Thesis PDF

Study of a Test Cell for Commercial Jet Engines
Gabriela Liani Beisl Ramos
Thesis to obtain the Master of Science Degree in

Mechanical Engineering
Supervisors: Prof. José Maria Campos da Silva André
Prof. Luı́s Filipe Galrão dos Reis
Examination Committee
Chairperson: Prof. Viriato Sérgio de Almeida Semião
Supervisor: Prof. José Maria Campos da Silva André
Members of the Committee: Prof. João Eduardo de Barros Teixeira Borges
Eng. Francisco José Toste de Azevedo
July 2015
ii
To my Father, for his words that never failed and silences that made words speak.
To my Mother, who always managed to build wings for each of my dreams.
To both I owe the airplane of my life.
Also, to all the family and friends it carries inside.
iii
iv
Acknowledgements
This dissertation results from many converging efforts and from the contribution of many people. I
would wish to enlighten the way it was made possible from its very beginning.
I am thankful to all the teachers I met during this course in Instituto Superior Técnico - IST - the
institution that geared and developed my reasoning. I would like to name Professor José Mendes Lopes
and Luı́s Toste de Azevedo, whose teachings I still keep in memory. My appreciation to Professor Mário
Costa for his patience all along the conversations held until the theme of this thesis was formulated. I also
much appreciate the help of Professors Virgı́nia Infante and Luı́s Reis without whose support I wouldn’t
have been able to find the theme of my dissertation.
If IST enabled me to load and equip the said aircraft now ready for takeoff, TAP represents the
engines empowering it. Both the Engineering and Maintenance Teams made part of my final way for
graduation and to start earning a higher perspective of professional working life.
Together with Engineer Francisco Azevedo I thank Engineer Ricardo Azevedo for all the availability
and supervising: the work and development of this test cell facility of TAP much results from their
dedication and hard work. Also my appreciation to Miguel Henriques for his full support. To Engineers
António Ferreira, João Vizinha, Joana Palmeiro, Ricardo Corrêa, Ricardo Réfega, Marta Verı́ssimo, João
Pita, Pedro Nunes, Tiago Silva and to my colleagues that recently started making part of this team,
special thanks for pushing my ideas forward through their questions, thoughts and presence.
To the Maintenance Team, namely Mr. Marques, Vitor and João: thank you for teaching me, and
supporting me on the project undertaken within this study. The extra work that it has brought leaves
me also grateful to Manuel Lourenço and António Furtado, who have passed the baton to Bernardo Silva
and José Freches currently leading the test cell team. Their tireless work and contribution to my thesis
was indeed exceptional; my special thanks namely to Tiago Godinho, João Marafuz, Pedro Duarte, Túlio
Trindade and Rodrigo Vaz. I must stress that all the profiency experienced as intern has undoubtedly
contributed to forge and motivate my work capability.
I am also indebted to my internship colleagues Miguel Batista, Julio Ridaura and Bernardo Pereira,
to whom I wish the most favourable winds and to whom I owe an important part of the support received
during this internship at TAP. Another significant part of that support I owe to Jorge Ramos from the
team of equipment calibration and Luı́s Silva: the patience, advice and service offered were truly essential
for my work and experience. I wish to acknowledge TAP - Transportes Aéreos Portugueses - Lisbon, who
provided opportunity and resources for the project developed in this study.
My sincere thanks to Professor José Maria André for sowing knowledge and encouraging me towards
- and along - each challenge found. His guidance was remarkable and to him goes my greatest gratitude.
v
vi
Abstract
Engine test cells simulate representative air flow conditions of the engines running in still open atmo-
sphere. TAP’s facility is able to operate intermediate-size engines, however, before it can handle bigger
ones, a detailed survey becomes essential.
This study estimates the air flow thermodynamics and fluid dynamics and displays an overall insight
of the facility. The computational model (TCC) relies on known engine data and measurements; further
experimental data was collected for this project, to validate the model. With the TCC tool it is possible to
foresee the test cell behavior operating with bigger engines and to quantify some alternative improvements
for the facility.
Less peripheral air flow being pumped may result in vortices, formed upstream of the engine intake,
that may lead to engine surge: increasing the CBR (cell bypass ratio) is a first approach to avoid this
problem. Nevertheless, other solutions are suggested, such as eliminating the wall of the augmenter inlet
section (upstream of the secondary air-intake) and installing ramps around the walls of the chamber,
near the engine bellmouth.
The test cell current configuration suffers from several sources of aerodynamic instability. Suggestions
are put forward, namely introducing changes in the diffusor and adding guide vanes at each turning of
the exhaust stack, for a better flow stabilization and resonance avoidance.
This research collects technical information and addresses a number of corrective measures. As a
conclusion, the discussion highlights TAP’s test cell potential to meet the actual and future demand of
maintenance operations of the aviation market.
Keywords: Jet Engine Test Cell, Computational Model, Large Turbofans, TAP.
vii
viii
Resumo
Os bancos de ensaio de motores a jacto simulam as condições de ensaios em espaço aberto sem vento.
O banco de ensaios da TAP opera com motores de tamanho intermédio mas precisa de uma análise
detalhada para ensaiar motores maiores.
Estudou-se o escoamento de ar e as propriedades termodinâmicas, para obter uma visão global da

instalação. O modelo numérico (TCC) utiliza parâmetros conhecidos do motor e medidas efectuadas
no banco de ensaios; acrescentaram-se outras medições no âmbito deste projecto, para validar o modelo
numérico. Esta ferramenta prevê o comportamento de motores maiores neste banco de ensaios e permite
quantificar algumas alternativas de melhoria da instalação.
Bombar menos ar periférico na sala pode libertar vórtices a montante da admissão do motor, que
podem danificar o motor: aumentar o CBR (razão de bypass na sala) poderia ser uma primeira solução.
No entanto, sugerem-se outras opções, tais como eliminar a parede da actual secção de entrada do
augmenter (a montante da segunda admissão de ar da sala) e instalar carenagens à volta das paredes da
sala, perto da admissão do motor.
A configuração actual do banco de ensaios apresenta fontes de instabilidade aerodinâmica. Sugere-se
nomeadamente alterar o difusor e adicionar pás de guiamento nas curvas do escape, com vista a estabilizar
o escoamento e evitar ressonâncias.
Este estudo coligiu informações técnicas e analisou uma série de medidas correctivas. Verificou-se o
potencial do banco de ensaios da TAP para atender a procura actual e futura do mercado da manutenção
de aeronaves.
Palavras-chave: Banco de Ensaios, Modelo Numérico, Motores a Jacto Comerciais, TAP.
ix
x
Contents
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
1 Introduction 1
1.1 Summary of the Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Typical Test Cell Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2 Typical Challenges of Test Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.3 Cell Depression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.4 Engine Bypass versus Test Cell Bypass . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.5 Potential Cores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.6 Correction Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.7 Engine to Agumentor Distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 TAP’s Test Cell 9

2.1 TAP’s Historical Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Head Loss Coefficients: Intake Section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.1 Inlet Head Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.2 Guiding Vane Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Intake Grid Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.4 Intake Silencer Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 The Augmenter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.1 Ring diffuser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Diffusion Section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.1 Equivalent Divergence Angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.2 Diffusor Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.3 Flow Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.4 Flow Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.5 Sudden Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.5 Head Loss Coefficients: Exhaust Section . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.6 Technical Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6.1 Selected Engines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6.2 Engine Bleeding and Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6.3 Test Cell Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6.4 Expected Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
xi
3 TCC Model 31
3.1 Target and Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Model Simplifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Mathematical Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.1 Energy Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.2 Test Room and Engine Air Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.3 Thermodynamic Flow Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3.4 Mean Flow Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.4 Engine Thrust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5 Entrainment Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5.1 Engine Exhaust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5.2 Flow Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.5.3 Ejector Pump Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.6 Harp Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.7 Augmenter Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4 Computational Model Results 55

4.1 Discussion of the Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5 Test Cell Upgrade 61

5.1 Summary of the Test Cell Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.2 Upgrade Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6 Conclusions 69
6.1 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Bibliography 73
A General Details 80
A.1 ISA - International Standard Atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
A.2 Model Basic Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
A.3 Koppers Harp Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
A.4 Probes Project Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
A.5 Probes Thermocouple Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
B Head Loss Coefficients Details 84
C Additional Figures 89
D Additional Tables 91
xii
List of Tables
2.1 Different approaches for the sudden expansion resistance cofficient estimate for the 7.9
equivalent diffusor divergence angle. The velocity distribution refers to the upstream section. 22
2.2 Engines analyzed in the study. Data for takeoff at sea level conditions were obtained from
http://www.jet-engine.net/civtfspec.html. (images from [1] and [31]) . . . . . . . . . . . . 24
2.3 Average result of the head loss coefficient for each test run. At last column: the global
mean value, k∞−2 = 7.73. Note: stable pressure readings at takeoff could not be found for the -3C
test run of 17-Set-14. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1 TCC model details. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.2 Molecular weight and gas constant Rexh for air and for products of combustion of lean
fuel-air mixtures. Fuel is (CH2 )n . Source: [5]. . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2 Engine exit velocities predicted with GasTurb and the TCC model. . . . . . . . . . . . . . 56
4.3 Cell bypass ratio CBR predicted from [16], from the TCC model and from TAP measure-
ments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.1 Engine test runs selected from the ones obtained between 20/Aug/14 to 14/Apr/15, at
TAP’s test cell facility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.1 Parametric study results using the TCC model. . . . . . . . . . . . . . . . . . . . . . . . . 66

5.2 Exhaust stack resonant modes for the engines analyzed in this study. . . . . . . . . . . . . 67
A.1 Radius and areas defining the rings of the Koppers harp. . . . . . . . . . . . . . . . . . . . 82
B.1 Resistance coefficients for the equivalent diffusors. Case i refers to a no insert configu-
ration: the thin boundary layer (b.l.) developing at the diffusor walls would be similar
to the one obtained with a short insert such as the augmenter tube (d.f. stands for fully
developed flow). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
B.2 Minimum resistance coefficients for a stepped-diffusor, for distinct approaches to find a
diffusor equivalent to TAP’s one ([22], Diagrams 5-9 to 5-11). . . . . . . . . . . . . . . . . 86
D.2 Specifications of civil turbofan engines tested at TAP in the past, 1967-1989. (Source:
http://www.jet-engine.net/). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
D.3 Specifications of civil turbofan engines: currently under maintenance test runs at TAP
(left) and targeted to be tested in the future (right). More details in http://www.jet-
engine.net/civtfspec.html. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
D.1 Test cell facility evolution from its inauguration and summary of recommendations fore-
seeing its upgrade. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
xiii
xiv
List of Figures
1.1 A typical test cell design (image from MDS - measured by the power of precision). . . . . 2
1.2 Test cell bypass ratio for the engines tested at TAP. . . . . . . . . . . . . . . . . . . . . . 4
1.3 Details about potential cores at a vertical centerplane of the exhaust of turbofan engines
with bypass ratio BPR = 5.0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Effect of the engine-to-augmenter distance on the cell air flow for relatively “large” turbofan
engines (Adapted from [25]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Test cell sections and stations (adapted from [24]). . . . . . . . . . . . . . . . . . . . . . . 9

2.2 Test cell geometry (From the 1:12 Scale Model Study [37], pp. 6-23). . . . . . . . . . . . . 11
2.3 Grid details. a) screens at stations SB and S2 ; b) test cell chamber; c) screens at station
S7B (left: side grids at the periphery of the exhaust; right: grid middle channel). . . . . . 12
2.4 Left: barriers distributed over the cross channels: screens made of circular metal wire.
([22], Diagrams 8-6). Right: drag coefficient CD for circular cylinders with infinite length
and smooth surface, function of the Reynolds number based on the hydraulic diameter,
Re Dh (From [3]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.5 Staggered tube bundle arrangement. Left: friction factor f and correction factor χ for
computation of the pressure drop. Right: flow conditions and geometric quantities for
staggered tubes ([7], pp. 437-443). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6 TAP noise reduction harp structure in the test cell augmenter. . . . . . . . . . . . . . . . 16
2.7 a) flat-diffusor stability map. (adapted from Fig. 6.26 of [38], page 382); b) diffusor with
a transitional cross-section, at TAP ([22], diagram 5-28). . . . . . . . . . . . . . . . . . . . 17
2.8 Velocity distribution at the inlet of diffusors, with a straight tube upstream ([22], p. 276). 18
2.9 Left - a) ideal pattern with good performance; b) pattern with boundary-layer separation
and resultant poor performance (Fig. 6.27 from [38]). Right - Flow losses in a conical
expansion region (source: [38], Fig. 6.23, p. 372). . . . . . . . . . . . . . . . . . . . . . . . 19
2.10 a) exhaust stack section. b) exit short 180° turn (head loss coefficient taken for a ratio of
bx /a0 = 3.17/5.28 = 0.6). c) Hood with three-sided exit (with an exit lenght of b = 4.8
and thus a ratio of b/a0 = 4.8/3.17 = 1.5, which can be extrapolated from the upper limit
of b/a0 = 1.2 from the author’s data). (Source: [22], diagram 9-31 a) and diagram 9-31 b).
All coefficients apply for the exit configuration). . . . . . . . . . . . . . . . . . . . . . . . . 23
2.11 Turbofan engine stations and hardware. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.12 Test cell static probes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.13 Top: Static pressure readings and the head loss coefficient for the cell intake (30/Nov/14,
cfm56-5B5/P test run). Bottom: idem, filtered readings. (LH: left-hand side pressure
probe; RH: right-hand side pressure probe (ALF)). . . . . . . . . . . . . . . . . . . . . . . 27
2.14 Averaged values for the head loss coefficient from the static pressure readings of LH, RH
probes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.15 Echeloned static pressure readings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
xv
2.16 Temperature and static pressure measurements at the test cell facility at TAP (from a
cfm56-3C engine). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1 I) Static cell pressure at each station S; II) Algorithm used to compute the cell mass flow
rate ṁ1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Left: head loss coefficients scheme, considering the exhaust grids (S7A − S7B ) and the
turnings along the exhaust channel (S7B − S7C ). Right: diffusion section scheme. . . . . . 38
3.3 Thermodynamic properties: air and combustion products specific heat and exhaust gas
constant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4 Engine momentum balance control volume CV. a) The test cell cross section is taken for
the control volume; b) CV coinciding with the engine outer walls. . . . . . . . . . . . . . 43
3.5 Left: Exit velocities depending on engine air flow and density. Right: Exhaust flow veloc-
ities, from GasTurb (BPR=5.28). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.6 A cfm56 engine scheme with the primary, secondary and entrained flow streams. The
radius defining the potential cores are also shown. . . . . . . . . . . . . . . . . . . . . . . 46
3.7 CBR trendline comparison between different author studies for turbojet engines. . . . . . 48
3.8 Sectioned view of the 2D ring harp structure (obtained from CAD Design Software Solid-
Works r ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.1 From left to right: engine rear configuration of the -3C, -5B, -C2 and -E1 engines. . . . . 57
4.2 Exponential trend of temperature T7A along time: T7A = T∞ − (T∞ − T0 ) exp [−β (t − t0 )],
where the fluid temperature near the probe is T∞ = 73.5 ◦C, T0 = 70.2 ◦ C is the probe
temperature at t0 = 14 : 57 : 52 (hh:mm:ss), taken as the initial instant of time, t − t0
is the time since the initial instant of time measured in s and β = 0.7 s−1 is a constant,
characteristic of the response of this probe. . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.1 Inlet ramp structures (images from [11]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

5.2 Effect of harp rings removal on the cell mass flow rate and CBR. . . . . . . . . . . . . . . 65
6.1 Separation in diffusors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

6.2 Pressure and mass flow rate as a function of the velocity uniformization ζ along the aug-
menter (results obtained with the TCC model for the -3C engine, based on the test run
data of 03-Oct-2014, with the Koppers harp installed). . . . . . . . . . . . . . . . . . . . . 71
6.3 Secondary air intake (adapted from [24]). . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
B.1 Sudden expansion details. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
C.1 TAP’s test cell design. Left: a 3D insight. Right: geometric altitude of TAP test cell facil-
ity, Lisbon’s Airport; the prevailing winds are from North and West. (from the Aeropho-
togrammetric Survey document of TAP). . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
C.2 TAP’s exhaust section plant (from [24]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
xvi
Nomenclature
Abbreviations
ALF After Looking Forward
b.l. Boundary layer
d.f. (Fully) developed flow
LH Inlet chamber pressure probe (Left Hand)
LPT Low Pressure Turbine
RH Inlet chamber pressure probe (Right Hand)
SE Sudden Expansion
TAP Transportes Aéreos Portugueses
TCC Test cell computational model
Greek Symbols
α Diffusor divergence angle
ǫ Wall roughness
γ Ratio of specific heats
σ Exhaust flow split coefficient
ξ Koppers harp uniformization factor
ζ Efficiency parameter of the augmenter
Roman Symbols
Ra Outer radius of the entrained flow
AFR Air to Fuel Ratio
BPR Engine bypass ratio
CBR Cell bypass ratio
Cp Frictionless diffuser pressure recovery coefficient
f Friction factor
n Diffusor area ratio
xvii
Re Reynolds number
S Any cross-section of the test cell
sto Stoichiometric
p′ Static pressure (gauge)
p′0 Total pressure (gauge)
p′din Dinamic pressure (gauge)
A Cross-sectional area
D Diameter
d Wire diameter of a screen
EF Energy flow rate
FN Engine thrust
L Length
MM Momentum flow rate (longitudinal component)
M Molecular weight
N1 Engine fan speed
PS2 Engine inlet static pressure
PT54 Total pressure of engine station 54
Q Volume flow rate
TT54 Total temperature at engine station 54
U Head loss factor
v Flow velocity (longitudinal component)
W Width
Subscripts
1A − 1B Acoustic treatment intake
1B − 1C Turning baffles section
1C − 1D Bird shield screen
1E − 1G Channel wall acoustic treatment
1H − 2 Flow conditioning screen
4A Engine fan discharge cross-section
4B Engine exhaust cross-section
4C Centerbody endtip cross-section
xviii
aug Augmenter section
avg Average value
bar Barometric pressure
dif Diffusor section
exh Exhaust thermodynamic quantity
fu Fuel
nw Diffusor with no wedge
p Engine primary (core) flow
per Test cell peripheral flow
s Engine secondary (fan) flow
sep Flow separation
w Diffusor with wedge
xix
xx
Chapter 1
Introduction
1.1 Summary of the Study

This study estimates the air flow thermodynamics and fluid dynamics for turbofan engine testings and
displays an overall insight of TAP’s test cell facility. A technical investigation was held refering to the test
cell enclosure; it was found that the current configuration of the test cell suffers from several sources of
aerodynamic instability. Also, engine surge associated to vortex formation can occur with specific engines
during runtime. The challenges that TAP has faced since 1969 and the ones TAP is facing currently are
issued in the literature found; typical challenges of test cells are presented in Section [section:state of art].
Hardware information and data from test reports were collected for a selected set of engines currently
tested at TAP. Some pressures and temperatures were measured, thus the velocities and air flow rates
could be determined by means of the Test Cell Computational (TCC) model, developed in MATLAB.
This model recreates the upstream to downstream characteristics of the test cell air flow and is a useful
tool to foresee the test cell behavior operating with bigger engines and to quantify some alternative
improvements for the facility.
The Chapter 2 provides an insight about TAP’s test cell facility characteristics, about the engines
tested and the test cell instrumentation available. Test Cell Computational model (TCC) relies on known
engine data and measurements; however, considered as insufficient to validate the model, additional ex-
perimental data was sought within two projects developed in the framework of this work and is presented
in Section 2.6.3.
The actual layout of the exhaust section may not allow the test cell to run more powerful engines. To
support this, the CBR (cell bypass ratio) as well as the front cell approach velocity and the mixed gas
temperature through the exhaust were predicted for each of the selected engines. All the simulations done
refer to steady-state regimes at takeoff power1 . The TCC model methodology is described in Chapter 3,
to which follows a summary of the results on Chapter 4.
Once validated through comparison between its estimates and the experimental data collected, the
TCC model enables flow behaviour prediction for an engine with higher dimensions. An upgrade study
for the test cell considers the effect of reducing the chamber room’s cross sectional area; the effects of
changing the engine-to-augmenter distance as well as rearranging some of the structures at the exhaust
room section are also evaluated and presented in Chapter 5.
This research collects technical information and addresses a number of corrective measures. A set of
1 Amongst all ratings of the engine, the takeoff one assures the most demanding conditions for the test cell performance,
thus to be taken as the one to represent the flow behaviour in this study.
1
recommendations for improvement of this test cell is available in Section 6.1, which fulfills the ultimate
section of this work. As a conclusion, the discussion highlights TAP’s test cell potential to meet the
actual and future demand of maintenance operations of the aviation market.
1.2 Literature Review
1.2.1 Typical Test Cell Structures
Test cells are designed to recreate within its enclosure the flow conditions of still air. As shown in
Fig. 1.1 any test cell comprises three main sections: the intake, the test main chamber and the exhaust
region. The intake provides a uniform air flow to the engine and absorbs part of the noise generated; the
smooth air flow it delivers is important to cool the room’s ambience before it is entrained and pumped
through the augmenter. The designs of the intake and exhaust acoustic treatments should provide a
uniform air flow at the chamber room, which is where the engine performance measurements are taken.
The former should also be aligned with the prevailing wind direction to prevent re-ingestion of the engine
combustion products emitted from the exhaust stack.
Figure 1.1: A typical test cell design (image from MDS - measured by the power of precision).
Because it determines the engine back pressure, the peripheral air flow at the chamber affects directly
the accuracy of measurements; the exhaust region configuration is also of great importance since it
contributes to lowering the noise emitted into the outside of the cell structure. The latter consists more
often of an augmenter, a diffusor section followed by a boot section and an exhaust stack, similarly as
Fig. 1.1 illustrates. The exhaust stack allows the diluted combustion streams to be expelled through the
atmosphere with an exit flue velocity low enough to minimize aerodynamic noise pollution. A range of
30 − 40 m/s is the maximum range prescribed by Jacques [23]. The modelling features of the exhaust
section of this cell are described in Section 2.5. Also, Fig. C.1 illustrates the TAP’s test cell facility,
located at the airport in Lisbon, Portugal.
2
1.2.2 Typical Challenges of Test Cells
Test cells present some challenges e.g. flow separation, sink vortices ingestion, re-ingestion of the
exhaust gases and resonance phenomena. These problems, briefly described below, compromise the
engine performance and have to be addressed when aiming to test more powerful engines.
Once the air in the chamber room is accelerated into the engine, wall-flow separation may occur if the
cell remaining air flow is not high enough to prevent low flow velocities at the cell walls. Also, since air
is drawn from all the directions nearby of the engine inlet, the components of the velocities may cancel
beneath it: the flow angular velocity and the stagnation point originated are a perfect opportunity for
vortices to form. This phenomenon depends on air and engine speed and its height above the floor. It is
known that bigger engines lead to a lower peripheral air flow available thus increasing the risk of vortex
formation2 . On the other hand, for a given test cell size large engines present problems related to 1)
excessive noise and forces applied to the cell structures due to high air velocities; 2) thrust correction
increase with increasing peripheral air flow.
Turbulence may happen just above and at a forward cell cross section of the bellmouth. For a
horizontal intake, a venetian blind vane (as suggested at [28], p. 3) directs correctly the air flow entering
the test cell; it uniformizes the flow and thus prevents the velocities near the test cell walls from being
too low.
Inlet ramps reduce the cross-sectional area of the test cell. Placed along the four sides of the chamber
(walls, ceiling and floor), this solution for vortex formation avoidance is presented by Clark and al. [11]
and is analyzed in Section 5.2.
Given the power rating of some engines, the size of the chamber may result inadequate. When large
engines are tested, secondary air intakes maintain the proper air flow needed for cooling the exhaust
section. Positioned vertically rearwards of the engine, secondary intakes need flow pattern monitoring
to ensure air is not drawn forward into the engine intake. According to Al-Alshaikh and al. [2], it is
however a good solution to test more powerful engines in already existing test cells3 .
Engine re-ingestion of the exhaust gases result from the pressure at the augmenter or at the exhaust
stack being excessively high; low peripheral airflows cause such hot gas recirculation. This is another
reason why the ratio between the cell peripheral air flow and the engine air flow needs to be within certain
limits.
Resonance may also arise whenever the frequencies of the flow match any of the test cell’s exhaust
resonant frequency. Infrasound problems due to large scale turbulence is another challenge; those may
be reduced by means of installing a ring diffusor, also known as a Koppers harp, a topic discussed in
Section 2.3.1.
1.2.3 Cell Depression

The depression in the test cell should be limited, otherwise the air flow stability and engine mea-
surements are compromised. Jacques [23] states that a difference between the ambient and the chamber
static pressure up to 150 mm H2 O (1470 Pa) is unlikely to be a problem for the engine work conditions
and the correction factor of the measurements.
The components of the cell that contribute to such depression are mainly the protective screens that
prevent hazardous objects to reach the test cell chamber. The accoustical treatments along the intake
section consist on silencer baffles and introduce additional friction to the flow. Also the wakes left in the
flow by inffluence of the guiding vanes increase the turbulence level of the flow. All these effects on the
overall cell depression were taken into account through head loss coefficients based on experimental data.
2 These topics are studied in details at the references [18] and [17].
3 The shutters on the secondary intakes may be kept closed for low power engine testings.
3
1.2.4 Engine Bypass versus Test Cell Bypass
The ratio of flow bypassing the engine relative to the flow that enters the engine is called the cell bypass
ratio, defined as CBR = ṁper /ṁeng . This is distinct from the engine bypass, which is the ratio between
the fan to core flows (herein called as the secondary and primary flows) defined as BPR = ṁs /ṁp .
The common curve shape of CBR’s changing with the augmenter diameter is shown at Fig. 1.2a.
Hastings [16] found that the smaller the augmenter tube was, the lesser the space for peripheral4 flow
and the CBR was smaller. Based on the engine primary diameter Dp , the CBR presented in Table 1.2c
of the cmf56-3C/-5B and cf6-80C2/-80E1 engines could be extrapolated for the (hydraulic) diameter of
the test cell chamber at TAP, Dcell . These were compared with the estimates from the TCC model.
Vortex formation [2] may be related to the deceleration of the peripheral air flow due to static pressure
rise along the chamber. The peripheral-to-front cell velocity ratio vper /v2 should always be higher than
0.4 − 0.5 (see Fig. 1.2b) to avoid vortex ingestion into the bellmouth.
(a) Primary exit diameter Dp and augmenter diameter (b) Bellmouth-ingested vortex formation (adapted from [2]).
Daug effect on CBR for several engine to cell diameter ra-
tios Dp /Dcell and for a fixed engine-to-augmenter distance
(adapted from [16]).
(c) CBR and cell depression of the engines considered in the study. * Length measured at TAP, against the 2.530 m from the
-5B correlation report. ** Assuming a same length between the fan’s discharge and the augmenter entrance.
Figure 1.2: Test cell bypass ratio for the engines tested at TAP.
The peripheral flow is related to the CBR and should thus be within certain limits depending on the
size and bypass of the engine being tested:
4 Also known in the literature as secondary flow. However this last term is used in this study to designate the engine
bypass air flow referred above. Together with the primary flow, the engine efflux gases emerge from the propulsion nozzle
towards the exhaust sector.
4
If the air flow speed near the walls is too low, vortices formed eventually may cause engine surge.
Low CBRs may also originate recirculation of the exhaust gases, which can cause significant flow
distortion, an increase of the cell temperature and engine surge as well. Moreover, low peripheral
flows mean a higher exhaust velocity and thus higher sound levels at the cell exhaust.
Letting the CBR be too high means the opposite: high flow velocities at the chamber’s periphery
affect the correction factor too much, resulting on less reliable correlations. The cell intake average
velocity may also increase and generated too much noise within the test chamber.
Experience has shown that bypass ratios greater than 0.75 or 0.8 are acceptable, i.e., to test an engine
of a given thrust the air flow that must be handled by the test cell should at least be 1.8 times the
engine air flow[11]. A large engine usually operates at cell bypass ratios between 1.0 and 2.0; for small
engines this can be greater than 5.0[18]. The CBR should be within the range of 0.8 and 2.0. The values
predicted for the cfm and cf6 engines considered in this study are presented at Chapter 4.
1.2.5 Potential Cores

The inflection points of the radial velocity profile shown at Fig. 1.3a characterize the interplay between
the two engine streams. Papamoschou [29] concluded that the secondary core ends at an axial location
normalized by the primary exit diameter of xs /Dp ≈ 4 − 7. With a higher momentum, the primary core
may extend up to (13.3 − 15) Dp , as illustrated in Fig. 1.3.
From Fig 1.3b II it may be seen that the secondary stream velocity is constant at nearly 2.5 Dp . As
such, the velocity profile does not change significantly from the fan’s discharge duct to the core nozzle
exit. Also, the same figure shows that a velocity ratio of vs /vp = 0.7 is typical for commercial turbofan
engines at takeoff power. The axial velocity of the jet plume is referred as v (m/s).
(a) Basic elements of mean flow in a dual-stream jet. (b) I) Potential core lengths; II) Distribution of velocities
(adapted from [30]). defining the primary and secondary shear layers for both
streams.
Figure 1.3: Details about potential cores at a vertical centerplane of the exhaust of turbofan engines with
bypass ratio BPR = 5.0.
The high velocity and thermal energy of the exhaust gases interact with the peripheral cold and low
5
in energy air flow: while mixing the latter yields some of the high energy of the former until a balance is
achieved - the fluids are then fully mixed. This produces a complex ejector pump effect, essential for the
test cell. Gullia [15] concluded that the entrainment ratio defined as ER = ṁper /ṁeng is maximized for
augmenter to engine gaps of about four engine nozzle diameters L4B−aug > 4 Dp .
Walsh and Fletcher [36] gives some guidelines for undesirable flow phenomena avoidance, such as
designing a test cell with an augmenter to engine nozzle diameter ratio of Daug /Dp = 3.0 , i.e., a 9:1
area ratio Aaug /Ap . Similarly as increasing the augmenter gap (as discussed in the Section 1.2.7) raising
this area ratio will increase the CBR up to a limit, according to a study by Choi and Soh [9]. Fig. 1.2c
summarizes the adimensional quantities defining the engines analyzed at TAP.
1.2.6 Correction Factors

Engines should perform in the test cells as indicated by the manufacturers; the guidelines given by
manufacturers allow to detect irregularities on the engine operation5 . Test cell correction factors are
applied to specific measurements (such as the engine air flow ṁeng , the engine fan speed N1 and thrust)
and allow the observed data to be comparable to other engine test runs.
The measurements at indoor engine tests are distinct to outdoors ones due to wall-presence effect.
The thrust measured inside a test cell is often smaller and this is because the pressure at the back of the
engine is lowered due to the acceleration of the peripheral air. This uneven static pressure distribution
is particularly important at the engine exhaust nozzle and it is related to the entrainment ratio. When
a different test cell is used as a reference for the measurements correction (e.g. the test cell at the
manufacturer’s site), the test run measurements are corrected by applying facility modifiers; these are
obtained from a correlation study. The hardware that equips the engine during a test run also affects
the thrust measured. The thrust measurements are affected by the peripheral velocity at the chamber’s
room; according to Rudnitski [34], depending on the test cell size, the peripheral velocity should be less
than approximately 10 m/s to meet acceptable (low) correction factors.
The measured values in a particular test cell are also standardized, to compare with the reference
testing ambient (standard day conditions) , i.e., regarding pressure, humidity and temperature in open
still air at the sea level. These reference conditions follow the ISA model (details at A.1). This study
however mainly seeks to analyze the flow characteristics and the observed engine thrust of a specific set
of engines tested at TAP.
1.2.7 Engine to Agumentor Distance

The engine to augmenter distance L4B−aug is related to the pressure around the engine nozzle, which
determines the pressure depression responsible for peripheral air flow entrainment. Ashwood [6] found
that higher peripheral airflows lead to a greater difference between the calculated thrust and the measured
thrust. The more distant the engine is from the augmenter the lower the depression around the nozzle
is. However the CBR may increase for that situation6 up to a certain limit, the one at from which the
exhaust plume spreads radially through a range that is wider than the augmenter inlet size Daug . Up to
that limit generally the flow pumping (and so the CBR) decreases with decreasing distances: the increase
of the augmenter gap gives an initial increase in entrainment and then a levelling off [2].
Results obtained by Karamanlis et al. [25] indicate that reducing this length is a solution to avoid
noise problems. Fig. 1.4 shows a 1.8% reduction on the cell air flow ṁcell by moving the engine 101
5 The test cell correction factors are adjusted according to the engine rated speed N1R, obtained from the manufacturer.
All information concerning the requirements for engine tests acceptance and concerning engine hardware (exit area of the
engine fan’s discharge and engine exhaust nozzle) are provided by the manufacturer; these were taken from the Engine
Shop Manuals [26].
6 For engine nozzle diameters which are much smaller than the augmenter diameter.
6
cm closer to the augmenter intake; the range of the cell air flow comprise those predicted for the -3C
and -5B engines. For larger turbofan engines the cell airflows are higher but the nozzle diameter is also
higher and the entrainment ratio ER reduction with increasing L4B−aug may be not as significant.
Figure 1.4: Effect of the engine-to-augmenter distance on the cell air flow for relatively “large” turbofan
engines (Adapted from [25]).
There is an optimal distance between the engine nozzle and the augmenter collector to attain the cor-
rect peripheral air flow rate. The cfm56-3C engines are installed at L4A,aug = 3.100 m of the augmenter;
cell static depression at takeoff is 350 Pa. These quantities are shown for the cfm56-5B and cf6-80C2
engines at Table 1.2c.
7
8
Chapter 2
TAP’s Test Cell
The majority of the topics studied herein are related with structures or equipments offering hydraulic
resistances to the air flow. The onedimensional model developed in this thesis considers several stations
(e.g. S2 for the inlet room station of the cell) and designating any test cell segment between two named
stations as sections (as S∞ − S2 for the intake section). Fig. 2.1 schematizes and identifies these sections.
Figure 2.1: Test cell sections and stations (adapted from [24]).
The TAP’s test cell is a test facility having a cross section of approximately H2 = 8.50 m x W2 = 9.75
m. The facility uses a down-draft inlet with turning vanes. The inlet passes through a bird screen,
noise reduction splitters and a flow conditioning screen before entering the cell testing section. Exhaust
gases pass rearward through conventional cylindrical augmenter with a wedge diffusor section, upwards
through a folded vertical exhaust stack.
9
2.1 TAP’s Historical Review
The thrust stand at TAP was designed to withstand engines performing test runs up to a hundred
thousand lbf thrust, i.e., nearly 445 kN. The engine specifications shown at Table 2.2 could induce one
to focus mainly on thrust ranges. This work shows how airflows can be even more important in a test
cell.
Formerly this cell was used to test the P&W JT3 and JT8 engine families. This means that reasonable
thrusts were observed at relatively low mass flow rates (80 kN, 204 kg/s). When high-bypass engines
arrived at TAP test cell facility, the inlet grating (similar to the current one at station S2 ) had to be
enlarged and a basket tube, such as Fig. 2.2a shows, was installed to increase the static pressure and
allow the exhaust gases to make the 180 turns of the exhaust stack, at S7C and S7D . The velocities
within the test chamber room were higher than recommended for good test cell design; it was recognized
that the challenge was on controlling the total air flow at the test cell[28]. With a blast basket 1 (see
Fig. 2.2a) this solution worked for the engines that Table D.2 shows. The referred document condenses
usefull information about the velocities reached at the time the P&W JT9D engines started to be tested
at TAP, at 1972; it enabled to conclude that the cell velocities were close to the maximum it could
undertake as designed in 19692 . The fact is that massive alternating flow separations at the diffusor gave
rise for the resonance experienced at the exhaust.
Table D.1 summarizes the cell upgrade evolution since 1969. After many years working under those
new conditions, the test cell had to be improved for bigger high-bypass engines as the cf6-80C2 : the
basket installed could not withstand the higher mass flow rates, which couldn’t expand and decelerate
properly anymore. In 1989 the cell was again redesigned. A system that could avoid flow separation
became a need one more time and that could be granted by reducing the diffusor area ratio. This is why
an inner fariring at the diffusor’s inlet (from S6 to S7A , forming a wedged mechanism) was set within it
in replace of the basket. Likewise, it prevents the exhaust diffusor velocity from rising too much. Despite
introducing extra friction losses, this was the best solution to avoid resonance meanwhile keeping the
diffusor unmodified.
A complex set of grids was also installed aft of the wedge. Likewise the vertical supports that work
as dividing walls (at S7B up to S7C ), both shown at Fig. 3.2, these are a way to moreover assure that the
most uniform velocity profile is reached downstream of the diffusor. They guide the flow analogously to
the perforated basket did, and moreover reduce the losses associated to the sudden expansion that the
figure mentioned shows as well. Together with the wedged-diffusor, by introducing significant pressure
losses, the grids lessen the pressure recovery from S5 to S7C - thus reducing the cell air flow - and
then increase the risk of wall-flow separation at the chamber. It should be kept in mind that without
a wedge, guide vanes or similar mechanisms as the one suggested at Chapter 6.1, resonance at this cell
will be always present due to the diffusor divergence angle being too high. Whatever suggestions of cell
improvement may be, they must undergo means of both peripheral air flow and diffusor flow control.
Fig. 2.2a shows the structures that were replaced at the time that the actual commercial aviation engines
considered in Table D.1 (left column) came to test cells. The figure mentioned also schematizes some
sections of the current test cell3 operating as such since 1990, when the redevelopment works ended.
1 The perforated section at the tip of the steel tube extending from the augmenter through the diffusor. Its rear was
buffered with a steel sheet inside and allowed the flow to be uniformized. More details at [2], section 2.11.
2 Peak velocities of about 17 m/s were measured for relatively low power settings of the JT9D engine and therefore the
flow velocities would be even higher when at the takeoff power rate.
3 Study developed by Ohio State University with the support of GE Aircraft Engines. Images taken from the resulting
document.
10
(a) Test cell initial geometry, when inaugurated: inlet system
quarter view and augmenter extension and basket.
(b) From left to right: intake and pressure drop screen; thrust frame engine installation; augmenter, diffusor and exhaust system;
diffusor wedge.
Figure 2.2: Test cell geometry (From the 1:12 Scale Model Study [37], pp. 6-23).
In order to validate the TCC model some pressure measurements were taken at the chamber’s entrance
section and at the diffusor exit station. This enabled to quantify the cell air flow. The design, installa-
tion, data collection and the post-treatment procedure applied to the results obtained are discussed in
Section A.4.
2.2 Head Loss Coefficients: Intake Section
The topics covered in this section go further than theoretical considerations and focus on presenting
basic experimental data that fit as criteria for estimating head loss coefficients. Results obtained by
Idelchik [22] were found a reliable source for such estimates; those were taken from the reference studies
pointed out or collected by the author itself. The majority of the head loss coefficients are computed
at the TCC code because depend (slightly) on the Reynolds number. Table. 3.1c presents some details
about each head loss coefficient.
The following sections present the information found keeping in mind the sequency of stations detailed
in the Fig. 2.3.
2.2.1 Inlet Head Losses
Head losses due to the intake acoustic devices, at section S∞ − S1B can be estimated from [22].
These accoustic baffles align the flow by splitting it into 23 parallel streams as shown at Fig. C.1.
This TAP’s intake structure has rounded inlet edges and therefore flow separation is not severe.The
head loss coefficient kinlet was estimated from Diagram 8-3 of [22] based on a Reynolds number of
Redh = v×dν
h
∼ 105 , being v the upstream flow velocity, dh = 4 πf00 = 0.767 m the hydraulic diameter
as defined by Idelchik and f0 = 3.66 m2 and π0 = 19.12 m the area and the perimeter of one channel
11
Figure 2.3: Grid details. a) screens at stations SB and S2 ; b) test cell chamber; c) screens at station S7B
(left: side grids at the periphery of the exhaust; right: grid middle channel).
respectively4 . This coefficient is based on the inlet velocity, which corresponds to station SA . Expressed
2
as a function of the inlet chamber’s velocity v2 each coefficient ki is affected by an area ratio (A2 /Ai ) .
2.2.2 Guiding Vane Losses

The head loss coefficient presented corresponds to a rectangular cross section elbow with thin guide
vanes installed (Diagram 6-26, p. 396 of [22]); the flow is turbulent5 (ReDh > 2 × 105 ) and the inner
corner is rounded in such a way that the relative radius is LrA = 8.530
0.47
= 0.055 (a smooth turn). An area
ratio of A B
= 1.08 ≃ 1.00, a stagger angle of α = 45°, a chord of c = 1.4 m and a total vane’s length of
p AA
2
S = L A + HB 2 = 12.15 m define geometrically the guiding vanes at station S ; H is the height of the
1 B
passage after the 90° turn, at SB .
Referring to Diagram 6-26 case 2), the head loss coefficient is obtained as k vanes = kloc × kRe + kf r =
0.385, under the conditions stated in Section B.
It should be remarked that the chord of the thin vanes is distinct from the reference value suggested
by Idelchik; in any case the head loss coefficient is probably within the range of values estimated above,
this is, it should be between 0.2 and 0.4.
2.2.3 Intake Grid Losses

The resistance coefficients associated to screens barriers available in [22] are valid for the range of
Reynolds number Re expected at the takeoff rates. Some grid structures of TAP are peculiar and some
approximations were done to model them.
4 Thishead loss coefficient depends on the free-area f¯ = Σf
A
0
≈ 0.57, being Σf0 the sum of the orifice areas and A = 84.5
m 2 the cross sectional area of the intake, which is equal to the one at station SA . The same stands for the screen at S2 .
The referred coefficient is also dependent on the friction coefficient λ = 0.013, obtained as suggested by the author and the
same way as described below for the guiding vane losses.
5 The hydraulic diameter comes for this section as D = 4 R = 2 (L×W )inlet = 8.834 m.
h h (L+W )
inlet
12
Idelchik [22] refers that the installation of screens downstream of guiding vanes is advisable, as occurs
at TAP at sections SC−D and SH−2 . These bird shields condition the income flow. Fig 2.4, left side,
refers to circular metal wire screens, used as a reference for the head loss coefficients estimates.
The main correlations associated with these coefficients are presented at Section 2.5. Head loss
coefficients of screens could also be obtained by taking screen wires as infinite cylinders as a first estimate.
Fig. 2.3 represents the main sections of the test cell and points out all stations were the flow barriers,
such as the one shown in Fig. 2.4.
Figure 2.4: Left: barriers distributed over the cross channels: screens made of circular metal wire. ([22],
Diagrams 8-6). Right: drag coefficient CD for circular cylinders with infinite length and smooth surface,
function of the Reynolds number based on the hydraulic diameter, Re Dh (From [3]).
The screens located before stations S1D and S2 are framed by steel grids that hold them. Our
MATLAB code models these screens as made of circular stainless steel wire. It follows that Atot,or =
j × k × Aor is the total free area through that section. The code assumes the screens are a set of wires
approximated as infinite cylinders perpendicular to the flow. The total area of the orifices is obtained as
a function of the wire spacing and the wire diameter dwire .
Coefficients associated to head losses due to entry and exit of orifices are presented at the Diagram 8-6
at p. 522 of [22] which gives estimates valid for Reynolds numbers6 of Re dB ≥ 103 and a coefficient of
A
screen clear-area to upstream-area f¯scr,B = tot,or
AB = 0.78. The upstream areas were obtained subtracting
the frame areas that support the screens; a head loss coefficient of kscr,B = 0.364 was obtained. As referred
above the screen wire at station SC has a non-circular cross section (it is a diamond wire mesh raised
expanded metal) and the resistance may be slightly higher.
The second screen at the intake of the test cell, at section S1H − S2 , is a flow-conditioning screen7
similar to the previous one but made of perfectly circular metal wire; with a smaller wire diameter and
smaller openings (the wire mesh is more refined), giving an area ratio of f¯scr,2 = 0.45 and a resistance
coefficient of kscr,2 = 2.26, much higher than for the previous screen (as expected, refined meshes exhibit
higher head loss coefficients).
The coefficients were obtained from the Diagram 8-6 of [22] and are based on the upstream flow
velocity.
6 The velocity is expected to be vB ≥ 6.0 m/s at SB and the velocity at the orifice vor comes from the conservation
equation: vor = vB × A AB (m/s), with an area ratio of A AB > 1; thus the Reynolds number is expected to be
tot,or tot,or
dwire ×vor
always higher than Red = νB
∼ 103 at takeoff rates for any engine at this cell. A typical kinematic viscosity of
−5
νB@T ref ≈ 1.51 × 10 m2 s−1 for dry air at the reference ambient temperature, 288K was taken from [35]; the results
shown in this thesis, however, are based on the viscosity values obtained from the TCC model.
7 Diagram 8-6 of [22] was used, valid for the range 50 <Re ∼ 102 < 103 verified at the current station.
d
13
2.2.4 Intake Silencer Assembly
The screen at the station S1D is followed by a flow channel with staggered walls as shown in Fig. 2.5,
right side. The cross section of the walls are assumed as cylindrical. Tube banks are typical in heat-
exchangers and the first estimate obtained for the pressure loss coefficient at this section is based on such
correlations: the pressure drop is expressed8 as:
2 2 2

ρvmax ρvmax ρvmax
∆p = NL χ f = f NL χ = kbundle , (2.1)
2 2 2
where NL = 2 is the number of rows (counted vertically), f is the friction factor and χ a factor that
extends the correlation results to staggered arrangements distinct from the equilateral triangle, as plotted
in the inset image of Fig. 2.5a). The dimensionless longitudinal and transverse pitches are:

P = S /D = 4.96
L L
(2.2)
P = S /D = 2.05 ,
T T
where the longitudinal and transverse lengths are SL = 2.016 m and ST = 0.832 m; D is the wall
diameter. The Reynolds number is based on the maximum flow velocity vmax 9 and it is expected to be
greater than 105 . For a pitch ratio of PT /PL = 0.413 and Remax > 105 , it follows χ = 1.45, f = 0.19 and
the coefficient kbundle may be thus attained with the MATLAB code.
Another estimate may be based on10 a loose equilateral triangle array with a ratio of SD /D = 1.46,
corresponding to a resistance coefficient of kbundle = 1.934 in laminar flow; for turbulent flows, kbundle =
0.64; however, these values come specifically for a ratio of SD /D = 1.00 (equilateral triangle) and the
previous method is the one implemented in the TCC model, because it allows to correct for the real
staggered arrangement. On the other hand, since the cross section of these walls are elliptic and are
followed by a fairing structure, the wakes at the wall endtips are smaller and the drag coefficient due to
flow separation is lower than the one from the correlation experiments. Therefore, the head loss coefficient
kbundle is being obtained in excess.
Figure 2.5: Staggered tube bundle arrangement. Left: friction factor f and correction factor χ for
computation of the pressure drop. Right: flow conditions and geometric quantities for staggered tubes
([7], pp. 437-443).
8 At [7], page 450. More details at pp. 437-443.
9 This velocity will occur through the diagonal direction if 2(SD − D) = 0.403 < (sT − D) = 0.416, which is true. This
leads to vmax = 2(SST−D) v1E > 25 m/s, function of the velocity at S1E . This equation, as many others, is implemented
D
in the MATLAB code. If v1E ∼ 10 m/s, then vmax ∼ 20 m/s and ReD ∼ 5 × 105 .
10 From Diagam 2-9 of [22], and pp. 91 and 125.
14
2.3 The Augmenter
The hot exhaust gases entrain the peripheral flow that surrounds the engine; both streams enter the
augmenter which then acts as an ejector pump. This peripheral air flow should be high enough: with
the exhaust temperatures ranging between 400 ° C and 700 ° C [23], an efficient dilution is necessary. The
velocity profile uniformization also enables the noise attenuation at the chamber and in the exhaust stack.
On the other hand, the peripheral flow must not be too high, to avoid a large static pressure gradient
between the bellmouth and the engine’s exhaust nozzle which would result on too large correction factors.
2.3.1 Ring diffuser

High-speed jet plumes contain large quantities of energy and constitute great sources of noise. The
harp structure, presented in Fig. 2.6a, has been installed to reduce low frequency noise by means of
increasing the flow uniformization. The smaller scale turbulence resulting from the harp presence solves
the infrasound problem. Fig. 2.6a (right side) represents a simplified arrangement of the harp structure as
a two-dimensional set of rings, with a total cross sectional area of 2.16 m 2 (see Appendix A.3 for details).
The scheme referred is merely representative; the mathematical approach described in Section 3.6 assumes
that the rings are spaced along the x direction; in fact, the ring blockage effect (which is lower than if
the harp was modeled bi-dimensional) is not being considered. However, the peripheral air flow is the
first to be uniformized, redirecting some of the air flow to the remaining augmenter ring-free area. The
velocities on the remaining rings are thus higher than those considered in the TCC model; at the time
the flow reaches the harp rings, the model considers the flow profiles as being just the same as in station
S5 , which is only an approximation when modelling the harp effect on the flow.
The harp uniformizes the velocity and temperature profiles, but it also affects the pressure at the
augmenter’s inlet and alters the aerodynamics of the system and in particular of the augmenter tube [27].
The effects that this structure exerts over the flow from S5 to S5B is presented at Section 3.5.3. The
harp support plate shown at Fig. 2.6b induces a large pressure drop and highly reduces the momentum
of the primary stream. A head loss coefficient of CD = 1.18 can be obtained from [19] (figure 28, p. 3-16)
for flat rectilinear plates.
2.4 Diffusion Section

2.4.1 Equivalent Divergence Angle
Transition sections are a way of connecting sections within a system and are of two different kinds:
their variable cross section along the flow may be shaped constant (these can be then of circular or of
rectangular cross sections) or may vary (both circular and rectangular cross sections are present). As
Fig. 6.1b right side shows, the diffusor section S6 − S7A at TAP consists on a transtition piece connecting
an augmenter tube to a rectangular room downstream (boot section).
For the case of the rectangular cross section presenting a small aspect ratio11 , the head loss coefficients
may be computed from the literature provided the divergence angle is equivalent to the real diffusor with
a transitional cross-section: αeq,rect = 16.2°. This came directly from12
11 Which is indeed veritied, since H7A /W7A ≃ 0.4 < 2.0.
12 [22], page 275, para. 15.
15
(a) Noise reducer harp picture (images taken from [24]), modeled with 3D CAD Design Software Solidworks. Right side: two -
dimensional approach.
(b) Harp support plate.
Figure 2.6: TAP noise reduction harp structure in the test cell augmenter.
p
2 H7A W7A /π − Dinlet
tan (αeq,rect /2) = . (2.3)
2 Ldif
Obtaining equivalent diffusors to the one installed at TAP is meaningful when the knowledge of
pressure losses due to their flow resistance becomes a need. This subject is again considered at Section 5.2,
when it comes to evaluate other solutions in order to optimize the test cell performance.
2.4.2 Diffusor Performance

Losses associated with diffusors were evaluated with correlations presented in White and Idelchik.
The first author offers quick calculations based on simple charts (Figures 2.9 and 2.7) and supports that
a fair criteria to predict the performance of a diffusor relies on the area ratio ndif = Aexit /Ainlet , called
the (static) pressure-recover (frictionless) coefficient
Cp = 1 − n−2
dif . (2.4)
It represents the amount of pressure recovered attained thanks to the divergence angle α characterizing
it. Too high divergence angles enhance the flow turbulence whilst the boundary layer separates from the
walls and vortices are formed ([22], Chapter 5, para. 2). The main parameters of a diffusor are connected
by equation 2.5, being Dinlet the inlet diameter.
√
Ldif  ndif −1 (conical)
= 2 tan /2
α
(2.5)
Dinlet  ndif −1 (plane).
2 tan α/2
L
dif L
Considering the diffusor at TAP, with a length to width ratio of Winlet ≡ Wdif6
≈ 3.50 that diffusor
is close to the transitory stall region referred in Fig. 2.7, scheme c), applicable to flat-walled diffusors; it
16
nearly coincides with the limiting line aa. This may not allow a sufficient static pressure rise of the flow
along the diffusor because it induces oscillating flow separations. With a stall pattern that jumps from one
wall to the other, with poor global performance13 (as Fig. 2.9 b) shows), the main issue of this alternating
separations is the low-frequency and high-intensity pressure variations. They produce resonance and may
induce engine surge. The original diffusor at TAP had all characteristics of separation-prone diffusors
(represented at Fig. 6.1b, left side).
Figure 2.7: a) flat-diffusor stability map. (adapted from Fig. 6.26 of [38], page 382); b) diffusor with a
transitional cross-section, at TAP ([22], diagram 5-28).
Actually, there are a set of different ways of promoting better flow behaviour within short diffusors with
large divergence angles, including (see Fig. 6.1a) splitters, guiding vanes and inner fairings. With means
of preventing flow separation or attenuating vortex generation, they decrease the head loss and provide
a more uniform velocity distribution14 . Inner fairings also introduce pressure losses; Idelchik estimates
them for particular cases, not applicable to TAP. For this reason the head loss coefficient attained for
the diffusor considers an equivalent angle αeq computed from the area ratio n that is changed due to
installing the said structure.
Originally the performance parameter for frictionless flow would be Cp = 1 − 3.54−2 = 0.92 (al-
most full pressure recovery), but despite this coefficient is defined by excess15 it does not take into
account separation phenomena. Indeed, the eq. 5-13 from [22] defines the diffusor efficiency as n =
−1/2
1 − kdif − cpdif . It follows ndif,ideal = 1.2 (with Cp,dif = 100%) versus the effective area ratio
ndif,T AP = 2.0 (corresponding to Cp,dif = 45%) for a uniform flow velocity distribution at the inlet and
exit diffusor sections.
13 [38],
page 383.
14 [22],
pp. 259-265, para. 60-72. Comments about flow splitters: page 262, para. 66. Moreover, para 81-82 at pp. 267-268
and diagram 5.12 refer to inner fairings.
15 Comparing to the reality studied with experiments, [38], p. 382.
17
If the diffusor area is restricted, the pressure-recover coefficient Cp comes smaller but flow separation
is avoided. A length-to-diameter ratio of Ldif,real /Dinlet = 10.370
3.353 = 3.09 and an area ratio of ndif,w =
A7A
A6 = 2.03 describe the current diffusor with wedge. This amounts to reducing the divergence angle
and the area ratio (for the same inlet geometry and the same length Ldif ) closer to the values at which
best performances are reached (Cpmax ). Accordingly, equation 2.3 gives an angle of αeq,rect = 7.9° and
equation 2.4 indicates a theoretical frictionless pressure-recover coefficient Cp = 0.76.
According to White the diffusor with wedge operates at no stall conditions and behaves as a separation-
free diffusor according to Idelchik. The reason why such mechanism makes part of the diffuse is thus
mathematically reasoned. Table B.1 summarizes all the results obtained for this diffusor for each different
points of view studied.
2.4.3 Flow Separation

The diffusor originally placed at TAP had an area ratio of ndif,nw = 3.54, and so the limit above which
separation could occur is αsep,nw ≈ 9°(see Fig. 6.1b). With an equivalent divergence angle of αeq,rect ≈ 16°
such diffusor was inappropriate, which is in agreement with the conclusions presented above. Concerning
flow separation at the wedged diffusor, separation occurs around αsep,w ≈ 22, for ndif,w = 2.03. With
αeq,rect ≈ 8°, this means that the flow does not separate at any section of the diffusor given its current
configuration, but it definitely would if the wedge was not installed. Summarizing,

 16.2° , αsep,nw ≥ 9° no wedge
αeq,rect = (2.6)
 7.9° , αsep,w ≥ 22° wedge.
2.4.4 Flow Patterns

Diffusors can be preceded by a straight segment of constant cross section known as an insert, such
as the augmenter section at TAP. This allows the velocity profiles of the flows to uniformize; the flow
uniformization depends on the augmenter size. For turbulent flow and for a single stream, it could be
expected a velocity ratio of vvmax
avg
= 1.05 ≈ 1 at the diffusor’s entrance16 . With no flow separation, the
insert helps to recover the static pressure even if it somewhat adds a flow head loss due to friction. The
shorter the augmenter is, the lesser the boundary layer is - which delays an eventual flow separation at
17
the diffusor - but the pressure recovery along the augmenter is also lessened .
Figure 2.8: Velocity distribution at the inlet of diffusors, with a straight tube upstream ([22], p. 276).
Fig. 2.7 a) refers to diffusors with different geometries. When compared with conical diffusors, the ones
with non-circular cross sections present an increase of the flow head loss due to the additional influence
16 For a relative length Laug /Dinlet = 3.66 and Re= 2 × 106 , from [22]. See Fig. 2.8.
17 Such uniformization decreases sharply for α > 30 ([22], para. 23 of Chapter 5).
18
of the corners 18 . Section 2.5 refers to the computation of the head loss coefficients and makes clear that
conical diffusors perform better than rectangular ones: they need lowerdivergence angles which together
with smoother cross sections help to avoid flow separation. The flow conditions upstream of the diffusor
affect the coeffcients mentioned, as Fig. 2.9 illustrates.
Although separation may occur locally, as a rule19 all flows separate for divergence angles up to 40,
due to diffusor or inlet velocity profile asymmetries. Idelchik [22] concludes that velocity profiles develop
as asymmetric for those situations and that local separations at one side of the diffusor impede flow
separation on the opposite wall. If both the diffusor and the velocity profile are symmetric instead,
separation occurs alternately at one side and the other side of the diffusor20 . The velocity profile at the
inlet of the current diffusor at TAP was this way assumed to be symmetric, as shown in Fig. 2.9 a).
In practice, the velocity profile at the inlet of the current diffusor at TAP is not uniform because
there are three potential cores developing from the engine’s rear and being blurred 21 along the augmenter
section. That way, the velocities profile will tend to be symmetric and convex shaped22 . Even though
turbulence is high and promotes flow streams mixing, the velocity profile v6 (r) could be considered as
being uniform at the inlet of the diffusor only as a first approximation.
Figure 2.9: Left - a) ideal pattern with good performance; b) pattern with boundary-layer separation
and resultant poor performance (Fig. 6.27 from [38]). Right - Flow losses in a conical expansion region
(source: [38], Fig. 6.23, p. 372).
2.4.5 Sudden Expansion

For reasons of limited overall size of a system, a sudden expansion or aperture may follow a diffusor.
This is also called a stepped diffusor; Fig. 3.2 schematizes it. The head loss of a diffusor is always lower
than that of a sudden expansion itself provided αdif ≤ 40−50 23 and thus large divergence angle diffusors
are preferable whenever uniform velocity profiles are attainable downstream. If the latter is not expected,
dividing walls or perforated plates e.g. grids or screens should be installed24 , as they are at TAP. The
18 However, an upstream insert lessens that rise in the resistance coefficient ([22], Chapter 5, para. 28).
19 [22], Chapter 5, para. 8
20 Idem, para. 9
21 The term blurred is used by Idelchik [22] and will be recalled at Section 3.5.2, where the velocity profiles are presented.
22 [22] , Chapter 5, para. 24
23 Large divergence angle diffusors are characterized by α > 50. Higher values mean an increase the area ratio n
dif if the
length Ldif is kept constant. [22], Chapter 5, para. 42.
24 [22], Chapter 5, paragraphs 42-43
19
head loss of a stepped-diffusor is greatly lower than that of simple diffusors25 if its divergence angle obeys
to the relation of equation 2.7. The flow resistance associated to sudden expansions are presented at the
following section.
 √
tan αdif ≤ ndif −1
2 2 Ldif /Dinlet (circular)
(2.7)
tan − αdif ≤ ndif −1
(plane).
2 2 Ldif /Winlet
2.5 Head Loss Coefficients: Exhaust Section

The head loss from S6 to the S7A station are estimated as discussed below. The velocity at S7A can
be obtained from mass conservation. If ρ̄6 is kept constant, the challenge consists on obtaining a fair
estimate for the global head loss coefficient between S7A and S8 ; that should consider a sudden expansion
downstream of the diffusor section (S7A to S7B ). There is also a pressure drop along the diffusor (S6 to
S7A ) which is computed to estimate the pressure p′6 . Finally, pressure losses due to flow deflection at the
baffles of the exhaust stack are also taken into account; these coefficients are normalized by the dynamic
pressure based on the upstream flow velocity.
The fact that the diffusor begins conical and finishes rectangular is noticed as relevant since the
equivalent diffusor obtained by Idelchik [22] has its flow characteristics changed by such geometry. The
head loss coefficients for the exhaust room’s grids were obtained similarly as for the screens of the test
cell’s entrance section. Again, Idelchik [22] presented the basis to estimate these coefficients.
The Diffusor
The head loss coefficients of diffusors depend on the divergence angle α, on the area ratio ndif = AA7A
6
and on the length Ldif ; the inlet Reynolds and Mach numbers (Re and Ma) are parameters related to
the flow regime and the displacement thickness δ ∗ - which is connected to the ratio Laug /Daug - stand
for the inlet flow conditions26 .
The diffusor placed in the test cell does not remain conical along the x direction: its cross section
expands along the z plane only - hence also known as a plane diffusor - but its shape also changes from
circular to rectangular (see Fig. 2.7 b)). An equivalent diffusor to this one may be obtained by relating
the area ratio to the diffusor length. Once achieved α of diffusors equivalent to the one installed at TAP,
it is possible to analyze the resistance coefficients obtained for the one that best represents that diffusor,
at the test cell section S5 − S7B .
Both White [38] and Idelchik [22] present results of head loss coefficients for a conical expansion.
Fig. 2.9 right side,
summarizes
some results obtained experimentally for conical
diffusors with a length-to-
Ldif Ldif
radius ratio of D1 ≃ 3.2, which is close to the one at TAP, D1 ≃ 3.1. The resistance
W HIT E T AP
coefficient may be thus directly obtained, even when considering the effect of an insert27 . A conical
equivalent diffusor could be obtained with an angle of aperture α defined as
25 [22],p. 266, para. 76.
26 [22], pp. 240 and 243, paragraphs 1 and 11. The ratio mentioned stands for the hypothesis of a straight segment - an
insert - preceding the diffusor.
27 The flow is turbulent and thus the inlet boundary layer should be thinner than if it was laminar; however, when placed
ahead of separation-prone diffusors, inserts agitate the flow and for L/D ≥ 10 they thicken even more the boundary layer,
anticipating separation; if the insert has a sufficient length, that should correspond to the fully developed flow of the last
figure mentioned. Furthermore, that raises the flow resistance and this effect is more intense for conical diffusors than for
rectangular ones. In the absence of such inserts, the opposite is verified when comparing cases iii, v and vii and cases iv, vi
and viii of table B.1.
20
2
2 Ldif
A2 = πR22 = π × (tan α/2 × Ldif + R1 ) = π × tan α/2 × + 1 × R12 , (2.8)
R1
which is the way that eq. 2.5 is deduced for the conical diffusor. Subscripts 1 and 2 refer to the inlet
and exit sections respectively. Re-writing the previous equation and considering that Ldif /R1 ≃ 6.2, it
follows αconic,w = 7.9°, attained for a diffusor with the wedge inside. Without the wedge the angle would
come as αconic,nw = 16.2. Based on the inlet velocity of the flow v6 , this last aperture angle lead to the
flow resistance coefficient kconic,nw shown at Table B.1. It should be noted that White [38] does not take
into account the flow regime (Reynolds number Re) and the values obtained refer to a thin boundary-
layer; Idelchik [22] does consider the flow regime and also allows to correct for the nonuniformity of the
flow due to an insert upstream. This way, the coefficients obtained from this author were found more
reliable.
The last equation presented allows to get an equivalent divergence angle also for a plane diffusor,
which came as being αplane,nw = 39.9° and αplane,w = 16.8°28 . The resistance coefficient predicted
is higher than for the conical one when there is no insert preceding it and vice-versa; as was already
expected. It is interesting to note that both the diffusors equivalent to the one at TAP - conical and
plane - are conducive to flow separation, for the reasons stated this far29 .
Finally, the eq. 2.3 gives an equivalent divergence angle of αrect,nw = 16.2° or αrect,w = 7.9° for a
rectangular equivalent diffusor for a diffusor without and with wedge, respectively. The resulting
head loss coefficient 30 is also presented at Table B.1. Because the height of the diffusor at S7A is equal
to the the augmenter’s diameter, the equivalent divergence angle is the same as for the conical diffusor.
The rectangular diffusor, without wedge, is close to the limit characterizing separation-prone diffusors;
these generally present divergence angles greater than 14°. That diffusor is also close to the transitory
stall region presented by White [38] in Fig. 2.7, which agrees with the conclusions already discussed.
This is the diffusor selected to represent the one at TAP, with a head loss coefficient of kdif,nw = 0.30,
leaving the flow effects associated to the stepped wall and downstream grids to be evaluated separately.
Further considerations concerning equivalent diffusors and the corresponding head loss coefficients are
summarized in Section B.
Sudden Expansion
White [38] presented the Fig. B.1a as an indication to estimate a head loss coefficient for the sudden
expansion (SE ) at S7B : based on the velocity head v7A it was obtained KSE = 0.621 31 ; considering
friction and the Reynolds number Re, it was evaluated as being kSE ≃ 0.97. Section B clarifies the way to
estimate these values; Table 2.1 presents the resistance coefficients for the sudden expansion depending on
the upstream flow velocity distribution. The TCC model considers an exponential velocity distribution.
28 From eq. 2.5 and letting the cross section be defined as Aaug = H6 × W6 = H62 ⇒ H6 = 2.972 m.
29 Also,in terms of the divergence angle, the placement of the wedge approaches the plane diffusor to that of a conical
one without wedge: αplane,w = 16.8° ∼ αconic,nw = 16.2°. But in terms of resistance, the situation of the former is greatly
improved. Both the diffusors are close to the original real one (transitional diffusor without wedge).
30 Page 290, Diagram 5-4 of [22].
31 Taking the upstream and downstream diameters d and D as the effective passage lengths of the flow bypassing the
W −W
wedge at S7A and at S7B respectively, the diameter ratio comes as D d
= 7A W wedge = 8.544−3.662
15.617
= 0.31.
7B
21
Table 2.1: Different approaches for the sudden expansion resistance cofficient estimate for the 7.9 equiv-
alent diffusor divergence angle. The velocity distribution refers to the upstream section.
From Table 2.1 we may see that installing baffles could be a measure to improve the test cell diffusor’s
performance. Their effect is noticeable: almost 40% of relative difference; this estimate holds for a uniform
velocity distribution but a similar conclusion follows for the remaining types of the velocity profile. If to
be installed, the optimal baffles arrangement remains the only concern.
Exhaust Stack Grid
Similarly to the intake section, the grid structure installed at station S7B at the back of the diffusor
consists on a set of thickened lath grids32 . In fact, properly (uniformly) distibuted screens along the cross
section downstream of a diffusor regulate the flow not only along the channels following it but also along
the diffusor itself; by preventing flow separation this allows for the head loss to become lesser than if no
screens were installed at all.
The mesh-screen placed at station S7B of the cell is shown at Fig. 3.2 and is made of two types of
grids: two ones (grids a and b) made of thick laths fixing a fine-mesh screen that direct the income flow
through the peripheries of the exhaust stack, and a single grid (set as being of type d ) in the middle of
the just mentioned ones. Also, parallel to the flow, two more grids extend from the diffusor exit up to
the grids b located downstream. The latter are equal to grids a-b without screenings and were named as
grids c. Fig. 3.2 clarifies this section of the cell.
The head loss coefficients of the screen structures were estimated from Diagram 8-6 of [22] for the
corresponding Reynolds numbers defined as Re d = dwire ×vor
ν7B . The TCC model takes into account the
range of the Reynolds number to which empirical correlations from Idelchik [22] are applicable by setting
the adequated corrections suggested by the author. The TCC model sets a head loss coefficient correction
factor based on the Reynolds number at section S7B , which in turn is dependent - besides the kinematic
viscosity ν 7B ≃ ν 5 - on a characteristic dimension and on the screen orifices local flow velocities, v or . 33 .
The TCC model computes the head loss coefficient from the coefficient f¯ obtained with the global orifice
areas of each grid. Such areas are the screen clear areas through which the exhaust mixture may flow,
as detailed in Section B.
This way, Idelchik [22] gives an expression to obtain such coefficient, that stands for circular metal
32 This grid was made visible by sketching over the drawings kept at TAP prior to the last construction works, at 1989;
see Fig. 3.2.

33 This velocity is based on the expected upstream flow velocity v
7B , obtained by mass flow conservation: v or =
v7B AA
7B
m/s.
or
22
screens such as the ones of grids a, b and d :
" 2 #
1
kscr = kRe × 1.3 (1 − f¯) + ¯ . (2.9)
f −1
As for the two grids referred as c, the head loss coefficient was considered based on Diagram 8-3 of [22].
When considering the diffusor with a grid installed downstream, a slight change was verified relative
to a simple diffusor head loss coefficient (from [22], p. 259, Chapter 5, para. 59).
Flow Deflection Losses

A loss coefficient of kturn,7C = 14.5 and kturn,exit = 7.0 could be obtained respectively for the first 180°
turn and for the last flow deflections through the remaining stack (S7D to S8 ) that Fig. 2.10 illustrates;
the reference velocities are v7B ′ and v7C , respectively. For both coefficients, the width of each turn (4.9
m) was assumed to be equal to the height of the baffles (Wturn = 4.9 m≈ a0 ). For the last turn, the
distance between the baffle tips to the wall (4.28 m) is close to the height of the baffles a0 = 3.17 m and
thus the diagrams shown at Fig. 2.10 b) and Fig. 2.10 c) may be used.
Figure 2.10: a) exhaust stack section. b) exit short 180° turn (head loss coefficient taken for a ratio
of bx /a0 = 3.17/5.28 = 0.6). c) Hood with three-sided exit (with an exit lenght of b = 4.8 and thus a
ratio of b/a0 = 4.8/3.17 = 1.5, which can be extrapolated from the upper limit of b/a0 = 1.2 from the
author’s data). (Source: [22], diagram 9-31 a) and diagram 9-31 b). All coefficients apply for the exit
configuration).
2.6 Technical Survey
2.6.1 Selected Engines

There is some variety on the engine types and on the thrust ratings of the engines that undergo bench
tests at TAP’s test cell facility. The engines targeted for this study are presented as shown in Table 2.2.
Aircraft engines are classified according to their manufacturers (CFM International , General Electric
Aviation, Pratt Whitney, Rolls-Royce) and according to the families they belong. Among the cfm56
engines, the -3C and -5B are present in this study; the -C2 and -E1 engines from the cf6 family are
also analyzed.
23
Table 2.2: Engines analyzed in the study. Data for takeoff at sea level conditions were obtained from
http://www.jet-engine.net/civtfspec.html. (images from [1] and [31])
The engines still fall into rating or power setting categories depending on the thrust delivered and thus
on the aircraft type they are installed. This study comprises the cfm56-3C1 , cfm56-5B1/3, cfm56-5B4/P,
cfm56-5B5/P , cfm56-5B6/P and cf6-80C2A2 engines, all tested at TAP. Finally, the flow conditions
within the test cell are estimated for the cf6-80E1A3 engine, one of the engines currently powering the
long-haul A330 fleet.
2.6.2 Engine Bleeding and Hardware

The engine bypass stream allows the cooling of the engine itself at the same time that small amounts
of fluid are picked along intemediate stages for e.g. cooling purposes. When running in test cells, the
vanes that control the engine bleeds at takeoff are positioned in such a way such bleeds are nearly null;
all engine testings considered in this study conform with this usual procedure34 . Therefore the air mass
flow rate is constant through the engine and only the fuel mass flow rate must be taken into account for
momentum balances. See Fig. 2.11a for cfm engine-type detailed sections.
All data about engine geometry and measurements from test runs were collected to compute the
engine exhaust fluid and thermodynamic properties. The exit fan cowl and core nozzle areas, As and
Ap 35 , are known from the manufacturer data. Any additional data needed to define the engine intake and
exhaust geometry was taken in loco at TAP. As Fig. 2.11b schematizes, the engine hardware available in
the test runs consists on (1) the bellmouth for an air flow smooth intake, (2) the cowlings defining the
fan’s enclosure, working as the nacelle of the engine during flight, (3) the nozzle and plug36 that set the
engine exhaust configuration and through which the primary exhaust gases expand at the cell’s room
pressure p′4 , (4) the boattail (working as for the pylon structure which fixes the engine and transmits its
thrust to the aircrafts’ wings), and the optional (5) variable fan nozzle that allows the exit fan cowl area
to be adapted according to the target engine’s thrust specifications.
The engine’s outside envelope - nacelle - defines the configuration of the secondary flow streamline.
The convergent section of the engine exhaust nozzle allows the primary flow to exit the engine smoothly.
34 The angle of the variable stator vanes VSV (°) and the variable bleed vanes VBV (%) ratio were set to the shutoff
position during takeoff testings.
35 A and A , if using the manufacturer’s nomeclature.
8 18
36 Plug: also known as centerbody.
24
(a) The cfm56 main engine stations and components. (adapted image; (b) Turbofan scheme and the components of the dress
courtesy of TAP) kit commonly used in test cells (images from SAFRAN
- Cenco International T M )
Figure 2.11: Turbofan engine stations and hardware.
Together with the outlet guide vanes at the engine’s exhaust exit, vortices due to the rotating turbine
blades - swirl - are mostly eliminated this way and allow the velocity exhaust profile to be approximately
uniform. The secondary flow at station S4A and the exhaust primary flow at station S4B were modeled
considering no diffusion until station S4C (end tip of the engine exhaust nozzle). The engine exhaust
streams were thus offset to station S4C and modeled at this same cross-section, for a given engine mass
flow rate and for the velocities predicted at each engine exit section. Since each stream density is constant,
vs,4C = vs,4A and vp,4C = vp,4B may be computed directly from mass-conservation (equation A.5), as
described in Section 3.5.1.
The area left for the flow to bypass the engine is referred to as Aper ; this is relevant only for the thrust
calculations within a control volume encompassing the peripheral flow. The effects of the boattail over
the engine and peripheral flows were not considered in this study37 . To compute the cell velociy at the
cell station S4C , the cross sectional area of the thrust stand Astand ≈ 4.56 m 2 was taken into account
and then subtracted to the area available at the station mentioned.
2.6.3 Test Cell Instrumentation

The thrust stand38 holds an adapter mechanism that makes the interface between the engine and the
coupling plate; the adapter is specific to each engine model and its instrumentation also allows to obtain
engine data needed to acknowledge its performance status, analyzed from the test cell control room.
TAP makes use of other instrumentation devices to obtain cell standard parameters: two gauge
measurements for static pressure, one at a forward section of the engine, p′ f orward (about 2 m upstream
of the engine bellmouth) and one at its rear, p′ rear (at station S4C ); see Fig. 2.12b. The barometric
pressure is measured outdoors at the elevation of the engine centerline. Also available, there are four
(static) temperature probes at the exit of the cell (station S8 ).
To estimate the test cell intake head loss, new probes were installed at the chamber’s inlet, station
S2 : two probes for total pressure measurements and eight probes distributed circumferentially for static
37 However, some detail about such effect may be found at [21]
38 Also known as framework , thrust frame, cradle or even bed.
25
pressure readings. These probes are embedded in a stainless steel structure, as Fig. 2.12a shows. Also,
two more probes were installed at the exhaust and fixed to a stainless steel support, as shown in Fig. 2.12c:
the exhaust static pressure and the exhaust static temperature, at station S7A , can now be known at
TAP. All probes of the test cell are signalized at the corresponding stations, as shown in Fig. 2.12.
(a) Intake pressure probes, at (b) Forward (left) and rear (right) (c) Exhaust pressure and temperature
S2 . chamber static pressure. (static), at S7A .
Figure 2.12: Test cell static probes.
Chamber Inlet Pressures
The pressure probes at station S2 are connected to the software acquisition system by means of
stainless steel hoses. The airtightness of the whole setup has been verified39 . The pressure loss coefficient
of the intake may be infered from the pressure measurements at station S2 :
△p∞−2 p′0,2
k∞−2 = = , (2.10)
p′0,2 − p′2 p′din,2
where p′din,2 is the dinamic pressure40 .

39 The pressure transducers were used with the most narrow range available, from -5 to 5 psi (-34.5 to 34.5 kPa). However,
this range is still quite large for the typical pressures at station S2 , limiting the accuracy of the pressure readings at this
section.
40 Difference between the total pressure p′ ′
0,2 and the static pressure p2 . All these pressures are gauge pressures, i.e.
obtained relative to the local hydrostatic pressure, denoted by the superscript ’ .
26
The head loss coefficient computed from eq. 2.10 is subject to errors from p′0,2 and mainly from
p′din,2 ; the accuracy increases with v2 . Transient flow regimes also affect the pressures and lead to some
error in k∞−2 . The statistical analysis of the scatter of k∞−2 provided a better information about its
correct vaule. The average of a great number of k∞−2 evaluations was retricted to steady-state high-flow
rate regimes. Outliers were also disregarded (see Fig. 2.14b), such as pressure fluctuations leading to
k ∞−2 > 10 ,and k ∞−2 < 6, and very low dinamic pressures (p′din,2 = P2′ − p′2 < 0.002 psi ≈ 14 P a).
Fig. 2.13 illustrates how the pressure readings became smoother by filtering them. The output static and
total pressures p′2 and P2′ were selected from the engine report data applying the same criteria as for the
head loss coefficient k∞−2 ; an average between the LH and RH probes was then obtained.
Figure 2.13: Top: Static pressure readings and the head loss coefficient for the cell intake (30/Nov/14,
cfm56-5B5/P test run). Bottom: idem, filtered readings. (LH: left-hand side pressure probe; RH: right-
hand side pressure probe (ALF)). .
It was noticed a higher fluctuation on the left probe LH readings than on the RH probe, which may
be related to the lack of accuracy associated to static pressure probes, which are greatly sensitive to
the incoming flow direction. However, the maximum uncertainty found was about ±8.60 P a; Fig. 2.14
illustrates this. Therefore, the engine test runs with insufficient representative data along time or those
at which the uncertainty of rake LH was out of acceptable bounds were eliminated.
Fig. 2.14a shows that the LH probe uncertainty is higher than the RH probe, mainly at low flow
velocities (higher k ∞−2 ). As Table 2.3 presents, a mean value of k∞−2 = 7.64 is obtained with the RH
probe; it is close to the average value of k∞−2 = 7.73, obtained from the mean values of both probes LH
and RH. However, some differences are observed between both probes at some test runs. Fig. 2.15 shows
27
(a) Uncertainty of pressure probes at S2 . (b) Frequency histogram of k∞−2 .
Figure 2.14: Averaged values for the head loss coefficient from the static pressure readings of LH, RH
probes
how the probe readings at station S2 are obtained in specific echelons (ranks). This way, the results from
the RH probe were selected to compare with the computational model estimates for each test run. The
static pressure at the chamber’s inlet p′2 is presented in Table 4.1.
Table 2.3: Average result of the head loss coefficient for each test run. At last column: the global mean
value, k∞−2 = 7.73. Note: stable pressure readings at takeoff could not be found for the -3C test run of 17-Set-14.
Figure 2.15: Echeloned static pressure readings.
The Fig. 2.16a presents the evolution of the static pressure probes, recently installed at S2 and S7A .
As for the inlet pressure readings p′2 , the exhaust static pressure readings p′7A present some fluctuation
also. p′7A varies inversely with p′2 : higher mass flow rates in the test cell are related to a higher p′7A and
result on higher pressure losses at the intake section (smaller p′2 ).
28
(a) Static pressure measurements at the cell intake and exhaust (b) Temperature evolution along an engine test run (03/Out/14,
sections, respectively S2 and S7A (03/Out/14, cfm56-3C1 ). cfm56-3C1 ).
Figure 2.16: Temperature and static pressure measurements at the test cell facility at TAP (from a
cfm56-3C engine).
General Data Post-treatment Process
The information of any test run is collected and treated by a test cell data acquisition software. This
system is able to keep all data given from any sensor installed in the test cell and commonly applies a filter
to disregard high variations in the probes readings. To achieve more stable readings for the pressures at
the inlet of the test cell, a filter similar to this one was applied.
In this study, the post-treatment process consists on selecting stable-takeoff data from each engine
test run at TAP collected between 20-Aug-14 and 14-April-15. This allows to compute the values of the
engine and the flow properties measurements that are most representative of a takeoff power setting.
Generally, takeoff rates have a range of three minutes and the maximum fan speeds of the -3C and
-5B engines are 4942 and 4836 rpm, respectively. The code developed with MATLAB searches for a set of
consecutive readings where the observed fan speed N1 is higher than a specific limit (e.g. above 5000 rpm
at takeoff for the -3C ); then the readings where N1 has a maximum deviation of △N 1 = 0.5% N 1 =
25 rpm are selected. This criterion is stringent enough if a stabilized takeoff rate is pretended.
As Fig. 2.16b shows, the temperature downstream of the intake section is nearly constant but the one
at the test cell exhaust section could not find a stabilized value given the short length of a takeoff test.
This way, the last 10 measures at takeoff were selected to compute the average temperature outputs T7A
and T8 .
2.6.4 Expected Results

All the information follows the Système International d’Unités (officially abbreviated in all languages
as SI 41 ). However, units for airflows and thrust generated by engines are occasionally presented in the
English Engineering system as well, in virtue of most aviation terms and quantities following the latter.
For this one, the unit of force is the pound force (lbf ) and the unit of mass is the pound mass (lbm ). The
pressures are by default expressed as gauge values.
Some results based may be foreseen from the information collected this far. The cell mass flow rate
ṁcell and the pressure at the augmenter’s inlet are the two major quantities to be obtained from the
TCC model. The former may be estimated in advance from the cell bypass ratio (CBR).
41 American Society for Testing and Materials, ASTM Standard for Metric Practice, E380-97. Conshohocken, PA: ASTM,
1997.
29
A commonly used rule of thumb is that a cell must have a bypass ratio of more than about 80% to
avoid vortex formation; typically, cells are designed with CBRs up to 200% [18]. This means that it
will be expected a range of 0.8 < CBR < 2.0 for this cell; higher values for small engines such as the
cfm56-3C will not be a surprise.
On the contrary, CBRs surpassing the lower limit indicated may be attained for some of the cf6
engine family, which is undesirable. The maximum intake air flow rate ever found at TAP was around
784.79 (kg/s), for a cf6-80C2A2 42 . From the CBR range mentioned, this leads to a cell air flow between
1412.6 < ṁcell < 2354.4 (kg/s), which is related to an intake and exit test cell flow velocities within the
ranges 13.9 < v2 < 23.2 and 19.5 < v8 < 32.5 in m/s, respectively43 . The exit flow velocities, already
close to the ones prescribed at 1.2.1, will be too high for an -E1 engine test run exceeding a CBR = 1.7.
42 From the 02-Oct-08 test run data report.

43 This is based on the mass flow rate definition ṁ = ρ A v and on an expected value of ρ2 = 1.2 and ρ8 = 1.0 (kg m−3 )
for the density respectively, which are not far away from the ones obtained with the computational model.
30
Chapter 3
TCC Model
3.1 Target and Scope
The TCC model is a useful tool to understand the flow and evaluate the main characteristics as
a function of the test cell and the engine parameters. The flow properties with larger engines can be
antecipated with this model and the effect of test cell improvements can also be accessed.
The set of equations of the model is solved iteratively for the unknowns ṁcell and p′5 ; a relaxation
coefficient λ together with Aitken’s acceleration were applied for the test cell air flow rate:
2
[ṁ(3) − ṁ(1)]
ṁ(a) = ṁ(3) − , (3.1)
ṁ(3) − 2 ṁ(2) + ṁ(1)
where ṁ(i = 1, 2, 3) are the three values needed to evaluate the new value computed with Aitken’s
formula, ṁ(a). The relaxation coefficient is defined as
ṁ(a) − ṁ(1)
λ= . (3.2)
ṁ(2) − ṁ(1)
Fig. 3.1 shows the static pressure along the test cell and also the algorithm used. The Harp Model
is discussed in the Section 3.6. The Augmenter Model is addressed in Section 3.7.
31
Figure 3.1: I) Static cell pressure at each station S; II) Algorithm used to compute the cell mass flow
rate ṁ1 .
Table 3.1a summarizes the constant quantities related to the engine hardware geometry. The model
input data refers to the front cell temperature T2 (in ° C), relative humidity, barometric pressure pbar
measured at the engine centerline elevation and the engine performance parameters: fuel and engine
mass flow rates ṁf u and ṁeng , the LPT total pressure and temperature, P T 54 and T T 54,. The cell
bypass ratio CBR and the measured thrust F Nmeas are the main TCC model outputs. The CBR is
obtained by the model prediction of the volume occupied by the peripheral flow. The model outputs
are the key-parameters defined at Table 3.1b, which serve to control the model accuracy when compared
with the experimental data.
The TCC model includes a factor ζ that quantifies the augmenter’s performance, i.e., a measure
of the flow diffusion achieved at the end of the augmenter; its value was adjusted to fit the available
experimental data and then kept constant for all numerical simulations.
In Table 3.1b, the parameters denoted as fixed refer to quantities depending on the engine type being
32
tested; the ones denoted as variable refer to quantities that change between test runs, depending on
the atmospheric conditions and other contingencies. The quantities shown at Table 3.1b were directly
compared with measurements and provided the basis for the model calibration: the pressure probes p′2
and P2′ , the forward and rear static pressure p′f orward and p′rear , the static temperature and static pressure
at the exhaust boot section T7A and p′7A and the exit cell temperature T8 . Some of the parameters of
the TCC model were obtained from the engines manufacturers e.g. the theoretical BPR (engine bypass
ratio). Other fixed input data are the distance from the engine primary flow exit to the augmenter’s
inlet L4B−aug , the outer bellmouth diameter and throat area (D3,per and Abell , respectively), the engine
secondary and primary exit areas, As,4 and Ap,4 and the engine nozzle area Anozzle = Acb + Ap,4 . Acb is
the cross-sectional area of the centerbody at station S4B .
Some correlations were found in the literature for the secondary core length, xs /Dp , and the head
loss coefficients summarized in Table 3.1c. The head loss coefficients of the model are also based on
correlations available in the literature. The overall head loss coefficient for some stretches of the plant,
namely k ∞−2 and U7A−8 , can be compared with the corresponding experimental values; as the Reynolds
numbers are very high, these values are characteristic constants of the cell, independent of the engine
being tested. The experimental pressures and flow rates were used to adjust the ζ parameter of the
augmenter. With an appropriate value of ζ, the numerical model agreed well with the experimental data.
Table 3.1: TCC model details.
(a) Geometric input data for the TCC model. (b) The different types of parameters defining the computational model TCC.
(c) Intake and exhaust head loss coefficients.
The following parameters must be within specific ranges: these are the radial distance at which the
moved flow enters into the augmenter Ra , the cell bypass ratio CBR and the global head loss coefficients
at the intake and exit of the test cell sections. The following relations have to be checked to simulate the
correct (standard and safe) operation of the test cell:
33

Ra < Raug




v
3,per > (v3,per )min

(3.3)


 vm,5 < (vm,5 )max


0.75 < CBR < CBR

,
max
where (v3,per )min represents the minimum peripheral velocity recommended at the engine’s inlet plane
(to avoid flow separating from the cell walls); (vm,5 )max represents the maximum moved flow velocity
recommended to avoid significant changes in the engine’s rear static pressure (resulting on significant
thrust changes); CBR max is the maximum CBR recommended (around 1.1 and 2.6 for large and small
engine respectively [2]).
Also, the thrust estimated should be close to the one measured. The head loss coefficient k ∞−2 and
the head loss factor U 7A−8 are computed and validated through comparison to the values obtained from
measurements. The parameters ζ and ζT allow to calibrate the TCC model by adjusting its estimates to
the measured static pressure p′7A and cell mass flow rate, and to the static temperature T7A , respectively.
3.2 Model Simplifications

Geometrical Approaches of the Plant
- The guide vanes losses are estimated according to Idelchik [22] as a function of the Reynolds number
and a standard surface roughness;
- The screens of the cell intake section are modeled as circular wire screens using experimental corre-
lations presented by Idelchik [22];
- The diffusor section shape changes from circular to rectangular: it is modeled as an equivalent
diffusor of rectangular cross-section;
- The diffusor and the expansion immediately downstream could be modeled as a single feature,
but the approach of considering each component of the cell independently and assembled in series was
preferred;
- The cross section of the frames that support the screens reduce the available upstream cross-sectional
area. Both the intake and the exhaust screens, at S1D , S2 and S7B , were modeled subtracting the frames
area, as suggested in the correlations;
- The head loss of a smooth bellmouth as well as the augmenter intake are nearly null and thus were
neglected (see the Appendix Section B).
Thermodynamic and Fluid Mechanic Approaches
- Changes in pressure along the intake and exhaust stacks, due to differences of height, are small. The
geopotential term in Bernoulli equation was neglected;
- Swirl effects of the engine jets fade with distance from the engine’s rear, along the x direction; at
station S4 , short after the engine exit, all velocity and density profiles are assumed uniform both in the
radial and circumferential direction;
- The flow screens at TAP introduce pressure losses but create a more uniform flow. Indeed, the
engine suction slightly distorts the incoming flow. Nevertheless, this effect is small at station S2 , given
the ratio between the distance to the bellmouth and the bellmouth diameter. According to [13] a cell
depression of 75 mm H 2 O (735 Pa) corresponds to a flow distortion v0 = vmaxvavg
−vmin
< 0.2 (considered
low enough) at station S2 . The velocities vmax , vmin and vavg are the maximum, minimum and average
34
flow velocities measured at the front cell station S2 . TAP’s test cell produces a depression about 400 Pa
for the -3C and -5B engines;
- The flow between the chamber inlet station S2 and the engine intake S3,eng is compressible and isen-
tropic. The fluid velocity is approximately uniform and radial over the hemispherical surface surrounding
the bellmouth;
- There are four types of streams at the cell’s chamber: the primary, the secondary, the entrained and
the moved flows. The sum of two first ones constitute the engine flow and the last ones the peripheral flow.
It was assumed the fluid temperature is the same along the chamber for each stream type: Ti,5 = Ti,4 = T2 ;
the subscript i stands for each flow stream and T2 is the average temperature at the test cell room’s
entrance;
- The velocity profile does not change significantly from the fan’s discharge duct to the core nozzle
exit; thus the secondary potential core initial portion was offset to the cell station S4C , which is away
from station S4A for about Dp ;
- The flow properties e.g. pressure, velocity and temperature, are constant at any cross section along
the test cell stations;
- The BPR used in the TCC model for the -3C engines was set equal to the value estimated from
GasTurb (BPR = 5.28). The remaining engine BPRs were fixed according to reference range values
obtained from the manufacturers1 .
- The flow properties are modeled according to the profiles presented in Section 3.5.2;
- The curvature of the streamlines at the augmenter’s intake are neglected when applying Bernoulli
equation between stations S2 and S5 ;
- The exhaust temperature at station S6 is assumed to be fully uniformized and thus the density
equals the average density upstream, at S5 . It is also assumed to be constant along the exhaust section,
from S6 − S8 ;
- The exhaust static pressure p′7A is dependent on the augmenter and the diffusor performance, i.e.,
on the amount of pressure recovered. Friction should be accounted as a pressure drop and the Darcy
friction factor f comes from the Haaland equation2 .

△paug,f = f L/D 21 ρv 2 .

(3.4)

1.11
ε/d
 √1f = −1.8 log10

3.7 + 6.9
Re .
A roughness of ε = 0.15 mm was taken as suggested in [22] (metal tubes, after several years of
operation under different conditions), for the equivalent roughness ε/d;
- No friction at the test cell chamber walls was considered.
3.3 Mathematical Formulation

The peripheral air flow is defined by the augmenter inlet pressure; when p′5 is arbitrated, it allows
to discretize the velocity profile at S5 for a fixed cell air flow ṁcell . From the energy equation, the flow
conditions at the intake are computed. The velocity vm,5 is re-computed until the pressure mentioned
converges. The flow properties at the diffusor inlet S6 are computed from an energy balance. The static
pressure at the exhaust room p′7A is then obtained from eq A.7.
At the same time, the flow conditions downstream of the diffusor are dependent on the pressure loss
at the exhaust. Both stations S1 and S8 are at the atmospheric hydrostatic pressure; this enables to
1 More details about civil turbojet and turbofan specifications at http://www.jet-engine.net/.
2 This equation is a solution found by Haaland for the implicit Colebrook-White equation known as √1 =
f
ε 2.51
−2 log10 3.7D + Re √
f
for Re > 4 × 103
h
35
define p′5 and p′7A easily from energy equation again. The cell air flow ṁcell may thus be known iteratively
according to both the intake and the exhaust head losses estimates; those were obtaiend also from the
measures taken at TAP within the fieldwork project; these enabled to validate the TCC model. The
criteria is to match the pressures p′2 and p′7A . The augmenter pressure recovery is adjusted accordingly,
as discussed in Section 3.7.
3.3.1 Energy Equation

The takeoff steady flow is incompressible for the cell intake, S1 − S2 , and compressible from a section
upstream of the bellmouth to the entrance of the engine itself, S2 − S3 . There are two different air flows
within the test cell: the equations stating these characteristics are summarized at A.2.
For the incompressible flow considered at the intake and the exhaust of the cell, eq. A.7enables to
know the static pressure (relative to the hydrostatic pressure) downstream of the cell intake, p′2 :

p′ = p′ + 1 ρ v 2 − 1 ρ v 2 (1 + k∞−2 )
2 2
2 2 2
∞ ∞
2 2 2 2 2
k∞,2 = k1 A A A2 A2 A2 A2
A1
2
+ kA AA2 + kvanes A A
+ kscr,B A B
+ kbundle A ′′
+ kscr,2 A ′
,
B C
(3.5)
being k∞,2 the intake head loss coefficient. Here, v2 is the reference velocity according to which the
coefficients k1 , kA , kvanes , kscr,B , kbundle and kscr,2 were obtained as described in Section 2.5; the Intake
Model referred in the algorithm scheme is based on eq. 3.5. Since these coefficients refer to a change
in pressure within a particular section, associated to a change in velocity, a correction must be made for
the latter one through an area ratio thatrelates the reference velocity and the velocity at each station
with the constant mass flow: v2 = A Ai
2
v1 . The areas, as it can be deduced, refer to passage areas
perpendicular to the flow local income, at each station from S∞ to S2 . The barometric pressure pbar at
the intake stack entrance and also at the exhaust stack exit (at heights h∞ and h8 , respectively) are null
when relative to the local hydrostatic pressure, i.e., p′∞ = p′8 = 0. It is expected the fluid static velocity
to be v∞ = 0, since the fluid is accelerated from rest when no ambient disturbances such as wind are
assumed. The pressure losses along the intake section of the test cell change the density slightly, but we
may assume that the density is constant along the cell intake and chamber sections: ρ = ρ2 ≡ ρ∞ .
The eq. A.8 allows to compute the pressure at the engine’s rear cross section: p′4 = p′2 + 21 ρ2 (v22 −vm,4
2
),
′
being that vm,4 = v2 A2 /Aper,4C . Similarly, for a known pressure p5 , the velocity at the augmenter inlet
section is obtained as: s
p′2 − p′5 + 21 ρ2 vxm,2
2
vm,5 = 1 . (3.6)
2 ρm,5
This pressure, along with p′2 and vm,5 , allows to estimate the mass flow rates and momentum quantities
that characterize the flow along the constant test cell, ṁcell = ṁeng + ṁf u .
As for the static pressure at the end of the diffusor, the flow faces a set of tight mesh screens followed
by a series of baffles that make it turn and decelerat before it exits the test cell; it comes that:
1 2 1 2
p′7A = p′8 + ρ v − ρ v (1 + k7A−8 ) , (3.7)
2 8 2 7A
where ρ is the mean density at the exhaust section. The flow at S8 exits as a jet at the ambient pressure
and so v8 6= 0; which also means that the pressure within the exhaust stack, from S6 − S8 , is lowering
36
till it matches the exterior one, p8 , which is thus lower. Again, the areas of the previous equation are
interpreted as being the ones perpendicular to the flow, at stations S6 , S7A , S7B and S8 . The coefficient
k7A−8 refers to the sudden expansion head loss coefficient followed by a global coefficient one due to
the screens through which the air flows in parallel; it accounts also for the stack head loss, from the
station immediately after the screens (S7B ′ ) to the exit. Fig. 3.2 presents a scheme of the exhaust section
modelling. The last coefficients schematized in series, k7C and k8 , are based on the local velocities already
discussed in Section 2.2: the velocity v7C stands for station S7C and v7D is the reference velocity for the
last two ones (S7D and S7E ).
At station S7B , the flow is divided in three parallel streams, as illustrated in Fig. 3.2. The TAP’s test
cell exhaust plant (Fig. C.2) is available in the Attachments section. For every stretch it can be defined
a loss factor U (kg.m−7 ) such that △p = U Q2 , where Q (in m 3 /s) is the volume flow rate.
1
U= ρ k/A2 , (3.8)
2
is related with the head loss coefficient k and with the local cross-sectional area A. The global head
loss factor U7A−8 relates with the individual head loss factors of the sections in parallel and in series.
The individual Ui of the sections in series add together as the individual pressure drops of the sections
in parallel add together. In our plant, the global term U7A−8 is thus expressed as:
h
−0.5 −0.5 −0.5
−1 i2
U7A−8 = Udif + USE + Uper + Umid + Ustep + U7C + U8 , (3.9)
where Udif and USE are the head loss factors associated with the diffusor and the sudden expansion
downstream; Umid and Uper stand for the head loss factors of several sets of grids of the middle and side
channels after the diffusor. These grids do not span the whole height of the room and Ustep is the head
loss factor of the escape path around such grids. U7C and U8 are the head loss factors of the labirinth
section of the exhaust stack. The exit (gauge) pressure is p′8 = 0; therefore, according to the energy
equation, p′7A is a function of the test cell volume flow rate Qcell and the density at the exhaust modeled
as the average density at station S6 , ρ = ρ̄6 :

1 1 2 1 1 1 1
∆p7A−8 ≡ p′7A = ρ v82 − ρ v7A + ρ Q2cell U7A−8 = ρ Q2cell − 2 + U7A−8 . (3.10)
2 2 2 2 A28 A7A
The head loss factor U7A−8 was estimated from pressure and flow rate measurements at TAP and
compared with the TCC model results. The Exhaust Model referred in the algorithm scheme is based
on eq. 3.9.
3.3.2 Test Room and Engine Air Flows

From the modelling point of view, the section S2 is divided into two simple regions, once the flow
behaves differently from this section on along the cell: part of it is accelerated from up to the engine’s
inlet which leaves the remaining flow to bypass the engine itself. The first one suffers the effects of
compressibility due to a great reduction in pressure imposed by the engine spinning turbofan blades; these
accelerate the fluid up to Mach numbers that are higher than Ma=0.3 and thus the density is affected.
37
Figure 3.2: Left: head loss coefficients scheme, considering the exhaust grids (S7A −S7B ) and the turnings
along the exhaust channel (S7B − S7C ). Right: diffusion section scheme.
The area occupied by this high-speed flow may be estimated from the mass flow rate of the engine:
A2,eng = ṁeng /(ρ2 v2 ). The remaining part of the intake air flow was again treated as incompressible;
the peripheral flow area is defined respectively as A2,per = A2 − A2,eng and A3,per = A3 − Abell,inlet for
stations S2 and S3 , respectively.
Part of the flow is accelerated towards the bellmouth and the remaining flow bypasses the engine,
being this the reason why the chamber room was sectioned the way as illustrated by Fig. 2.1.
Bernoulli equation supports that the static pressure relative to the hydrostatic pressure changes along
the test cell room in as much as the dynamic pressure changes. This is true for the moved flow which,
together with the entrained flow, constitute the peripheral flow:
ṁper = ṁe + ṁm . (3.11)
The entrained flow denoted by subscript e accelerates by diffusion from the engine jet momentum. It
is of great interest to have both pressure and velocity fields defined and then sketch the flow streamlines
along the cell. However, to perform such a task it would be prompted a considerable number of new
instrumentation. A way to model this section (S2 to S5 ) is to take fluid mechanics basic knowledge to set
a basis that enables to predict the amount of air flow bypassing the engine and the one being entrained
into the augmenter. The Section 3.5 develops this issue in detail.
The peripheral air flow is incompressible along the chamber until it enters and uniformizes through
the augmenter, but the flow entering the engine through the bellmouth duct suffers compressible effects.
The TCC model computes the thermodynamic quantities needed to perform a momentum balance to
estimate the engine’s thrust. It makes use of the flow isentropic relations3 and also of aitken’s acceleration
process for the Mach number M 3 ; the engine intake flow velocity, temperature and density depend on
this Mach number.
3 The isentropic flow properties relation between any generic station of the cell (M
S ) and the stagnation conditions
(M = 0) or even the critical conditions of any station (M S = 1) and as well as may be found at [4], p. 25.
38
3.3.3 Thermodynamic Flow Properties
The specific heat for humid air treated as an ideal gas comes as

c = c
p p,air + w × cp,vap
(3.12)
 = γ cp = γ c = R + c → R = cp (1 − γ1 ) .
γ v v
being w the humidity measured in grains (gr = 0.14176 × 10−3 kg water vapor / kg dry air)4 ; cv
is the specific heat at constant volume and γ the specific heat ratio. The TCC model evaluates the
specific heat ratio γ according to the fluid temperature by means of interpolation from the table of dry
air properties at atmospheric pressure 5 . The γ dependence on the air humidity is studied in detail by
Wong [39] and is outside the scope of this work. The gas constant R depends this way on air humidity
and temperature. The specific heat at constant pressure cp and the ratio of specific heats γ are set equal
between the secondary, entrained and moved flows.
The exhaust hot gases are pre-heated air burned with jet fuel6 . The high temperature of the com-
bustion products as well as the fuel injection change the gas constant Rexh . The stoichiometric air-fuel
ratio AFR sto may be obtained remembering that it will be needed a mass of x kg/s of air to burn all
the fuel mass flow ṁf u ; given the excess of air during the combustion process of turbofan engines when
cruising or idling (this is, continuous or low power demands), a mass quantity of y kg/s of air is used
to burn the hydrocarbon fuel and the mixture can reach 400% of theoretical air. The chemical equation
representing its combustion7 is

C H + a (O + 3.76N ) → x CO + y H O + 3.76 a N
x y sto 2 2 2 2 2 sto 2
(3.13)
(CH ) = C H = C H y
2 n x y 12 24 ⇒ asto = x + 4 = 18.
The stoichiometry coefficient (100% theoretical air) asto relies on the molecular formula of the Jet A1
fuel presented above, and it follows that

mair MO2 + 3.76 MN2 Mair kgair
AF Rsto = = asto = 4.76 asto ≈ 14.76 . (3.14)
mf u sto Mf uel Mf uel kgf uel
The TCC code computes the actual AFR from each test run data collected at TAP and the diluition
is obtained as
AF R ṁp
φ= = (%). (3.15)
(AF R)sto ṁf u AF Rsto
4 From the url
://ftp.demec.ufpr.br/disciplinas/TM182/CLIMATIZACAO/apostila/2 PROPRIEDADES%20DO%20AR%20ATMOSFERICO.pdf,
p. 4, eq. 10.
5 Available at http://www.engineeringtoolbox.com/dry-air-properties-d 973.html. This is applicable also to TAP because
the pressure at the test cell is close to the atmospheric pressure.

6 Of type Jet A-1 is a 100% kerosene fuel. Its surrogate is 1-dodecene (C H ) thus with a M = 168 kg/kmol molecular
12 24 f
weight. Details from p. 5 at http://webserver.dmt.upm.es/˜isidoro/bk3/c15/Fuel%20properties.pdf.
7 From [12] eq. 2.19, p. 30. The molecular weight in kg/kmol of oxygen and nitrogen is approximately M
O2 = 16 and
MN2 = 28 and the air composition was assumed to be 79% N2 and 21% O2 by volume, giving a molecular weight of air
of Mair = 28.97. However, burned gases have a decreasing molecular weight which also depends on the diluition of fuel in
air, this is, on the AFR. More information about aviation hydrocarbon fuels at [5] at chapter 8, pp. 543-556.
39
The specific heat and the gas constant of the exhaust gas cp,exh and Rexh were obtained from inter-
polation of the Table 3.2.
Table 3.2: Molecular weight and gas constant Rexh for air and for products of combustion of lean fuel-air
mixtures. Fuel is (CH2 )n . Source: [5].
The flow properties of the combustion products were obtained from [5]. In accordance with the
authors, from the TCC model it was verified that the temperature rise due to combustion changes very
little the gas constant of the combustion products but rise the specific heats cp and cv significantly.
The specific heats ratio is distinct from the standard air at steady-state conditions value of γ = 1.4
and is about γexh = 1.3 for a -3C takeoff test8 . The effect of the humidity on the gas constant of the
combustion products Rexh may be evaluated considering the gas constant R (computed for humid air)
and the diluition. If R(φ = 100%) = 287.62 is the gas constant for stoichiometry if follows R(φ =
200%) = (R(φ=100%)+R)
2 and R (φ = 400%) = (R(φ=100%)+3×R)
4 . The TCC model interpolates the gas
constant from these values and computes the specific heat by interpolation from Fig. 3.3b.
(a) Specific heat of air and water vapor dependence on (b) Influence of temperature and mixture ratio (diluition) on the
air flow temperature. cp of equilibrium fuel-air combustion products (trendline based
on [5], Fig. 4-2-5.)
Figure 3.3: Thermodynamic properties: air and combustion products specific heat and exhaust gas
constant.
The total measured temperature T2 ≡ T02 is always constant and depends on the static and dynamic
quantities:
v2
T0 = T + = const. (3.16)
2 cp
The static temperature T , like the static pressure p, is affected by the air flow speed and does not
change much at the periphery of the chamber. This way, the error of assuming the flow properties of
the engine secondary stream equal to the one obtained at station S2 is negligible. The same applies
8 The 1
specific heat ratio γ is again expressed by the relation γ = 1−R/Cp
.
40
for the engine inlet air flow (S3,eng ), for the peripheral flow along the chamber (S3 and S4 ) and for the
augmenter’s inlet as well (S5 ).
2
Combining equation 3.16 with the continuity equation (A.5) it follows T0 − T = 2 cpṁ(ρA)2 . Combining
γ−1
γ
γ−1
even further with the following isentropic relation TT0 = Pp0 = ρρ0 , it comes:
" γ−1 #
ṁ2

2 ρ
ρ 1− = , (3.17)
ρ0 2 cp T0 A2
where ρ0 is the total density.
The stagnation adiabatic temperature T T 54 and isentropic stagnation pressure P T 54 at the engine
LPT are missing in the test run data of the -5B and -C2 engines. The TCC model estimates those values
based on two assumptions:
- the product of the air-fuel mass flow ratio ṁair /ṁf u by the stagnation temperature is constant and
the -3C engine test run of 03-Oct-2014 is taken as reference for that constant:

ṁair ṁair
T T 54 = T T 54 . (3.18)
ṁf u ṁf u ref
This approximate equality is used to evaluate the stagnation temperature T T 54 of the primary flow. The
pγ (γ+1)
2 − 2(γ−1)
factor M R 1 + γ−1
2 M , where M denotes the local Mach number, is assumed to be the
same for all engines operating at takeoff regimes. From the stagnation pressure P T 54, from The mass
flow rate per unit area ṁ/A and the ratio of isentropic stagnation properties √P T 54 are equal to that
T T 54
factor (eq. 3.19, from [4], p. 28):
√ r (γ+1)
− 2(γ−1)
ṁ T T 54 γ γ−1 2
=M 1+ M . (3.19)
A P T 54 R 2
Again, the -3C engine test run of 03-Oct-2014 is taken as reference for the right-hand side of this
equation, to isolate the stagnation pressure P T 54 as:
√

P T 54
P T 54 = ṁ/A T T 54 √ . (3.20)
ṁ/A T T 54 ref
According to this assumption, the Mach numbers of different engines are approximately equal (except
for the influence of the differences in the γ and R values).
3.3.4 Mean Flow Properties
The energy equation may be also used to compute the gauge pressures of any station of the exhaust
stack, especially of S7A , the one where the rake shown at Fig. 3.2 was installed. Hence, this is the
reference station taken for the computation of the exhaust global head loss. Once assuming the flow to
be fully developed at S6 the mean flow properties at that station are a known: the mean flow velocity,
density, temperature and specific heat may be obtained from integration of the profiles presented at
Section 3.5. This way, the average values of the said quantities come as:
41
 h ´ Rs2 ´R 2
i


 v̄5 = 2π vp R2p1 /2 + Rp1 2
v(r)ps + vs (Rs1 2
/2 − Rs2 /2) + Rs1a v(r)se + vm (Raug /2 − R2a /2) /Aaug

 h ´ Rs2 2
i
ρ̄5 = 2π ρp R2p1 /2 + Rp1 ρ(r)ps + ρs (R2a /2 − Rs2
2
/2) + ρm (Raug /2 − R2a /2) /Aaug

h ´ Rs2 2
i


 T̄5 = 2π Tp R2p1 /2 + Rp1 T (r)ps + Ts (Ra2 /2 − Rs22
/2) + Tm (Raug /2 − R2a /2) /Aaug

 h i
c ¯ = 2π c R2 /2 + ´ Rs2 c (r) + c (R2 /2 − R2 /2) + c
 2
2
p,5 p,p p1 Rp1 p ps p,s a s2 p,m (Raug /2 − Ra /2) /Aaug .
(3.21)
p6 ṁ6 2
On the other hand, ρ̄6 = R6 T6 , v6 = Aaug ρ̄6 , T6 = (E5 /ṁ6 − vx6 /2)/cp,6 , being the energy flow rate
defined as



 Ē5 = Ē5,1 + Ē5,2 + Ē5,3 + Ē5,4 + Ē5,5


Ē5,1 = π ρp vp (cpp Tp + vp2 /2) × Rp1
2






Ē5,2 = π ρs vs (cps Ts + v 2s /2) × (Rs1
2 2
− Rs2 )


(3.22)
2 2


 Ē5,3 = π ρm vm (cpm T m + vm /2) × (Raug − Ra2 )

 ´ Rs2
ρ(r)ps × v(r)ps × cp(r)ps T (r)ps + v(r)2ps /2

Ē5,4 = 2π Rp1





 ´R
Ē5,5 = 2π Rs1a ρe × v(r)se × cpe Te + v(r)2se /2

.


Finally, since the exhaust air cooling is attained through dilution of the peripheral air flow being
entrained thanks to the engine exhaust gases, there is a need for estimating the mean values for the
specific heat, specific heat ratio and gas constant. From the diffusor’s inlet at S6 , the flow has already
been developed within a profile defined by the hot exhaust gases and - the major term - peripheral air.
Assuming complete mixing (which is an over-simplification) between all the streams the themodynamic
quantities from that station on are referred as the averaged values presented above. This way, for the
cell section S6 − S8 the values obtained from integration of the profiles of cp (r) and ρ(r) defined below
were kept the same. That is how the average exhaust density was found (around ρ6 ≈ 1.0 kg/m3 ) and
assumed to be constant along the exhaust, i.e., ρ̄ ≡ ρ6 .
3.4 Engine Thrust

Thrust measurements within test cells need calibration to achieve engine performance comparison
between indoor test cell and outdoor free-field facilities. However, atmosphere ideal conditions in still
air are sporadic and thus some alternatives to obtain the gross thrust from corrections of the net thrust
measured have been developed along the latest years.
The momentum equation is a statement of Newton’s Second Law. Fig. 3.4 schematizes an aircraft
engine equipped for a test run, subject to a uniform flow of constant velocity. Considering a control
volume defined between the stations S3 and S4C , the force the engine transmits to the load cells can be
numerically obtained and compared to the thrust measured at TAP. For a steady-state flow regime it
follows that
ˆ ˆ
− pnx dS + fx = ρvx (~v .~n) dS. (3.23)
S S
There is some uncertainty on the peripheral flow properties, once wall friction and streamlines deflec-
tion due to test cell handrails, engine monorail etc. are not considered in this model; this increases with
the peripheral mass flow rate. Taking the control volume defined by the contours of the engine itself, the
thrust estimated incurs a less error than if it is defined as in Fig. 3.4 a). This gives an estimate closer
42
to the real engine thrust. The fact that the static engine pressure estimated with the TCC model was
nearly equal to the one measured at TAP (p3,eng = P S2), supports this.
On the other hand, the velocity profile at S3 may be considered as uniform, this is, the curvature of
the streamlines at the bellmouth entry may be neglected: considering the semispherical geometry for the
control volume at the engine intake shown in Fig. 3.4 b), the approach flow direction is radial; the static
pressure and the velocities along this boundary are thus uniform. For a control volume as in case b1) the
momentum comes as:
π π
ˆ ˆ ˆ
M Mx,3 = ρvx (~v .~n) dS = −ρvr2 sin θ sin2 ϕR2 dθdϕ = ρvr2 πR2 , (3.24)
S 0 0
where vx (θ, ϕ) = vr sin θ sin ϕ, (~v .~n) = vr and dS = R2 sin ϕdθdϕ. The momentum is thus defined as
the product of ρvr2 with the projected area πR2 ; it follows that M Mx,3 = ṁ2 /(ρπR2 ), which is the same
as taking the bellmouth vertical plane at S3 for the control volume, i.e., taking the flow velocity profile
at the engine intake as shown in Fig. 3.4 b2).
Figure 3.4: Engine momentum balance control volume CV. a) The test cell cross section is taken for the
control volume; b) CV coinciding with the engine outer walls.
The thrust developed is a sum of the pressure forces with the momentum of the flow within a control
volume; if this is defined as in Fig. 3.4 b) - considering a control volume that coincides with the engine
itself - the engine measured thrust comes as:



 F Nmeas = Feng − Fstand = Fpressure + Fmomentum − Fstand


F ′ ′ ′
pressure = p4 (As + Abell,ext ) + p54 Anozzle − p3 A3

(3.25)
F 2 2

 momentum = ρp,4 vp,4C Ap,4C + ρs,4 vs,4C As,4C


= −ρ3 v3 A3 ,


with Anozzle = Ap + Acb , where Acb is the centerbody cross-sectional area at station S4B . A3 is the
area of the bellmouth inlet, based on its outer radius, Dper,3 . Fig. 3.4 c) illustrates the engine exhaust
areas.
The (static) pressure p′4 is approximately uniform along stations S3 to S4C (p′3,m ≈ p′4C ≈ p′4B ≈ p′4 )
and may be computed from eq. A.8 applied between S2 and S4C ; the moved flow velocity is obtained
from the mass conservation law: vm,4C = v2 Aper,2 /Aper,4C and thus p′4 = p′2 + 12 ρ(v22 − vm,4
2
).
43
Finally, the aerodynamic force acting along the area surface of the thrust stand affects the thrust
measurement; considering a drag coefficient of C D = 1.05, taken from [3], p. 16,this may be estimated
from
1
Fstand = CD × ρv42 Astand . (3.26)
2
3.5 Entrainment Model

There are two streams exiting the engine: a primary mass flow mixed with burned fuel, with a
velocity vp,4 , and a bypass mass flow with a velocity of vs,4 . Subscript 4 stands for station S4 , where
the exhaust gases leave the engine and are expelled through the test cell exhaust section. Similarly,
subscript 3 represents the bellmouth intake station S3 , again through a cross-section of the test cell.
The density changes as the flow accelerates towards the engine intake. This way, subscript m stands for
the peripheral flow that is moved by means of the pressure difference between the augmenter and the
chamber. At the meantime, part of the peripheral air flow is entrained into the engine exhaust stream;
subscript e represents this air flow.
The density of the engine air flow should be as much trustful as possible; this is because the engine exit
velocities are highly dependent on the stream density. The TCC model computes either the specific heat
at constant pressure cp (J kg −1 K −1 ) and the ideal gas constant R (J kg −1 K −1 ) based on the temperature
and the relative humidity; the computation of these quantities is already described in Section 3.3.3.
3.5.1 Engine Exhaust

The exhaust flow exits the engine LPT (station 54) and meets the chamber’s room pressure at S4 .
The secondary stream pressure is assumed to equal the cell rear pressure p4 (Pa). The primary flow total
temperature at station S4 , (T0,p )4 , may be obtained considering an isentropic evolution from the engine
LPT station:
(γexh−1 )/γexh
p0,4
(T0,p )4 = T T54 × , (3.27)
P T54
being γexh the specific heats ratio for the exhaust (primary) flow and T T54 and P T54 respectively
the total temperature (in K) and pressure (in Pa, absolute) measured at TAP. The total pressure at
the engine’s rear (T0,p )4 (in K) is computed from the total pressure defined as p0,4 = p4 + 12 ρvm,4
2
. It
2
vp,4
follows that Tp,4 = (T0,p )4 − 2 cp,exh and ρp,4 comes from eq. 3.17; the secondary stream density comes
p4
as ρs,4 = R Ts,4 , with Ts,4 ≃ T2 ; the gas constant R is obtained for the air inlet stream at station S2 .
The total density of the primary flow may be computed from the ideal gas law and from the engine
P T 54
total pressure and total temperature measurements, ρ0,4 = Rexh T T 54 . Rexh is the ideal gas constant of
the combustion products, where the air humidity is taken into consideration. Solving eq. 3.17 for ρ, the
static density of the primary flow ρp,4 may be estimated, and γ and cp would refer to the exhaust flow
specific heats ratio and specific heat, respectively.
Using eq. 3.16, the computational model also computes the static temperature T54 at the engine exit
station S4B based on the engine exhaust temperature T T54. The static pressure at S4B is computed
from the ideal gas law, p54 = ρp,4 Rexh T54 . The exhaust pressure of the primary stream and around the
engine convergent nozzle section9 is assumed to equal the static pressure p54 (Pa).
9 Along the centerbody, shown in Fig. 2.11b.
44
The engine bypass ratio BPR changes slightly with time. One reason is the average chord of the fan
blades varying slightly along the engine’s life. Fig. 3.5 shows the engine exit velocities evolution with the
thrust of a cfm56-3C engine, predicted with GasTurb 10 . Based on the guidelines from [33] (pp. 43-44),
the velocities vp4 ≈ 420 m/s and vs4 ≈ 305 m/s at takeoff could be estimated for the -3C engine. The
BPR is 5.28. Similarly, the velocities at the engine exhaust of vp4 ≈ 397 m/s andvs4 ≈ 313 m/s were
obtained for the -C2 engine. The GasTurb simulations rely on specific engine data collected from test
reports available at TAP and on fluid properties at standard conditions11 . The TCC model, however,
considers the effects of the air humidity and temperature on ideal gas properties and predicts different
engine exit velocities, e.g. vp4 ≈ 428 m/s and vs4 ≈ 315 m/s at takeoff conditions for the -3C engine.
Figure 3.5: Left: Exit velocities depending on engine air flow and density. Right: Exhaust flow velocities,
from GasTurb (BPR=5.28).
The primary flow accelerates from the end of the engine nozzle and, since the mass flow rate ṁp is kept
constant, it occupies a smaller annular area downstream of the nozzle than if there was no centerbody.
However, the development of the engine exhaust flows at station S4C is simplified by computing the radial
distances to the jet centerline taking the same exit areas and velocities estimated for the corresponding
ṁ ṁs
nozzles, this is vp,4C ∼ vp,4B = ρp ×App,4B with Ap,4C ∼ Ap,4B ≡ A8 and vs,4C ∼ vs,4A = ρs ×A s,4A
with
As,4C ∼ As,4A ≡ A18 . By setting the downstream areas equivalent to the previous ones, the model
can estimate the parameters that define the development of the exhaust gases at the augmenter’s inlet,
namelly Ra , the radius that corresponds to the moved stream being pumped along the cell test room.
This parameter, computed below, is critical when analyzing cf6 engines. q
Ap
The radii defining the outer region of each engine exit flow are then computed as Rcore = π = Dp /2
q
As 2
and Rf an = π + Rcore . The area occupied by the secondary flow at S4C comes as the difference
between Af an = π Rf2 an and the primary flow area, Ap,4C ≡ Acore = π Rcore
2
≡ A8 , this is, As,4C =
Af an − Acore .
3.5.2 Flow Profiles

The pressure p′5 is obtained iteratively in the TCC model applying eq. A.8 between S2 and S5 . The test
cell air flow ṁcell gives an estimate for the flow velocities v2 and vm,5 (eq. 3.6). The engine exit velocities
10 A software developed by Joachim Kurzke that allows the evaluation of the thermodynamic cycle and of the engine
performance data, for a wide set of gas turbines. The following results are obtained for the engine Design Cycle Point
simulation, according to GasTurb nomenclature, consisting on iterating unknown output quantities in such a way that the
given value (target) is achieved. These target values are given from the engines correlation report, accessed at TAP.
11 The gas constant R, specific heat at constant pressure c and specific heats ratio γ are for dry air at the reference
p
ambient temperature (15 º C) and sea-level pressure.
45
vp,4 and vs,4 depend on the engine BPR and also allow to define the flow profiles at the augmenter inlet,
S5 .
Fig. 3.6 illustrates a cfm56 engine scheme; the fan’s discharge at station S4A and the exhaust nozzle
at S4B are similar to the one at in-flight configuration. The way found to define the profile of any flow
property was to set a polar coordinate frame of reference at station S4C , i.e., x (S4C ) = 0. This allows to
study the evolution of the flow from the engine’s rear to the entrance of the augmenter. The ratio x/Dp
up to which flow velocities are kept constant within the velocity profile along the x direction is between
4.0 and 6.8 for the secondary flow and 13.3-15.0 and 16.0 for the primary one depending on the engine
bypass ratio, according to Papamoschou [30], p. 4 and [29], p. 8) 12 .
The primary potential core develops within the range 0 < r < Rp1 and the secondary one within
the range Rs2 < r < Rs1 . The radius Ra borders the air flow entrained to the augmenter by means of
momentum diffusion from the engine streams. The air flow moved by means of the ejector pump effect
accelerates between Ra < r < Raug . Outside the potential cores, the air flow properties change along the
r coordinate; a cosinus function define their profile.
The TCC model evaluates if the potential cores have ceased at the augmenter’s entrance S5 and
also at its exit station S6 : if so, the velocity, the density, the specific heat and temperature profiles are
defined as in case A shown at Fig. 3.6, otherwise the model assumes they develop as presented in case
B . However, case A would only occur if the engine-to-augmenter gap was increased up to a certain limit
which is never verified at TAP; this way, it remains as an hypothetical case, possible of being studied
using the TCC model, e.g. within the framework of a parametric study.
Figure 3.6: A cfm56 engine scheme with the primary, secondary and entrained flow streams. The radius
defining the potential cores are also shown.
For the actual engine-to-augmenter distances L4B−aug at TAP, both the primary and the potential
cores are present at station S5 and are blurred along the augmenter. This way, the velocity profile
develops as in case B :
12 The ratios x /D = 4.0 and x /D = 13.3−15.0 were obtained for an engine bypass ratio of 6.0; the ratios x /D = 6.8
s p p p s p
and xp /Dp = 16.0 refer to a BPR = 5.0.
46



 vp,5 = vp,4 0 < r < Rp,1
 h i
 (v −v ) R −r
vps (r) = vp,5 − p,5 2 s,5 cos π Rs,2s,2 + 1 Rp1 < r < Rs,2



 −R p,1
v5 (r) = vs,5 = vs,4 Rs,2 < r < Rs,1 (3.28)

 h i
vse (r) = vs,5 − (vs,5 −ve,5 )

Ra −r
cos π + 1 Rs,1 < r < Ra



 2 Ra −Rs,1


v Ra < r < Raug .
m,5
Assuming complete flow uniformization and neglecting the Koppers harp additional effect, the velocity
profile at S6 would be as sketched in Fig. 3.6. If a momentum balance is applied at a control volume
defined by S5 and S6 , the average flow velocity v̄6 can be obtained. The pressure along the control volume
boundary cross sections is constant, i.e., pS5 (r) = p5 and pS6 (r) = p6 . Similarly, pS7i (r) = p7i , being
subscript i each station of the exhaust test cell section.
The remaining flow properties (density, specific heat and temperature) are constant between Rs,2 and
Raug ; these were modeled according to the following profiles:



ρ(r)ps = ρp − (ρp − ρs )/2 × (cos [π (Rs2 − r)/(Rs2 − Rp1 )] + 1)

T (r)ps = Tp − (Tp − Ts )/2 × (cos [π (Rs2 − r)/(Rs2 − Rp1 )] + 1) (3.29)


cp (r)ps = cp,p − (cp,p − cp,s )/2 × (cos [π (Rs2 − x)/(Rs2 − Rp1 )] + 1) .

3.5.3 Ejector Pump Effect

The jet momentum creates a depression at the augmenter inlet, thus the peripheral air flow accelerates
from the engine’s rear towards the augmenter at station S5 . The volume flow rate per unit area entrained
in the augmenter is
Qper vper AR Aeng

= = vper AR (AR + 1)−1 , (3.30)
Atot Atot
where AR = Aper /Aeng is the peripheral area to engine flow area ratio. Fig. 3.7c, based on the
numerical simulations of Choi and Soh [9], presents the expected trend for the turbofan engines tested
2
at TAP. In that figure, Aeng /Atot = (Deng /Daug ) . Turbofans have two distinct flow streams instead
of a single primary flow. This favours the momentum diffusion. The turbofan engines considered in
our study are expected to behave approximately as in Fig. 3.7c. This figure is in agreement with [2],
D
who presented empirical relations for the entrainment ratio, respectively ER = 0.22 Daug
n
+ 1.07 and
Daug
ER = 0.9304 Dn + 1.189213.
Fig. 3.7b plots the augmentation performance on mixing augmenter length for parametric levels of
primary temperature and pressure ratios. From that figure, if (L/D)aug = 8 corresponds to 100%, a value
of (L/D)aug,T AP ≈ 4.1 leads to a performance factor of ψ = 3.2/4.4 = 73%. The performance factor is
function of the mass flows and on the primary-to-ambient temperature ratio θ: ψ = ψ(Qper /Qeng ; θ), as
defined by Quinn [32]. The geometry of the augmenter determines the amount of peripheral cool air flow
being entrained: the depression it allows is key factor controlling that air flow.
The uniformization coefficient ζ defined at the TCC model guarantees that the increase in pressure
along the augmenter △p5−6 is such that the static pressure at the cell exhaust room p′7A is close to the
13 The constant 1.07 in the first correlation appears in the original document as 10.7, but we believe this is a typo,
according to all the other information provided by the authors.
47
(a) Pumping characteristics of ejector nozzle at various (b) Augmentation performance on mixing
area and nozzle pressure ratios (Numerical study from augmenter length. Here, π is the primary-
[9]). to-ambient pressure ratio. (source: Quinn,
[32]).
(c) Trendline for the entrainment flow ratio as function of the en-
gine size for the test cell at TAP.
Figure 3.7: CBR trendline comparison between different author studies for turbojet engines.
measured one at TAP. As for the Koppers harp, ζ comes as a factor defining the velocity profile at the
end of the augmenter, which was found to be not totally uniformized: the TCC model predicted a value
5% higher than the experimental one for the pressure p′7A 14 . Fig. 6.2 presents the effect of ζ on the cell
air flow. The TCC model predicts a value of ζ = 0.70.
3.6 Harp Model

The challenge about modelling the impact of the harp on the flow consists on quantifying how much
momentum is converted into static pressure; at the meantime, the harp introduces a head loss that also
needs to be quantified.
The Koppers harp (ring diffuser) was modeled as being two-dimensional. Each harp ring affects the
flow streams differently, depending on its position within the augmenter section and on the engine size.
The position of each ring is defined at Table A.1 and illustrated by Fig. 3.8. Ri indicates the mean
14 For the -3C engine if the exhaust flow was fully mixed (ζ = 1) at S6 : p′7A,matlab /p′7A,meas = 1845/1751 = 1.05.
48
position of each ring, where subscript i identifies the rings effectively installed at TAP.
Figure 3.8: Sectioned view of the 2D ring harp structure (obtained from CAD Design Software Solid-
Works r ).
The flow conditions (vring , νring , Red,ring and ρ) are computed to each ring and the resistance force
it exerts on the flow may be estimated. The ring thickness taken for the length scale, dring , is the ring
cross-section diameter and is equal for all rings. The TCC model evaluates the effect of each ring on
each flow stream, concerning pressure drop and diffusion. The force being applied to the flow may be
expressed in terms of dynamic pressure affected by a resistance drag coefficient C D,ring , associated to
the cylindrical cross section of the rings; this coefficient is obtained from a correlation from Hoerner [19]
(Fig. 12, p. 3-9)15 . The total force Ftot applied on the flow is a sum of the forces shown in eq. 3.31:

F = ΣFi + Fplate
tot
(3.31)
F
ring,i = CD,i Ai pdin,i , Fplate = CD,plate Aplate 12 ρ5 v52 ,
.
1
with pdin,i = 2 ρi vi2 ; subscript i refers to each ring of the Koppers harp. The flow properties (kine-
matic viscosity νi or the density ρi ) are evaluated at the radius of each ring.
The flow uniformization depends on the velocity profile v5 (r) of the streams at S5 (illustrated in
Fig. 3.6) and can be quantified by means of an uniformization factor ξ that expresses the departure from
a uniform velocity v̄5 defined in eq. 3.33. The velocity profile downstream of the harp structure, at S5B
is:
v5B (r) = ξ v5 (r) + (1 − ξ)v̄5 , (3.32)
ṁ5
v̄5 = ´ . (3.33)
ρ dA
A 5
15 Based vi dring
on the Reynolds number defined with the characteristic length dring , Re dring ,i = vi
.
49
The expression 3.32 applies for the primary, secondary and moved flows. The profiles defining the
intermediate regions (between the primary and secondary potential cores and between the secondary and
the entrained flows) are modeled as in station S5 (similar to equation set 3.28). For each longitudinal line,
the density is assumed constant from S5 to S5B : ρ5B (r) = ρ5 (r). The mean density is ρ̄5B = ρ̄5 = ṁ Q5 .
5
For ξ = 0 (full uniformization) the velocity after the harp v5B (r) degenerates into the constant term v̄5 ;
when ξ = 1 (no uniformization), the velocity v5B (r) takes exactly the shape of v5 (r). Applying a balance
of forces and momentum between station S5 and S5B , expanding v5B (r) and combining terms, it comes
(p′5 − p′5B ) Aaug − |Ftot | = −M Mx,5 + M Mx,5B = (ξ 2 − 1) [M Mx,5 − v̄5 ṁ5 ] , (3.34)
Applying the energy equation along a streamline of the primary flow, of the secondary flow and of
the moved flow, assuming local energy dissipation related with the drag force of the corresponding harp
rings, the velocity after the harp in the primary flow region can be obtained, in the secondary flow region
and in the moved flow region. These three velocities vj,5B (subscript j denotes each flow region: p, s or
m) enable an approximate reconstruction of the whole v5B (r) profile at stationS5B .
" 2 #
1 2 v̄5 CD Aring,j
(p′5 − p′5B ) = ρvj,5B ξ + (1 − ξ) + − 1 + ǫj , (3.35)
2 vj,5B Aj,5
where Aring,j is the frontal area of the rings facing each flow region (p, s or m), Aj,5 is the cross-
sectional area occupied by each of these flow regions and v̄5 is the constant velocity term in expression 3.32,
which is a peculiar average velocity, in the sense of expression 3.33. An error term ǫj has been supple-
mented to each Bernoulli equation since, together with eq. 3.34, they constitute an overdetermined set
of equations. The TCC model shows that the dynamic pressure term is one order of magnitude greater
than the the remaining terms; the error term ǫj is thus (relatively) small.
The pressure difference term (p′5 − p′5B ) can be eliminated in the set of equations and ζ can be
evaluated to minimize the square-norm error
E 2 (ξ) = ǫ2p + ǫ2s + ǫ2m . (3.36)
Expanding ǫp , ǫs and ǫm from equations 3.35, E 2 (ξ) is a second order polynomial of ξ: E 2 (ξ) =
aξ 2 + bξ + c. The minimum square-norm error occurs for the ξ value:
d b
E2 = 0 → ξ = − .

(3.37)
dξ 2a
For all possible values of the variables, the minimum occurs in the range 0 < ξ < 1. The a and b
coefficients of E 2 (ξ) can be found by evaluating E 2 (ξ) at three particular values of ξ (denoted ξ1 , ξ2 and
ξ3 ):

a = E 2 (ξ1 ) (ξ2 −ξ3 )+E 2 (ξ2 ) (ξ1 −ξ3 )+E 2 (ξ3 ) (ξ1 −ξ2 )
(ξ2 −ξ3 ) (ξ1 −ξ3 ) (ξ1 −ξ2 )
(3.38)
b = [E (ξ2 )−E (ξ3 )] ξ1 +[E (ξ3 )−E (ξ1 )] ξ2 +[E (ξ1 )−E (ξ2 )] ξ3
2 2 2 2 2 2 2 2 2
(ξ2 −ξ3 ) (ξ1 −ξ3 ) (ξ1 −ξ2 ) .
Similarly, the temperature after the harp structure is defined as:
50
∗
T5B (r) = ξT T5 (r) + (1 − ξT )T5B (r) (0 < r < Raug ), (3.39)
where ξT is an uniformization factor of the temperature, not necessarily equal to ξ, defined to give
∗
some flexibility to the computational model. The value T5B is the one that makes the energy flow rate of
station S5B (evaluated with temperature T5B (r) from eq. 3.39) verify the energy conservation between
stations S5 and S5B :
3
v5B
ˆ ˆ
EF5B = (cp,5B ρ5B T5B dA) v5B dA + ρ5B dA = EF5 , (3.40)
A A 2
from where
´ ´ v3
∗ EF5 − ξT c ρ v T
A p,5B ´5B 5B 5
dA − A ρ5B 5B
2 dA
T5B = , ∀ξT . (3.41)
(1 − ξT ) A cp,5B ρ5B v5B dA
where EF5B (in W) is the energy flow rate at station S5B , equal to the energy flow rate at the aug-
menter inlet station S5 . The temperature profile T5B (r) can be used to compute the average temperature
at the diffusor’s entrance section, T6 , and from this, the exhaust temperatures T7A and T8 . If however
the flow uniformization is not fully attained at S6 , the energy balance equation provides the most correct
way to estimate those temperatures (represented as Ti ):
(QE5 /M F5 ) − v 2i /2)
Ti = . (3.42)
cp,6
This work also analyzes the effects of completely removing the harp structure. It is recalled that the
velocity at S5B is distinct from the one at S5 , since not only there is some pressure recovery due to the
diffusion from the flow with high momentum to the remaining flows but only there is some pressure loss
due to the harp rings presence; the velocity at S5B may be obtained as
ṁ5B
v5B = ´ Raug . (3.43)
0
ρ5B r dr
3.7 Augmenter Model

The flow streams entering the augmenter would attain a velocity of v̄6 if they were fully mixed. How-
ever, the pressure recovery along the augmenter is limited by its length. As suggested by Papamoschou,
if the primary potential core length may be Lp /Dp = 13.3 − 15.0, it can be concluded that the primary
potential core ends at half of the augmenter’s length for the -3C engine. In fact, the augmenter’s exit
normalized by the engine primary diameter isLS6 /Dp = 26 and 18 for the -3C and -C2 engine respec-
tively. Once the augmenter exit section is rather close to the engine exhaust, the flow uniformization is
quite small.
Being ζ a factor associated to the lack of flow uniformity, the flow velocity at S6 may be defined as
in eq. 3.44.. It seems plausible to model the flow variables at the end of a short augmenter (station S6 )
as a weighted average between a mean value (ρ̄ and v̄) and the value at the previous station (ρ5B (r) and
v5B (r)) assuming the augmenter starts immediately downstream of the harp:

ρ6 (r) = ζ ρ̄ + (1 − ζ) ρ5B (r) v6 (r) = µ ζ v̄ + (1 − ζ) v5B (r) (0 < r < Raug ). (3.44)
Factor 0 ≤ ζ ≤ 1 represents an uniformization rate,ζ = 0 denotes the total absence of diffusion and
ζ = 1 denotes fully uniformized profiles. It is assumed that ζ affects the density profile the same way as
51
the velocity profile, considering a turbulent Prandtl number close to 1. Factor µ ≈ 1 introduces a small
correction to preserve the mass flow rate along the augmenter.
Defining
1 ṁ
ˆ
ρ̄ = ρ5B dA and v̄ = ´ . (3.45)
A A ρ
A 5B
dA
The cross-sectional area of the augmenter will be denoted by A (= Aaug ). Expanding ρ6 (r) v6 (r), the
mass flow rate in station S6 , it comes
ˆ ˆ ˆ
ṁ6 = µ{ ζ 2 ρ̄ v̄ A − ρ̄ v5B dA − v̄ ρ5B dA + ρ5B v5B dA
A A A
ˆ ˆ ˆ ˆ
+ζ ρ̄ v5B dA + v̄ ρ5B dA − 2 ρ5B v5B dA + ρ5B v5B dA }. (3.46)
A A A A
The average quantities ρ̄ and v̄ are uniform throughout the cross section A and could be taken out
of the integrals. One can also substitute their values according to the definition, and can further remark
´
that A ρ5B , v5B dA = ṁ is the mass flow rate across station S5B . After some manipulations,
´ ´
A ρ5B dA A v5B dA

ṁ6 = µ ζ (ζ − 1) ṁ − + ṁ . (3.47)
A
Since the mass flow rate must be conserved, ṁ6 = ṁ, µ can be isolated:
1
µ= ´ ´
ρ5B dA A v5B dA
. (3.48)
1 + ζ (ζ − 1) 1 − A
A ṁ
Factor
´
µ is´ zero in the limiting cases of ζ = 0 and ζ = 1 and is generally very small. Factor
ρ
A 5B
dA A v5B dA
1− A ṁ is usually a small negative number (about −0.1 or less); the highest value of
|ζ (ζ − 1)| is 0.25, therefore µ is in the range between 0.98 and ∼ 1. This way, the mass conservation law
is verified.
Similarly, the temperature at the end of the augmenter is defined as:
T6 (r) = ζT T6∗ + (1 − ζT )T5B (r) (0 < r < Raug ), (3.49)
where ζT is an uniformization factor of the temperature. ζT is calibrated by the measured temperature

T7A . The value T6∗ is the one that makes the energy flow rate of station S6 , evaluated with temperature
T6 (r) from eq. 3.49, verify the energy conservation between stations S5B and S6 :
v63
ˆ ˆ
EF6 = (cp,6 ρ6 T6 dA) v6 dA + ρ6 dA = EF5B (3.50)
A A 2
from where
´ ´ v63
EF5B − (1 − ζ) c ρ v T dA − ρ6 2 dA
T6∗ = ´ p,6 6 6 5B
A A
, ∀ζT , (3.51)
ζ A cp,6 ρ6 v6 dA
where EF (W) is the energy flow rate of the streams at the augmenter section. For a particular value
of ζT , the mean temperature T̄6 ≈ T7A ≈ T8 may be estimated from eq. 3.52. It was assumed it will
be nearly constant along the remaining exhaust section, i.e., from the diffusor to the exit station S8 .
The TCC predictions for the mean density of the exhaust, obtained as ρ̄6 = ṁ
Q6 are consistent with the
6
p6
ideal gas law, ρ̄6 = Rexh T̄6
, where the mean temperature at the end of the augmenter comes as the ratio
between the enthalpy flow rate and the heat capacity:
52
EF5 − A 21 ρ6 v63 dA
´ ´
c ρ v T dA
T̄6 = ´ p,6 6 6 6
A
= ´ . (3.52)
A cp,6 ρ6 v6 dA A cp,6 ρ6 v6 dA
The temperature T7A,T AP was measured near the diffuser wall, just before the sudden expansion of
station S7B . If the flow is not fully uniformized, it is expected that the mean temperature estimated
with the TCC is higher than the measured temperature: T̄7A,TCC & T7A,T AP . The temperatures along
the diffusor are nearly constant and the TCC model is able to match the moved flow temperature
T7A,m ≈ T6,m with the experimental one, T7A,T AP , by adjusting factor ζT .
53
54
Chapter 4
Computational Model Results
4.1 Discussion of the Simulation Results
This chapter presents the results obtained with the computational model for the current test cell
conditions. Table 4.1 summarizes the main experimental measurements and the TCC model estimates.
In each row, the measured values (above) may be compared with the estimated ones (below).
NOTE: Recently, it was remarked that, refining the convergence settings, for a number of cases, the
TCC model predictions agreed more closely with the field measurements, at the expense of a significant
increase in computational time. We decided not to update all the tables since the effort was not worth-
while, because the differences are already acceptable (lower than 10% for all results except for the -5B
engines, for the reasons explained later).
Chamber and Engine Inlet
Concerning the cell intake head loss coefficient, the TCC model predicts a mean value of k ∞−2 = 7.56
(from Table 2.3) against the k ∞−2 = 7.73 obtained from experimental data1 . The relative difference of
2.2% is small and the intake computations of the TCC model are thus validated. The maximum difference
between the estimates and the experimental measurements about the chamber’s gauge inlet pressure is
4.5% and 17.7% for the -3C and -5B engines, respectively. There is no experimental information about
this pressure for the -C2 engine, since this engine testings occur less often at TAP; the cross marks in
Table 4.1 signalize such missing data.
The absolute static pressure at the bellmouth throat section yields a 0.7% difference between the
measured values and the model estimates, which validates the compressible flow model between the cell
stations S2 and S3 .
Engine Exhaust
With a secondary potential core length set as xs/Dp = 4, the TCC model predicted a primary potential
core length of xp/Dp = 15 for the -3C engines (for a BPR of 5.28, which is close to the reference value
used by Papamoschou, BPR = 5.00).
1 The average value is k

∞−2 = 7.64 from the RH probe and k ∞−2 = 7.83 from the LH probe. The results shown at
Table 4.1 refer to the data obtained from the RH probe, which is slightly different from the LH probe.
55
Table 4.2: Engine exit velocities predicted with GasTurb and the TCC model.
As shown in Table 4.2, the primary and secondary velocities are similar to the ones predicted with
GasTurb simulations for most of the engines selected in this study. It was not possible to achieve
convergence for the -5B simulations with GasTurb software, and therefore the results displayed for this
engine (denoted in Table 4.2 with *) are not reliable. Concerning the computational model, there is some
uncertainty about the exit velocities due to the lack of input data for the TCC model (stagnation pressure
and temperature at the engine exit). Also, there was no sufficient input data (e.g. engine pressures at
the combustor’s inlet and engine exhaust nozzle) to simulate the –E1 engine with GasTurb.
A secondary-to-primary velocity ratio of about vs,4 /vp,4 = 0.7 was obtained for the -3C engine, which
matches the typical velocity ratio of commercial turbofan engines at takeoff power[29]. The TCC model
predicts a velocities ratio vs,4 /vp,4 of about 0.6 for the -5B engines tested at TAP2 . For the -C2 and -E1
engines, the velocity ratios estimated are about 0.7 and 0.8, respectively.
Engine Thrust
The thrust predictions of the TCC model are close to the measured values at TAP: (0.4 − 0.8)% for the
-3C engines, (3 − 9)% for the -C2 engines and 1% for the -E1 engines. These results are underestimated
for all cases except for the -C2 engines; for the -5B engines, they differ (17 − 22)% from the measured
values. This is because all engines have a different geometric configuration at the rear, namely the -5Bs
ones.
The present implementation of the TCC code considers only an engine rear with a centerbody cone
inside the primary flow, as in Fig. 4.1, images a), c) and d ). The configuration of the -5B engines is
significantly different, as seen in Fig. 4.1b). However, the purpose of this study is to predict the cell mass
flow rate and the CBR of larger engines and the thrust predictions stand mainly to verify the model’s
physical consistency. The discrepancy relative to the thrust estimates is also associated with the lack of
accuracy of the experimental data, namely the stagnation temperature T T 54 and the stagnation pressure
P T 54. Since the stagnation pressure and stagnation temperature are unavailable for the -5B , -C2 and
-E1 engines, these values were calibrated with the test run data from a -3C engine, dated of October
3rd of 2014, as discussed in the Section 3.5.1. The estimate for P T 54 may be inadequate to evaluate the
thrust (mainly the -5B engines).
Another remark is that the interaction between the engine jet and the boattail (presented in Fig. 2.11b)
is not considered in the TCC model.
Augmenter Flow Mixing

The model shows that the harp is responsible for an important part of the flow mixing. The square-
norm error of the energy balance around the harp is minimized for ξ = 0.35 for the -3C engine, meaning
2 For an -5B4/P test run at 8-Oct-2014 and an -5B1/3 testing at 23-Oct-14.
56
Figure 4.1: From left to right: engine rear configuration of the -3C, -5B, -C2 and -E1 engines.
that the harp produces 65% of velocity uniformization. The uniformization rate decreases with increasing
engine dimensions: 1 − ξ = 0.48 and 1 − ξ = 0.44 for the -C2 and -E1 engines, respectively. It can be
seen that the harp mixing capability becomes smaller when larger engines are tested: the harp produces
only 44% of velocity uniformization for the -E1 engine.
The second stage of the velocity uniformization occurs along the augmenter tube. The augmenter
uniformization rate was set ζ = 0.70 to fit the measured flow rate ṁcell and the static pressure p′7A . An
augmenter uniformization rate of ζ = 0.70 shows that the flow is not fully developed at the end of the
augmenter. A slightly higher pressure recovery could be obtained with a longer augmenter. Also along
the augmenter tube and after the harp’s section, the temperature uniformization rate is estimated to be
ζT = 0.70 to fit the experimental data (T7A ). A value of ζT lesser than 1 shows that the thermal diffusion
is still occurring at the exit of the augmenter. The kinetic diffusion is close to the thermal diffusion
(turbulent Prandtl number Pr turb ≈ 1), as expected3 .
At the augmenter’s entrance, the peripheral flow ṁper occupies an annular cross-sectional area limited
by an inner radius Ra and an outer radius Raug ; for a given engine jet (secondary and primary flow), a
wider augmenter tube provides more area for the peripheral flow and also a better augmenter efficiency,
i.e., a higher pressure recovery since the reduction in momentum flow rate between the inlet and exit
sections is greater. A higher pressure recovery along the augmenter reduces p′5 and therefore increases
the cell mass flow rate.
The static pressure increases at the harp’s section due to a loss in momentum; but the harp also
introduces a head loss. For larger engines the net pressure rise is smaller. For the -3C engine, the
relative pressure p′5B (after the harp) is 46% higher than p5 ’ (before the harp); for the -E1 engine, that
difference is 5% only. This reduces the amount of peripheral air flow pumped by the augmenter and
increases the risk of vortex formation when testing -E1 engines. A suggestion would be to revise the
number of harp rings to increase the peripheral flow rate.
Predicted CBRs
As Table 4.3 shows, the CBRs predicted with the TCC model are close to the ones obtained from
the experimental data collected at TAP. The deviations of the TCC model estimates from Hastings’
predictions [16] are about 1.8%, 20.5%, 21.3% and 8% for the -3C, -5B, -C2 and -E1 engines, respectively.
The deviation of the TCC model estimates from the experimental data of TAP is about 1.1% and 8%
for the -3C and -5B engines; there is no experimental information concerning the -C2 and -E1 engines
running in the TAP test cell.
3 The Prandtl number is defined as the viscous diffusion rate over the thermal diffusion rate, Pr = ν/α = c µ/k, where
p
ν is the kinematic viscosity and α is the thermal diffusivity in (m2 /s), cp (J/(kg.K)) is the specific heat , µ is the dynamic
viscosity (N s/m2 ) and k (W/(m K)) is the thermal conductivity.
57
Table 4.3: Cell bypass ratio CBR predicted from [16], from the TCC model and from TAP measurements.
Test Cell Exhaust

The head loss coefficients estimated with the TCC model provide a fair overall insight of the perfor-
mance of the exhaust section. A remark should be made about the fact that the head loss estimate for
the whole exhaust section does not consider the interaction between its separate parts (e.g. the sudden
expansion does not take into account the diffusor upstream) 4 . However, the overall head loss factor
between stations S7A and S8 , U7A−8 (see eq. 3.8), obtained with the measurements at TAP is close to
the one predicted by the TCC model (from eq. 3.9). The experimental and numerical values can be
compared in Table 4.1. The experimental head loss coefficients can be evaluated, given measured values
of p7A , T7A and ṁcell , from eq. 3.10 and the ideal gas equation of state, ρ7A = p7A /Rexh /T ,

p7A ρ7A 1 1
U7A−8 = ρ27A + − 2 , (4.1)
ṁ2cell 2 A27A A8
The cell mass flow rate predicted with the TCC model is within the uncertainty range obtained
with the pressure measurements at station S2 of TAP’s test cell. The TCC model estimates are highly
dependent on the engine mass flow rate ṁeng , namely the engine exit velocities and temperatures, the
cell mass flow rate and the thrust. There is some uncertainty about the measured cell mass flow rate,
because of the inaccuracy associated to the pressure readings. Moreover, this mass flow rate ṁeng is
computed at TAP through correlations that loose validity with time, due to changes in the test cell and
the instrumentation. Unfortunately, small errors on ṁeng may lead to significant errors in the momentum
balance used by the TCC model to evaluate the engine thrust. It was a curious finding to verify that the
measured values of thrust globally agreed well with the numerical predictions.
EF6 v2
The average static temperatures T6 = 89.6 ◦ C and T7A = ṁ6 cp,6 − 2 c7A
p,6
= 92.2 ◦ C (obtained from the
TCC model for the -3C engine) differ from each other due to a change in the velocity from station S6 to
S7A , v6 ≈ 2 v7A . The exhaust temperature predicted by the numerical model is T8 ≈ T7A = 92.2 ◦ C while
the measured temperature is slightly smaller (T8,T AP = 80.6 ◦ C). One reason may be the fact that the
three thermocouple probes at station S8 loose some energy by conduction to the walls and by radiation
to the ambient air. Also, the takeoff steady-state regime is not actually attained, as Fig. 4.2 shows. A
higher temperature would be expected for a takeoff regime lasting more than the actual 3 minutes.
The moved-flow temperature predicted by the TCC model for the -3C engine at the end of the
augmenter, T6,m = 87.2 ◦ C, is higher than the temperature measured near the walls, T7A,T AP = 73 ◦ C
for similar reasons: there is some heat conduction to the steel support and the temperature is not yet
stabilized. After 3 minutes at takeoff, the measured temperature T7A,T AP would be probably higher
than the actual measured value. Since the temperature readings did not reach the steady-state takeoff
regime, the temperature parameter ζT was left equal to the momentum parameter ζ, assuming a turbulent
Prandtl number of Pr turb = 1.
4 [22], p. 549, para. 26.
58
Figure 4.2: Exponential trend of temperature T7A along time: T7A = T∞ − (T∞ − T0 ) exp [−β (t − t0 )],
where the fluid temperature near the probe is T∞ = 73.5 ◦ C, T0 = 70.2 ◦ C is the probe temperature at
t0 = 14 : 57 : 52 (hh:mm:ss), taken as the initial instant of time, t − t0 is the time since the initial instant
of time measured in s and β = 0.7 s−1 is a constant, characteristic of the response of this probe.
With the present diffusor, the test cell is able to provide enough air flow for the current engines being
tested. But the wedge inside the diffusor represents a problem to bigger engines, because it reduces
too much the pressure recovery along the diffusor. For this reason, the peripheral air flow is insufficient
for the -E1 engines and the risk of flow separation within the test chamber is high for those engines;
the Chapter 5 presents some suggestions that could enhance the test cell performance and broaden its
capability.
59
Table 4.1: Engine test runs selected from the ones obtained between 20/Aug/14 to 14/Apr/15, at TAP’s
test cell facility.
60
Chapter 5
Test Cell Upgrade
5.1 Summary of the Test Cell Study

As already illustrated in Fig. 2.7 at Section 2.4, the original design of the diffusor may have suffered
from unstable flow separation. To avoid low-frequency pressure oscillations induced by the unstable
flow regime, a wedge has been inserted. Reducing the divergence angle, brings the diffusor into the
no-stall region of Fig. 2.7 a). The wedge leads to an equivalent diffusor with a smaller divergence angle.
Idelchik [22] presents an estimate of head loss coefficients for a rectangular diffusor equivalent to this
one. Conical and plane diffusors were also analyzed and compared. Higher head loss coefficients are
expected for plane diffusors, as compared with rectangular and conical diffusors, for a given area ratio
and length. The TCC model considers a rectangular diffusor equivalent to TAP’s one to estimate the
head loss coefficient k dif .
The wedge allows the diffusor to operate with a lower exit-inlet area ratio n and this affects the
pressure recovery. The unsteady phenomena occurring at the diffusor (at 1972 with the JTD-9 engines)
was probably related with the diffusor divergence angle being too high. [28] considered that throttling
the flow after the diffusor could fix this problem, e.g. a tube basket as shown in Fig. 2.2a. Moreover, at
that time, this reference considered that the peripheral air flow was was too high: a peak flow velocity of
v = 17.1 m/s was measured with a portable annemometer 6 m ahead of the engine’s inlet, for a thrust
of 25000 lbf; [28] states that the flow velocity for a JTD-9 engine at takeoff power (45000 lbf) would be
bigger than is recommended for good test cell design.
Resonance occurred again at TAP in 1989, when the cf6-80C2 engines were tested. The previous
basket tube introduced an exagerated head loss at the higher air flow rates of the new engines. Threfore,
it was decided to re-install a diffusor suplemented with an inside wedge to prevent flow instability. The
concrete wedge inside the diffusor showed signs of deterioration due to exposure to high-temperature
gases. Probably, this occurred mainly during acceleration regimes. Also, the relatively short length
of the augmenter1 could not assure a sufficient temperature uniformization. For both these reasons, a
Koppers harp was installed. The performance of the augmenter dropped as a consequence, since there is
less velocity uniformization to produce after the harp.
Regarding the -E1 engines, [20] pointed out the possible need of a wider augmenter tube. Nevertheless,
this reference considers that much is contingent on the engine exhaust proximity to the augmenter
entrance plane (see Section 1.2.7). Indeed, TAP augmenter’s diameter is about Daug ≈ 3.4 m and foreign
test cells in which the -E1 is tested have augmenter diameters in the range of 4.3 to 4.5 m. However,
1 According to Quinn [32], the optimal length-to-diameter ratio of the augmenter is about L
aug /Daug ≈ 8.0. The
length-to-diameter ratio of the TAP’s augmenter is about 4.0, probably insufficient to ensure complete flow mixing for large
furbofan engines.
61
this chapter summarizes other solutions that could be implemented, to allow large engine testings such
as the -E1.
5.2 Upgrade Study

Once validated (k ∞−2 and U 7A−8 ) and calibrated (ζ and ζT ) the parameters of the computational
model (see Table 3.1b), there is only one chance this test cell can test the cf6-80E1 : if all the following
parameters are checked:



 Ra < Raug


v
3,per > (v3,per )min

v

 m,5 < (vm,5 )max


0.75 < CBR < CBR

,
max
The first relation, quite intuitive, is always checked in the TCC model. The second relation refers to
the fact that there is a pressure gradient along the test chamber’s walls, and at an upstream section of
the engine’s inlet; this should not be too much high, meaning that an extreme adverse pressure gradient
may lead to fow separation. The third relation recalls that a high moved flow velocity will reduce the
engine’s thrust. The CBR’s relation offer a guideline that quantifies the test cell performance as a whole.
The TCC model allows to foresee whether the requirements stated above will be verified for the
current configuration of the test cell. Whenever those relations proved to fail, new solutions were sought.
In that sense, some test cell structures were virtually changed to analyze how much the TCC model’s
output parameters are sensitive to test cell variations. Therefore, configurations distinct from the current
one could be studied. The scenarios presented in this section refer to specific configurations capable of
running cf6-80E1 engines, such as an exhaust system with e.g. a longer augmenter tube with the same
diameter; a wider augmenter; a different engine-to-augmenter distance; the removal of some harp rings;
an exhaust room without grids and with turning vanes installed and a wider intake cross-section.
Inlet Ramps
The flow streamlines are deflected downstream of the engine bellmouth intake, once the available
cross-sectional area increases. The formation of vortices is a risk due to the flow deceleration associated
to this. The shape of the streamlines can be studied using the potential flow theory. The flow can be
represented by the sum of a uniform unbounded flow with an infinite number of sink-singularities. The
x-axis would be aligned with the incoming flow with the origin at the bellmouth focus. The sink-images
would be located at the corners of a rectangular mesh extending to infinity: infinite sinks repeating
themselves in the horizontal direction, with a step hz = 8.484 m, equal to the width of the test cell;
infinite sinks repeat themselves in the vertical direction, with a step hy = 9.760 m, equal to the height of
the test cell. For the purpose of such analysis, besides of the central sink repesenting the engine intake,
a small number of sinks is enough to represent the flow. The sink strength corresponds to an engine
volume flow rate of 879 m 3 /s and the uniform flow can be adjusted to model an incoming flow with
vx = 14.5 m/s, which is the velocity predicted for the -E1 engine.
The most important simplifications consist on assuming incompressible flow overall and replacing the
bellmouth inlet by a single sink point, located at its focus. Since one we are mainly concerned with
flow separations that might arise over the solid wall, the compressibility effects (that occur only near
the engine inlet) do not have a significant impact in the velocity and pressure near the walls of the
chamber. Furthermore, from the viewpoint of the peripheral flow, the exact shape of that flow entering
62
the bellmouth is not important and it can be modeled conveniently by a singularity located at its focus.
This way, a three-dimensional model help to understand how the velocity (and thus the pressure) changes
along the test chamber’s walls, from the screens of the inlet station S2 to the engine entrance station S3 .
Ramp structures can be designed to avoid flow separation by reducing the longitudinal pressure gradient
at the walls of the chamber. The pressure gradient will be approximately null if the ramps follow the
curvature of the streamlines of the flow entering the bellmouth.
(a) Inlet ramp structures. (b) Vortex formation (up). Vortex forma-
tion avoidance with inlet ramps (bottom).
Figure 5.1: Inlet ramp structures (images from [11]).
The installation of the inlet ramps shown in Fig. 5.1 along the four sides of the chamber (walls, ceiling
and floor), introduce a partial blockage of the flow area (that can go to from 30 % to 50 %) and accelerate
the peripheral flow thus helping to avoid the flow decelerating too much. The walls of the TAP’s test
cell already reduce the cross-sectional area by 34%, at un upstream section of the engine inlet. When
testing larger engines, an even more reduced cross-section undoubtedly helps to prevent flow separation
to occur between the chamber’s inlet and the engine intake (stations S2 − S3 ).
The great advantage on using such ramps is that the velocity at the augmenter’s inlet will not rise
as it would with other test cell configurations, e.g. the ones discussed later. The ramps establish flow
separation at a specific location near the walls (away of the engine inlet). The ramps used in [10] are
located adjacent to the engine inlet, with the upstream faces at about a 45° angle (see Fig. 5.1a); this
originates however a vortex higher than the ramp itself, downstream of the ramp. To obviate this effect,
the ramps can be continued longitudinally by a horizontal plate. With a test cell chamber width W and
a test cell chamber height H, [10] states that the installation of inlet ramps should occur at a location
less than 0.1H or 0.1W relative to the engine intake [10].
63
A Different Augmenter Tube
The flow rate admitted at the test cell varies with the pumping efficiency of the augmenter and the
diffusor. A too short augmenter provides an insufficient pressure recovery and the mass flow rate through
the plant is relatively small. A longer augmenter tube helps to achieve a better flow uniformization. A
wider augmenter tube increases the augmenter’s performance for any engine: it helps to obtain bigger
peripheral air flows - which help avoid vortices forming within the chamber - and originates less uniform
velocity profiles at the inlet of the augmenter; this increases the pressure recovery along the augmenter.
If a wider augmenter is installed, a wider diffusor will be needed, but its area ratio may be kept the
same. If the available length of the cell exhaust section is mantained, the diffusor will thus be shorter,
however a new diffusor with circular arc walls can boost its performance2 . Resonance at the exhaust
stack is related to the length of that exhaust room, Lstack , but if the diffusor operates correctly, this will
never be a problem. However, there are other (and more simple) ways available to raise the cell air flow,
as will be seen.
A Bigger Engine-to-Augmenter Distance

The distance between the engine exit and the augmenter, L4B−aug , for the cf6-80E1 at TAP should
assure the most even static pressure distribution between an upstream section of the engine inlet and the
engine’s rear: p′f orward ≈ p′rear . The pressure at the augmenter inlet (station S5 ) is defined by the moved
air flow; that pressure can be known by applying Bernoulli equation, since the flow is incompressible at
this section of the test cell. The close the engine is to station S5 , the lower the pressure is at the engine’s
rear, thus affecting its thrust significantly. A higher L4B−aug minimizes this effect and leads to a small
thrust correction factor. Also, the entrained air flow is higher for high engine-to-augmenter distances
L4B−aug ; increasing this distance may be a solution when low CBRs are expected in test cells.
This distance L4B−aug also affects the end of the potential cores. The TCC model considered the
same L4B−aug for the -E1 engine as for the -C2 engine, since the characteristics of the jet flow (exit
velocities, pressures and temperatures) are similar. An increase in this distance leads to an earlier fading
(by diffusion) of the primary and the secondary potential cores. The case A schematized in Fig. 3.6
represents the velocity profile along the augmenter for a limit situation where the secondary potential
core of the jet ends before entering the augmenter section, as a result of increasing L4B−aug . However,
this is not a practical solution.
An interesting solution is to break down the outer wall of the augmenter inlet section; the plant view
of such section is presented in Fig. C.2. The flow velocity may be defined as v = Q/A = Q/(πr2 ), where r
is the radial distance from the engine to the focus located at the centerline of the augmenter’s inlet plane.
The dinamic pressure at the rear of the engine may be considerably reduced with small increments on
its distance to the augmenter’s inlet, because p′din = f (r−4 ). This measure allows to lessen the impact
of the augmenter on the thust measurements and also to increase the air flow entrainment ratio.
Harp Rings Removal - a Parametric Study

The temperatures at the exhaust section (S5 − S8 ) are quite high (above 70 ◦ C for the -3C engines).
However, during an engine acceleration, these temperatures can be even higher. In the case of TAP, this
has severe consequences on the concrete wedge placed in the diffusor: it deteriorates along time. During
such transient regime (engine acelerations), there is no enough peripheral air flow to allow the dilution of
the hot exhaust gases leaving the engine. However, flow mixing can be improved by means of a Kopper
harp, as Fig. 5.2a schematizes.
2 [22], p. 264 para. 69.
64
(a) Harp structure: rings effectively installed and (b) Harp rings removal with -3C engines
rings already removed at 1989 (adapted from [24]).
(c) Harp rings removal with -C2 engines. (d) Harp rings removal with -E1 engines.
Figure 5.2: Effect of harp rings removal on the cell mass flow rate and CBR.
Fig. 5.2 shows the effect of removing some harp rings for -3C , -C2 and -E1 engines. The abcissa of
this figure represents the number of rings removed beginning from the outer ones. Case 0 denotes the
actual configuration of the harp; case 1 denotes one ring less, case 2 denotes two rings less and so forth;
case 10 refers to the complete removal of the 9 rings of the harp. From the periphery to the center of
the axis, the rings currently installed at TAP are referred sequentially as #15, #12, #9, #7, #5, #4,
#3, #2 and #1 (see Fig. 5.2a). It can be concluded that the outer three rings are always ineffective
(cases 0 through 3); for -3C engines, rings #5 and #6 are also ineffective (cases 4 and 5). In summary,
the outer harp rings do not contribute much to uniformize the flow. Indeed, as shown in Fig. 5.2, they
introduce a head loss that balances the pressure gain due to the reduction of the momentum flow rate.
The pressure rise along the augmenter, responsible for pumping the mass flow rate ṁcell through the
plant, is maximum when the harp is fully removed.
There is no experimental data to support the TCC model of the augmenter without harp, however
the uniformization rate after the harp is mainly a characteristic of the augmenter and its value is not
expected to change significantly. The value of ζ = 0.7 was kept from the baseline model described in
Chapter 3. The cell bypass ratio CBR is a direct function of the cell mass flow rate; as this test cell is
short of bypass flow rate, taking off at least the five outer rings of the harp increases the overal flow and
may consent the testing of larger engines. If the five outer rings are removed, the CBR always falls into
the interval 0.75 < CBR < 2.0, typical of test cells. The removal of all the rings leads to CBRs in excess
of 2.0 for small engines like the -3C ones. However, this should not be a concern, since the velocities
predicted at the chamber’s inlet (v2 ≈ 11 m/s) are still small, even smaller than the current velocity
65
(v2 ≈ 13 m/s) predicted for the -C2 engines presently tested at TAP.
Table 5.1: Parametric study results using the TCC model.
From Table 5.1, it is concluded that the wedge removal or the removal of the five outer rings have
a similar effect on the increase of the mass flow rate and consequently the CBR: any of these solutions
could be sufficient to allow an -E1 engine test run. The minimum CBR allowable at the test cell is
highlighted in bold at Table 5.1. We remark that removing the rings #15, #12, #9, #7 and #5 is safer
to achieve at least a cell bypass ratio of CBR = 0.75; this may be the most attractive solution to TAP.
Reduced Resistance Diffusor

Different internal arrangements (as presented in Fig. 6.1a) may be selected for TAP’s diffusor: dividing
splitters and guide vanes (baffles) can reduce its head loss coefficient by 35%. For such arrangements,
it follows k dif,new ≈ 0.65 k dif = 0.195 from Idelchik’s [22] correlations (Diagrams 5-12 and 5-13). This
value is close to the minimum value presented in Table B.2 for the rectangular diffusor with wedge). The
exhaust pressure p′7A and the cell mass flow rate ṁcell increase by 24.5% and 7.2%, respectively, for the
-3C engine; it is predicted a slightly lower rise for the -E1 engine: 23.9% and 6.9% with respect to p′7A
and ṁcell . With the configuration being discussed, the CBR increases by 11.5% and 28.9% for the -3C
and -E1 engine.
The number of flow splitters N of the diffusor is chosen depending on the divergence angle α; there-
fore, a number of N = 2.7 ≈ 3 splitters could be installed at TAP, placed with strictly equal a′0 spacings
(distance between the splitters); the distance between splitters at the diffusor exit a′1 should be approxi-
mately the same3 . The ratio a′0 /a0 = a′1 /a1 = 1/(N + 1) = 0.25 comes for a diffusor with α > 30º . As
shown in Fig. 6.1a a), the splitters extend before and after parallel to the diffusor’s axis in such a way
that l > 0.1 a0 and l > 0.1 a1 (in m).
As shown in Fig. 6.1a, Idelchik [22] suggests the guide vanes to be placed ahead of the diffusor’s
entrance section and behind it up to some extent; they can be made of sheet metal having constant
curvature and a constant chord of about c = (20 − 25%) Daug and therfore 0.67 m < c < 0.84 m. The
channels between the walls and those vanes contract, as a rule.
3 The schemes in Fig. 6.1a illustrate these solutions.
66
The TCC model computes the head loss coefficient k dif of this configuration (diffuser with guide
vanes or splitters, no wedge installed); since the wedge is removed, the screens placed at its rear (named
as screens of type c in Section 2.5) are also suppressed in the computational code.
Exhaust Grids and Turning Vanes

Some velocity uniformization at the exhaust section can be obtained by means of the grids placed
at station S7B : these grids direct the flow more evenly through the middle and the peripheries of the
exhaust room. Some of the static pressure recovery through the augmenter and the diffusor is lost at
these grids. If not exagerated, this dissipation of energy is good for flow throttling, but loosing too much
energy at the exhaust is unwanted, thus the placement of grids with higher free-area meshes are a benefit
for testing larger engines. TAP may be interested in evolving its test cell up to meet engine air flow rates
even higher than those for the cf6-80E1, e.g. the Trent 1000 engines that empower the A350-800/-900
aircrafts (see Table D.3 for more details). If higher mass flow rates are a need, the current exhaust grids
may be partially removed. Since the exhaust head loss is lowered, a higher mass flow rate can be pumped
through the augmenter. However, if the cell mass flow rate is too high, the moved flow velocity may
be excessive and may undesirably lead to more significant thrust corrections. An even more interesting
solution would be to build a system that could adjust the head loss to the engine being tested.
Besides supplying air flow, the augmenter tube supplies turbulent energy to the base of the test cell
exhaust stack. That energy is converted to noise at infrasound wave lengths through a resonant process
and the resonance modes may be expressed as
2n + 1
f =c
4L
where n = 0, 1, 2..., c is the sound speed and L = 15 m is the characteristic length of the exhaust
section. Tab. 5.2 summarizes the first resonant frequencies predicted for the engines analyzed in this
study; the speed of sound is based on the exit temperature T8 , obtained from recent measurements at
TAP.
Table 5.2: Exhaust stack resonant modes for the engines analyzed in this study.
Vertical exhaust stacks reduce the size of the test cell plant; also, the surroundings of the test cell are
less affected by noise and high temperatures. However, the flow at the exhaust stack should turn along
the baffles with the less local flow separations possible; fairings at the endtips of the baffles are installed
in this sense, at TAP, but the head losses between stations S7C and S7E represent about 68 % of the
exhaust head loss downstream of the diffusor (for the -3C and -E1 engines). Turning vanes installed
similarly to the intake section of the test cell can reduce the head loss at the exhaust stack. For a head
loss coefficient similar to the intake guide vanes, that coefficient could be reduced in more than 30 times,
P1 2 1 2
resulting on a new head loss of ∆p = 2 ρvi = 5 × 2 ρv = 5 × 500 = 2500 Pa, where v = 50 m/s is the
exhaust velocity (for the -3C engine).
67
Guide vanes placed along the exhaust stack avoid flow separation along the baffles that deflect the
exhaust gases. The dimensions of any vortex formed along the exhaust section (e.g. at the diffusor), can
be reduced. Resonance phenomena can be avoided this way. This can bring a great improvement on the
test cell pumping.
68
Chapter 6
Conclusions
Historical information about TAP’s test cell since 1972 until its current configuration has been col-
lected. This survey was useful to understand the thresholds of the test cell capacity. This background
information was important to develop the numerical model TCC of the whole plant, with which it was
possible to identify a set of practical provisions that would make its upgrade achievable. Table D.3
summarizes some engines that could be tested at TAP with the adequate cell adjustments.
If the geometric dimensions were not a constraint, flow separation and nonuniformity of the velocity
profile could be easily overcome1. The original diffusor, without the wedge addition, had probably an
exagerated divergence angle inducing strong low-frequency aerodynamic resonance. An upgrade design
of the diffusor, already outlined in the previous section, could obviate this inconvenience.
With baffles or flow splitters installed, a diffusor with a greater cross section could have about the
same length of the current diffusor and operate free of flow separation. Its present area ratio could be
preserved, providing a diffusor efficiency (performance) of CpT AP = 45%. For the available length of the
exhaust section, a wider diffusor with guide vanes could be made even shorter than the current diffusor.
A new diffusor with circular walls can improve its performance2 .
We have verified that the current diffusor’s divergence angle (wedge installed) satisfies the optimal
value (see Table. B.2) for an open diffusor without guide vanes, if the flow was uniform; but indeed
the flow at the end of the augmenter is not uniform, not even fully developed. Flow splitters or baffles
(see Fig. 6.1a) could be installed to avoid the flow separation in the diffusor. If the wedge is kept, its
deterioration along time will always be a concern, since there are always transient processes along any
engine test run; such deterioration is expected to be even higher with the -E1 engine testings, because
the jet engine exhaust temperatures are higher (T6 = 96 ◦ C for the -E1 engine versus T6 = 81 ◦ C for the
-C2 engine).
The diffusor is one of the critical sections of the test cell either because of flow separation (at its
original configuration) either because it limits the test cell air flow too much (with the wedge currently
installed). Also, the augmenter length is too short to obtain a fully uniformized flow velocity profile.
Increasing its length would be a benefit for testing larger engines where CBRs < 0.75 are expected, such
as the -E1 . The computational model shows that the augmenter and the diffusor would benefit from a
larger cross section. However, some other solutions have been considered to run larger engines at TAP,
with time and investment economy.
1 [22], Chapter 5, para. 49
2 [22], p. 264 para. 69.
69
(a) a) dividing splitters; b) baffles; c) inner fairing ([22]). (b) Separation for different dif-
fusor area ratios and diver-
gence angles ([22], p. 242).
Figure 6.1: Separation in diffusors.
The present research verified that, as engine size increases, the amount of peripheral air flow becomes
critical for the same size and geometry of the test cell. With the current test cell configuration, a test
run of the cf6-80E1 would yield a very small CBR value, as can be deduced from the Fig. 1.2a presented
in Section 1.2.4. Less peripheral air flow being pumped may result in vortices formed upstream of the
engine intake, that may lead to engine surge. Therefore, increasing the CBR seems a first solution to the
present test cell limitations. Nevertheless, other solutions were found, such as placing ramps around the
walls of the chamber’s station S3 .
According to this approach, a smaller CBR would be acceptable, with the advantage that the pressure
downstream of the engine and at the augmenter’s inlet would be closer to atmospheric conditions. This
change would reduce the amount of correction needed when measuring the engine thrust.
A suggestion is to eliminate the wall in the plane of the current inlet section of the augmenter to
increase the distance between the engine and the augmenter. This change would improve the thrust mea-
sured (less correction needed) and increase the CBR without lowering the inlet pressure of the augmenter.
Furthermore, separating the engine from the augmenter’s inlet would compensate an hypothetical need
of a greater CBR.
This study shows that the augmenter does not fully uniformize the velocity and temperature profiles
along its length, not even with a Koppers harp installed. With an uniformization rate of ζ = 0.7, the TCC
model matches the measured ṁcell and p′7A . Fig. 6.2 shows the cell mass flow rate ṁcell , the augmenter
inlet and exit static pressures (p′5 and p′6 ) and the exhaust static pressure p′7A evolution as function of
the augmenter performance (represented by ζ). Accordingly, we may see from Fig. 6.2 that for ζ = 1
(full velocity uniformization), the static pressure p′7A and the cell air flow rate estimated are about 5.4%
and 1.6% higher than with ζ = 0.7. The TCC model is thus weakly dependent on ζ, as any model should
be. According to [32], full uniformization would be achieved approximately with a length-diameter ratio
of (L/D)aug ≈ 8 .
Despite the flow velocity at the chamber’s inlet station S2 being within acceptable ranges for the
engines currently tested at TAP, a bigger intake could correct eventual flow distortion occurring in a
larger engine test run. The original (higher) diffusor area ratio ndif = 3.54 was also studied with the
TCC model, i.e., for a diffusor without wedge and with baffles (or flow splitters) instead. Other solutions
were found, such as placing inlet ramps at the chamber’s station S3 , installing guide vanes along the
exhaust stack’s turnings or increasing the free-area of the grids at station S7B of the exhaust section.
However, experimental data with new exhaust configurations at TAP would be needed to calibrate the
model.
70
Figure 6.2: Pressure and mass flow rate as a function of the velocity uniformization ζ along the augmenter
(results obtained with the TCC model for the -3C engine, based on the test run data of 03-Oct-2014,
with the Koppers harp installed).
The parametric study considering the harp rings removal, presented in Section 5.2, allows to conclude
that from the CBR viewpoint, there is no great advantage on using a Koppers harp as currently installed,
because it negativelly affects the augmenter pumping performance [27]. With a Koppers harp, the large
scale turbulence is reduced and resonance avoided for a particular set of engines’ type, but this solution
could be revised to run larger engines.The flow in the diffusor would improve replacing the wedge by
a set of metallic guide vanes. The present wedge avoids a bi-stable flow separation (that would occur
with the wide-open geometry prior to the wedge) and reduces an otherwise excessive pressure recovery,
however this is an inefficient way of stabilizing the flow and does not provide an adjustable control of
the peripheral flow rate. The removal of the Koppers harp would allow the augmenter to pump a higher
peripheral flow as needed by the cf6-80E1 .
Another set of guide vanes installed at each curve of the exhaust stack would provide a better flow
stabilization and avoid resonance. Some test cell flexibility would allow to test a wider variety of engines.
The flow conditions for the -E1 engine were predicted for the current configuration of the test cell
and for other test cell configurations. Results of the simulations showed that removing the five outer
harp rings lead to a test cell bypass ratio acceptable (CBR ≥ 0.75) for the -E1 engine testings; the wedge
removal presented also a similar result. Inlet ramps close to the engine inlet station may be an alternative
solution. If the wedge is kept, other solutions for flow uniformization should however be implemented
additionally to prevent its deterioration; if flow mixing along the augmenter is a problem, using a second
air inlet at the augmenter’s entry section (see Fig. 6.3) is an option that could be preferred to constructing
a new test cell, and make it possible for the upgraded cell to test even more powerful turbofan engines
than the -E1. However, this option would probably need an augmenter with a wider cross-section.
The current exhaust system limits the capability of the test cell to pump higher mass flow rates
needed for engines with higher intake air flows, such as the ones that Table D.1 presents (right column).
71
Figure 6.3: Secondary air intake (adapted from [24]).
According to the thrust stand manufacturer, the test cell might be structurally able to withstand thrusts
of the order of 105 lbf (∼ 44500 daN), but aerodynamically it is not yet prepared to assure adequate flow
conditions for bigger engines requiring higher mass flow rates:
- the CBR was found quite low (below 0.7);
- the velocity of the peripheral flow near the engine inlet was also found to be low. From the chamber’s
inlet (station S2 ) to the engine’s inlet (station S3 ), the velocity is reduced by 53% and 75% for the -3C
and the -E1 engines, respectively. The pressure gradient is more adverse and the risk of flow separation
is higher for this last engine. The provisions suggested in this report evaluate the correct amount of
peripheral air flow that should be handled with the -E1 engines: ṁper = ṁcell − ṁeng ? 1560 − 880 =
680 kg/s.
6.1 Recommendations
The cell depression must be acceptable for any engine operation; a limit range between −735 Pa and
−1470 Pa is considered acceptable and thus there will not be any inconvenience for the cf6-80E1 engine
(a cell depression of about −1000 Pa is expected for the current test cell configuration, and of about
−1300 Pa with the harp removed). Starting from low power rates, if a test trial would be done for the
-E1 , the static pressure probes p′2 or even p′f orward and p′rear can indicate whether the cell depression
will fall within or out the bounds mentioned. The flow conditioning screen should only be revised (as
suggested in [20]) if an excessive pressure drop at the chamber was expected, thus affecting the thrust
measurements. This may happen for engines larger than the -E1s; for such engines, the mesh of the
chamber’s inlet screen (station S2 ) could be enlarged.
Air flow streamers enable visual flow monitoring at the test cell chamber; the ones installed at TAP’s
chamber walls revealed good flow conditions (stability) for the -3C and -5B engines. These should
however be monitored to check the flow behaviour when testing the -E1 engines.
Larger engines as the -E1 need larger cell airflows and exhaust larger amounts of hot gases. If the
peripheral air flow is small, the mean temperature of the exhaust is too high and may damage some of its
components. Special attention should be made about the temperatures attained at station S6 : there is a
sheet metal placed at the leading edge of the wedge to prolong its life by preventing the degradation of
the material. Enough secondary air flow rates should always be assured to reduce the high temperatures
occurring in the diffusor3 ; air flow streamers are a good indicator of flow reversal aft of the engine’s rear,
3 Temperature gradients here are cyclic and, with time, may lead to a quality loss of the components of the wedge
material. Even though concrete does not have a melting temperature, high temperatures make it “crumble”.
72
a situtation that can damage the structures installed in the test cell exhaust.
One of the greatest challenge about upgrading this test cell remains the correct operation of the
diffusor. With a diffusor without a wedge inside, the high temperatures experienced at the augmenter’s
exit section will no more constitute a problem. If however the wedge remains installed, maintenance
checkings to its integrity should be done regularly. Even if the present wedge avoids flow separation, it
reduces the diffusor’s performance. To outcome this disadvantage, guide vanes can be installed to direct
the flow and at the same time enabling a higher pressure recovery.
The exhaust room aft of the diffusor is also critical: there are large forces being applied and there
might be some flow separation at the exhaust stack’s baffles, thus reducing the cell mass flow rate; the
head loss may be lowered by replacing the grids with turning vanes; these help to make the exhaust flow
turn as it decelerates through the turnings of the exhaust stack section. If uncontrolled flow separation
along the exhaust is avoided, the infrasound problems already experienced at TAP cease to exist.
As for the engine hardware equipment, the engine cowlings, the exhaust nozzle and the bellmouth
specifically needed for the -E1 engine were assumed to be similar to the -C2 engine; the augmenter
distance to the exhaust nozzle’s discharge L4B−aug however is not necessarily the same for the -E1 engine.
If different, the adapter mechanism connecting the engine to the thrust stand will not be compatible with
the -C2 engines and a new one should be provided for the -E1 engine testings.
Summarizing, the improvement measures suggested in this work should be considered provided that
the CBR and the peripheral flow velocity are kept between the ranges specified. These two criterions
are distinct to one another and may lead to different scenarios (measures) that could be implemented to
allow TAP to test the -E1 engines.
6.2 Future Work

An acoustic study could evaluate the near field and far field noise levels associated with cf6-80E1
engine operation. Acoustic measurements could provide further information about the flow stability (the
main resonance aerodynamic frequencies) and foresee whether the test cell would acoustically operate
properly for the cf6-80E1 engine.
When testing larger engines, a reduced cross-sectional area usptream of the engine intake helps to
avoid flow separation from the chamber’s walls. The potential theory discussed in Section 5.2 could be
developed even further and implemented in the TCC model to evaluate the best cross-sectional area
upstream of the engine’s inlet and therefore to define the adequate sketch of the inlet ramps.
The use of CFD (Computational Fluid Dynamics) programmes may evaluate the best arrangement
of the guide vanes that could be placed in the diffusor or along the exhaust stack.
A computational model for this test cell has been developed for its current configuration and its
estimates could be compared with experimental data. The TCC model provides a useful tool to analyze
the impact of parametric changes of the test cell. Unfortunately, the flow conditions during a -C2
engine testing could not yet be measured. It would be of great value to obtain that data, to further
substantiate the accuracy of the TCC model. The engine test runs sampling is quite short. Obtaining
a more representative number of test runs (for the -3C and -C2 engines mainly) could also improve the
accuracy of the TCC model.
The TCC model can quantify the cell behavior when operating with -E1 engines, with P&W4168 and
even larger engines, such as the Trent-1000 engines. Table D.1 summarizes the TAP’s test cell facility
evolution and some provisions that could be implemented, looking forward to the near future of TAP’s
maintenance services fostering its competivity.
73
74
Bibliography
[1] R.A. Ahti, E. Bouchard, and M.A. Umney. Holding device for gas turbine rotor blades and machine
tool incorporating such a device, Jun. 2005. EP Patent App. EP 20040257658.
[2] A. Al-Alshaikh and al. An experimental and numerical investigation of the effect of aero gas turbine
test facility aspect ratio on thrust measurement. Phd thesis, Cranfield University, Cranfield, U.K.,
Aug. 2011.
[3] José Maria C. S. André. Hidrodinâmica - Apontamentos sobre Análise Dimensional. Pdf document,
2015.
[4] José Maria C. S. André. Mecânica dos Fluidos I - Apontamentos de Escoamentos Compressı́veis.
Pdf document, 2015.
[5] R.D. Archer and M. Saarlas. Introduction to Aerospace Propulsion. Electronic Publishing Services
Inc., Jan. 1996. ISBN: 0-13-120496-3.
[6] P.F. Ashwood. Operation and performance measurement on engine in sea level test facilities. pages
1–177, Mar. 1984. AGARD Lecture Series No. 307.
[7] A.S.; Incropera F.P. Bergman, T.L.; Lavine and D.P. Dewitt. Fundamentals of Heat and Mass
Transfer. John Wiley & Sons, 6th edition, 2007. ISBN: 0-470-05554-5 X.
[8] V. Brederode. Aerodinâmica Incompressı́vel: Fundamentos. IST Press, 1st edition, 2014. ISBN:
978-989-8481-32-0.
[9] Y. H. Choi and W.Y. Soh. Computational analysis of the flowfield of a twodimensional ejector
nozzle. In AIAA 90-1901 26th Joint Propulsion Conference, Orlando, FL, Jul. 1990. AIAA 90-1901.
[10] M. W. Roberts J. H. Muller G. L. Nikkanen J. P. Clark, T. A.; Peszko. Gas turbine engine test cell,
Mar 1994. US Patent 5293775 A.
[11] M.W.; Roberts J.H.; Muller G.L. Clark, T.A.; Peszko and J.P. Nikkanen. Gas turbine engine test
cell. United States Patent, Mar. 1994. US Patent 5293775.
[12] P. Coelho and M. Costa. Combustão. ORION, 1st edition, Oct. 2007. ISBN: 978-972-8620-10-3.
[13] R.J. Freuler and R.A. Dickman. Current techniques for jet engine test cell modelling. In 11th
AIAA/CEAS Aeroacoustics Conference, Cleveland, Ohio, Jun. 1982. Joint Propulsion Conference.
AIAA 82-1272.
[14] S. Gerlafingen. European wide flange beams.

www.stahl-gerlafingen.com/EN/LinkClick.aspx?fileticket=v8VZrPjv%2Fy4%3D&tabid=1429& language=de-CH,
Web link last accessed at 28-Nov.-14.
75
[15] P. Gullia, A.; Laskaridis and K.W. Ramsden. Ejector pump theory applied to gas turbine engine
performance inside indoor sea-level test cell-analytical and cfd study. In AIAA 25th Aerodynamic
Measurement Technology and Ground Testing Conference, San Francisco, California, Jun. 2006.
AIAA 2006-3152.
[16] R.R. Hastings. A Simulation of a Jet Engine Test Cell. Division of Mechanical Engineering, 1983.
Canada.
[17] H.; Jermy M.C. Ho, W.H.; Dumbleton. Effect of upstream velocity gradient on the formation of sink
vortices in a jet engine test cell. In IMECS, editor, Proceedings of the International MultiConference
of Engineers and Computer Scientists, volume 2, Hong Kong, Mar. 2008. ISBN: 978-988-17012-1-3.
[18] J.; Henry D. HO, W. H.; Mark. Formation of sink vortices in a jet engine test cell. Engineering
Letters, 16(3), Aug. 2008. EL 16-3-20.
[19] S.F. Hoerner. Fluid-Dynamic Drag - Practical Information on AERODYNAMIC DRAG and
HIDRODYNAMIC RESISTANCE. Hoerner, S.F., 1965.
[20] ASE Holdings. Aerosystems scope of work - analytical and acoustical study. Pdf document, property
of TAP Portugal, 2004.
[21] R.H.; Abdol-Hamid K.S. Hunter, C.A.; Thomas and S.P. Pao. Computational analysis of the flow and
acoustic effects of jet-pylon interaction. In 11th AIAA/CEAS Aeroacoustics Conference, Monterey,
California, May 2005. NASA Langley Research Center, USA.
[22] I.E. Idelchik. Handbook of Hydraulic Resistance. JAICO Publishing House, 3rd edition, 2008.
ISBN:81-7992-118-2.
[23] R. Jacques. Operation and performance measurement on engines in sea level test facilities, Mar.
1984. AGARD Lecture Series No.132.
[24] Arcelia Villanueva Jonathan Santiago. TAP Maintenance & Engineering - Test Cell Correlation for
the CF6-80C2 Model Engine. GE Aviation, Jun. 2013. GE property; all containing information is
confidential and under U.S. Government authorization.
[25] J.S.; Bellomy-D.C. Karamanlis, A.I.; Sokhey and T.C. Dunn. Theoretical and experimental investi-
gation of test cell aerodynamics for turbofan application. In AIAA/ASME/SAE/ASEE 22nd Joint
Propulsion Conference, Huntsville, Alabama, Jun. 1986. AIAA 86-1732.
[26] Jorge M. Duarte Leite. CFM56-3 Basic Engine - B737-300, May 1992. Property of TAP Portugal -
Formação Profissional.
[27] D.F Long. Jet engine test cell structure. Airo Systems Engineering, Inc., Mar. 2002. Application
number: EP 1995119399.
[28] Pratt & Whitney Aircraft Division of United Aircraft Corporation. JT9D stall problem - TAP
testcell. This document is property of TAP Portugal, Jun. 1972.
[29] D. Papamoschou. A new method for jet noise reduction in turbofan engines. In 41st Aerospace
Sciences Meeting & Exhibit, Reno, Nevada, Jan. 2004. AIAA 2003-1059.
[30] D. Papamoschou. New method for jet noise reduction in turbofan engines. AIAA - 92697-3975,
42(11), May 2004.
[31] G.A. Pauley. For use in an aircraft turbofan engine, Oct. 1993. US Patent 5.251.435.
76
[32] B. Quinn. Ejector performance at high temperature and pressure. volume 13, Wright-Patterson Air
Force Base, Ohio, Mar. 1976.
[33] Julio Antonio Rubio Ridaura. Correlation analysis between hpc blade chord and compressor effi-
ciency for the CFM56-3. Master’s thesis, Instituto Superior T’ecnico, Nov. 2014.
[34] D.M. Rudnitski. Performance derivation of turbojet and turbofans from tests in sea-level tests cells,
1984. AGARD-LS-132.
[35] The Engineering Toolbox. Dry air properties.

www.engineeringtoolbox.com/dry-air-properties-d 973.html,
Web link last accessed at 16-Dec.-14.
[36] P.P. Walsh and P. Fletcher. Gas Turbine Performance. Blackwell Science, 1998.
[37] Richard J. Freuler Werner Hoelmer. Test cell aero evaluation scale model test program, May 1989.
[38] F.M. White. Fluid Mechanics. McGraw-Hill, 4th edition.
[39] G.S.K. Wong and T.F.W. Embleton. Variation of specific heats and of specific heat ratio in air with
humidity. The Journal of the Acoustical Society of America, 76(2):555–559, 1984.
77
78
Attachments
79
Appendix A
General Details
A.1 ISA - International Standard Atmosphere

ISA is a model for calculations, simulations and analysis. It stands for International Standard Atmo-
sphere and – to be physically correct – it obeys to the equation of state (eq. A.4) and pressure increase
due to air gravity changing with z. It ignores the weather effects (and so, both convection and water
vapour and gas formed, that are present in the real atmosphere), thus consisting on a clear blue sky in
equilibrium conditions.
The physical properties ([8], pp. 661-663) at reference altitude h = 0 (average sea level) are presented
for temperature, sound speed, pressure, density, dynamic and kinematic viscosities: T0 = 288.15 K,
p0 = 101325 Pa, µ0 = 1.789 × 10−5 kg/(m.s), c0 = 340.294 m/s, ρ0 = 1.225 kg/m 3 ν0 = 1.4607 × 10−5
m2 /s.
As for different altitudes the properties presented do change; such evolution has been modeled in ISA
by sub-dividing the atmosphere in layers where variations obey to the following equations:
T = Tb + β(H − Hb ), (A.1)
p = pb [1 + β/Tb (H − Hb )](−g0 /βR) , (A.2)
for β 6= 0. The dynamic viscosity can be obtained from
h i
µ = βS T (3/2) /(T + S), (A.3)
being subscript b the values at the base (inferior limit) of each layer of the atmosphere and β =
dT /dH = −6.5 (K/m) the vertical temperature gradient for the 0 − 11 km atmosphere layer (the one at
which commercial planes operate and the one comprising TAP’s altitude); for that layer, the specific heat
ratio at constant pressure and volume γ = 1.401 remains unchanged. The dynamic viscosity is assumed
to follow Sutherland’s empiric law, with βS = 1.458 × 10−6 kg/(m.s.K 1/2 ) and the constant S = 110.4
K. H is the geopotential altitude1 . Finally, R is the ideal gas constant for air and equals R = 287.05287
J/(kg.K), needed whenever using the ideal gas equation for the altitude layers considered in ISA:
1 See [8], pp. 661-663 for details.
80
p = ρRT. (A.4)
A.2 Model Basic Principles
The following equation - known as the equation of continuity - states that mass flow rates within a
flow tube2 do always remain the same, this is, mass is conserved:
ṁin = ṁout ↔ ρAv = const. (A.5)
This is the basis for the computations done along the test cell with TCC model. When there are no
compressibility effects, the eq. A.5 is simply written based on the volume flow rate definition, Q = Av =
const. For incompressible flows the energy conservation equation states that the energy of a fluid element
is constant along a streamtube. Bernoulli’s equation derives from this and rules that, for ideal fluids:
1
p + ρv 2 + ρgh = const. (A.6)
2
where h is the coordinate along y direction, in m, relative to the reference plane (corresponding to a
geometric height), p is the static absolute pressure at the level chosen in Pa and v is the flow velocity
along the x coordinate, in m/s. This equation, relating these three quantities, is valid along a streamline
of a steady, incompressible frictionless flow.
The geometric height is the same for the chamber’s and exhaust rooms; the remaining test cell sections
(intake and exhaust stacks), it is is nearly constant. This way, Bernoulli equation may be written in terms
of the static pressure relative to the local hydrostatic pressure measured at the engine centerline height,
referred as p′bar . The same equation still holds for the intake and exhaust sections (S∞ − S2 and S5 − S8 ),
and from equation A.6 it follows that:
1 1 1 1 2
p′Si + ρvS2 i = p′Sj + ρvS2 j + ∆pSi −Sj = p′Sj + ρvS2 j + ρvref × KSi,j , (A.7)
2 2 2 2
where vref is a reference velocity, based on which the head loss coefficient KSi,j is obtained (from
experimental data). If the the changes in pressure due to differences of height were considered, the
eq. A.7 still holds provided the geopotential term is included. When no head losses are present, Bernoulli
equation is simply written as
1
p′ + ρv 2 = const. (A.8)
2
2 This section aims to present the mathematical and physical formulations that served as basis for the dissertation,
referring to its computational model. Explanations and demonstrations from which those arise are not intended to be
considered in much details. Whenever considered practical or meaningful, source references come along in footnotes.
81
A.3 Koppers Harp Geometry
When designed, the harp structure was thought to take a total of 15 rings. However, further studies,
developed in loco, lead to implement only 9 specific rings (rings number 6, 8, 10, 11, 13 and 14 ones were
not installed). The area facing the exhaust gas flow, this is, the projected area, due to all rings and the
support mechanism of the structure sum a total of Aharp = Arings + Asupport,x = 2.16 + 0.78 = 2.94 m 2 ,
the total amout that Table A.1 allows to deduce, once containing the radius defining each ring over the
harp structure3 . Each ring area, projected on xy plane, results from the difference between the areas
2 2

based on the outside and inside radius, Ri,out and Ri,out . This is, Aring,i = π Ri,out − Ri,ins , which
would be the same as taking the product of each perimeter with its “thickness”, defined at a forth section.
The area of the structure supporting the set of rings mentioned accounts for the y-shaped gusset plate
and the three flange beams fastened to it4 . Its projected area on plane xy, Asupport,x , was obtained with
the aid of the Design Software SolidWorks r , by taking advantage of the evaluate features it offers to
compute projected areas of both isolated parts and assembled ones.
Table A.1: Radius and areas defining the rings of the Koppers harp.
A.4 Probes Project Design

Respective to the structure manufacturing and installation, the images shown at Fig. 2.12 illustrate
the probes at station S2 . The tower-structure holds two pressure probes aligned with the flow able to
measure the total (gauge) pressure (p′0 ), and the static pressure at the chamber’s room entrance. It is
made of stainless steel plates, 1 mm thick, welded according to the configuration presented in Fig. 2.12.
The design of this structure was designed to avoid aerodynamic resonance phenomena.
A.5 Probes Thermocouple Calibration

J-type thermocouples5 , with a tolerance of 2.2 ◦ C, are reliable for temperature measurements about
750 ◦ C but calibration is needed to assure accuracy. The thermocouple installed at station S7A has been
calibrated by comparing a set of measurements taken from fondant ice and a portable dry kiln equipment
available at TAP for thermocouple calibration. This way, a set of known temperatures called standard
temperatures, Tstd , were targeted within the interval shown at Fig. A.1b and compared with the readings
obtained with the thermocouple.
The deviation of the measured temperatures from the standard (targeted) values are shown at the
table below. The calibrating correlation is shown in the figure; it relates any temperature reading
3 This information is available at TAP test cell facility, and it comes from the noise reducer harp technical drawings.
4 Complete information about ipb 100 beams can be found at [14].
5 The thermocouple operation is based on the Seebeck effect, which relates the electric current with the thermal variation
between the two metallic wires that constitute the thermocouple itself. Based on the material they are made of (iron-
constantan for the type J thermocouples), those two extremities generate an electromotive force Felec (mV). When dipped
on fondant ice or on an properly isolated kiln, that force is directly obtained from the rake readings and is automatically
converted into temperature. The tables that relate the thermoelectric voltage as a function of temperature may be found at
http://www.pyromation.com/Downloads/Data/temperature-tables.pdf. The voltmeter sensivity is 5.1 µ V for a measure-
ments’ accuracy of 0.1 C. More details at http://www.fisica.ufmg.br/˜labexp/roteirosPDF/Calibracao de um termopar.pdf.
82
Tmeas with the true temperature T7A . The results given at section 4.1 are corrected accordingly to the
referred correlation. The latter allows to correct the measurement deviation from the standard (reference)
temperature: T7A = a Tmeas + b. The coefficients are a = 0.9937 and b = 0.1211, as presented in table
shown in Fig. A.1b.
(a) The type J thermocouple installed was

an exposed Iron/Constantan alloy.
(b) Thermocouple calibration equation.
83
Appendix B
Head Loss Coefficients Details
Guiding Vanes
The head loss coefficient given for thin shaped vanes is dependent on the chord taken as the arc of the
√ √
inner curvature of the elbow, this is c = r 2 = 0.47 2 = 0.665 m. A value of nv = 18 for the number of
vanes is the one corresponding to TAP’s test cell intake section; it is close to the minimum value expected
−1
for the reference chord of c = 0.665 m (according to the author, nv ≈ 0.9 Lr − 1 = 15.4≈ 16)1 . This
way the Diagram 2-26 was selected to obtain an estimate for this head loss coefficient. It defines a
1
local head loss coefficient as being kloc = r
5.41 L = 0.465, the correction for the Reynolds number as
+1.85
4
4.02×10
≈ 0.807 and the friction coefficient as kf r = 1 + 1.57 Lr λ.

kRe = 0.8 + ReL
Diagram 2-6 ([22], p. 118) offers estimates for the friction coefficient of rectangular cross section
elbows λ ≡ λnon−c = knon−c × λcirc , where knon−c is the correction factor that takes into account the
effect of the rectangular cross-sectional shape and where the friction coefficient λcirc is determined as for
circular tubes (at the same Reynolds number). Letting a0 be L and b0 = W , it follows b0 /a0 = 0.931 and
interpolation from the table presented for turbulent regime gives a correction factor of knon−c = 1.00345.
However, the head loss coefficient k may also depend on the surface roughness, which in turn is dependent
on the flow regime.
If the grain roughness is uniform the surface of the cell intake may be considered as hydraulically
¯ ≤
smooth, which is true whenever the relative roughness satisfies the following relation ([22], p. 78). ∆
¯
∆lim = 181×lgRe−16.4 −4 2
=∼ 10 . For a typical equivalent roughness of ∆ = 0.5 mm , and if Dh = 8.834
Re
¯ = ∆ ∼ 10−5 ≤ ∆
m, the relative equivalent roughness should be ∆ ¯ lim . Therefore, the intake can be
Dh
considered as hydraulically smooth and the friction coefficient could be obtained from the curve c of
1
diagram 2-1: λcirc = f ( Re )≈ (1.8lgRe−1.64) 2
= 0.0090. This way, λnon−c = 0.009031.
Smooth Intakes
If to consider resistance coefficients for smooth contractions, these would come as kbell,inlet = 0.03
(for free standing circular bellmouth inlet 3 ) provided the bellmouth curvature to hydraulic diameter ratio
rbell
verifies the relation Dbell
≥ 0.20. Similarly the resistance coefficient of the augmenter inlet followings
the same reasoning4 and would also be negligible: kaug,inlet = 0.03 + 0.47 × 10−7.7r̄ = 0.05, where the
raug
adimensional radius comes as r̄ = Daug = 0.17.
1 Being r −1

that the most advantageous number of vanes would be nv ≈ 1.4 L − 1 = 24.5 ≈ 25.
2 From page 109 of [22], for the “carefully smoothed air-placed concrete” that the slab walls of the test cell intake are
made off.
3 Diagram 3-4 of [22], p. 164, wall mounted setting.
4 For the entry with limited volume arrangement indicated at Diagram 3-9, in[22].
84
Equivalent Diffusors
One should be mainly interested in comparing cases iii (conical) and v (plan) with vii (combined
diffusor). The latter takes the benefits from both the conical and the plane diffusors5 . The presence of
an insert impairs6 the performance of all diffusors: they turn to be less efficient since the flow resistance
is raised.
When comparing the insert and no insert cases, the Table B.1 allows to conclude that the wedge
mostly affects (raises the flow resistance) the plane and the combined diffusors. Conical diffusors, when
preceded by an insert, are more prone to separation: they tend to separate easier and thus at an upstream
section.
The wedge reduces the flow head loss coefficient, endowing the diffusor effectiveness due to flow
separation avoidance: the diffusor is then able to operate away from the transitory-stall flow regime
characterized by alternating flow separation.
The short and long inserts refer to a short or a long augmenter tube. When short, the boundary
layer is thinner and the velocity profile is more uniform and able to withstand more intense favourable
pressure gradients at the diffusor. That is why the head loss coefficient is lesser for that situations. The
opposite situation corresponds to the head loss coefficients presented in bold at Table B.1. A diffusor
head loss coefficient of k dif = 0.14 could be expected with the wedge inside. However, since the flow
entering the augmenter section tube is not uniform, the velocity profile will behave even worse than with
a long augmenter. Moreover, the wall friction at both sides of the wedge cannot be neglected and the
diffusor head loss coefficient is somewhat higher than the ones estimated by Idelchik [22]. For a null
pressure gradient dp′ /dx = 0, the shear stress along the turbulent boundary layer gives a drag force of
1
D = C D A ρUe2 , (B.1)
2
where the longitudinal area is an average between the diffusor’s inlet and exit cross-sectional areas,
(10.5+N ×12)×Ldif
A= 2 ≃ 190 m 2 ; N = 2 is the number of channels that the wedge creates at the diffusor.
Ue is the non-disturbed flow velocity (e.g. varying between Ue,6 = 100 m/s and Ue,7A = 50 m/s and giving
−1/5
an average of < Ue >= 75 m/s for the -3C engine); the drag coefficient is defined as C D ≃ 0.0726 Re x ,
where x is the distance from the wedge leading edge, in m. For a kinematic viscosity of ν = 2×10−5 m 2 /s
and a density of ρ = 0.9 kg/m 3 , this gives a rough estimate of the wedge friction head loss:
p D/At
kf = 1 2
= 1 2
= 0.02 , (B.2)
2 ρvref 2 ρv6
A6 +A7A
where At = 2 = 13.4 m 2 is the area perpendicular to the flow along the diffusor. In fact, the
pressure gradient is not null; also, there is an additional term due to the non-uniformity of the inlet flow
velocity profile (which is non fully-developed, nfd ). The diffusor head loss coefficient is thus quite higher
than the previous value: k dif = kidelchik + kf + knf d > 0.14. The simulations done with the TCC model
agree best with the experimental data obtained at TAP, at station S7A , with k dif = 0.30 and this is the
value used for all engine simulations.
5 Plane diffusors allow to connect systems separated by short distances without flow separation even at high divergence
angles; the conical ones are known for offering the lower flow resistances.
6 As whished (for more detail, do recall Section 2).
85
Table B.1: Resistance coefficients for the equivalent diffusors. Case i refers to a no insert configuration:
the thin boundary layer (b.l.) developing at the diffusor walls would be similar to the one obtained with
a short insert such as the augmenter tube (d.f. stands for fully developed flow).
Stepped-wall diffusors are also covered by Idelchik [22] (Diagram 5-9 through 5-11). The author
presents limits for the divergence angle α up to which the stepped diffusor has low head losses due to
friction at the walls. Table B.2 summarizes presents estimates for minimum head losses, i.e., considering
an optimal divergence angle αopt or an optimal relative length (Ldif /Dh )opt .
For a given relative length it is recommended that optimal divergence angles are used at stepped
diffusors; however, the head loss coefficient also depends on the area ratio n. The area ratio nstep =
A7B /A6 comes as nstep,nw = 82.3/8.83 = 9.32 and nstep,w = 61.3/8.83 = 6.95 for a diffusor without
wedge and with wedge, respectively. The diffusor with wedge at TAP is close to the diffusor with optimal
divergence angle.
Table B.2: Minimum resistance coefficients for a stepped-diffusor, for distinct approaches to find a diffusor
equivalent to TAP’s one ([22], Diagrams 5-9 to 5-11).
Sudden Expansion
According to Idelchik [22] , shock losses result from abrupt enlargements of tube or channels; the
author offers different formulae for computing the head loss coefficients depending on the flow velocity
profile; practical results show that such profiles upstream abrupt expansions are never uniform7 and for
large Re, there is a generalized formula that stands for such nonuniformities: kloc = n21 + Ψ − 2 ×
SE
Φ A0
nSE = 0.599 ([22], Diagram 4-3), where nSE = A6 = 3.09 is the cross-sectional area ratio; the area
downstream of the sudden expansion is A0 = 55.3 m 2 ; the Boussinesq coefficient Φ and the Coriolis
coefficient Ψ may be computed for a rectangular diffusor of cross section aspect ratio valid within the
7 In [22], pp. 189-190, para. 4.
86
(a) Sudden expansion losses, based on velocity head in the (b) Top: velocity distribution in plane diffusors with
small pipe (Fig. 6.22 of [38], p. 372). divergence angle up to 8 ◦ and comparison with thee
power law. Bottom: vortices generated along length l2
(p. 191, [22]).
Figure B.1: Sudden expansion details.
D
interval 0.3 < ab00 = W7A −W
aug
wedge
= 0.627 < 3.0. This is applicable for a velocity distribution over the
cross section following the power function law,
1/m
v(r) r
= 1− , (B.3)
vmax Rh
typical of α ≤ 8 − 10° divergence angle diffusors with a developed turbulent velocity profile, i.e., in which
A6
Q
v(r) is the velocity at a given point and Rh = 2 × is the hydraulic radius; vmax is the maximum
6
velocity of the flow developing in the cross section; the exponent m varies between 1 and ∞, being that
for m = ∞ the profile equals the rectangle profile shown at Fig. B.1b, i.e., a uniform velocity distribution.
A7A
For long diffusors of area ratio ndif = A6 = 2.031 > 2, that corresponds8 to an exponent of m = 3; it
2
(m+1) (m+1)3
follows that Φ = m(m+2) = 1.07 and Ψ = m2 (m+3) = 1.19 and a local resistance coefficient kloc was this
way attained.
The estimate for the head loss coefficient comes as ktot = kloc + kf r , where
l7B 1
kf r = λ × 2 (B.4)
Dk,7B nSE
stands for friction losses along the system and l7B is the length over which the jet separates and generates
vortices during the time the flow speeds over the cross section. Similar to the intake acoustic vertical
stack, it comes that λ ≡ λnon−c = knon−c × λcirc , but the diffusor nonuniform grain roughness leads to
¯ ¯ + 68 0.25 = 0.01945. The equivalent roughness and relative

a coefficient of λcirc = f ( Re , ∆)≈ 0.11 ∆ Re
equivalent roughness were assumed respectively as ∆ = 4 mm9 and ∆ ¯ = ∆ = 9.70 × 10−4 , with
Dh
4A
Dh = Dh,7A = Q 7A = 4.122 m10 .
7A
1 0.0635
∼ 103 , based on the

The correction factor knon−c = 1.01865 is valid for Re W > 2090 ¯
∆
8 In[22], para. 7 of p. 191.
9 [22]p. 108: for “used with corroded and wavy surface” concrete, characterizing quite damaged walls full with soot.
¯ >∆
10 This section couldn’t be considered as hydraulically smooth, since ∆ ¯ lim = 181×lgRe−16.4 = 9.52 × 10−5 .
Re
87
Daug
upstream velocity and with a cross-section length ratio Wdif f,out = 0.627. This way, λ ≡ λnon−c = 0.0198
for a stabilized turbulent flow11 .
Lastly, for the Re assumed above it comes that ktot ≈ 0.604 ≡ kSE , where from eq. B.4 kf r = 0.005
answers for an available length l7B set equal to Lexhaust = 17.68 m.
Exhaust Stack Grids Head Loss Coefficients

The grids a and b have high mesh screens and thus they hamper the flow to go through them: only a
portion of the flow is obliged to go straightforward from station S7A through these grids. The grids c are
parallel to the flow and are connected to the grids d downstream. Most of the incoming mass flow rate
crosses the grid d as it goes through grids c. The flow split coefficient 0 < σ < 1 represents that amount
in percentage terms and may be computed from eq. B.5. The velocity and the Reynolds number of the
flow at station S7B are computed based on the air flow rates developing in paralel; the flow velocity at
station S7B is always higher than 103 for takeoff regimes regardless of the coefficient σ; this way, the
correction coefficient comes as k Re = 1.0 for all test runs of any engine.
1/2 1/2
U7A−8 U7A−8
Q7B,per = (1 − σ) Qcell = Qcell Q7B,mid = σ Qcell = Qcell . (B.5)
Uab Ucd
For a given cell mass flow Qcell = Q7B,per + Q7B,mid , the mass flows at the periphery and at the
middle of station S7B could be known. The cross section areas available at the middle of the exhaust
room are A7B,mid−c = 25.86 m 2 (direction normal to z ) and A7B,mid−d = 20.92 m 2 (direction normal to
x), respectively where grids c and d are located. The remaining flow goes through the room’s peripheral
area, A7B,per = 57.45 m 2 . Grids a and b together are at a peripheral path of the exhaust boot section
but the most outer grid (grid a) was modeled as facing the flow perpendicularly, this is, aligned along
the z coordinate (grid a makes a 125º angle with that axis).
The available orifice areas for the screens a and b are Aor,ab = 6.12 m 2 ; for the grids at the middle
region of the exhaust room those are Aor,c = 11.43 m 2 and Aor,d = 7.99 m 2 . Astep = 1.63 m 2 is the
cross-sectional area at S7B that is free of screens.
2
The head loss coefficients are defined as k = 2 UA
ρ for parallel flows. For a density for the flow
−3
mixture of ρ = 1.00 kg.m it comes that kmid−c = 3.37, kmid−d = 3.43 and kper = 71.39. The head loss
is higher at the peripheral region of the exhaust room: as expected, more meshed structures introduce
higher pressure losses.
11 The limiting value for the beginning of turbulent flow is issued at para. 29, p. 83 of [22]. Together with Diagram 2-4
this allows to obtain a value for the friction coefficient λcirc .
88
Appendix C
Additional Figures
Test Cell Facility
Figure C.1: TAP’s test cell design. Left: a 3D insight. Right: geometric altitude of TAP test cell facility,
Lisbon’s Airport; the prevailing winds are from North and West. (from the Aerophotogrammetric Survey
document of TAP).
89
Figure C.2: TAP’s exhaust section plant (from [24]).
90
Appendix D
Additional Tables
91
Table D.2: Specifications of civil turbofan engines tested at TAP in the past, 1967-1989. (Source:
http://www.jet-engine.net/).
92
Table D.3: Specifications of civil turbofan engines: currently under maintenance test runs at TAP (left)
and targeted to be tested in the future (right). More details in http://www.jet-engine.net/civtfspec.html.
93
Table D.1: Test cell facility evolution from its inauguration and summary of recommendations foreseeing
its upgrade.
94

Thesis PDF

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Thesis PDF

Diunggah oleh

Hak Cipta:

Format Tersedia

Study of a Test Cell for Commercial Jet Engines

Gabriela Liani Beisl Ramos

Thesis to obtain the Master of Science Degree in

To my Mother, who always managed to build wings for each of my dreams.

To both I owe the airplane of my life.

Also, to all the family and friends it carries inside.

Estudou-se o escoamento de ar e as propriedades termodinâmicas, para obter uma visão global da

Palavras-chave: Banco de Ensaios, Modelo Numérico, Motores a Jacto Comerciais, TAP.

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii

2 TAP’s Test Cell 9

4 Computational Model Results 55

5 Test Cell Upgrade 61

B Head Loss Coefficients Details 84

3.1 TCC model details. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

5.1 Parametric study results using the TCC model. . . . . . . . . . . . . . . . . . . . . . . . . 66

2.1 Test cell sections and stations (adapted from [24]). . . . . . . . . . . . . . . . . . . . . . . 9

5.1 Inlet ramp structures (images from [11]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6.1 Separation in diffusors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

B.1 Sudden expansion details. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

ALF After Looking Forward

b.l. Boundary layer

d.f. (Fully) developed flow

LH Inlet chamber pressure probe (Left Hand)

LPT Low Pressure Turbine

RH Inlet chamber pressure probe (Right Hand)

TAP Transportes Aéreos Portugueses

TCC Test cell computational model

α Diffusor divergence angle

γ Ratio of specific heats

σ Exhaust flow split coefficient

ξ Koppers harp uniformization factor

ζ Efficiency parameter of the augmenter

Ra Outer radius of the entrained flow

AFR Air to Fuel Ratio

BPR Engine bypass ratio

CBR Cell bypass ratio

Cp Frictionless diffuser pressure recovery coefficient

n Diffusor area ratio

S Any cross-section of the test cell

p′ Static pressure (gauge)

p′0 Total pressure (gauge)

p′din Dinamic pressure (gauge)

d Wire diameter of a screen

EF Energy flow rate

MM Momentum flow rate (longitudinal component)

N1 Engine fan speed

PS2 Engine inlet static pressure

PT54 Total pressure of engine station 54

Q Volume flow rate

TT54 Total temperature at engine station 54

U Head loss factor

v Flow velocity (longitudinal component)

1A − 1B Acoustic treatment intake

1B − 1C Turning baffles section

1C − 1D Bird shield screen

1E − 1G Channel wall acoustic treatment

1H − 2 Flow conditioning screen

4A Engine fan discharge cross-section

4B Engine exhaust cross-section

4C Centerbody endtip cross-section

avg Average value

bar Barometric pressure

dif Diffusor section