Aircraft Icing: Computational Studies Team

1
Aircraft Icing Computational Studies Team

• ASE-463 Q
• Fall 2006
• Advisor
• Dr. Graham F. Carey
• Members
• Vikram Garg
• Cuong Tran

2
Team Information

• Vikram Garg Team leader, finite element
  formulation, code development and analysis of
  results
• Cuong Tran Parametric studies, data
  presentation and analysis of results

3
Presentation Outline

• Background and Motivation - Cuong
• Finite Element Method - Vikram
• Meshing Cuong
• Results Validation - Cuong
• Steady Flow Results Cuong
• Unsteady Flow Results Vikram
• Future Work Recommendations - Cuong

4
Background

• Icing on wings rapidly frozen water droplets
  accumulate on wings
• Negative effects loss in lift, increase in
  drag, lower stall angle, and loss of stability
• A comprehensive, reliable fluid-structure code
  to enhance the study of icing in our department

5
Motivation

• Experimental approach
• limited choice of parameters
• time-consuming
• Numerical approach
• input parameters easily modified
• data extraction is simple and robust
• fast design modifications are possible for
  optimization

6
Objectives

• Formulate and implement a Finite Element solution
  of the mathematical model that governs the
  physics
• Extract appropriate data from the simulations to
  compute performance parameters (CL, CD, etc)
• Parametric studies effects of Re, AOA on a
  system operating at an off-design condition

7
Finite Element Method

• Numerical method to solve flow PDEs.
• Discretize domain meshing
• Introduce approximations to variables, integrate
  against test functions variational
  formulation
• C2 basis functions, allow recovery of velocity
  gradients
• Nonlinear PDE Use Newtons method,

8
Governing Equations

• Basic equations governing incompressible, viscous
  (Newtonian) fluid flow
• Re Reynolds number - different flow regimes
• 0 lt Re lt 10 Stokes flow
  
• 10 lt Re lt 1000 Laminar flow
• 1000 lt Re ... Turbulent
  flow

9
Impact of Re number

• Transition to turbulence difficult to capture
  numerically, element size requirement 1/Re
• Numerical stabilization use non-symmetric test
  functions, add , a stabilization term
  introduce artificial dissipation
• Need fewer elements, smaller solve times but
  need to validate

10
Solution Methodology

• First prepare an unstructured triangular mesh
  Matlab TRIANGLE
• Assemble and solve linear system (or seq.) to get
  - Libmesh
• Post-process P distribution, vel. Profiles,
  aerodynamic co-efficients

11
Airfoil Specs

• Swept GLC-305 airfoil
• 18.72 inch chord length
• Chord length used for non-dimensionalization (Re,
  L, etc.)
• Standard Ice horn shape acquired from Bragg et
  al

12
Meshing

• Wrote Matlab code to extract x y coordinates
  for clean and iced airfoil
• TRIANGLE uses nodal points along airfoil
  bounding box to create a mesh
• Libmesh uses triangular elements
• Linear (with 3 nodal pts) for pressure
• Quadratic (with 6 nodal pts) for velocity

13
Meshing Results
TRIANGLE meshing output
Clean Airfoil
Zoomed in on Ice Horn
The fine, detailed meshing near the surface of
each airfoil enables accurate data acquisition.
14
Adaptive Mesh Refinement

• More action More elements needed for accuracy
• Compute a crude solution, use to decide where we
  need more elements
• Capture viscous boundary layer

15
Validation
1) Domain size sensitivity results obtained
must be independent of the domain size
y
y
vs
x
x
2) Mesh detail sensitivity results must also be
independent of mesh size
y
y
vs
x
x
16
Validation

• Smaller domain and mesh resolution fewer dofs
  faster computational speeds
• Large domain minimize diffusive effects
• High mesh resolution - accurate solution
• Must minimize computational time ensure -
  integrity of the results
• Achieved by using parallel processing adaptive
  refinement

17
Domain Sensitivity
100 Increase in Domain Size with Re1000, a10,
clean
Original Domain Size with Re1000, a10, clean
Clearly, the results are nearly identical. Thus,
we can be sure that data acquired at the original
domain size is reliable.
18
Pressure Distribution

• Re 1000, steady flow
• Flow attached for the most part

19
Re number dependence
20
Angle Of Attack Dep.
21
Unsteady Calculations

• Computed unsteady flow fields for a 0, 3 6 and
  8 and verfied that they match the ss.
• Iced airfoil has lower drag!!!! Why

22
Unsteady Calculations

• At low Re, viscous effects dominate
• Breakdown viscous and pressure drag

23
Unsteady Flow Results

• At aoa 10 the 2 separation bubbles merge

24
Pressure distribution
25
Higher AOA

• At higher AOA, even the clean airfoil had a fully
  separated flow field, flow field is possibly
  turbulent
• Movie clean 16 deg

26
Conclusions

• Few quantitative differences between clean and
  iced at low aoa
• But a 10, sep. bubbles merged and no stationary
  s-s
• At higher aoa, flowfield is possibly turbulent
• Some implications for design of micro-aircraft in
  this viscous dominated regime

27
Future Work

• All the dirty work done, formulation, jacobian
  derivation, most of the debugging
• Turbulence model, k-omega already coded but any
  2-equation model should be easy to implement
• Flow structure coupling for aero-elasticity,
  need N-S in an inertial frame to avoid moving mesh

28
Questions

• Do you have any?