Prediction And Design Of Fuel-Air Mixing in a Combustion Wave Rotor Using Two-Dimensional Unsteady Moving Mesh

1
Prediction And Design Of Fuel-Air Mixing in a
Combustion Wave Rotor Using Two-Dimensional
Unsteady Moving Mesh Flow Computation

• Arnab Banerjee
• Mechanical Engineering
• IUPUI
• MSME Thesis Presentation
• Advisor Prof. Razi Nalim
• November 27, 2005

2
Objectives of the present work

• Develop a methodology to study multidimensional
  effects of wave rotors and apply to NASA
  four-port pressure exchanger using commercial CFD
  code
• Predict the fuel-air mixing in an internal
  combustion wave rotor (ICWR)
• Determine key parameters that affect the fuel-air
  distribution in a wave rotor and improve
  understanding to obtain desired fuel distribution

3
Introduction

• Wave Rotor A device for energy exchange
  efficiently within fluids of differing densities
  by utilizing unsteady wave motion
• Two configurations studied here
• NASA four-port pressure exchanger
• Internal combustion wave rotor (ICWR)

4
NASA four-port pressure exchanger
Inlet from the Burner








Exits to Turbine and Burner
Inlet from the compressor
Partially cut away 3D view
Schematic of a gas turbine topped by a four-port
wave rotor

• Turbine inlet pressure is 15 -20 more than
  compressor exit pressure ideally
• Increased overall engine thermal efficiency and
  specific work

5
Internal Combustion Wave Rotor (ICWR)
Wave Rotor
Compressor
Turbine
Schematic of ICWR

• Constant Volume Combustion

6
2D 3D view of wave rotor
7
Pre- and Post- Processing Package

• Developed in-house by Khalid (2004-05)
• Hexagonal unstructured grid
• Parametric geometry and grid
• Grid and geometry stored in small portable files
• Variable port/rotor channel counts and shape
• Tailored grid clustering
• Imports and exports STAR-CD files
• 3D and unwrapped simultaneous view
• Runs easily on laptops (windows)

8
Results of two grid packages
IUPUI in-house code
Star-Design
9
Past 1-D simulations
Paxson and Nalim 1-D code (1997)
Berrak and Nalim Detonation 1-D code (2004)
10
Past 2-D simulations
Piechna et.al (2004) wave rotor
Welch (1997) NASA 4-port
Kerem Nalim (2002) single channel
11
Solution Methodology

• Arbitrary Sliding Interface
• MARS (Monotone Advection Reconstruction Scheme)
  2nd order accurate
• PISO predictor-corrector algorithm
• Corrector stages below specified limit (20)
  indicates convergence reached for specified
  residual tolerance

12
Arbitrary Sliding Interface
13
Estimating Artificial Diffusivity

• Use shock tube with different grid resolutions
  representing the range of CFD simulations carried
  out
• Calculated artificial diffusion from known
  equation
• Compared these values with physical diffusivity
  in simulations

14
Cell size (cm) Artificial diffusivity (m2/s)
2.50 1.5
1.00 0.5
0.25 0.05
Shock tube
Ti
Physical Diffusivities Thermal diffusivity for
air 0.00002 Turbulent diffusivity for ICWR case
0.5
Distance along tube
15
Hardware Resources
15

• AVIDD Linux Cluster
• Huge Scratch space
• Batch Scheduling
• Accessible from outside of network (SSH)
• Dual CPU PC
• Quick turnaround
• Debugging
• Manual decomposition

16
Methodology Development

• Welch (1997) simulated NASA 4-port configuration
  using code validated against experiment
• 2D unsteady, laminar, compressible, ideal gas,
  adiabatic walls, no leakage
• IUPUI simulation
• Same as above and also included passage to
  passage leakage

17
Grid Resolution
18
Grid discretization comparable to Welch (1997)
Welch IUPUI
Rotor Passage Grid Dimensions (nodes) 115 x 41 123 x 41
Rotor Wall Tangential Spacing (in cm) 8.90E-03 9.00E-03
Rotor Wall Tangential Spacing (in cm) 6.40E-02 6.20E-02
Inlet Outlet Port Grid Dimensions (nodes) 85 x 151 85 x 151
Low Pressure Exhaust Port Dimensions (nodes) 85 x 165 85 x 151
Port Wall Tangential Spacing (in cm) 8.90E-03 9.00E-03
Rotor/Port Interface Axial Spacing (in cm) 6.40E-02 6.00E-02
Rotor Interior Axial Spacing (in cm) 0.25 0.25
19
Computed instantaneous total temperature
400 1200
IUPUI
Welch-2D
20

• Interface skewing between cold driven flow and
  hot driver flow not seen in one-dimensional
  computations

Hot driver gas coats the trailing end of the high
pressure exit port thus discharging more hot gas
to the burner
21
Computed instantaneous static temperature
contours showing close up view of passage gradual
opening process and 2D flow features
IUPUI
Welch-2D
22
Fuel-Air Mixing in an Internal Combustion Wave
Rotor (ICWR)

• Include multidimensional effects
• Include turbulence modeling (k-epsilon with wall
  functions)
• Include species transport equations
• Include property dependence on mixture
  composition and temperature
• Examine the effect of fuel-air distribution on
  combustion

23

• Boundary Conditions - from Alparslan, Nalim and
  Synder (2004)
• Inlet was specified as total conditions
• Total pressure at inlet segments ? 109 KPa
• Total temperature at inlet segments ? 291 K
• Exit port was specified as static conditions
• Static pressure at ? 72 KPa
• Hot gas injection port
• Static temperature ? 600 K
• Combustion using one-step reaction combined time
  scale model
• C3H8 5O2 ? 3CO2 4H2O
• the reaction time scale is the sum of the
  dissipation and chemical kinetics time scales.

24
ICWR geometry
Rotational speed of the rotor (rpm) 4100
Number of cycles per revolution 1
Rotor angular velocity (rad/s) 429.2
Number of passages 20
Passage length (meters) 0.7747
Mean passage width (meters) 0.062
Mean radius (meters) 0.199
Gap b/w rotor end wall blade (meters) 0.005
25
Grid Resolution
Ignition port
26
Inlet species compositions
Species Mass Fractions
Air Inlet
Fuel-air Inlet
Fuel or Air Inlet
Direction of Flow
27
Non-Combustion Pressure waves for time converged
solution
10.5 KPa 182.6 KPa
28
Fuel distribution for one-dimensional and
two-dimensional

Red indicates stoichiometric fuel-air mixture,
the desired fuel fraction for the ignition region
29
Shape of fuel-air interface

• Fuel-air interface at the middle of the inlet has
  expected skew (tangential non-uniformity) due to
  passage opening to fuel over time
• Fuel-air interface forming at the beginning of
  the inlet is less skew
• The skew of interface maybe something useful to
  control

30
Close-up view of first inlet segment opening to
rotor passagetufts indicate flow vectors
relative to rotor
Vabs
Vrel
General velocity diagram
Vabs
Vrel
Modified relative velocity diagram for present
case
31
Developing more uniform fuel-air interface

• All the inlet port segments have the same total
  pressures
• First inlet segment has higher static pressure
  than other segments due to higher pressure from
  rotor passage
• Thus absolute velocity in the first inlet segment
  is lower than other segments
• Non-axial relative velocity forces more fuel into
  the trailing side of the passage

32
Reduced total pressure at first inlet segment
33
Increased total pressure at first inlet segment
34
Results of varying total pressure at first inlet
segment

• Decreasing total pressure at first inlet segment
  has backflow ? not helping in the fuel
  distribution shape in other passages
• The fuel-air interface is skewed similar to fuel
  air interaction in middle of inlet ports
• Increasing total pressure at first inlet segment
  causes no backflow
• The fuel-air interface is skewed too

35
Adding air-buffer as first inlet segment
36
Results from air buffer case

• The non-axial relative velocity in the first
  inlet segment which doesnt have fuel ?
  doesnt influence the filling of fuel in passage
• The fuel-air surface is skewed similar to the
  fuel-air surface in the middle of the inlet port

37
Close-up view of inlet port opening to rotor
passage with without air buffer
Fuel sent in from first inlet segment
Air sent in from first inlet segment
38
Setup - combustion case

• Boundary conditions obtained from 1-D detonation
  model.
• The present case is studied for deflagration and
  2-D ? incompatible with 1-D BCs
• Modified BCs to velocity ? high flow causing
  choke exhaust
• Used case to study general effect of fuel-air
  distribution on combustion

39
Combustion with fuel-air coming in from first
three inlet segments
Ignition port
40
Combustion air coming in from first inlet segment
acting as air-buffer
Ignition port
41
Results of combustion case

• Premature ignition when fuel-air mixture from
  first three inlet segments due to hot products
  from previous cycle
• Presence of air buffer as first inlet segment
  prevents premature combustion

42
Skewness (tangential non-uniformity)
43
Comparison of penetration of fuel for both
configurations
44
Conclusions

• Developed methodology for 2-D wave rotor
  simulation
• Compared with published 2-D simulation results by
  Welch (1997)
• Used commercial solver for CFD simulations
• Applied methodology to ICWR
• Studied multidimensional factors affecting
  fuel-air distribution on few configurations
• With no air buffer skew can be affected by
  timing, total inlet conditions
• Premature ignition can be prevented by air-buffer
• To do a higher fidelity simulation, of a given
  wave rotor configuration, include a finer grid
  based on NASA 4-port wave rotor and geometry and
  boundary conditions obtained from one-dimensional
  deflagration.