Report Verification of 2D Fitting Results

As seen from figure 5.9 and 5.10, a different length of the inlet domain may result in a different inlet
boundary condition as a result of the spurious reflection caused by downstream potential effect. This
effect is more pronounced for the Mach number as compared to the static pressure. As such, the
simulation domain which has an inflow/outflow domain length of 150% of the axial chord is used for
the remaining simulation.

5.4.2 Comparison of Different Coordinate Sources for Profile 1865


In this subsection, the simulation results of profile 1865 belonging to three different coordinate sources
will be compared. As mentioned in chapter 4, the distributed points of each control profile was first
extracted from the CGNS mesh that was provided. After that, a geometry fitting algorithm was de-
veloped to fit the BSpline curve of the pressure/suction side to the distributed points along the blade
profile contour. Finally, a BSpline surface was generated to fit across all the control profiles on both
the pressure and suction side of the blade. Hence, each blade profile can originate from 3 different
coordinate sources depending on the stage of the construction process. The profile generated directly
from the distributed points of the original CGNS mesh will be known as ’actual point profile’. The
profile generated from the fitting of the BSpline curve using the geometry fitting algorithm will be
known as the ’2D fitted profile’. Finally, the profile generated from the slice of the 3D blade belonging
to the BSpline surfaces will be known as the ’surface fitted profile’.

The comparison of the blade profiles will be made based on results from the Mach and Pressure
contour distribution. For the first test case, a comparison was made between the ’actual point’ profile
and the ’2D fitted’ profile. This is illustrated in figure 5.11 and 5.12.

Figure 5.11: Comparison of 1865 Mach Contour Figure 5.12: Comparison of 1865 Pressure Contour
between 2D fitted and Actual Point Profile between 2D fitted and Actual Point Profile

In figure 5.11, the thick lines emanating from the surface of the blade contour represents the generation
of a shockwave as the flow velocity exceeds Mach 1. A visual comparison of the shockwave position
between both profile types shows that the overall shape of both profile is very similar although there
are slight differences in the trailing edge wake pattern and the exact position of each shockwave. In
the case that both profiles have exactly the same shape, the periodicity of the solution from both
profiles will be preserved. This implies that the pressure and mach contour pattern of both profile
will be exactly the same.

49
Master Thesis Report Verification of 2D Fitting Results

For the second test case, a comparison was made between the ’2d fitted’ and ’surface fitted’ pro-
file. This is illustrated in figure 5.13 and 5.14 respectively.

Figure 5.13: Comparison of 1865 Mach Contour Figure 5.14: Comparison of 1865 Pressure Contour
between 2D fitted and Surface fitted Profile between 2D fitted and Surface fitted Profile

It can be seen that the results are very similar for both profile although there are minor differences in
the position of the shockwaves. Comparing figure 5.11 and 5.13, it can be concluded that the results
of the 2d fitted profile matches more closely to the actual point profile than the surface fitted profile.
For the third test case, a comparison was made between the actual point profile and the surface fitted
profile. This is illustrated in figure 5.15 and 5.16.

Figure 5.15: Comparison of 1865 Mach Contour Figure 5.16: Comparison of 1865 Press. Contour
between Surface fitted Profile and Actual between Surf. fitted Profile and Actual point
point profile profile

For this test case, the pressure and mach contour of the actual point and surface fitted profile are
very similar, making it comparable to the first test case. A comparison of the blade loading of all the
different test cases will reveal their similarity and differences more closely.
A comparison of the various blade loading in figure 5.17, 5.18 and 5.19 reveals that the actual point
profile and the surface fitted profile actually have the closest fit. The kinks present in the various

50
Master Thesis Report Verification of 2D Fitting Results

Figure 5.17: Comparison of 1865 Blade loading Figure 5.18: Comparison of 1865 Blade loading
between 2D fitted Profile and Actual between 2D fitted Profile and Surface
point profile fitted profile

Figure 5.19: Comparison of 1865 Blade loading


Figure 5.20: Comparison of 1865 Blade Profiles
between Surface fitted Profile and Actual
from 3 Coordinate sources
point profile

blade loading indicate the position of the shockwaves since an increase in static pressure is expected
after each shockwave. A first look at the profiles from the 3 coordinate sources shows that the profiles’
geometry are almost similar for all the testcases. However, a close zoom on the leading and trailing
edge region reveals the difference in geometry between the different profile type. This is illustrated in
figure 5.21 and 5.22 respectively.

51
Master Thesis Report Verification of 2D Fitting Results

Figure 5.21: 1865 Blade Profile leading Figure 5.22: 1865 Blade Profile Trailing
edge zoom edge zoom

During the blade construction process, error builds up from a variety of sources. The first source of
error is human error arising from the difficulty of obtaining a perfect fit for the camberline. Secondly,
the geometry fitting algorithm cannot ensures a perfect interpolation for all the points distributed
across the blade profile contour. Thirdly, the generated BSpline surfaces does not necessarily go
through the curves of all the control profiles.

5.4.3 Comparison of Different Coordinate Sources for Profile 1265


In this subsection, a similar comparison of profile 1265 belonging to the different coordinate sources
will be made. Profile 1265 is a relatively thick profile that is located at a radial position close to the
hub. For the first test case, a comparison is made between the actual point profile and the 2D fitted
profile.

Figure 5.23: Comparison of 1265 Mach Contour Figure 5.24: Comparison of 1265 Pressure Contour
between 2D fitted and Actual Point Profile between 2D fitted and Actual Point Profile

52
Master Thesis Report Verification of 2D Fitting Results

It can be seen that the mach and pressure contour patterns are very similar for both profile types.
The position of their shocks almost coincide and the trailing edge wake pattern is identical. A closer
examination of their blade loading will reveal more information about their similarity and differences.
For the second test case, a comparison was made between the 2d fitted and surface fitted profile.

Figure 5.25: Comparison of 1265 Mach Contour Figure 5.26: Comparison of 1265 Pressure Contour
between 2D fitted and Surface fitted Profile between 2D fitted and Surface fitted Profile

It can be seen that the 2d fitted and surface fitted profiles are different in the sense that the position of
their shocks are shifted with respect to each other. However, both profile types have almost identical
trailing edge wake pattern. For the third test case, a comparison is made between the actual point
and surface fitted profile.

Figure 5.27: Comparison of 1265 Mach Contour Figure 5.28: Comparison of 1265 Pressure Contour
between Actual point and Surface fitted Profile between Actual point and Surface fitted Profile

It can be seen that the actual point and surface fitted profiles are different in the sense that the position
of their shocks are shifted with respect to each other. However, both profile types have almost identical
trailing edge wake pattern. This suggest that the blade loading would be similar except for a shift in
the kink which represents the shockwave. A comparison of the blade loading and profile geometry for
the different profile types are shown in figure 5.29, 5.30 and 5.31.

53
Master Thesis Report Verification of 2D Fitting Results

Figure 5.29: Comparison of 1265 Blade loading Figure 5.30: Comparison of 1265 Blade loading
between 2D fitted and Actual point profile between 2D fitted and Surface fitted profile

Figure 5.31: Comparison of 1265 Blade loading Figure 5.32: Comparison of 1265 Blade Profiles
between Surface fitted and Actual point profile from 3 Coordinate sources

From the plot of the various blade loadings, it can be seen that the 2d fitted and actual point profiles
have the closest fit. The surface fitted profile is noticeably different as compared to the other blade
profiles. This is verifiable through an inspection of the blade profile geometry and blade loading plot
comparing the surface fitted profile to both the 2D fitted and actual point profile respectively. This
conclusion is similar to the previous 1865 blade profile comparison. A significant error as compared
to the original profile geometry was produced during the process of generating a 3D blade shape with
the BSpline surfaces. Further improvement to the BSpline surface generation algorithm is required.

5.4.4 Comparison of Different Coordinate Sources for Profile 1565


In this subsection, a comparison of profile 1565 belonging to the different coordinate sources will be
made. For the first test case, a comparison was made between the actual point profile and the 2D

54
Master Thesis Report Verification of 2D Fitting Results

fitted profile. It can be seen that the results from the 2D fitted and Actual point profile are different.
The position of the shockwaves between both profiles are quite far apart.

Figure 5.33: Comparison of 1565 Mach Contour Figure 5.34: Comparison of 1565 Pressure Contour
between 2D fitted and Actual Point Profile between 2D fitted and Actual Point Profile

For the second test case, a comparison was made between the 2D fitted and Surface fitted profile. It
can be seen that the results from the 2D fitted and Surface fitted profile are very similar.

Figure 5.35: Comparison of 1565 Mach Contour Figure 5.36: Comparison of 1565 Pressure Contour
between 2D fitted and Surface fitted Profile between 2D fitted and Surface fitted Profile

For the third test case, a comparison was made between the Actual point and Surface fitted profile. It
can be seen that the results from the Actual point and Surface fitted profile are very different in terms
of the shockwave position. However, both profile types share a similar trailing edge wake pattern.
Finally, a comparison of the respective blade loadings and profile geometry is shown in figure 5.39,
5.40 and 5.41.
The general trend of the blade loadings are similar for all three profile types. In particular, the 2D
fitted and Surface fitted profiles have the closest fit. As compared to the previous blade profiles
comparison, the main source of error for profile 1565 originates from the 2d fitting process. A close
zoom onto the blade profiles’ leading and trailing edge region (see figure 5.43 and 5.44) confirm this

55
Master Thesis Report Verification of 2D Fitting Results

Figure 5.37: Comparison of 1565 Mach Contour Figure 5.38: Comparison of 1565 Pressure Contour
between Actual point and Surface fitted Profile between Actual point and Surface fitted Profile

Figure 5.39: Comparison of 1565 Blade loading Figure 5.40: Comparison of 1565 Blade loading
between 2D fitted and Actual point profile between 2D fitted and Surface Fitted profile

observation.

56
Master Thesis Report Verification of 2D Fitting Results

Figure 5.41: Comparison of 1565 Blade loading Figure 5.42: Comparison of 1565 Blade Profiles
between Surface fitted and Actual point profile from 3 Coordinate sources

Plot of Blade Profile From the 3 Coordinate Sources Plot of Blade Profile From the 3 Coordinate Sources

0.008 2D Fitted 2D Fitted


-0.045
Actual Points Actual Points
Surface Fitted Surface Fitted

0.006

0.004
y

-0.05

0.002

0
-0.055

0.005 0.01 0.075 0.08 0.085


x x

Figure 5.43: 1565 Blade Profile leading Figure 5.44: 1565 Blade Profile trailing
edge zoom edge zoom

5.4.5 Comparison of Different Coordinate Sources for Profile 2265


In this subsection, a comparison of profile 2265 belonging to the different coordinate sources will be
made. As compared to the other blade profiles, this profile is comparatively much thinner and the
curvature of the pressure/suction side is almost symmetrical. For the first test case, a comparison was
made between the actual point profile and the 2D fitted profile. It can be seen that both the 2d fitted
and actual point profile have similar mach and pressure contour pattern, except for a slight shift in
the shockwave position. The mach contour close to the leading edge region is also different for both
profile types.

57
Master Thesis Report Verification of 2D Fitting Results

Figure 5.45: Comparison of 2265 Mach Contour Figure 5.46: Comparison of 2265 Pressure Contour
between 2D fitted and Actual Point Profile between 2D fitted and Actual Point Profile

For the second test case, a comparison was made between the 2D fitted and Surface fitted profile. It
can be seen that the mach and pressure contour are almost identical for both profiles except for a
slight shift in the shockwave position.

Figure 5.47: Comparison of 2265 Mach Contour Figure 5.48: Comparison of 2265 Pressure Contour
between 2D fitted and Surface fitted Profile between 2D fitted and Surface fitted Profile

58
Master Thesis Report Verification of 2D Fitting Results

For the third test case, a comparison was made between the Actual point and Surface fitted profile.
Observable differences can be found in the exact position of the shockwave as well as the mach contour
pattern near the leading edge region of the blade.

Figure 5.49: Comparison of 2265 Mach Contour Figure 5.50: Comparison of 2265 Pressure Contour
between Actual point and Surface fitted Profile between Actual point and Surface fitted Profile

Finally, a comparison of the respective blade loadings and profile geometry is shown in figure 5.51,
5.52 and 5.53.

Figure 5.51: Comparison of 2265 Blade loading Figure 5.52: Comparison of 2265 Blade loading
between 2D fitted and Actual point profile between 2D fitted and Surface Fitted profile

59
Master Thesis Report Verification of 2D Fitting Results

Figure 5.53: Comparison of 2265 Blade loading Figure 5.54: Comparison of 2265 Blade Profiles
between Surface fitted and Actual point profile from 3 Coordinate sources

The general trend of the blade loadings are similar for all three profile types. In particular, the 2D fitted
and Surface fitted profiles have the closest fit in terms of blade loading. However, a close zoom on the
leading and trailing edge region reveals that small positional error exists on the actual leading/trailing
edge position of the surface fitted profile as compared to the other two profiles. This suggests that
the position of the leading/trailing edge predominantly affects the position of the shockwave since it
changes the incidence and deviation angle. Whereas, the curvature of the blade primarily affects the
general shape of the blade loading.

Plot of Blade Profile From 3 Coordinate Sources Plot of Blade Profile From 3 Coordinate Sources

2D Fitted Profile -0.05 2D Fitted Profile


Actual Point Profile Actual Point Profile
0.02 Surface Fitted Profile Surface Fitted Profile
-0.055

-0.06

-0.065
0
y

-0.07

-0.075

-0.02 -0.08

-0.085

0.01 0.02 0.03 0.04 0.05 0.055 0.06 0.065 0.07


x x

Figure 5.55: 2265 Blade Profile leading Figure 5.56: 2265 Blade Profile trailing
edge zoom edge zoom

60
Master Thesis Report Verification of 2D Fitting Results

5.5 Comparison of 2D Stagnation Pressure Loss and Blade Channel


Area Distribution
Another way of comparing the same blade profile belonging to three different coordinate sources is to
look at the stagnation pressure loss coefficient. The stagnation pressure loss coefficient is a measure
of the loss in stagnation pressure across the blade profile due to viscous effect. As the simulations
were carried out without any rotation of the mesh, this implies that no work is added to the fluid and
hence the total enthalpy is constant across the profile. However, the fluid experiences a total pressure
drop due to viscous effect. This behaviour is similar to the case of a stator blade. An illustration of
the thermodynamic process is presented in figure 5.57.

Figure 5.57: Enthalpy-Entropy diagram of a Stator Blade

The stagnation pressure loss coefficient can be defined using equation 2.45b. This is re-presented in
equation 5.20.

P01,rel − P02,rel
δloss = (5.20)
P01,rel − P2

Using equation 5.20, the stagnation pressure loss coefficient is computed for all the profiles. In order
to ensure a fair comparison between the different profile types, a fixed location on the inlet and outlet
simulation domain was used to compute the mass averaged quantities. A comparison of the stagnation
pressure loss coefficient of each profile belonging to the different coordinate sources is presented in
table 5.9.

Table 5.9: Comparison of Stagnation Pressure Loss Coefficient between different profile types

Profile Type 1865 1565 1265 2265


2D fitted 0.08112 0.14034 0.29563 0.04628
Actual Point 0.08629 0.13098 0.29418 0.04729
Surface fitted 0.09055 0.13957 0.31320 0.04564

61
Master Thesis Report Verification of 2D Fitting Results

The results in table 5.9 suggests that the surface fitted profile of 1865 and 1265 are shaped differently
from their respective 2d fitted and actual point counterparts. This was a similar conclusion reached
when a comparison of profile 1865 and 1265 blade loadings and Mach/Pressure contour were made
earlier. As for profile 1565, the results of the stagnation pressure loss coefficient reveal that the Actual
point profiles are shaped differently from its respective 2d fitted and surface fitted counterparts. This
is in line with the results concluded from the blade loading plot and Mach/Pressure contour of profile
1565. For these results, it was also found that the 2d fitted and surface fitted profile have the closest
fit. As for profile 2265, the results of the stagnation pressure loss coefficient reveal that the surface
fitted profile is shaped closely to the 2d fitted profile. This was a similar conclusion reached when a
comparison of profile 2265 blade loadings and Mach/Pressure contour were made earlier.

When the simulations were carried out for the various 2d blade profiles, it was found that profile
2265 behaves more like a turbine instead of a compressor since the static pressure increases across the
blade profile. Therefore, an investigation was carried out to understand this physical phenomenon
by looking into the area distribution across the blade profile. An illustration of the respective area
distribution belonging to the different blade profiles is shown in figure 5.58.

Plot of Blade to Blade Area Distribution


0.04
Area Dist 1265 The start of the Kink signifies
Area Dist 1565 the start of the outlet region
0.038 Area Dist 1865
Area Dist 2265

0.036

0.034

0.032
y

0.03

0.028

0.026

0.024
0 0.2 0.4 0.6 0.8 1
x

Figure 5.58: Area Distribution for various blade profiles

In figure 5.58, it should be pointed out that the step change in area distribution (or start of a kink) for
the various blade profiles represents the beginning of the outlet region for the blade to blade domain.
Firstly, it can be seen that the actual blade to blade domain becomes smaller with increasing stagger
angle. Secondly, the total change in area distribution also becomes smaller for blade profiles located
closer to the tip than the hub. Thirdly, it can be seen that profile 2265 exhibits the smallest change
in area distribution within its blade to blade region. Therefore, it is possible that profile 2265 behaves
more like a turbine instead of a compressor since the flow converges for about half of the total blade
channel before it diverges. The resulting effect would still be an increase in the velocity despite the
flow diverging for a small section of the blade channel.

62
Master Thesis Report Development of Salome 2D & 3D Meshing template and Grid Generator

6|Development of Salome 2D & 3D Mesh-


ing template and Grid Generator

In this chapter, the report will discuss about the development of SALOME based 2D and 3D meshing
template. SALOME is an open-source meshing tool developed by EDF, a french based research and
development company. SALOME is equipped primarily with python based geometry and meshing
modules that can be run in both batch mode and in Graphic User Interface (GUI) mode. The
main aim in developing these meshing template is to be able to generate mesh automatically for
similar geometries with minimal effort. The secondary aim is the need for open-source tools that
can be coupled with the in-house blade modeller After the meshes were generated in SALOME, it
is necessary to convert them into SU 2 based mesh before the simulation can be executed using the
SU 2 CFD solver. Therefore, a SALOME-SU 2 grid generator was developed. The first section of this
chapter will touch on the development of 2D meshing template. The second section of this report
will compare the CFD results from SALOME and UMG 2 generated mesh. The third section of this
chapter will look into the development of 3D meshing template. The fourth section of this report will
compare the CFD results from SALOME and UMG 3 generated mesh. Lastly, the fifth section of this
report will look into the development of a grid generator which extracts meshing information from
SALOME and writes SU 2 based mesh.

6.1 Development of 2D Meshing Template


In order to develop a SALOME based meshing template that have similar features to UMG 2, it is
necessary to investigate the specific requirements needed for the generation of a good quality 2D mesh.
An illustration of the required features for a 2D mesh is shown in figure 6.1.

Figure 6.1: Features of a good quality 2D mesh

63
Master Thesis Report Development of Salome 2D & 3D Meshing template and Grid Generator

Basically, a good quality 2D mesh should comprise of finer discretization within circular zone 1 and
2. This will enable the constructed mesh to produce a good curvature around the leading and trailing
edge zone. Secondly, the growth rate of the unstructured mesh within the circular zone should ensure a
smooth transition between the finer discretised mesh within the circular zone and the external domain.
Thirdly, a reasonably thick and sufficiently discretized boundary layer should be present around the
entire blade wall. Lastly, the projection 1D-2D algorithm is used to produce a periodic boundary
condition between the upper and lower edges of the simulation domain. An illustration of the 2D
mesh developed in Salome which fulfils this criteria is presented in figure 6.2, 6.3, 6.4 and 6.5.

Figure 6.2: 2D Mesh Generated by Salome

Figure 6.3: Leading edge zoom on Salome 2D MeshFigure 6.4: Trailing edge zoom on Salome 2D Mesh

Figure 6.5: Boundary Layer zoom on Salome 2D Mesh

Firstly, the geometry is exported into Salome as a ’step’ or ’igs’ file. Thereafter, the geometry is
subdivided into edges for which they are used to identify the inflow outflow, periodic 1, periodic 2
and the blade wall of the meshing domain. This is a prior step to be taken for the construction
of SU 2 based mesh. The development of this meshing template involves the use of Netgen 1D-2D
meshing algorithm. The growth rate of the mesh is controlled through the fineness ratio, a setting
adjustable by the user in the meshing template. In addition, the user is able to control both the
minimum and maximum allowable discretization of the mesh. The viscous layer option in SALOME
enable the users to control the total thickness, number of layer as well as the stretch factor of the

64
Master Thesis Report Development of Salome 2D & 3D Meshing template and Grid Generator

boundary layer. This algorithm automatically generates a finer discretization around geometry which
have distinct curvature, such as the leading/trailing edge. In addition, the user can also manually
specify the vertex or curve for which finer discretization should be constructed within the settings of
the Netgen 1D-2D meshing algorithm.

6.2 Comparison of 2D CFD Results between SALOME and UMG


2 Mesh
In this section, the 2D CFD results from the Salome generated mesh will be compared to the UMG 2
mesh. In particular, the Mach/pressure contour and blade loading will be compared.

Figure 6.6: Comparison of Profile 1265 Salome Figure 6.7: Comparison of Profile 1265 Salome and
and UMG 2 Mesh Mach Contour UMG 2 Mesh Pressure Contour

Comparison of Profile 1265 UMG 2 and Salome Blade Loading

200000

180000 1265 UMG 2


1265 Salome

160000

140000
Pressure

120000

100000

80000

60000

0 0.02 0.04 0.06 0.08 0.1


x

Figure 6.8: Comparison of Profile 1265 Salome and UMG 2 Mesh Blade Loading

As seen from figure 6.6, 6.7 and 6.8, the CFD results generated from SALOME mesh and UMG 2
mesh is almost identical. Except for a very small shift in the shockwave, the magnitude of the shock

65
Master Thesis Report Development of Salome 2D & 3D Meshing template and Grid Generator

as well as the trailing edge wake pattern is identical for both the mach and pressure contour. This
implies that the use of either Salome or UMG 2 generated mesh will give identical results. Since the
CFD results are dependent on the mesh quality as well as obtaining an converged solution, differences
in results between both mesh sources can result at times if either criteria are not met.

6.3 Development of 3D Meshing Template


The development approach of the 3D template is similar to the 2D template. The Netgen 1D-2D
Algorithm is used to generate these meshing template. However, this algorithm is applied specifically
to the surfaces belonging to the blade wall as well as the surfaces above and below the blade geometry
(see figure 6.10). Within the 3D simulation domain, it is also necessary to ensure that fine discretized
3D mesh is located around the discretized 2D leading and trailing edge surface mesh. Lastly, the
projection 1D-2D algorithm is used to produce a periodic boundary condition between the periodic
faces of the simulation domain. The algorithm basically projects the nodes from one side of the
periodic face to the other side. An illustration of the NASA Rotor 67 Salome mesh which fulfils this
criteria is presented in figure 6.9, 6.10, 6.11 and 6.12. Other than the NASA Rotor 67 mesh, this
approach was also applied to a centrifugal rotor geometry. This is illustrated in figure 6.13, 6.14 and
6.15.

Figure 6.9: Features of a good quality 3D mesh Figure 6.10: Fine Discretized Mesh across the
Leading/Trailing edge and the Top/Bottom Blade
Surfaces

Similar to the 2D mesh development approach, the geometry is first exported into Salome as a ’step’
or ’igs’ file. Thereafter, the geometry is subdivided into faces for which they are used to identify the
inflow outflow, periodic 1, periodic 2 and the blade wall of the meshing domain. This is a prior step
to be taken for the construction of SU 2 based mesh. Instead of using edges to define the boundary
markers, faces are used to identify the boundary markers for 3D geometries. The development of this
meshing template also involves the use of Netgen 1D-2D meshing algorithm. The growth rate of the
mesh is controlled through the fineness ratio, a setting adjustable by the user in the meshing template.
In addition, the user is able to control both the minimum and maximum allowable discretization of the
mesh. The viscous layer option in SALOME enable the users to control the total thickness, number of

66
Master Thesis Report Development of Salome 2D & 3D Meshing template and Grid Generator

Figure 6.12: Clipped Horizontal View of NASA Ro-


tor 67 Internal Mesh Domain
Figure 6.11: Clipped Vertical View of NASA Rotor
67 Internal Mesh Domain

Figure 6.13: 3D mesh of Centrifugal Rotor with tip clearance

layer as well as the stretch factor of the boundary layer. Boundary layer is applied to faces instead of
edges for 3D geometries. This Netgen 1D-2D algorithm automatically generates a finer discretization
around geometry which have distinct curvature, such as the leading/trailing edge. In addition, the
user can also manually specify the vertex or curve for which finer discretization should be constructed
within the settings of the Netgen 1D-2D meshing algorithm.

67
Master Thesis Report Development of Salome 2D & 3D Meshing template and Grid Generator

Figure 6.15: Clipped Cross-Section View of Cen-


Figure 6.14: Clipped Vertical View of trifugal Rotor Internal Mesh Domain
Centrifugal Rotor Internal Mesh Domain

6.4 3D CFD Results of Salome Generated Mesh


In this section, the 3D CFD results from the Salome generated mesh will be presented. Since SU2 was
not able to simulation periodic boundary conditions, the SALOME-SU2 mesh generator was used to
generate a wing mesh and to simulate it in SU2 using a far field and symmetric boundary condition.
The visualization of the 3D wing Mach streamlines in the cross sectional and top view is presented in
figure 6.16 and 6.17.

Y X

Mach
1.2
1.13333
1.06667
1
0.933333
0.866667
0.8
0.733333
0.666667
0.6
0.533333
0.466667
0.4

Figure 6.16: Mach Streamlines of Wing Profile


Figure 6.17: Top view of Mach contour of Wing
Domain

A plot of the pressure coefficient at a specified wing span position is shown in figure ??.

6.5 Development of SALOME-SU 2 Grid Generator


The development of the SALOME-SU 2 mesh begins with understanding the required format of SU 2
mesh and then figuring out how this required information can be extracted from SALOME directly.
A format of the current SU 2 based mesh belonging to the 2D and 3D geometry was used for the
investigation. From these meshes, the required information was determined by analysing the format

68
Master Thesis Report Development of Salome 2D & 3D Meshing template and Grid Generator

Plot of Pressure Coefficient (Cp) at a Span Position of Y = 7

-1.5

Pressure_Coefficient -1

-0.5

0.5

5 6 7 8 9 10 11
x

Figure 6.18: Pressure Coefficient Plot of Salome Meshed Wing Slice

of the mesh. In order to construct the format of the SU 2 based mesh, the following information is
required:

• VTK Number for each Element Type

• Connecting Nodes for a given Element ID

• Element Counter

• Nodal Coordinates

• Node Counter

• Marker Tag

• No. of Marker Elements

• Connecting Nodes for a given Element ID for each Boundary Marker

For a 2D triangle and quad mesh, there is a total of 3 and 4 connecting nodes respectively. When
the information is being transferred from SALOME to SU 2 , certain information such as the total
number of elements has to be modified before they can be printed onto the SU 2 based mesh. Within
SALOME, the total number of elements for a 2D geometry includes the edge elements as well as
the face elements. However, the SU 2 based mesh only requires the face elements during the element
generation process. Therefore, the edge elements are removed for the generation of the SU 2 based
mesh. As for a 3D geometry, only the volume type mesh is required. Therefore, the edge and face
elements are removed during the generation of the SU 2 based mesh. Specifically, they are removed
from the total number of element counter as well as from the element generation process during the
first part of the SU 2 Mesh writing sequences. An illustration of the SU 2 Mesh format template is
presented in figure 6.19.

69
Master Thesis Report Development of Salome 2D & 3D Meshing template and Grid Generator

Figure 6.19: SU 2 Mesh Format

70
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

7|3D Flow Simulation of the NASA Ro-


tor 67

In this chapter, the results from the 3D simulation of the NASA Rotor 67 will be analysed and
discussed. In particular, the 3D flow simulation of the rotor under uniform and non uniform inflow
condition as well as the flow simulation of the rotor with and without casing will be compared. The first
section of this chapter will present on the boundary conditions which were obtained from literature
and the information extracted from it. The second section of this chapter will discuss about the
development of the actual 3D non uniform boundary condition. The non uniform boundary condition
was basically produced by quantifying the circumferential averaged total pressure at each normalized
radial position and then applying them onto the same normalized position of the rotor blade. The
actual 3D uniform inlet boundary condition was then produced by using the radial averaged total
pressure. The outlet static pressure was then quantified based on the total to static pressure ratio
found from [16]. The third section of this report will then present on the simulation settings used
in ANSYS CFX. The fourth section of this chapter will compare the simulation results between a
uniform and non uniform inflow condition. Finally, the fifth section of this chapter will compare the
simulation results of a rotor blade under uniform inflow inlet condition with and without a casing.
The case of a simulation with and without casing was considered since it was necessary to understand
the effect of using either a non slip boundary condition or a non slip boundary condition with counter
rotating radial velocity setting for the shroud wall in ANSYS CFX.

7.1 3D Uniform Inflow Boundary Conditions


In this section, the uniform inflow boundary conditions extracted from the NASA Technical Report
TP2879 [16] are illustrated in figure 7.1, 7.2 and 7.3

71
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Spanwise Rotor Total Pressure Spanwise Rotor Total Temperature


Upstream Total Pressure Upstream Total Temperature
25 Downstream Total Pressure 25 Downstream Total Temperature
Spanwise Radial Position [cm]

Spanwise Radial Position [cm]


20 20

15 15

10 10

5 5
90000 100000 110000 120000 130000 140000 150000 160000 170000 280 290 300 310 320 330 340 350
Pressure [Pascal] Temperature [Kelvin]

Figure 7.1: Spanwise Total Pressure Figure 7.2: Spanwise Total Temperature
at Inlet/Outlet Boundary at Inlet/Outlet Boundary

Spanwise Rotor Static Pressure


Upstream Static Pressure
25 Downstream Static Pressure
Spanwise Radial Position [cm]

20

15

10

5
80000 90000 100000 110000 120000 130000
Pressure [Pascal]

Figure 7.3: Spanwise Static Pressure


at Inlet/Outlet Boundary
Figure 7.4: Measurement Points at Inlet and
Outlet Boundary

An illustration of the measurement points taken at both the inlet and outlet boundary condition is
shown in figure 7.4. Using this boundary condition, the total to static pressure ratio at the mid radius
was obtained. A value of approximately 1.2 was obtained.

7.2 3D Non Uniform Inflow Boundary Conditions


In order to produce a non uniform inflow boundary condition, a previous study on the design of op-
timum inlet configuration for BLI was used as a reference. Since the stagnation pressure loss across
the inlet is non-uniform due to the fact that fuselage boundary layer exists primarily on the bottom
segment of the aerodynamic interface plane, the circumferential averaged total pressure at the aero-

72
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

dynamic interface plane was quantified as a function of the radial position.

As previously described, distortion is by definition a region of the flow with a lower stagnation pres-
sure. It is a result of the incomplete mixing of the freestream and the boundary layer flow. The
resulting total pressure distortion is essentially an axial velocity distortion which leads to a change in
the angle of attack of the flow approaching the compressor blade. As the axial velocity reduces due
to inlet distortion and the rotor speed remain constant, the reduction in axial velocity leads to an
increase in the angle of attack (reduced surge margin) and hence the blade loading. At the extreme
end, the blade is driven towards stall if the flow incidence angle becomes too large. For the overall
rotor assembly, the inlet distortion environment causes some blade to do more work than other blades
while some blades might stall leading to instability of the entire compressor/fan.

A visualization of the total pressure and velocity contour across the inlet and outlet domain of the
finalized inlet configuration is presented in figure 7.5 and 7.6. It can be seen that a region of lower
stagnation pressure is found at the bottom segment of both the inlet and outlet plane. This is ex-
pected since the ingested fuselage boundary layer tend to accumulates at the bottom segment of both
planes. The intake boundary condition was obtained based on the velocity profiles obtained over the
aft fuselage of the Blended Wing Body. As for the Blended Wing Body, it was assumed to be operating
under cruise flight condition with the following parameters given in table 7.1 [17].

Figure 7.5: Total Pressure Contour at the Inlet and Outlet Plane of the Inlet Configuration

The mass flow averaged total conditions across the inlet and outlet plane of the finalized inlet con-
figuration is given in table 7.2. These were extracted from the CFD results of the finalised inlet
configuration.
As shown above, the fluid flow through the finalized inlet design leads to a total pressure loss and
velocity deficit of 3.59% and 13% respectively with reference to the inlet plane. For the case of this
inlet design and operating condition, the reduction in velocity is more critical than the total pressure
loss. In order to quantify the overall loss with respect to a case without boundary layer ingestion, the
calculated mass flow averaged total condition at the outlet plane shall be compared with respect to

73
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.6: Velocity Contour at the Inlet and Outlet Plane of the Inlet Configuration

the condition without BLI. The condition without BLI was obtained by taking the average values over
the ’undisturbed’ region of the inlet plane. It should be pointed out that the velocity over the aft top
section of the fuselage may be greater than the free stream Mach number since the flow is accelerating
over the top section of the fuselage.

74
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Table 7.1: BWB Flight Operating Condition

Parameters Values Units


Flight Regime Cruise -
Altitude 10000 m
Mach Number 0.82 -
Static Pressure 26500 N/m2
Free Stream Velocity 245.6 m/s
Air Density 0.41 kg/m3
Dynamic Viscosity of Air 0.0000147 Pa-sec
Heat Transfer Adiabatic -
Molar Mass of Air 29 kg/kmol

Table 7.2: Mass flow Averaged Total Condition Across Inlet Configuration

Parameters Values Units


Inlet Stagnation Pressure 12241.2 Pascal
Outlet Stagnation Pressure 12001 Pascal
Percentage Differences (w.r.t Inlet) 3.59 %
Inlet Velocity 272.005 m/s
Outlet Velocity 236.665 m/s
Percentage Differences (w.r.t Inlet) 13 %

As demonstrated above, the total pressure loss and velocity deficit are caused primarily by losses
due to BLI as well as losses incurred within the S Duct (with or without BLI). For the purpose of
simulation and obtaining a non uniform boundary condition, the total pressure and velocity sampled
points at the outlet plane (Aerodynamic Interface Plane) were extracted and transformed into their
polar coordinates using the following transformation:

Ymax + Ymin
Yc = (7.1)
2

Zmax + Zmin
Zc = (7.2)
2
q
R= (Y − Yc )2 + (Z − Zc )2 (7.3)

!
Z − Zc
R = arctan (7.4)
Y − Yc

For the computation of equation 7.4, the ’atan2’ function in python is used so that the angular coor-
dinate over the entire 360◦ sector is taken into account. After transforming the coordinates, both the
total pressure and velocity values were then sampled along specified radius. Subsequently, the plot of
total pressure and velocity against the circumferential position for each radial position are produced
and compared. This is illustrated in figure 7.7, 7.8 and 7.9.

As seen in figure 7.8, the total pressure distortion is greatest closest to the tip radius. This is primar-
ily due to the fact that fuselage boundary layer accumulates primarily at region close to the bottom
segment of the outlet plane, leading to a greater pressure drop. Secondly, the greatest pressure drop

75
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Table 7.3: Comparison of Outlet Plane Total Condition vs. Aft Section Total Condition without BLI

Parameters Values Units


Outlet Plane Stagnation Pressure 12001 Pascal
Stagnation Pressure (no BLI) 15550 Pascal
Percentage Differences (w.r.t no BLI case) 22.823 %
Outlet Plane Velocity 236.665 m/s
Velocity (no BLI) 297 m/s
Percentage Differences (w.r.t no BLI case) 20.31 %

Table 7.4: Comparison of Inlet Plane Total Condition vs. Aft Section Total Condition without BLI

Parameters Values Units


Inlet Plane Stagnation Pressure 12441.2 Pascal
Stagnation Pressure (no BLI) 15550 Pascal
Percentage Differences (w.r.t no BLI case) 19.99 %
Inlet Plane Velocity 272.005 m/s
Velocity (no BLI) 297 m/s
Percentage Differences (w.r.t no BLI case) 8.416 %

Plot of Sampled Points


90°in Polar Coordinates
R = 0.2
135° 45° R = 0.3
R = 0.4
R = 0.5
R = 0.6
1.6 1.8 R = 0.7
1.0 1.2 1.4 R = 0.8
0.8 R = 0.9
0.2 0.4 0.6
R = 1.0
180° 0°
R = 1.1
R = 1.2
R = 1.3
R = 1.4
R = 1.5
R = 1.55
R = 1.6
225° 315°

270°

Figure 7.7: Sampled Points along the Outlet Plane of inlet configuration

is found at a sectoral position of 270◦ , right where the fuselage boundary layer accumulates. Using
this result, the plots are then curved fitted and then integrated numerically (using the Riemann Sum
Approach) to obtain the circumferential average pressure for a given radius. As for the calculation of

76
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

18000 Plot of Total Pressure versus Angular Position

16000
R = 0.7
14000
R = 0.8
12000 R = 0.9
Total Pressure [Pascal]

R = 1.0
10000 R = 1.1
R = 1.2
8000 R = 1.3
R = 1.4
6000
R = 1.5
4000 R = 1.55
R = 1.6
2000

0
0 50 100 150 200 250 300 350 400
Angular Position [ ◦ ]

Figure 7.8: Plot of total pressure versus angular position for varying radial position

300 Plot of Velocity versus Angular Position

250 R = 0.7
R = 0.8
200 R = 0.9
R = 1.0
Velocity [m/s]

R = 1.1
150 R = 1.2
R = 1.3
100 R = 1.4
R = 1.5
R = 1.55
50 R = 1.6

0
0 50 100 150 200 250 300 350 400
Angular Position [ ◦ ]

Figure 7.9: Plot of velocity versus angular position for varying radial position

the circumferential distortion intensity, the mid radius was used as a reference radius for comparison
with other inlet configuration. The calculation procedures are based on equations 2.52, 2.53, 2.54,
2.55 presented in chapter 2. The results are illustrated in figure 7.10.

77
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

18000
Plot of Total Pressure versus Circumferential Location at Mid Radius

16000

14000

12000
Total Pressure

10000

8000

6000
Data
4000 Curve Fit
P Average
P Average Low
2000
0 50 100 150 200 250 300 350 400
Circumferential Location [ ◦ ]

Figure 7.10: Calculation of distortion intensity at mid radius

The calculated values of the circumferential distortion intend (θi )¯ and circumferential distortion inten-
sity ( ∆p c ◦
p )i were found to be 103.482 and 0.69135 respectively at the mid radius location. Applying
the curve fitting and numerical integration approach to all radial positions, the circumferential aver-
aged pressure was then plotted as a function of the spanwise radial position. This is illustrated in
figure 7.11.

78
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

1.8
Plot of C.A. Outlet Plane Total Pressure versus Radial Span Position

1.6

1.4

1.2
Radial Position [m]

1.0

0.8

0.6

0.4

0.2 C.A. Total Pressure for each R.


Radial Average Total Pressure
0.0
4000 6000 8000 10000 12000 14000 16000 18000
Total Pressure [Pascal]

Figure 7.11: Plot of Circumferential Averaged Total Pressure at various Spanwise Radial Position

As seen in figure 7.11, the circumferential averaged total pressure reduces as we approach the tip of the
spanwise radial position. This is expected since the fuselage boundary layer accumulates primarily at
region close to the bottom segment of the outlet plane, leading to a greater pressure drop circumfer-
entially. The non uniform boundary condition would then be applied onto the NASA Rotor 67 based
on a scaling rule so as to fit the exact geometry of the blade. The radial averaged total pressure was
then computed using the Riemann sum approach. The computed value of the radial averaged total
pressure was found to be 12609.15 Pascal.

The obtained results from figure 7.11 will be used as the non uniform inlet boundary condition while
the uniform inflow condition will be based on the radial averaged total pressure. The outlet static
pressure condition of the fan was then obtained by using the total to static pressure ratio found from
[16]. With the inlet total pressure as well as the outlet static pressure quantified, the 3D simulations
were performed.

7.3 ANSYS CFX Solver Settings and Mesh Convergence Studies


In this section of the chapter, the simulation settings used in the ANSYS CFX solver will be explained.
As the CFD solver SU 2 was not able to simulate 3D geometries with periodic boundary condition
despite extensive testing, ANSYS CFX was employed as a backup CFD solver for this part of the
simulation. Other than ensuring a ’good’ mesh for a converged simulation, it is also necessary to have
an optimal set of solver parameters for the simulation. The most commonly considered parameters
used for the simulation will be described in this section.

79
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

7.3.1 Simulation Setup


The following parameters were considered for the simulation setup. Firstly, a steady state simulation
is being considered. The rotation speed specified is based on the design rotation speed of 16043 rpm.
The actual direction of the rotation speed is determined from the relative inflow angle (derived from
the rotor blade geometry) as well as the specified absolute inflow angle. As for the inlet and outlet
boundary condition, the total and static pressure were specified respectively. At the hub of the blade
domain, a no slip wall boundary condition where the velocity of the fluid at the wall boundary is set
to zero is specified. CFX by default considers that the wall is rotating with the grid and hence they
are moving with respect to the stationary reference frame. For the case of a fan rotor without casing,
a no slip wall with zero rotation velocity will be specified at the tip of the blade domain. In the case of
a fan rotor with casing, the casing wall is stationary in the non rotating reference frame and therefore
a counter rotating velocity, −ωR is specified at the tip wall boundary. In this simulation, the case of
a rotor with and without casing was considered.

For the uniform inflow boundary condition, the inlet total pressure is specified based on the radial
averaged total pressure found from the outlet plane of the finalized inlet configuration. As for the non
uniform inflow boundary condition, the values were specified based on the circumferential averaged
total pressure values found for each radial position (See figure 7.11). In order to apply the circumfer-
ential averaged pressure value found at each radial position onto the actual scale of the NASA Rotor
67, the radial values of both the inlet configuration and the rotor blade was normalized before the
total pressure values found at the outlet the inlet configuration can be applied across the span of the
rotor blade inlet. The two main turbulence model available in ANSYS CFX is the k-epsilon and SST
model. The k-epsilon turbulence model uses a scalable wall function for near wall treatment. The
main advantage of the scalable wall functions is that they can be applied on arbitrary fine meshes.
In the case that the boundary layer is not fully resolved, CFX relies on the logarithmic wall func-
tion approximation to model the boundary layer without affecting the validity of the scalable wall
function approach. In the case that detailed investigation of the boundary layer is required, the SST
turbulence model is used in conjunction with the automatic near-wall treatment to take advantage of
the additional effect in the viscous sublayer. For this simulation, the SST turbulence model is being
considered. For the solver control settings, a residual target of 1e-7 as well as a maximum of 500
iterations is specified for each test run. This is necessary in order to ensure convergence of the various
residuals.

7.3.2 Mesh Convergence Studies


In order to ensure that the finalized mesh density does not influence the accuracy of the simulation
results, meshes of varying density were created and simulated. The mass averaged pressure change
across the inlet and outlet face of the rotor was compared for all the mesh densities created. The chosen
mesh density should ensure convergence in the results of the chosen simulation parameter while keeping
the computational cost to a minimal. Due to the limit imposed on the allowable minimum discretized
mesh size by SALOME, the setting which gives the densest mesh and provides the converged solution
was chosen. A mesh with a total of 686328 elements and 131847 nodes was finally chosen for this
simulation.

80
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

7.4 3D Uniform and Non Uniform Inflow Simulation Results


In this section of the chapter, the 3D results will be presented and discussed. In particular, the focus
of the results will be on the performance comparison of the rotor under both uniform and non uniform
inflow boundary condition as well as the visualisation of the blade to blade domain for the velocity
vector, pressure as well as the entropy loss across the blade channel. Finally, the streamlines view of
the meridional channel will also be presented. An illustration of the isometric view of the 3D blade
assembly (22 blades) as well as the meridional view of a single blade is presented in figure 7.12 and
7.13.

Figure 7.12: Isometric View of 3D Blade Assembly Figure 7.13: Meridional View of Single Blade

7.4.1 Blade Loading Comparison


A comparison of the blade loading plot for the case of a uniform and non uniform inflow condition is
presented in figure 7.14 to 7.19. For this comparison, the ducted fan configuration was used in the
comparison of the results between the uniform and non uniform inflow condition. The blade loading
at the 20%, 50% and 80% span are presented. Since non uniform inflow results in a change in the
circumferential averaged total pressure at each radial position, a change in the inflow incidence angle
is expected. The resulting effect is a change in the blade loading and the overall performance of the
rotor in terms of efficiency and entropy loss. As the change in total pressure is more abrupt near the
tip and hub section of the blade, greater differences in the blade loading at the 80% and 20% span
position is expected.
In figure 7.14, 7.15, 7.18 and 7.19, it can be seen that the blade loading plot at the 20% and 80% span
position exhibits more differences as compared to the blade loading plot at the 50% span position.
Also, it is noticeable that a larger change in the blade loading occur at the trailing edge pressure
side of the blade. At the 20% span position, the shock (abrupt change in pressure) is shifted towards
the trailing edge while it is shifted towards the leading edge at the 80% span position. Secondly, the
strength of the shock also reduces as a result of the lower inlet total pressure for the non uniform
boundary condition.

7.4.2 Comparison of the Blade to Blade Domain Visualization


The visualization of the velocity vectors as well as the entropy change across the blade to blade domain
is presented below. The plot of the velocity vector is illustrated in figure 7.20, 7.21, 7.22, 7.23, 7.24

81
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.14: 20% Span blade loading Uniform Figure 7.15: 20% Span blade loading Non Uniform
Inflow Inflow

Figure 7.16: 50% Span blade loading Uniform Figure 7.17: 50% Span blade loading Non Uniform
Inflow Inflow

and 7.25
As seen above, the effect of the non uniform inflow is a change in the inlet incidence angle as well
as the magnitude of the relative inlet velocity due to reduced axial velocity. This is illustrated very
clearly in the velocity vector plots at the 80% span position. In figure 7.24 and 7.25, other than a
reduced magnitude of the relative inlet velocity, the change in the inlet incidence angle resulted in
a forward shift of the shockwave near the TE as well as slight difference in the wake pattern. The

82
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.18: 80% Span blade loading Uniform Figure 7.19: 80% Span blade loading Non Uniform
Inflow Inflow

Figure 7.20: 20% Span Uniform Inflow Blade Figure 7.21: 20% Span Non Uniform Inflow Blade
to Blade Domain Velocity Vector to Blade Domain Velocity Vector

differences in the wake pattern is illustrated more clearly in the entropy plot of the blade to blade
domain. This is presented in figure 7.26, 7.27, 7.28, 7.29, 7.30 and 7.31.

83
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.22: 50% Span Uniform Inflow Blade Figure 7.23: 50% Span Non Uniform Inflow Blade
to Blade Domain Velocity Vector to Blade Domain Velocity Vector

Figure 7.24: 80% Span Uniform Inflow Blade Figure 7.25: 80% Span Non Uniform Inflow Blade
to Blade Domain Velocity Vector to Blade Domain Velocity Vector

From the entropy plots shown in figure 7.26, 7.27, 7.28, 7.29, 7.30 and 7.31 , it can be concluded that
the increased incidence angle leads to a reduced wake trail. Finally, the pressure contour of the blade
to blade domain is presented in figure 7.32, 7.33, 7.34, 7.35, 7.36 and 7.37. From these plots, the shift
in the TE shockwave as well as the change in the TE wake pattern can be clearly visualized. This is
especially clear in the 80% span position plot of the blade to blade domain static pressure contour.

7.4.3 Comparison of the Streamlines Visualization


The visualization of the streamlines between the case of a uniform and non uniform inflow is presented
in figure 7.38 and 7.39. It can be clearly seen that the effect of a non uniform inflow is a change in
the magnitude of the velocity vector near the tip of the blade. Although the change in incidence angle
at the blade’s LE cannot be observed clearly, the effect of the increased incidence angle is reflected in
the change in flow pattern at the trailing edge of the blade.

84
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.26: 20% Span Uniform Inflow Blade Figure 7.27: 20% Span Non Uniform Inflow Blade
to Blade Domain Entropy Change to Blade Domain Entropy Change

Figure 7.28: 50% Span Uniform Inflow Blade Figure 7.29: 50% Span Non Uniform Inflow Blade
to Blade Domain Entropy Change to Blade Domain Entropy Change

85
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.30: 80% Span Uniform Inflow Blade Figure 7.31: 80% Span Non Uniform Inflow Blade
to Blade Domain Entropy Change to Blade Domain Entropy Change

Figure 7.32: 20% Span Uniform Inflow Blade Figure 7.33: 20% Span Non Uniform Inflow Blade
to Blade Domain Static Pressure to Blade Domain Static Pressure

86
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.34: 50% Span Uniform Inflow Blade Figure 7.35: 50% Span Non Uniform Inflow Blade
to Blade Domain Static Pressure to Blade Domain Static Pressure

Figure 7.36: 80% Span Uniform Inflow Blade Figure 7.37: 80% Span Non Uniform Inflow Blade
to Blade Domain Static Pressure to Blade Domain Static Pressure

87
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.38: Visualisation of Uniform Inflow Streamlines

88
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.39: Visualisation of Non Uniform Inflow Streamlines

89
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

7.4.4 Performance Comparison of the Rotor Under Uniform and Non Uniform
Inflow

In this section, the performance of the rotor under both uniform and non uniform inflow will be
compared. In particular, the main aspect of the performance comparison would be to look into the
isentropic efficiency of the rotor, static entropy change, the total to total pressure ratio as well as
the enthalpy change across the rotor. It is noted that these quantities are the mass flow averaged
quantities obtained at the mid radius of the rotor.

Table 7.5: Performance of the Rotor under Non Uniform Inflow Boundary Condition

Parameters Units Inlet Outlet Ratio or Change


Temperature [K] 76.6 108 -
Total Temperature [K] 86 129 -
Pressure [Pascal] 6780 12848.1 -
Total Pressure [Pascal] 10281.9 24518.4 -
Enthalpy [m2 /s2 ] -222521 -191051 -
Total Enthalpy [m2 /s2 ] -213084 -169730 -
Total to Total Pressure Ratio - - - 2.385
Static Entropy [J/kgK] -587.58 -430.49 157.1
Total to Total Isen. Efficiency - - - 0.5814

Table 7.6: Performance of the Rotor under Uniform Inflow Boundary Condition

Parameters Units Inlet Outlet Ratio or Change


Temperature [K] 94.76 123 -
Total Temperature [K] 105 144.2 -
Pressure [Pascal] 8805.63 15197.3 -
Total Pressure [Pascal] 12608 27077.6 -
Enthalpy [m2 /s2 ] -204286 -175925 -
Total Enthalpy [m2 /s2 ] -193998 -154651 -
Total to Total Pressure Ratio - - - 2.15
Static Entropy [J/kgK] -449.891 -345.2 104.691
Total to Total Isen. Efficiency - - - 0.627926

As seen in table 7.5 and 7.6, the effect of the non uniform inflow is an increase in entropy production
and a decrease in the isentropic efficiency. For the non uniform boundary condition, the mass average
total pressure at the inlet is lower as compared to the uniform boundary condition. Therefore, the
outlet static pressure for both simulation cases are different.

7.5 3D Uniform Inflow Simulation Results With/Without Casing

In this section of the chapter, the 3D results will be presented and discussed. In particular, the focus
of the results will be on the blade loading and performance comparison of the rotor with and without
a casing as well as the visualisation of the streamlines view of the meridional channel. In addition,
the plot of the circumferential average total pressure at the inlet and outlet domain will be presented.

90
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

7.5.1 Blade Loading Comparison


The blade loading at the 20%, 50%, 80%, 95% and 99% span position is presented in figure 7.40, 7.41,
7.42, 7.43 and 7.44.

Figure 7.40: 20% Span Blade Loading Figure 7.41: 50% Span Blade Loading
Comparison Comparison

Figure 7.42: 80% Span Blade Loading Figure 7.43: 95% Span Blade Loading
Comparison Comparison

From the above figures, it can be seen that the blade loading of the rotor with and without casing
exhibits very similar blade loading near the mid region of the blade. However, close to the trailing

91
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.44: 99% Span Blade Loading


Comparison

edge of the blade, the blade loading of both cases exhibits differences in terms of the pressure rise due
to the shock near the tip region of the blade (from 95% span position of the blade onwards). From the
20% span to the 80% span position of the blade, the pressure rise near the TE appears to be slightly
higher and shifted towards the TE for the case of the rotor without the casing.

7.5.2 Comparison of C.A. Total Pressure across Normalized Span Position


In this subsection, the comparison of the circumferential averaged total pressure versus the normalized
span at the inlet and outlet domain will be presented. It can be seen that the total pressure is higher at
the tip position for the case of the rotor with casing. This is mainly because of the reduced tip losses as
compared to the unducted rotor. It is also noted that the outlet total pressure in the stationary frame
is higher for the case of the rotor without the casing near the mid span of the blade. In addition, there
is a small increase in total pressure near the tip region of the blade (without casing). Apparently, the
lower inlet total pressure results in a lower axial and relative velocity approaching the blade. A lower
axial velocity leads to an increase in the inlet incidence angle and this could be beneficial in terms
of increased blade loading and enthalpy change over the rotor. As a result, a small increase in total
pressure exists on the outlet plane of the blade tip zone as compared to the ducted rotor configuration.

7.5.3 Comparison of the Streamlines Visualization


The visualization of the streamlines between the case of a rotor with and without casing is presented
in figure 7.47 and 7.48. As a result of the reduced total pressure at the tip zone for the unducted
rotor, the maximum velocity present at the tip of the rotor blade is also smaller for the same inlet
and outlet boundary conditions used in the simulation.

92
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.45: Circumferential Average Total Pres-Figure 7.46: Circumferential Average Total Pres-
sure Versus Normalized Span Position at Inlet sure Versus Normalized Span Position at Outlet

Figure 7.47: Visualisation of Rotor Streamlines (with casing and Uniform Inflow)

93
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

Figure 7.48: Visualisation of Rotor Streamlines (without casing and Uniform Inflow)

94
Master Thesis Report 3D Flow Simulation of the NASA Rotor 67

7.5.4 Performance Comparison of the Rotor Under Uniform Inflow with/without


Casing
In this section, the performance of the rotor under uniform inflow for the case with and without casing
will be presented. In particular, the main aspect of the performance comparison would be to look into
the isentropic efficiency of the rotor, static entropy change, the total to total pressure ratio as well
as the enthalpy change across the rotor. It is noted that these quantities are the mass flow averaged
quantities obtained at the mid radius of the rotor. The results are presented in table 7.7 and 7.8.

Table 7.7: Performance of the Rotor (with casing) under Uniform Inflow Boundary Condition

Parameters Units Inlet Outlet Ratio or Change


Temperature [K] 94.76 123 -
Total Temperature [K] 105 144.2 -
Pressure [Pascal] 8805.63 15197.3 -
Total Pressure [Pascal] 12608 27077.6 -
Enthalpy [m2 /s2 ] -204286 -175925 -
Total Enthalpy [m2 /s2 ] -193998 -154651 -
Total to Total Pressure Ratio - - - 2.15
Static Entropy [J/kgK] -449.891 -345.2 104.691
Total to Total Isen. Efficiency - - - 0.627926

Table 7.8: Performance of the Rotor (without casing) under Uniform Inflow Boundary Condition

Parameters Units Inlet Outlet Ratio or Change


Temperature [K] 77.4 104.43 -
Total Temperature [K] 86 125 -
Pressure [Pascal] 8723.38 15227.7 -
Total Pressure [Pascal] 12607.2 29052.3 -
Enthalpy [m2 /s2 ] -221715 -194573 -
Total Enthalpy [m2 /s2 ] -213082 -174221 -
Total to Total Pressure Ratio - - - 2.304
Static Entropy [J/kgK] -650.358 -511.3 139.06
Total to Total Isen. Efficiency - - - 0.58484

From the above table, it can be seen that the case of a rotor without the casing experienced a reduction
in the isentropic efficiency despite having a slightly higher pressure ratio for the same inlet and outlet
boundary conditions used in the simulation. As seen in the plot of the circumferential averaged total
pressure versus normalized span, it can be inferred that the loss in efficiency and gain in entropy is
due to the lower total pressure at the tip of the blade as a result of tip vortex formation.

95
Master Thesis Report Conclusion and Recommendation

8|Conclusion and Recommendation

In the literature study that was conducted at the early stage of this research, it was concluded that
boundary layer ingestion has the following influence on engine design and cycle performance; firstly,
the engine suffers from a reduced specific thrust, reduced thrust to weight ratio as a result of in-
creased mass flow, an increased fan diameter and consequently an increase in the core size of the
engine. In terms of cycle performance, the engine suffers from reduced thermal efficiency but an in-
crease in propulsive efficiency due to the lower exit velocity. However, the net effect is a reduction in
the overall engine efficiency. The reduction in thermal efficiency is mainly a result of the lower inlet
total pressure and reduced compressor efficiency caused by non-uniformities in the flow. In essence,
the ingestion of the boundary layer represent a reduction in the overall efficiency of the engine cycle.
However, the momentum deficit captured by the engine represent a drag reduction to the aircraft.
Therefore a trade-off exists between the increase in aircraft drag reduction (increased fuel efficiency)
and the decrease in overall engine efficiency (reduced fuel efficiency) as more boundary layer is ingested.

Using the result of the literature review, an effort was made to investigate the influence of the non
uniform inflow boundary condition on compressor efficiency. In order to do that, the original NASA
Rotor 67 geometry was reconstructed and parametrised in an in house Blade Modeller. Before a full
3D simulation can be carried out, the sub-goal is to verify the 2D fitting approach of the blade profiles.
As previously mentioned, the blade profiles can originate from three main sources. The first source is
the blade profile generated from the original geometrical coordinate. The second source is the blade
profile generated from the fitting of the blades using the geometry fitting algorithm. Finally, the last
source represent the blade profile generated from the slice of the blade surface. From the 2D results,
it was found that the 2D fitted and Actual point coordinate sources of profile 1265 and 1865 have the
closest fit in terms of blade loading, Mach/Pressure Contour as well as their stagnation pressure loss
results. However, in the case of profile 2265 and 1565, it was found that the 2d fitted and surface
fitted profile have the closest fit in terms of their blade loading, Mach/Pressure Contour and stagna-
tion pressure loss results. Therefore, this result suggests that it is difficult to pinpoint a particular
source of error for the blade fitting procedures. However, it can be seen that minor deviation in the
leading/trailing edge position and curvature of the blade profile can bring about a significant shift in
the shockwave position. It can also be concluded that the BSpline surface generation algorithm gives
a better fit to the 2d fitted profiles closer to the tip radius than the hub radius.

For the 3D simulation results comparing the case of a uniform and non uniform inflow boundary
condition, it was found that the effect of the non uniform inflow is an increase in entropy production
and a decrease in isentropic efficiency. In the blade loading comparison, it is noticeable that a larger
difference in the blade loading occur at locations close to the hub and tip section of the blade. This is
mainly because of the greater difference in inlet total pressure as compared to the uniform boundary
condition at such locations. In the blade to blade domain velocity vector comparison, it was found that
the effect of the non uniform inflow is a change in the inlet incidence angle as well as the magnitude
of the relative inlet velocity. This was illustrated clearly in the velocity vector plot at the 80% span

96
Master Thesis Report Conclusion and Recommendation

position. As a result of the change in inlet incidence angle, a shift in the position of the shockwaves as
well as slight difference in the wake pattern was expected. Close to the tip location, the pressure rise
is reduced as compared to the original case as a result of the reduced inlet total pressure. As for the
hub region, the pressure rise is larger as compared to the original case as a result of the increased inlet
total pressure. It is also noted that the shock shifts forward in the case of the reduced total pressure
as a result of the increased incidence angle and vice versa.

For the comparison of the 3D results between the case of a rotor with and without casing under
uniform inflow condition, it was found that the case of the rotor without the casing exhibits a reduc-
tion in isentropic efficiency. This is mainly because of the increased tip leakage losses as compared to
the case of a rotor with casing. However, it is noted that the pressure ratio over the unducted rotor is
slightly higher as compared to the ducted rotor. In this simulation, the total duct length is relatively
short and therefore the pressure induced by the shroud does not significantly add to the total pressure
ratio.

The results of the 2D simulation showed that the BSpline surface generation algorithm gives a bet-
ter fit closer to the tip section as compared to the hub section. Therefore, a more accurate surface
generation algorithm was needed. At the point of writing this report, a new algorithm was already
implemented in the later version of the Blade Modeller. The main goal of this new algorithm was to
ensure that the generated surface fits through the control points of all the control profiles. Secondly,
it was found that the accuracy of the 2d fitted profiles are very sensitive to the actual location of the
leading and trailing edge. Therefore, an accurate manual adjustment of the actual LE/TE position
from the closest selected distributed point is required at the early stage of the fitting process.

Finally, the analysis made from the 3D simulation showed that the blade profile can be better shaped
to improve the performances under non uniform inflow condition. Although it is difficult to quan-
tify the actual change in the blade angle at this stage of the research, some possible direction of
investigation will be discussed. As a result of the change in inlet total pressure and consequently an
increase/decrease in the inlet incidence angle, adjustment to the blade profile can be made in term of
the inlet incidence angle or the stagger angle. It is noted that changing the inlet blade metal angle
influences the enthalpy/pressure change across the blade profile and consequently the thermodynamic
properties of the engine cycle. Therefore, it is not recommended to change it if possible. Implement-
ing changes to the stagger angle at certain span position is a possibility. Detailed investigation by
numerical optimization is necessary in order to find the optimal stagger angle at each span position.
Master Thesis Report Bibliography

Bibliography

[1] A.P. Plas, M.A. Sargeant, V. M. D. E. G. and Hynes, T., “Performance of a Boundary Layer
Ingesting Propulsion System,” AIAA 2007-0450 , January 2007.

[2] Drela, M., “Power Balance Method in Aerodynamic Flows,” AIAA 2009-3762 , June 2009.

[3] Drela, M., “MIT, Aurora Flight Sciences, Pratt and Whitney, and NASA. Aircraft and Technology
Concepts for an N+3 Subsonic Transport - Phase 2,” Year 3 Status Review Presentation, February
2014.

[4] Hanlon, C., “Engine Design Implication for a Blended Wing-Body Aircraft with Boundary Layer
Ingestion,” Tech. rep., Master Thesis, Department of Aeronautics and Astronautics, 2010.

[5] Boxer, E., “Influence of wall boundary layer upon the performance of an axial flow fan rotor,”
NASA Langley Research Center, Technical Report TN-2291 , February 1951.

[6] Kimzey, W., “An analysis of the influence of some external disturbance on the aerodynamic stabil-
ity of turbine engine axial flow fans and compressor,” Tech. rep., Arnold Engineering Development
Center, Arnold Airforce Station, 1977.

[7] R.C.W.de.Koning, “Development of a Parametric 3D Turbomachinery Blade Modeller,” Tech.


rep., Master Thesis, Faculty of Aerospace Engineering, Delft University of Technology, 2015.

[8] Verstraete, T., “CADO: A Computer Aided Design and Optimization Tool for Turbomachinery
Applications,” 2nd International Conference on Engineering Optimization, VKI , September 2010.

[9] Verabhi, “Secondary Flow Losses,” https://en.wikipedia.org/wiki/File:Secondary_flow_


losses.png, 2013, [Online; accessed 03-August-2015].

[10] Dixon, S. and C.A.Hall, Fluid Mechanics and Thermodynamics of Turbomachinery, Elsevier, 6th
ed., 2010.

[11] Ferrar, A. and O’Brien, W., “Progress in Boundary Layer Ingesting Embedded Engine Research,”
AIAA 2012-4283 , August 2012.

[12] H.J.M. Kok, M. V. and van Tooren, M. J., “Distributed Propulsion featuring Boundary Layer
Ingestion Engines for the Blended Wing Body Subsonic Transport,” AIAA 2010-3064 , April 2010.

[13] Francisco Palacios, Michael R. Colonno, A. C. A. A. C. S. R. C. T. D. E. A. K. L. T. W. L.


T. W. T., “Stanford University Unstructured (SU 2 ): An open-source integrated computational
environment for multi-physics simulation and design,” AIAA 2013-0287 , January 2013.

[14] Medida, S. and Baeder, J., “Numerical prediction of static and dynamic stall phenomenon using
the γ − Reθt model,” American Helicopter Society 67th Annual Forum, May 2011.

I
Master Thesis Report Bibliography

[15] Yong Sang Yoon, Seung Jin Song, H. W. S., “Influence of Flow Coefficient, Stagger Angle, and
Tip Clearance on Tip Vortex in Axial Compressor,” ASME Journal of Turbomachinery, Volume
128 , November 2006.

[16] Anthony J. Strazisar, J. R. M. D. H. and Suder, K. L., “Laser Anenometer Measurements in


a Transonic Axial Flow Fan Rotor,” Tech. rep., NASA Lewis Research Center; Cleveland, OH,
United States, 1989.

[17] Sharma, A., “Design of Inlet for Boundary Layer Ingestion in a Blended Wing Body Aircraft,”
Tech. rep., Master Thesis, Faculty of Aerospace Engineering, Delft University of Technology,
2015.

II

You might also like