Image Space Based Visualization of Unsteady Flow on Surfaces

1
Image Space Based Visualization of Unsteady Flow
on Surfaces Robert Laramee Bruno Jobard Helwig
Hauser Presenter Bob Armstrong 24 January 2007
2
What's This About?

• New Algorithm ISBV
• Generates dense representations of arbitrary
  fluid flow on computational fluid dynamics (CFD)
  surfaces
• Applied to unsteady flow on boundary surfaces of
  complex meshes from CFDs of more than 250K
  polygons, dynamic meshes, medical data
• Seeing is believing

3
(No Transcript)
4
Visualization of flow on the surface of an intake
manifold. The ideal intake manifold distributes
flow evenly to the piston valves.
5
Visualization of flow at the complex surface of a
cooling jacket a composite of over 250,000
polygons
6
The Baldwin-Lomax model is a classical algebraic
turbulence model which is suitable for high-speed
flows with thin attached boundary-layers,
typically present in aerospace and turbomachinery
applications.
7
What's the Problem?

• Traditional direct visualization of unsteady flow
  is computationally expensive
• No real-time production
• Animation/Simulation is expensive

8
Their Approach

• Build off of Past Work
• LEA
• IBFV
• Reduce complexity by generating 3D vector fields
  onto 2D image space

9
Past Work

• Foundation for the ISBV Algorithm
• Both produce dense representations of unsteady,
  2D vector fields
• LEA
• Lagrangian-Eulerian Advection
• IBFV
• Image Based Flow Visualization

10
Lagrangian-Eulerian Advection

• Produces animations with high spatio-temporal
  correlation
• All data stored in 2D arrays
• Each frame depicts instantaneous flow structure
• Generates animations at interactive frame rates

11
Image Based Flow Visualization

• Simulation of 2D fluid flow
• Each frame of a flow animation is a blend between
  a warped version of the previous image and a
  number of background images.
• Relies on graphics hardware

12
IBFV Method
image k
distort
render
blend
k k 1
Note Distortion Phase There is nothing to stop
the image from being advected outside of the
geometrical boundary.
13
Definition CFD

• Computational Fluid Dynamics
• Branch of Fluid Mechanics
• Relies on Numerical Methods Algorithms

14
More CFD

• How does one treat a continuous fluid in a
  discretized fashion on a computer?
• Widely used Discretize the spatial domain into
  small cells that form a grid or mesh
• Apply suitable algorithm to solve the equations
  of motion.
• www.cfd-online.com

15
Definition Advection

• Advection is transport of a conserved scalar
  quantity in a vector field.
• A good example is the transport of pollutants or
  silt in a river the motion of the water carries
  these impurities downstream.

16
Different Advections
Velocity
Vorticity
Vorticity mapped to helicity
More information here http//www.vrvis.at/scivis/
laramee/isa-streamsurface/
17
A hybrid stream surface-texture advection
visualization showing tumble motion inside a gas
engine cylinder. Texture is advected according
to the velocity field and color is mapped to
velocity magnitude.
18
The same stream surface geometry. Texture is
advected according to the vorticity field and
color is mapped to vorticity magnitude. The
relationship to the velocity field can be
explored in a novel fashion.
19
The same stream surface geometry. Texture is
advected according to the vorticity field and
color is mapped to helicity. Mapping color to
helicity indicates candidate vortex core regions.
20
Traditional Visualization Methods

• Maps one or more 2D textures to a 3D geometrical
  surface
• Textured geometry rendered to an image space

21
ISBV Algorithm

• Project surface geometry to image space
• Apply texturing
• Texture properties are advected in image space

22
ISBV Method (1)

• Associate the 3D flow data with polygons at
  boundary surface
• Project the surface and vector field onto the
  image plane
• Identify geometric discontinuities
• Advect texture properties according to the vector
  field in image space

23
ISBV Method (2)

• Inject and blend noise
• Apply additional blending along the geometric
  discontinuities previously identified
• Overlay all optional visualization cues such as
  showing a semi-transparent representation of the
  surface with shading

24
Flow Diagram
Dynamic Case k k1
Vector Field Projection
Edge Detection
Compute Advection Mesh
Image Advection
Static Case k k1
Noise Blending
Edge Blending
Image Overlay Application
k time as a frame
25
Vector Edge Projection

• Project vector field to the image plane
• Velocity vectors stored at the polygon vertices
• Encode velocity vectors as color values at the
  mesh vertices

26
Color Velocity Component

• Color assignment done as a scaling operation
• Each color component assigned velocity

27
Why Use Velocity Image Fields?

• Advection computation, noise blending simpler in
  image space
• Vector field and polygon mesh decoupled
• Hardware-accelerated occlusion culling
• Supported on graphics hardware

28
Velocity Vectors

• Decoded vectors used to compute advection mesh,
  then projected upon the image plane

29
Timing of Projection of Vectors

• Project vectors to image plane after velocity
  image construction
• Don't have to project all vectors
• Use original 3D vectors for the velocity mask

30
Example Velocity Image Field
31
Edge Detection

• Edge detection solves the problem of artificial
  flow continuity because of artifacts from 3D to
  2D image space projection
• Selectively detects only edges affecting flow
  texture properties

32
Problem with Sharp Edges
33
Geometric Edge Boundaries
34
Edge Blending

• During this phase of the algorithm
• Introduce discontinuities in the texture aligned
  with discontinuities from the surface
• Essentially, ensure that edges are accounted for
  visually

35
Geometric Edge Detection
Disabled
Enabled
36
Compute Advection Mesh

• Distort regular, rectilinear mesh according to
  velocity vectors stored at mesh grid intersections

p is a path line, k is a frame
Above ensures that distortion does not extend
beyond geometric boundaries.
37
Image Advection Texture Clipping
Coarse Resolution Advection Mesh, With Artifacts
Texture Clipping applied
38
Noise Blending

• Use IBVF method for noise injection and blending

39
Blending
Sample Noise Image
Sample Blended Image
40
No Edge Blending
With Edge Blending
41
Image Overlay Application

• Optional step
• Overlay enhances visualization with colorization,
  shading, etc.

Overlay Applied
Composite of All
42
Vector Projection

• The projection of the vectors to the image plane
  is done after velocity image construction for 2
  reasons
• not all of the vectors have to be projected, thus
  saving computation time
• can use the original 3D vectors for the velocity
  mask

43
Performance
44
Performance Discussed

• Authors looking for speedup specifically,
  increased frame rates
• up to 20 fps
• Performance tied to image resolution polygon
  count (unsteady state flow)
• Don't have any good comparative figures

45
79K polygons dynamic geometry dynamic
topology time-lapsed pics
46
221K Polygonal Intake Port
47
Intake Port Mesh Close-Up