Transport and Dispersion of Fire Extinguishing Agents

Home , Clutter (advertising), Eclipse 500

TRANSPORT AND DISPERSION OF FIRE EXTINGUISHING AGENTS

DOWNSTREAM FROM CLUTTER ELEMENTS OF

AIRCRAFT ENGINE NACELLES

Khaled Zbeeb

A Thesis Submitted to the Faculty of

The College of Engineering and Computer Science

in Partial Fulfillment of the Requirements for the Degree of

Master of Science

Florida Atlantic University

Boca Raton, Florida

April 2009

ACKNOWLEDGEMENTS

I would like to thank Dr. Chaouki Ghenai for his hard work and support in developing the needed resources for this thesis. Also, I would like to express my gratitude to Dr. W. Bober and Dr. D. Moslemian for their advice and contribution. I would like to thank the people who helped me reviewing this thesis.

iii

ABSTRACT

Author: Khaled Zbeeb

Title: Transport and Dispersion of Fire Extinguishing

Agents Downstream From Clutter Elements of

Aircraft Engine Nacelles

Institution: Florida Atlantic University

Dissertation Advisor: Dr. Chaouki Ghenai

Degree: Master of Science

Year: 2009

The combination of highly turbulent airflow, flammable fluids, and numerous ignition sources makes aircraft engine nacelles a difficult fire zone to protect. Better understanding of nacelle air flow and how it influences the spread of fires and fire extinguishing agents is needed to improve the efficiency of fire suppression. The first objective was to establish a CFD model for a flow field test section to analyze the transport and dispersion of fire extinguishing agents in the presence of various clutter elements. To validate the use of the CFD model, the simulation results of the CFD model were compared to the experimental data and they show an agreement with the experimental data. The second objective was to present parametric studies to show the effects of the coflow speed, turbulence intensity and agent droplet size on the transport and dispersion of the agent particles downstream from the clutter elements.

TRANSPORT AND DISPERSION OF FIRE EXTINGUISHING AGENTS

DOWNSTREAM FROM CLUTTER ELEMENTS OF AIRCRAFT

ENGINE NACELLES

LIST OF FIGURES...... viii

LIST OF TABLES ...... x

1. INTRODUCTION...... 1

1.1 Engine Nacelles Fire Extinguishing Agents ...... 1

1.2 Understanding the Whole Fire Suppression Challenge...... 5

1.3 Nomenclature...... 10

1.4 Test Section and Clutter Package...... 11

2. GOVERNING EQUATIONS...... 14

2.1 Continuity and Momentum Equations...... 14

2.2 Continuous Phase Modeling ...... 16

2.3 Turbulent Flow Equations ...... 16

2.4 Discrete Phase Modeling ...... 18

2.5 Turbulence Dispersion of Particles ...... 19

2.6 Coupling between the Discrete and Continuous Phases...... 19

3. NUMERICAL METHODS ...... 21

3.1 CFD Numerical Methods...... 21

3.2 Numerical Methods using Fluent ...... 24

4. VALIDATION OF THE CFD MODEL RESULTS...... 34 v 4.1 Experimental Results...... 34

4.2 CFD Simulation Results ...... 39

4.3 Validation of the CFD Simulation Results ...... 43

5. FLOW SIMULATION STUDY ...... 48

5.1 Coflow Speed Effects on Agent Particles Behavior...... 48

5.2 Turbulence Intensity Effects on Agent Particles Behavior...... 51

5.3 Injected Droplet Size Effects on Agent Particles Behavior...... 54

6. CONCLUSION...... 60

6.1 CFD Model Verification...... 60

6.2 Flow Simulation Study...... 60

6.3 Future Work...... 62

7. APPENDIX...... 64

7.1 Phase Doppler Anemometry System (Experiment) ...... 64

7.2 Experimental Strategy ...... 64

7.3 Experimental General Strategy ...... 65

7.4 Experimental Data Analysis Strategy...... 66

7.5 Suppressant Transport Experimental Results and Discussion...... 67

BIBLIOGRAPHY...... 72

LIST OF FIGURES

Fig. 1.2.1 (PhostrEx Agent Flame Poisoning Chemistry [11])...... 7

Fig. 1.2.2 (Front view of the test Section where new fire agents were tested on the

F-15 engine nacelle[12])...... 8

Fig. 1.2.3 (Side view of the test Section where new fire agents were tested on the

F-15 engine nacelle[12])...... 9

Fig. 1.2.4 (Back view of the test Section where new fire agents were tested on the

F-15 engine nacelle[12])...... 10

Fig. 1.4.1 (Test Section and Clutter Package)...... 12

Fig. 1.4.2 (3D view of the Test section and clutter package) ...... 12

Fig. 1.4.3 (Top view of the Test section and clutter package)...... 13

Fig. 3.1.1 (HTF7000 Nacelle Ventilation Design Intent [13])...... 21

Fig. 3.1.2 (Surfaces modeled (pylon side) and the nozzle surfaces used for injection

of Halon 1301 into the nacelle [13]) ...... 22

Fig. 3.1.3 (Comparison of Halon concentrations between CFD and test Data for

Probe1 [13])...... 23

Fig. 3.2.1 (3D View of the mesh created for the test section) ...... 31

Fig. 3.2.2 (Side view of the mesh created for the test section) ...... 31

Fig. 3.2.3 (Side view schematic of SF2 test section) ...... 31

Fig. 4.1.1 ((SF)2 Zone A Test Section)[4]...... 34

vii Fig. 4.1.2 (U Velocity Data at 2.0 D Downstream for Various UC, CS)[4] ...... 35

Fig. 4.1.3 (V Velocity Data at 2.0 D Downstream for Various UC, CS)[4] ...... 35

Fig. 4.1.4 (Sauter Diameter Data at 2.0 D Downstream for Various UC, CS)[4] ...... 36

Fig. 4.1.5 (U Velocity Data at 5.5 D Downstream for Various UC, CS)[4] ...... 36

Fig. 4.1.6 (V Velocity Data at 5.5 D Downstream for Various UC, CS)[4] ...... 37

Fig. 4.1.7 (Sauter Diameter Data at 5.5 D Downstream for Various UC, CS)[4] ...... 37

Fig. 4.2.1 (Particle Traces colord by Particle X-Velocity (m/s)) ...... 38

Fig. 4.2.2 (Particle Traces colord by Particle Y-Velocity (m/s)) ...... 39

Fig. 4.2.3 (Particle Traces colord by Particle Z-Velocity (m/s)) ...... 39

Fig. 4.2.4 (Particle Traces colord by Particle Diameter (m))...... 40

Fig. 4.2.5 (3D Air X-Velocity Distribution along the Centerline (m/s))...... 40

Fig. 4.2.6 (Air X-Velocity Distribution At Z=0 (Centerline) (ms/s)) ...... 41

Fig. 4.2.7 (Air X-velocity Distribution at 2D Downstream the Clutter (m/s)) ...... 41

Fig. 4.3.1 (Droplet X-Velocity Comparison 2D Downstream the Clutter) ...... 43

Fig. 4.3.2 (Droplet Y-Velocity Comparison 2D Downstream the Clutter) ...... 44

Fig. 4.3.3 (Droplet Diameter Comparison 2D Downstream the Clutter) ...... 45

Fig. 4.3.4 (Percentage Error of the Droplet Results by using Fluent)...... 46

Fig. 5.1.1 (Comparison of Particle Traces Colored by Particle X-velocity for different

Inlet Velocity (m/s)) ...... 48

Fig. 5.1.2 (Comparison of Particle Traces Colored by Particle Y-velocity for different

Inlet Velocity (m/s))...... 49

Fig. 5.1.3 (Comparison of Particle Traces Colored by Particle Z-velocity for different

Inlet Velocity (m/s))...... 49

viii Fig. 5.1.4 (Comparison of Particle Traces Colored by Particle Diameter for different

Inlet Velocity (m))...... 50

Fig. 5.2.1 (Comparison of Particle Traces Colored by Particle X-velocity(Vx) for

different Turbulence Intensity (m/s)) ...... 51

Fig. 5.2.2 (Comparison of Particle Traces Colored by Particle Y-velocity(Vy) for

different Turbulence Intensity (m/s)) ...... 52

Fig. 5.2.3 (Comparison of Particle Traces Colored by Particle Z-velocity(Vz) for

different Turbulence Intensity (m/s)) ...... 52

Fig. 5.2.4 (Comparison of Particle Traces Colored by Particle Diameter (Dp) for

different Turbulence Intensity (m))...... 53

Fig. 5.3.1 (Comparison of Particle Traces Colored by Particle X-velocity(Vx) for

different Averaged Injected Droplet Diameter (m/s))...... 55

Fig. 5.3.2 (Comparison of Particle Traces Colored by Particle Y-velocity(Vy) for

different Averaged Injected Droplet Diameter (m/s))...... 56

Fig. 5.3.3 (Comparison of Particle Traces Colored by Particle Z-velocity(Vz) for

different Averaged Injected Droplet Diameter (m/s))...... 57

Fig. 5.3.4 (Comparison of Particle Traces Colored by Particle Diameter (Dp) for

different Averaged Injected Droplet Diameter (m)) ...... 58

LIST OF TABLES

Table 1.1.1 (HFE-7100 Chemical and Physical Properties [10])...... 5

Table 1.3.1 (Nomenclature) ...... 11

Table 3.1 (Number of Cells, nodes and faces for the test section CFD Model) ...... 26

Table 3.2 (Test Section CFD Models Selection)...... 27

Table 3.3 (Test Section CFD Model Boundary Conditions) ...... 27

Table 3.4 (Test Section CFD Simulation Solvers election)……………………………28

Table 3.5 (Material Properties used in the CFD Model Simulation)...... 30

CHAPTER

1. INTRODUCTION

This study presents CFD simulations for a simplified model of a fire extinguishing agent distribution downstream from engine nacelle clutter elements. A three-dimensional model for the test section that was used in Reference [4] was established using Gambit Software [1]. The purpose of this study is to first validate the results of the three-dimensional computational fluid dynamics Simulation [2] by comparing them to those obtained by the actual test performed at Sandia national laboratory [4], and then perform complete computational fluid dynamics simulations to study the effects of the air inlet speed, turbulence intensity and injected agent droplet size on transport and dispersion of fire extinguishing agent in this complex geometry.

1.1 Engine Nacelles Fire extinguishing Agents

Modern aircraft fire protection has historically relied on chemicals known as

Halons, but these substances have been recognized since the 1980s as being particularly destructive to the stratospheric ozone layer. The end of global Halon production, uncertainties in the availability and quality of stockpiles, and the lag in adopting available substitutes raise the concern that civil aviation is unprepared for a future without Halons

[8]. According to a recent report prepared for the U.S. Environmental Protection Agency

(EPA) [9], transition in the United States to the next-generation Halon alternatives for aircraft fire protection systems is impeded by technical, regulatory and procedural issues.

The reports conclude that this situation is likely to pose a significant problem for airlines 1 worldwide. Over two decades ago, atmospheric scientists warned that the annual appearance of the Antarctic ozone hole and the probability of increasing depletion of the stratospheric ozone layer had serious consequences for all life forms. This came after discoveries in the mid-1970s that some man-made chemicals could destroy ozone, resulting in increased ultraviolet radiation reaching the Earth. Although ozone is a small component of the atmosphere, the ozone layer plays a vital role in shielding life on Earth from harmful ultraviolet (UV-B) radiation from the sun. Human exposure to UV-B is known to increase the risk of skin cancer, cataracts, and a suppressed immune system.

UV-B exposure can also damage terrestrial plant life, single-cell organisms and aquatic ecosystems. The evidence that human activities were destroying the stratospheric ozone layer was compelling. Emissions of certain manmade chemicals used in many common products such as refrigerators, air conditioners, cars, fire extinguishers, foams and cleaning solvents were reaching the stratosphere, between 10-16 kilometers and up to 50 kilometers above the Earth’s surface, and destroying ozone molecules. With the appearance of the ozone hole, many countries joined efforts for the first time to combat this global environmental threat. Over 180 countries to date have ratified the landmark international environmental treaty, the 1987 Montreal Protocol on Substances that

Deplete the Ozone Layer. Parties to the Montreal Protocol are committed to eliminating production of ozone-depleting substances by meeting specific phase-out deadlines.

Almost 20 years later, the parties to the Montreal Protocol are now gauging the success of the ozone protection policies and regulations that have been implemented, and are reviewing atmospheric measurements that could indicate the beginnings of a recovery of the ozone layer. In addition, they are identifying sectors that lag in moving away from the

2 use of ozone depleting substances and consequently have the potential to significantly delay or prevent this recovery. Owing to the high ozone depletion potential of Halons, the

Montreal Protocol called for an end to their production by 1994. Before production ceased, however, Halons found extensive use worldwide as clean, safe and very effective gaseous fire suppression.

The combination of highly turbulent airflow, flammable fluids, and numerous ignition sources makes aircraft engine nacelles a difficult fire zone to protect [5]. In the past, halogenated agents such as Halon 1301 have been used to protect engine nacelles from fires. Due to the ban of Halon 1301 under Montreal Protocol, replacement agents that can replicate Halon 1301 are needed. New liquid fire suppressants have been proposed as possible alternatives to Halon 1301 in certain fires [6]. The proposed system contains a liquid fire suppressant under high pressure that is atomized into droplets traveling with high velocities. Better understanding of nacelle air flow and how it influences the spread of fires and fire extinguishing agents is needed to improve the fire suppression efficiency of these candidate agents [5]. Presser and Avedisian (2006) [7] investigated liquid agent transport around unheated and heated horizontal cylinders under ambient conditions. Experimental results were presented for a well characterized, droplet- laden, homogeneous turbulent flow using water, methoxynonafluorobutane (HFE-

7100:C4F9OCH3), and1-methoxyeptafluoropropane (HFE-7000: C3F7OCH3). Phase

Doppler interferometry and visualization techniques were used to explore the thermal effects on spray surface-impingement, vaporization, and transport around and downstream of the cylinder. DesJardin et al. (2002) developed a particle impact model for the VULCAN fire physics code [3]. The model is based on the conservation of mass

3 and energy principles along with the breakup correlations for individual droplets. Three cases were tested using 1 mm HFE-7100 droplets to explore the use of a model spanning impact regimes ranging from droplet bouncing, to sticking, to shattering. Disimile and P.

Tucker, J.R., (2005) [5] examined the volume of suppressant that could pass through the clutter packages and reach a potential downstream fire zone. Results indicated that the amount of suppressant captured by the clutter was directly related to stream wise spacing of the clutter and coflow air speed at the clutter location. An experimental study was performed by Disimile and P., Tucker, J.R., (2005) [4] in an effort to help develop a spray transport model that would be used within the computational fire code currently under development at Sandia National Laboratory. The goal of this research was to enhance the fundamental knowledge of spray interactions with clutter (obstacles representing fuel and hydraulic lines, electrical wire bundles, etc…). Analysis of velocity and droplet diameter data consisted of polynomial regression and trigonometric analysis, which enabled the construction of mathematical correlations for the data. The objective of this numerical study is to extend the previous experimental results and to catalog the nature of the flow field (mean flow speed and turbulence intensity) in the presence of various clutter elements found within aircraft engine nacelles and assess its impact on the distribution of liquid droplets. Table 1.1.1 shows the chemical and physical properties for

HFE-7100. In this study, water is considered to be the fire extinguishing agent so that the computational fluid dynamics simulation results can be compared to those results obtained from the experimental test performed at Sandia national laboratory [4].

HFE–7100 is manufactured as a mixture of two inseparable isomers with essentially identical properties. The mixture is a clear colorless liquid. Appearance @20ºC and 101.3 ka

Boiling point: 58.34-58.59°C 3 Density: 1.5305 g/cm at 20°C Particle size: Not applicable Vapor pressure: 27.736 kPa at 25°C Water solubility: 8.47 mg/L at 20°C Partition co-efficient (n-octanol/water): Log P = 3.54 at 20°C ow Hydrolysis as a function of pH: T 1 day to 1 year, at 1/2 pH 4.0, 7.9, 9.9 (see comments below) Adsorption/desorption: log K 2.56 at 20°C oc Dissociation constant: Not provided Flash point: No flash point Surface Tension 13.86 mN/m Auto-ignition temperature: 397°C Explosive properties: Not explosive Flammability limits: Not flammable Reactivity/stability: Not reactive

Table 1.1.1: HFE-7100 Chemical and Physical Properties [10]

1.2 Understanding the Whole Fire Suppression Challenge

The main aspect of fire suppression that needs to be reconsidered is physical.

Current fire suppression approaches flood the engine compartment with an agent. Current certification of commercial jet engines by the FAA requires demonstration of Halon 1301

5 concentrations of not less than 6% for not less than 0.5 second as measured with normal airflow and no fire. This empirical requirement is based on extensive fire testing in the

1950s; however, engine and nacelle designs have changed significantly in the last half century [11].

It has been discovered that only certain zones within a partially enclosed space have the combined features of a fuel source, air flow, ignition, and flow pattern to support a sustained fire. These flame-holding regions vary in location for different installations and may also change over the flight envelope of a single aircraft. Agent delivered to non- flame holding regions has no influence on fire suppression and is therefore wasted. By combining modern tools of computational fluid dynamics, in-flight testing, and setting fires in a high fidelity simulation of the Eclipse 500 nacelle [11], the results of these tests have identified these flame-holding regions and targeted the distribution of the labile bromine agent and its decomposition products (primarily HBr) to them. As an added benefit, it has been found that ambient flows within the nacelle can be used to transport agent to flame-holding regions more effectively and with less engineering complexity.

Figure 1.2.1 shows the PhostrEx agent flame poisoning chemistry [11].

Figure 1.2.1: PhostrEx agent flame poisoning chemistry [11]

For the last two decades, many experimental tests were conducted to replace

Halon fire extinguishing agents and to study new environmental friendly fire extinguishing agents. The National Fire Protection Association and the Singapore

Aviation Academy [12] have also conducted many tests for new fire extinguishing agents. And they have concluded the followings:

 Wind milling only effective for very small fires

 CO2 usually successful if no wind (if discharge two simultaneously)

 Inlet attack with 30 knot head wind is most difficult

 No Clean agent equal to Halon 1211

 PKP equal to Halon 1211, but not clean

 High mass flow rate helps (need 2 PPS)

 Technique critical for agents that discharge as a liquid 7  Non vaporizing liquids (water mist or AFFF) not effective through

inlet.

Figures 1.2.2 to 1.2.4 show the tests for new fire extinguishing agents that was used on the F-15 engine nacelle.

Figure 1.2.2: Front view of the test section where new fire agents were tested on the F-15 engine nacelle [12]

Figure 1.2.3: Side view of the test section where new fire agents were tested on the F- 15 engine nacelle [12]

Figure 1.2.4: Back view of the test section where new fire agents were applied on the F-15 engine nacelle [12]

1.3 Nomenclature

Latin c Specific Heat Df Liquid film Diameter D Clutter element diameter[m] DD Distance downstream from clutter package[m] K Thermal conductivity Re Reynolds number T time V Impact velocity We Weber Number X Downstream distance from the test section center[m] Y Vertical distance from the center of the test section[m] Z Transverse distance from the section center line[m] U Stream wise droplet velocity[m/s] V Droplet velocity in the vertical direction[m/s]

10 W Droplet velocity in the lateral plane[m/s] D32 Sauter mean Diameter of droplet [m] Uc Nominal stream wise coflow airspeed[m/s] Greek symbols µ dynamic viscosity ν kinematic viscosity ρ density σ surface tension τ normalized time

Table 1.3.1: Nomenclature

1.4 Test Section and Clutter Package

The CFD simulation was performed in a three-dimensional test section with

2000 mm length, 915 mm width and 610 mm height [1] See Fig.1.4.1 to 1.4.3. The air inlet velocity was set at 1m/s with turbulence intensity of 10 %. The air stream serves as coflow for agent discharge. The fluid suppressant spray cone nozzle is located on center line of the test section at a distance of 600 mm of the test section and directed toward the clutter elements. The axisymetric is injected at the center of the test section from a nozzle. Water is used in the present study as suppressant agent. Evaporation of the droplet was neglected in this study. The main task of this study is to analyze the effects of the clutter package and gas flow turbulence on the droplet velocities, distributions and trajectories inside the test section. The water flow rate is kept constant at 17.1 l/min. The clutter package consists of 16 cylindrical elements. The elements have an outer diameter

D of 50.8 mm and assembled to form three separate arrays of 5, 6 and 5 elements. The stream wise spacing between each clutter array is set to 2D (101.6mm). These dimensions were exactly matched with test section that was used in Sandia National

11 Laboratory [4]. This will enable a comparison between the experimental and numerical results.

Test Section

y Outlet

Co Flow x* Fluid Suppressant Spray Nozzle Clutter Elements Nozzle .

Figure 1.4.1: Test section and clutter package

Figure 1.4.2: 3D view of the Test section and clutter package

Figure 1.4.3: Top view of the Test section and clutter package

CHAPTER

2. GOVERNING EQUATIONS

2.1 Continuity and Momentum Equations

For all flows, CFD numerical methods solve conservation equations for mass and momentum. For flows involving heat transfer or compressibility, an additional equation for energy conservation must be included. For flows involving species mixing or reactions, a species conservation equation must be included or, if the non-premixed combustion model is used, conservation equations for the mixture fraction and its variance are included. Additional transport equations are also solved when the flow is turbulent [2]. In this study, the conservation equations for turbulent flow are presented.

The conservation equations relevant to heat transfer and species transport will not be considered in this project. The k-e turbulence modeling will be discussed in later sections. The following is the continuity and momentum equations:

The equation for the conservation of mass, or continuity, can be written as follows:

Continuity:

 u i  0 (2.1.1) xi

Conservation of momentum in an inertial (non-accelerating) reference frame is described by momentum Equation, which is:

14 ui u j  P tij  ij     (2.1.2) x j xi x j

Where t ij is the viscous stress tensor defined as:

   ui u j 2 uk t ij        (2.1.3)  x x  3 x ij  j i  k 

 ij  if i  j and  ij 0 if i  j

 ij is the average Reynolds stress tensor defined as:

' '  ij   ui u j (2.1.4)

 u    ui j  2 uk 2  ij  t     ij    k  ij  (2.1.5)  x x  3 x 3  j i  k 

Where k is the average turbulent kinetic energy defined as:

1 k  u 'u ' (2.1.6) 2 i j

t is the turbulent eddy viscosity expressed as:

 k 2 t  c (2.1.7) 

Where C is constant (C = 0.09) and  is the average dissipation rate of the turbulent

u ' u ' kinetic energy and defined as:   i i (2.1.8) x j x j

15 2.2 Continuous Phase Modeling

The time-averaged gas-phase equations for steady, turbulent flow can be expressed as:

       (2.2.1) ui       S xi xi  xi 

Where  , and S are the representative dependent variable, the effective diffusion coefficient and the source term for the gas phase, respectively. Replacement of Ф with a value of 1 yields the continuity equation, while a substitution with u, v and w for Ф represents the momentum equations in the x, y and z directions, and substitution of Ф with k, ε yield respectively the equations of turbulent kinetic energy and dissipation rate of the turbulent kinetic energy [2].

2.3 Turbulent Flow Equations

Turbulent flows are characterized by fluctuating velocity fields. These fluctuations mix transported quantities such as momentum, energy, and species concentration, and cause the transported quantities to fluctuate as well. Since these fluctuations can be of small scale and high frequency, they are too computationally expensive to simulate directly in practical engineering calculations. Instead, the instantaneous (exact) governing equations can be time-averaged, ensemble-averaged, or otherwise manipulated to remove the small scales, resulting in a modified set of equations that are computationally less expensive to solve. However, the modified equations contain additional unknown variables, and turbulence models are needed to determine these variables in terms of known quantities [2]. In This simulation study, the Standard k-є turbulence model was used.

16 The standard k-є model is a semi-empirical model based on model transport equations for the turbulence kinetic energy (k) and its dissipation rate (є). The model transport equation for k is derived from the exact equation, while the model transport equation for

є was obtained using physical reasoning and bears little resemblance to its mathematically exact counterpart. In the derivation of the k-є model, it was assumed that the flow is fully turbulent, and the effects of molecular viscosity are negligible. The standard k-є model is therefore valid only for fully turbulent flows.

The turbulence kinetic energy, k, and its rate of dissipation, є, are obtained from the following transport equations:

Turbulent kinetic energy equation:

   k    t      x  k u j   k  j    Gk    (2.3.1) x j x j

Where k = 1 and Gk is the production of the turbulent kinetic energy defined as:

u u  u u  u   i j  i 2 i k GK t     ij t  k (2.3.2) x x  x 3 x x  j i  j j  k  Dissipation of the kinetic energy

       t      x 2   u j      j   C1 Gk  C 2  (2.3.3) xj k x j k

Where C1 = 1.44, C2 = 1.92, and = 1.3

2.4 Discrete Phase Modeling

Equation of motion of particles

In addition to solving transport equations for the continuous phase, we simulate a discrete second phase in a Lagrangian frame of reference. This second phase consists of

17 spherical droplets dispersed in the continuous phase. The trajectory of a discrete phase droplet is predicted by integrating the force balance on the particle. This force balance equates the droplet inertia with the forces acting on the droplet and can be written (for the x direction in Cartesian coordinates) [2] as du P  F u  u  g      F (2.4.1) dt D P x P P x

18 CD Re FD  (2.4.2)  p d p 24

Where FD (u - up) is the drag force per unit droplet mass; u is the gas phase velocity; up is the droplet velocity; ρ is the gas density, ρp is the density of the droplet, µ is the molecular viscosity of the gas, dp is the droplet diameter, Re is the relative Reynolds number, CD is the drag coefficient and gx is the gravitational acceleration. Equation (2) incorporates additional forces (Fx) in the droplet force balance that can be important, such as Brownian force (for submicron particles) and Saffman’s lift force (lift due to shear).

The trajectory equations are solved by stepwise integration over discrete time steps.

Integration in time of Equation (2) yields the velocity of the droplet at each point along the trajectory, with the trajectory itself predicted by: dx  u (2.4.3) dt P

Equations similar to (2.4.1) and (2.4.3) are solved for each coordinate direction to predict the trajectories of the discrete phase [2].

2.5 Turbulent dispersion of particles

The dispersion of droplets due to turbulence in the gas phase is predicted using the stochastic tracking model. This model includes the effect of instantaneous turbulent 18 velocity fluctuations on the droplet trajectories through the use of a stochastic method. In the stochastic tracking approach, we predict the turbulent dispersion of droplets by integrating the trajectory equations for individual droplets using the instantaneous fluid velocity ( u  u ' (t) ) along the droplet path during integration [2]. For small "tracer'' particles that move with the fluid (zero drift velocity), the integral time becomes the fluid

Lagrangian integral time, TL. For k-є model, the time scale can be approximated as

TL = 0.30 k/є

2.6 Coupling between the discrete and continuous phases

While the continuous phase always impacts the discrete phase, we also incorporate the effect of the discrete phase trajectories on the continuum. This two-way coupling is accomplished by alternately solving the discrete and continuous phase equations until the solutions in both phases have stopped changing [16]. The momentum transfer from the continuous phase to the discrete phase is computed by examining the change in momentum of a droplet as it passes through each control volume [15]. This momentum change is computed as

 18C Re  F   D u  u  F  m t (2.5.1)   2  P  other  P   P d P 24  where mp is the mass flow rate of the droplets, Δt is the time step and Fother is other interaction forces (Saffman lift forces, and Brownian forces). This momentum exchange appears as a momentum sink in the continuous phase momentum balance of the continuous phase flow field calculations [2].

CHAPTER

3. NUMERICAL METHODS

3.1 CFD Numerical Methods

There have been many CFD numerical simulations that were performed in order

to validate a CFD model for fire extinguishing agents around an engine nacelle. For

example, Ramnath Kandalla, Robert Hoover, Kiran Vannam and Dhinagaran

Ramachandran [13] have performed CFD simulations to establish a CFD model for the

fire extinguishing process around an engine nacelle. The objective of their work was to

develop a CFD process for simulating the injection of fire extinguishing agent into

aircraft nacelles. Halon 1301 (CBrF3), which was the fire extinguishant used, was

injected into the nacelle using a nozzle. Transient simulations were performed as the

Halon spread into the various regions of the nacelle. To validate the CFD process, the concentration of Halon was compared with the test data at certain critical locations in the

nacelle. There was good agreement between the test data and the simulations. Figures

3.1.1 and 3.1.2 show the HTF7000 Nacelle Ventilation Design Intent [13]. Figure 3.1.3 shows the comparison between the CFD results and the actual experimental results [13].

Figure 3.1.1: HTF7000 Nacelle Ventilation Design Intent [13]

Figure 3.1.2: surfaces modeled (pylon side) and the nozzle surfaces used for Injection of Halon 1301 into the nacelle [13]

Figure 3.1.3:.Comparison of Halon concentrations between CFD and test data for Probe 1[13]

3.2 Numerical Methods using FLUENT

FLUENT provides comprehensive modeling capabilities for a wide range of incompressible and compressible, laminar and turbulent fluid flow problems. Steady-state or transient analyses can be performed. In FLUENT, a broad range of mathematical models for transport phenomena (like the discrete phase model (DPM) is combined with the ability to model complex geometries [2].

To permit modeling of fluid flow and related transport phenomena in industrial equipment and processes, various useful features are provided. These include porous media, lumped parameter (fan and heat exchanger), streamwise-periodic flow and heat transfer, swirl, and moving reference frame models. The moving reference frame family of models includes the ability to model single or multiple reference frames [18]. A time- accurate sliding mesh method, useful for modeling multiple stages in turbo machinery 23 applications, for example, is also provided, along with the mixing plane model for computing time-averaged flow fields [2].

Another very useful group of models in FLUENT is the set of free surface and multiphase flow models. These can be used for analysis of gas-liquid, gas-solid, liquid- solid, and gas-liquid-solid flows. For these types of problems, FLUENT provides the volume-of-fluid (VOF) [14], mixture, and Eulerian models, as well as the discrete phase model (DPM). The DPM performs Lagrangian trajectory calculations for dispersed phases (particles, droplets, or bubbles), including coupling with the continuous phase.

Examples of multiphase flows include channel flows, sprays, sedimentation, separation, and cavitation. Robust and accurate turbulence models are a vital component of the

FLUENT suite of models [15]. The turbulence models provided have a broad range of applicability, and they include the effects of other physical phenomena, such as buoyancy and compressibility. Particular care has been devoted to addressing issues of near-wall accuracy via the use of extended wall functions and zonal models [20].

In our study, the FLUENT 6.3 [2] software was used to solve the governing equations in integral form. In this method, a control volume-based technique is used

[19].The procedure for the calculation of the dispersion of droplets in turbulent flow is: solve the continuous phase flow field prior to the introduction of the discrete phase equations, introduce the discrete phase by calculating the droplet trajectories for each discrete phase injection, recalculate the continuous phase flow using the inter-phase exchange of momentum and mass determined during the previous droplet calculation, recalculate the discrete phase trajectories in the modified continuous phase flow field, and repeat the last two steps until a converged solution is achieved in which both the

24 continuous phase flow field and the discrete phase droplet trajectories are unchanged with each additional calculation[18]. The computational domain of the three Dimensional geometry used in this study includes a mesh of triangular mesh elements (type pave) [21].

It is composed of 1,234,879 cells, 2,526,507 faces and 233,856 nodes (See Figures 3.2.1 and 3.2.2). The initial conditions and boundary conditions for the gas phase are: (1) inlet coflow (velocity = 1 m/s, turbulence intensity 10 %); (2) Outlet is set as an outflow; and

(3) Wall: u=v =w= 0 m/s; Clutter is set as a wall; the injector pipe is set as a wall. For the discrete phase (water droplets): injection position: x = 0.6 m, y = 0.305 m, and z=0; number of particle stream = 500 (group injection), injection angle = 45o, droplet diameter

= 180 to 280 µm, flow rate: 0.285 Kg/s, boundary condition: inlet and outlet (escape) and at the wall (rebound/stick) (See Figure 3.2.3). In all simulation cases of the flow, continuity, momentum in X, Y and Z directions, k-є turbulence and discrete phase equations were solved. Table 3.1 shows the number of cells, nodes and faces that were created for the test section. Table 3.2 shows the CFD models that are used to conduct the model simulation. Table 3.3 shows the Boundary conditions that were applied for the test section CFD model. Table 3.4 shows Test Section CFD Simulation Solvers Selection.

Table 3.5 shows the material properties used in the CFD Model Simulation. Figures 3.1 and 3.2 show the grid that was generated by using Gambit software. In principal, the continuity and three X, Y and Z momentum equations were solved to determine all three of the flow velocity components throughout the flow. Two k-є turbulence equations were also solved to determine the turbulence kinetic energy (k) and the dissipation rate (є). The

CFD model was set to be three dimensional, steady, viscous and k-є turbulent with standard wall functions. Five boundary conditions were set as seen in Table 3.3. The

25 relaxation factors of the CFD model computation were also set as seen in Table 3.4. The pressure discretization scheme was chosen to be Standard. The momentum, turbulent kinetic energy and dissipation rate discretization schemes were all set to be first order upwind [14] (See Table 3.4). The numerical iterations for the CFD model calculation were performed with limits for each flow variable as shown in Table 3.4. The material properties of the text section, air and the fire extinguishing agent (liquid Water) were also set in Table 3.5. Since the flow was considered incompressible the air and water densities were set to be constant. The Discrete Phase model (DPM) was also considered to determine the droplet x, y and z velocities and diameter downstream from the clutter elements. The most influential aspect of this selection was the initial conditions of the injected fire extinguishing particles. An inert water particle was considered at injection. A group injection scheme was applied with initial x, y and z velocities of 23.1 m/s with a symmetrical injection in three dimensional directions. A surface type injection was also considered. The particle injection surface was circular with 3mm diameter. The water was injected at a mass flow rate of 0.285 Kg/s. 500 injected particles were tracked. The water droplet was set to rebound when it hits the walls or the clutter elements. All the

Discrete phase model equations shown earlier in section 2 were used to determine the fire extinguishing agent droplet characteristics downstream from the clutter elements.

Level Cells Faces Nodes Partitions

0 1234879 2526507 233856 1

1 cell zone, 6 face zones.

Table 3.1: Number of Cells, Nodes and Faces for the Test Section CFD Model

Model Settings ------Space 3D Time Steady Viscous Standard k-epsilon turbulence model Wall Treatment Standard Wall Functions

Table 3.2: Test Section CFD Models Selection

Boundary Conditions ------

Zones

name id type ------fluid 2 fluid Outflow 6 outflow wall 3 wall Agent_inlet 4 mass-flow-inlet Clutter 5 wall velocity_Inlet 7 velocity-inlet Default-Interior 9 interior

Table 3.3 Test Section CFD Model Boundary Conditions

Solver Controls ------

Equations

Equation Solved ------Flow Turbulence

Relaxation

Variable Relaxation Factor ------Pressure 0.30000001 Density 1 Body Forces 1 Momentum 0.69999999 Turbulent Kinetic Energy 0.80000001 Turbulent Dissipation Rate 0.80000001 Turbulent Viscosity 1 Discrete Phase Sources 0.5

Linear Solver Solver Termination Residual Reduction Variable Type Criterion Tolerance ------Pressure V-Cycle 0.1 X-Momentum Flexible 0.1 0.7 Y-Momentum Flexible 0.1 0.7 Z-Momentum Flexible 0.1 0.7 Turbulent Kinetic Energy Flexible 0.1 0.7 Turbulent Dissipation Rate Flexible 0.1 0.7

Pressure-Velocity Coupling

Parameter Value ------Type SIMPLE

Discretization Scheme

Variable Scheme ------Pressure Standard 28 Momentum First Order Upwind Turbulent Kinetic Energy First Order Upwind Turbulent Dissipation Rate First Order Upwind

Solution Limits

Quantity Limit ------Minimum Absolute Pressure 1 Maximum Absolute Pressure 5e+10 Minimum Temperature 1 Maximum Temperature 5000 Minimum Turb. Kinetic Energy 1e-14 Minimum Turb. Dissipation Rate 1e-20 Maximum Turb. Viscosity Ratio 100000

Table 3.4: Test Section CFD Simulation Solvers Selection

29 Material Properties ------Material: water-liquid (inert-particle)

Property Units Method Value(s) ------Density kg/m3 constant 998.2 Cp (Specific Heat) j/kg-k constant 4182 Thermal Conductivity w/m-k constant 0.6

Material: air (fluid)

Property Units Method Value(s) ------Density kg/m3 constant 1.225 Cp (Specific Heat) j/kg-k constant 1006.43 Thermal Conductivity w/m-k constant 0.0242 Viscosity kg/m-s constant 1.7894001e-05 Molecular Weight kg/kgmol constant 28.966 Standard State Enthalpy j/kgmol constant 0 Reference Temperature k constant 298.14999 L-J Characteristic Length angstrom constant 3.711 L-J Energy Parameter k constant 78.6 Thermal Expansion Coefficient 1/k constant 0 Degrees of Freedom constant 0

Material: aluminum (solid)

Property Units Method Value(s) ------Density kg/m3 constant 2719 Cp (Specific Heat) j/kg-k constant 871 Thermal Conductivity w/m-k constant 202.4

Table 3.5: Material Properties used in the CFD Model Simulation

Figure 3.2.1: 3D view of the mesh created for the test section

Figure 3.2.2: Side view of the mesh created for the test section

Figure 3.2.3: Side view schematic of SF2 test section

CHAPTER

4. VALIDATION OF THE CFD MODEL RESULTS

4.1 Experimental Results

The actual experiment for this study that was done by Disimile and Tucker [4] in

Sandia National Laboratory (See Figure 4.1.1) [4] was presented as part of an effort to help develop a spray transport model that would be used within the computational fire code currently under development by Sandia National Laboratory. As part of a Halon replacement research program, new high-boiling-point chemical suppressants have been identified. These agents would discharge in a liquid state, breaking into liquid droplets, and be entrained within the flow passing through the nacelle, impinging on various objects prior to reaching the fire zone. The goal of this research effort is to enhance the fundamental knowledge of spray interactions with clutter (e.g., obstacles representing fuel and hydraulic lines, electrical wire bundles, etc). Three-dimensional velocity and diameter data was collected at two locations downstream for nine combinations of clutter spacing and coflow airspeed. Analysis of velocity and diameter data consisted of polynomial regression and trigonometric analysis, which enabled the construction of mathematical correlations for the data. The particle x-velocity, y velocity and droplet diameter results of this experiment are shown in Figures 4.1.2 to 4.1.4 [4]. It is noted that the experiment was performed for several values of coflow air speed and stream wise distance between clutter arrays elements. Complete details of how the experiment was conducted and what experimental strategies were used are listed in the Appendix section 33 of this study. The Appendix section also includes all the experimental data and it explains in great details how these data were imported, plotted, and analyzed [4].

Figure 4.1.1: (SF)2 Zone A Test Section

Figure 4.1.2: U Velocity Data at 2.0 D Downstream for Various UC, CS [4]

Figure 4.1.3: V Velocity Data at 2.0 D Downstream for Various UC, CS [4]

Figure 4.1.4: Sauter Diameter Data at 2.0 D Downstream for Various UC, CS [4]

Figure 4.1.5: U Velocity Data at 5.5 D Downstream for Various UC, CS [4]

Figure 4.1.6: V Velocity Data at 5.5 D Downstream for Various UC, CS [4]

Figure 4.1.7: Sauter Diameter Data at 5.5 D Downstream for Various UC, CS [4]

37 4.2 CFD Simulation Results

Computational Fluid Dynamics Simulations for this study were performed numerically by considering a coflow air speed of 1(m/s) and a stream wise distance between clutter elements arrays of 2D (101.6mm). Figures 4.2.1 to 4.2.7 show the results for the water droplets velocities in x, y and z directions at a distance of 2D (101.6mm) downstream from the clutter, the droplet velocities in x, y and z directions the droplet diameter distribution for 500 particles along the particles trajectory path lines.

41 4.3 Validation of the CFD Simulations Results

The simulation results obtained in the previous section contain tremendous amount of data due to the three-dimensional computational fluid dynamics performed numerically by using Fluent. Verification of the experiment data by comparing it to the simulation data can be performed by narrowing our data comparison options. The droplet

X, Y and Z Velocities and the droplet diameter were only considered at 2D downstream from the clutter elements. The number of the injected water droplets was 500 particles which resulted in 143 particles that crossed the clutter elements. The droplets results that were obtained by the experiment are for 12 to 15 particles, and they were also selected at certain points at 2D downstream from the clutter (See Figures 4.1.2 to 4.1.4). These experiment plots do not show the Z values of these points. That is why verifying 12 to 15 particles data that were obtained by experiment to 143 particles data obtained by simulation requires more depth.

The simulation data for each droplet was analyzed separately by exporting the tracked particle data from Fluent to Excel. Then the particles data at 2D downstream from the clutter elements were the only data that of interest to analyze. Comparison plots between the experiment results and the simulation results were performed (See Figures

4.3.1 to 4.3.4). These plots show that the simulation results fairly agree with the experiment results. The error percentages of the results were within 7 to 14%. The droplet

X-velocity simulation results were within 10% of the experimental results. The droplet

Y-velocity simulation results were within 14% of the experimental results. The droplet

Diameter simulation results were within 7.5% of the experimental results. This error

42 margin is fairly acceptable giving the nature of the three-dimensional study that is performed.

4.00 Vx (Experiment) 3.00 Vx (Simulation)

2.00

1.00

0.00

-1.00

-2.00

Y Diameter-D) position (clutter -3.00

-4.00 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 Droplet X-Velocity-Vx- (m/s)

Figure 4.3.1: Droplet X- Velocity Comparison (2D) downstream the clutter

43 3.00 Vy (Expirement) Vy (Simulation) 2.00

1.00

0.00

-1.00

Y Position (clutter Diameter-D)(clutter Position Y -2.00

-3.00 -2.40 -2.00 -1.60 -1.20 -0.80 -0.40 0.00

Droplet Y-Velocity-Vy- (m/s)

Figure 4.3.2: Droplet Y- Velocity Comparison (2D) downstream the clutter

44 3.00 Droplet Diameter 2.50 (Experiment) 2.00 Droplet Diameter 1.50 (Simulation)

1.00

0.50

0.00

-0.50

-1.00

Y Position (clutter Diameter-D)(clutter Position Y -1.50

-2.00

-2.50

-3.00 100 125 150 175 200 225 250 275 300 Droplet Diameter-Dp- (Micro- meter)

Figure 4.3.3: Droplet Diameter Comparison (2D) downstream the clutter

45 16 Droplet X-Velocity(Vx) 14 Droplet Y-Velocity(Vy) Droplet Diameter (Dp) 12

Percentage Error (%) Percentage 4

0 Figure 4.3.4:Percentage Error1 of the Droplet Results by using Fluent

CHAPTER

5. FLOW SIMULATION STUDY

The CFD simulation results obtained from the previous chapter were satisfactory so that the CFD model of the test section was validated and considered to perform other simulations of the flow. The results of these flow simulations can reflect the effects of different flow parameters on the droplet behavior downstream from the clutter. The study of the effects of different parameters of the flow could be used to optimize the engine nacelle clutter elements design.

5.1 Coflow Speed Effects on Agent Particles Behavior

The first flow parameter that is of interest is the coflow air speed. The coflow air speed has a direct effect on the particle speed and behavior throughout the flow.

Increasing or decreasing the coflow speed can have an impact on the results of the tracked droplets downstream the clutter elements. Acknowledging this fact can be helpful in optimizing the engine nacelle clutter elements design with respect to different airplane

Mach numbers. In this study 3 coflow speeds were considered (Uc =1, 3 and 5m/s).

Figures 5.1.1 to 5.1.4 show the effects of the coflow speed on the particles behavior downstream the clutter. It is noted that the X, Y and Z velocities of the tracked particles at 2D downstream from the clutter have increased with the coflow speed increases. On the other hand, by increasing the coflow speed of the flow, the number of tracked particles that cross the clutter elements of engine nacelle is decreasing. The droplet diameter of the tracked particles downstream from the clutter does not seem to be 47 affected by the change of the coflow speed. This phenomena is a determining factor in optimizing the clutter elements location and design because the speed of the tracked particles and the number of tracked particles that cross the clutter elements have a tremendous impact on the fire extinguishing process downstream from the clutter elements.

5.2 Turbulence Intensity Effects on Agent Particles Behavior

Turbulence intensity of the flow has a direct effect on the flow behavior. The injected water droplets trajectories are also tremendously affected by the turbulence intensity of the flow. This section studies these effects in great details. It is of extreme importance to find out if an increase in flow turbulence intensity affects the fire extinguishing process downstream from the clutter elements in a positive manner. Figures

5.2.1 to 5.2.4 show the effects of the flow turbulence intensity on the particles behavior downstream the clutter.

It is noted that the number of tracked particles that cross the clutter elements have increased with the turbulence intensity increase. This result reflects a positive impact on the extinguishing process downstream from the clutter elements. Moreover, the X, Y and

50 Z velocities and the droplet diameter of the tracked particles downstream from the clutter have not been fairly affected by the turbulence intensity change.

5.3 Injected Droplet Size Effects on Agent Particles Behavior

Like the coflow air speed and the turbulence intensity the injected droplet diameter has a direct effect on the behavior of the tracked particles downstream from the clutter elements. It is of great interest to find how a change of the injected agent droplet size can affect the behavior of these droplets downstream from the clutter elements.

Figures 5.3.1 to 5.3.4 show the effects of the injected droplet size on the tracked particles downstream from the clutter elements. It is noted that the X, Y and Z velocities of the tracked particles at 2D downstream from the clutter have increased with the injected droplet diameter decrease. It is also noted that by decreasing the size of the injected droplet diameter, the number of the tracked agent particles downstream from the clutter elements have increased. That reflects a positive impact on the extinguishing process downstream from the clutter elements. However, this particular aspect of this study can 53 not result in a conclusion or a fact that states by decreasing the size of the agent droplet diameter, a better extinguishing process can be established downstream fromthe clutter elements. There are different factors that also affect the extinguishing process. One of these determining factors is the heat transfer phenomena that occur between the droplet and the flame caused by the fire downstream from the clutter elements. One can say by increasing the agent droplet size downstream from the clutter elements can help in expediting the extinguishing process. This implies that the effect of the size of the agent droplet on the extinguishing process can not be determined by studying one aspect of the flow. A complete heat transfer analysis of the flow must be performed, so that a better understanding of the extinguishing process can be established. This recommended flow study should include a change in temperature of the clutter elements and an introduction of a fire flame in the flow downstream from the clutter elements. This recommended study will later be discussed in the conclusion part of this report.

CHAPTER

6. CONCLUSION

6.1 CFD Model Validation

The first task of this study was to verify the experimental results obtained from

Ref. [4]. Chapter 4 shows that the CFD model that was created to perform all the flow simulations was considered to be satisfactory when the simulation results were compared to the experimental data [4]. The error margin was within 7 to 13%. This margin is considered to be fairly satisfactory giving the type of the 3D flow that is being investigated. It is noted that there were few experimental data that were compared to the flow simulation results. There were only 13 sets of data for particles in the experimental study that was performed in Sandia National Laboratory, and the simulation results produced 143 sets of data for all the tracked particles that cross the clutter elements. It is preferable to have more experimental results for particles downstream the clutter elements. This could reflect better and deeper understanding of the particles behavior downstream from the clutter. Moreover a better comparison of data between the experiment and the simulation results could be accomplished.

6.2 Flow Simulation Study

The second task of this study was to generate variety of flow simulations to understand the effects of the coflow speed, the turbulence intensity and the droplet diameter size on the behavior of the tracked particles downstream from the clutter elements. The behavior of the particles would also in turn affect the fire extinguishing 59 process downstream from the clutter elements of an engine nacelle. Chapter 5 concludes that the X, Y and Z velocities of the tracked particles at 2D downstream from the clutter have increased with the coflow speed increases. On the other hand, by increasing the coflow speed of the flow, the number of tracked particles that cross the clutter elements of engine nacelle is decreasing. The droplet diameter of the tracked particles downstream from the clutter does not seem to be affected by the change of the coflow speed. This phenomena is a determining factor in optimizing clutter elements design and location because the speed of the tracked particles and the number of tracked particles that cross the clutter elements have a tremendous impact on a fire extinguishing process downstream from the clutter elements. In addition to that, It is concluded that the number of tracked particles that cross the clutter elements have increased with the turbulence intensity increase. That reflects a positive impact on the extinguishing process downstream from the clutter elements. Moreover, the X, Y and Z velocities and the droplet diameter of the tracked particles downstream the clutter have not been fairly affected by the turbulence intensity change. Moreover, It is noted that the X, Y and Z velocities of the tracked particles at 2D downstream from the clutter have increased with the injected droplet diameter decrease. It is also concluded that by decreasing the size of the injected droplet diameter, the number of the tracked agent particles downstream from the clutter elements have increased. That reflects a positive impact on the extinguishing process downstream from the clutter elements. However, this particular aspect of this study can not result in a conclusion or a fact that states by decreasing the size of the agent droplet diameter, a better extinguishing process can be established downstream from the clutter elements. The next section will explain more in details of what flow studies should

60 be performed to establish better understanding of what are the determining factors of the fire extinguishing process downstream from the clutter elements of an engine nacelle.

6.3 Future Work

As it was mentioned in the previous sections this CFD simulation study that was performed in this report is a step from several steps that are recommended to be analyzed in order to get a complete understanding of the nature of the behavior of a fire extinguishing agent particles downstream from clutter elements of an engine nacelle. The first recommended work is to reestablish a new three-dimensional CFD model with different geometry where the spacing between the clutter elements is changed to 0.25 D.

The simulation results of this CFD model should also be compared to the experimental data [4]. This would confirm our validation of use of the CFD simulations in this particular flow study. Moreover, additional experimental results are also needed to verify our CFD model results. On the other hand the CFD simulations that could be performed using the CFD model is really unlimited and could be very fruitful in understanding and describing the behavior of the fire extinguishing agent particle downstream from the clutter elements. Since the main objective is to study the fire extinguishing process and the behavior of the fire extinguishing agent particles downstream from the clutter elements, it is strongly recommended to establish a new CFD model that reflects a heated clutter elements and a fire flame downstream from the clutter elements. There are different factors that also affect the extinguishing process. One of these determining factors is the heat transfer phenomena that occur between the droplet and the flame caused by the fire downstream from the clutter elements. One can say by increasing the agent droplet size downstream the clutter elements can help in expediting the

61 extinguishing process. This implies that the effect of the size of the agent droplet on the extinguishing process can not be determined by studying one aspect of the flow. A complete heat transfer analysis of the flow must be performed, so that a better understanding of the extinguishing process can be established. This recommended flow study should include a change in temperature of the clutter elements and an introduction of a fire flame in the flow downstream from the clutter elements.

CHAPTER

7. APPENDIX

7.1 Phase Doppler Anemometry System (Experiment)

A three-dimensional Phase Doppler Anemometer (PDA) was used to simultaneously acquire velocity components and the diameter of the liquid drops exiting the suppressant spray nozzle. To position the PDA measuring volume in the center of the test section, 1000 mm lenses were 5 used on both the transmitting and receiving optics.

Based on a previous study by Davis and Disimile [3], this measurement configuration also utilized the largest spatial filter (Mask A) in the receiving optics. In this configuration the maximum drop diameter capable of being measured was 810.8 µm [4].

7.2 Experimental Strategy

Air drawn through a two-dimensional contraction is conditioned and passed through a turbulence generator section, producing a 10% turbulent intensity (T.I.) level.

2 This air stream enters the (SF) test section and serves as a coflow for agent discharge.

Agent discharge is generated using a dual fluid nozzle located on the centerline of the test section and directed downstream toward the clutter elements; however, the nozzle was slightly misaligned off-center, providing bias in the -y and +z directions. Water and air are mixed by the dual fluid nozzle and travel toward the leading edge of the clutter package. The suppressant droplets not captured by the clutter package proceeded downstream, where their three-dimensional velocity and diameters were measured

63 using the Phase Doppler Anemometry (PDA) system. The measurement volume of the

PDA system was located at 2.0 and 5.5 clutter diameters downstream of the trailing edge of the final clutter array. The measurement volume was traversed along the vertical axis from 140.00 ± 0.79 mm (5.51 ± 0.03 inches) above to 140.00 ± 0.79 mm below the horizontal centerline at each downstream measurement location [4].

7.3 Experiment General Strategy

The spray nozzle water flow rate was set and maintained at 17.1 ± 0.4 L/min (4.5

± 0.1 gal/min) with a corresponding nozzle water pressure of 158.0 ± 13.7 kPa (23.0 ±

2.0 psi). The water flow was monitored using a turbine type flow meter positioned directly upstream of a pressure gage. The incoming air was regulated to a pressure of

171.8 ± 3.4 kPa (25.0 ± 0.5 psi). The test matrix consisted of 18 experimental conditions.

This included three coflow speeds, three clutter densities, and two downstream measurement locations. Clutter package densities were varied by changes in the stream wise spacing between individual clutter arrays. Array spacing was selected to be 0.25 D,

1.00 D, and 2.00 D with an error of ± 0.06 D. In the current study D = 50.80 ± 0.79 mm

(2.00 ± 0.03 inches). Since the leading edge of the clutter package was fixed, changing the array spacing affected the location of the trailing edge of the last clutter element.

Therefore, to maintain a fixed downstream location of the PDA measurement volume with respect to the trailing edge of the clutter, the PDA measurement volume had to be moved correspondingly with the trailing edge movement. Previous experiments were conducted to determine the volume of water transmitted through the clutter package as a function of the stream wise spacing of the clutter array and coflow speed. During these experiments, a repeatable procedure for measuring the liquid water volume was followed

64 and previously reported [3]. Based on these studies, air speeds of 3.0 m/s, 4.0 m/s, and

5.0 m/s were chosen for the current investigation.

7.4 Experimental Data Analysis Strategy

The downstream three-dimensional velocity and diameter data were imported, plotted, and analyzed in Excel. Each data component (U, V, W, and D32) was grouped by each independent parameter: clutter spacing, downstream distance, and coflow speed.

Overall, the most appropriate polynomial regression of the data was of degree six. Thus, five observable changes in direction of the data were noticed; however, since this was the limit of Excel's built-in capabilities, no higher-order polynomial regression could be examined. The final organization of data resulted in variation of coflow speed at constant downstream distance and clutter spacing. This revealed similarities in each flow, and a parabolic fit was placed on each of the three segments of data, as defined by the periodic pattern of high and low peaks in both velocity and diameter data. Since a sixth-degree polynomial must be smooth and continuous across the entire domain, three separate parabolic fits allowed for discontinuities 8 where the data was neither smooth nor asymptotically stable—approaching a finite limit—from visual observation. These discontinuities in the data could be overcome by the parabolic approach. Each of the three parabolas for each test was manually fit to the data; rather than a regression, which would provide an average of values, the manual fit provided an upper bound to the data.

Resolving each of the six plots of U and V data at varying coflow speed, Uc, resulted in six groups of three parabolas; this data was then collapsed by taking the three parabolic coefficients a, b, and c at each Uc and creating a relation for each coefficient as a function of Uc. A second degree polynomial regression was used for each coefficient at the three

65 values for Uc, providing the functions a(Uc), b(Uc), and c(Uc) for each parabolic fit, reducing all U data to three parabolic functions for each clutter spacing. The same procedure was conducted for the V data; however, neither W nor D provided enough of a pattern for such parabolic fits. Once the periodic nature of the data was observed, an

Excel-based fast Fourier transform (FFT) algorithm was used to construct trigonometric models of all data. This method was needed only to build a mathematical function relating the vertical position (Y) to U, V, W, or D32. No regression (with the intention to further collapse the data) was provided from the FFT method.

7.5 Suppressant Transport Experimental Results and Discussion

Varying such parameters as coflow speed and clutter spacing produced varying results for U, V, W, and D32. However, similarities between all related data sets (e.g. all

U data at 2.0 D downstream) were noticed when these data sets were plotted together. For instance, the qualitative relationship of peaks in droplet velocity and diameter data along the vertical direction corresponded to the location of clutter elements in the vertical direction. Quantitatively, proportionality of the peaks to the element locations has not yet been established; however, the mathematical methods in use will continue to provide insight into these relationships. In Figures 4.1.2 through 4.1.7 below, all U, V, and D32 data at the 2.0 DD and 5.5 DD locations were plotted as a function of vertical position, Y.

The vertical positions of each clutter element were also placed on each figure, along with a representation of the clutter elements along the vertical span of the traverse.

In Figures 4.1.2 through 4.1.4 all data recorded 2.0 DD of the trailing edge of the clutter is presented. At this location, a pattern common to all data can be noticed: the data behaves periodically in correlation to the vertical location of the clutter elements.

66 For the U data (Figure 4.1.2), the various data all show peaks in magnitude at the locations of the front/rear clutter elements and low points at the middle clutter elements.

With the exception of the 1.0 m/s, 0.25 D CS case, every combination of horizontal spacing of clutter elements and nominal coflow speed examined shows this pattern. In all cases, the 0.25 D CS data showed the least conformity to the rest of the data at a given coflow speed, with the 5.0 m/s, 0.25 D CS data representing the upper extreme[4].

The V data (Figure 4.1.3) shows a similar trend to the U data: peaks in magnitude correspond to the front/rear clutter elements, while low points correspond to the middle elements. Notice that increasing U is plotted from left to right, whereas increasing V is plotted from right to left; taking this into account, the same trend observed in U (three peaks pointing toward 9 m/s) can be observed in V (three peaks pointing toward –5 m/s).

Very few data points show that droplets carried any upward positive velocity at 2.0 D downstream. Again, the 0.25 D CS data (at 5.0 m/s and also 3.0 m/s UC) showed noticeable deviation from the rest of the data. While U also increased toward the floor of the test section, the same observation of V is emphasized. The Sauter diameter, D32

(weighted mean derived from volume divided by surface area), showed a similar periodic trend (Figure 4.1.3) inversely related to the velocity data. Whereas U and V data both showed peaks at the front/rear elements and low points at the middle, D32 showed low points at the top of the front/rear elements and peaks at the tops of the middle. Most droplet diameters were measured in the 200 to 400 µm range at 2.0 D downstream. The noticeable change between 2.0 and 5.5 D downstream is noted below. Since the nozzle was biased downward (as noted in Experimental Strategy), this correlation between the peaks in the U and V data and the peaks in the D32 data may be misleading in the figures

67 above. Data at 5.5 DD was also analyzed (Figures 4.1.4 through 4.1.7). Upon initial inspection, the periodic pattern noted at 2.0 D downstream is no longer noticeable. The velocity profiles at this location have smoothed out to reveal a single peak in magnitude, occurring around the test section centerline. At 5.5 DD, the periodicity of the velocity data is reduced and symmetry along the vertical axis is increased. In Figure 4.1.4, two trends in U data can be observed; the more common trend involves a slight decrease in velocity near the centerline. The less common trend involves (as Y progresses from +3.5 to –3.5 D) a rapid increase in velocity and then a slight decrease in velocity; this is seen in the data at 0.25 D CS (at 1.0 and 3.0 m/s coflow speeds). Again, data at 0.25 D CS (for

3.0 and 5.0 m/s UC) follows similar trends to the rest of the data, but it is less pronounced.

Besides deviating in velocity, data at these two configurations also deviate in the pattern; rather than decreasing near the centerline, these two data sets increase strongly. Even though other data sets at 5.0 m/s deviate from the rest of the data, they still follow the same trend. The V data at 5.5 D downstream (Figure 4.1.5) showed double the number of data points at which droplets carried any upward velocity than did the data at 2.0 D downstream. Most of the data conformed to a trend of (from Y = +3.5 to –3.5 D) low negative velocity to increasingly negative velocity. The other trend is of very low velocity far from the centerline to very highly negative velocity near the centerline. The data to show this pattern are all at 0.25 D CS; the higher co flow speeds showed the greatest change in velocity along the vertical direction D Downstream for Various UC,

CS13 The diameter data at 5.5 D downstream (Figure 4.1.6) shows consistency. Except for 0.25 D CS data, all other data showed relatively little change in diameter along the vertical direction. 2.0 D CS data decreases slightly in diameter as Y decreases, while 1.0

68 D CS data increases slightly. The range of mean diameters has changed since 2.0 D downstream; also, the concentration of Sauter diameters has decreased from approximately 300 µm at 2.0 DD to approximately 250 µm at 5.5 DD. The data collected at 2.0 DD was analyzed using three methods (in the Data Analysis Strategy section); six- degree polynomials, multiple parabolas, and Fourier analyses were applied to the data in

Figures 4.1.2 through 4.1.4. The main goal was to collapse the data to few formulas, rather than individual data points. Little physical significance was yet found in the coefficients for the analysis methods; however, velocity and diameter data series could be reproduced with these methods, including the periodic trends and peaks in the data. So far, the parabolic/regression method has provided the best results, while the Fourier analysis represented the data series more selectively. Multiple data series have been collapsed into one set of formulas; for example, all of the U data at 1.0 D CS, for varying coflow speeds, can be represented by a set of formulas derived from the parabolic/regression method. In the future, models allowing clutter spacing and not only coflow speed will be developed. In short, one set of three parabolic equations has been developed for each clutter spacing. For example, at 1.0 DC (at 2.0 DD and between 3 and

4 m/s UC), the set of three equations for U [m/s] (each equation representing a separate peak in the U data seen in Figure 4.1.2) as functions of UC [m/s] and Y [mm] is shown below:

69 If physical significance cannot be identified in the coefficients of these formulas, then at least multiple sets of empirical data may be collapsed to very few formulas for development of Scandia’s fire code [4].

BIBLOGRAPHY

[1] Gambit (2.3.16), Full Tutorial. 3D Drawing and Mesh Design.

[2] Fluent (6.3), Tutorial 9-11-12-23-28-29.Turbulence and Discrete Phase Models.

[3] DesJardin, P.E., Presser, C., Disimile, P.J., and Tucker, R., (2002), Droplet impact

model for agent transport in engine nacelles, NIST Special Publication 984.

[4] Disimile, P., Tucker, J.R., Croswell, B., and Davis, J. (2003), the transport of water

sprays past generic clutter elements found within engine nacelles, Halon options

technical working conference.

[5] Disimile, P., Tucker, J.R., Stern, j., Mehl, l., and Croswell, B., (2005), Fire Safety

Journal, 40, pp. 65-78.

[6] King, M.D., Yang, J.C., Chien, W.S., Grosshandler, W.L., (1997), Evaporation of a

small water droplet containing and additive.

[7] Presser, C., Avedisian, C.T., (2006), Atomization and Sprays, Volume 16, Issue 6, pp.

627-656.

[8] Catchpole, D., Verdonik, D., Maranion, B., (2005), ICAO Journal, Issue 6.

[9] U.S. Environmental Agency, Fire Suppression. http://www.epa.gov/ozone/snap/fire/.

[10] Commonwealth of Australia., (2006), National Industrial Chemicals Notification

and Assessment Scheme. http://www.ag.gov.au/cca.

[11] Eclipse Aviation Corporation, (2006), the Eclipse 500 Fire Suppression System

Using the PhostrEx Agent. http://www.eclipseaviation.com.

71 [12] Scheffey, J., 2000, International Forum on Fire Protection and Risk Management.

The National Fire Protection Association and the Singapore Aviation Academy.[13]

Kandala, R., Hoover, K., Vunnam, D., Ramachandran, D., (2007), CFD analysis of

a fire extinguishing system for an aircraft nacelle, International Aerospace CFD

Conference.

[14] Anderson, J., (1995), Computational Fluid Dynamics, The basics with Applications.

[15] Hirsch, Charles, (1988), Numerical Computation of Internal and External Flows,

Vol.I: Fundamentals of Numerical Discretization.

[16] Hirsch, Charles, (1988), Numerical Computation of Internal and External Flows,

Vol.II: Methods of Inviscid and Viscous Flows.

[17] Fletcher, C.A, (1988), Computational Techniques for Fluid Dynamics, Vol.I:

Fundamental and General Techniques.

[18] Fletcher, C.A, (1988), Computational Techniques for Fluid Dynamics, Vol.II:

Specific Techniques for Different Flow Categories.

[19] Thompson, Joe, (1982), Numerical Grid Generation

[20] Thomson, Joe F., Z.V.A. Warsi, and C. Wayne Mastin, (1985), Numerical Grid

Generation: Foundations and Applications.

[21] Roberts, B.O., (1977): Computational Meshes for Boundary Layer Problems,

Lecture Notes in Physics, pp.171-177.