Knowledge-Assisted Visualization of Turbulent Combustion Simulations

rangebeaverMécanique

22 févr. 2014 (il y a 3 années et 1 mois)

54 vue(s)

Knowledge
-
Assisted
Visualization of Turbulent
Combustion Simulations

Chaoli Wang, Hongfeng Yu, Kwan
-
Liu Ma


Turbulent combustion simulations


Direct numerical simulations


Time
-
varying, multivariate data


800 * 686 * 217, 450MB


53 time steps


4 variables:
mixfrac
,
chi
,
HO
2
, and
OH


93GB in total


Visualization
-
specific task


Scientific interests


Observe variable relationships close to the
mixfrac

surface


Bring out more the lower values of other variables

The mixed rendering of the
mixfrac

(isovalue = 0.2) and the
HO
2

variables

Challenges


Data ranges of other variables close to the
surface are unknown


Only value
-
based transfer function may
bring out undesired visualization contents


Lack of control over the amount of
information shown around the surface

Our solution

create distance
volume
plot partial
histogram
specify transfer
function
preprocessing
compute voxel
importance
volume
rendering
distance threshold
d
runtime
Transfer function specification

d

= 0.02

d

= 0.05

Visualization result

d

= 0.10

d

= 0.20

Visualization result

Video demos

Summary


Knowledge
-
assisted visualization


Domain knowledge: isovalue, ranges of
interests


Derived knowledge: distance volume and
partial histogram


Importance
-
driven visualization


Future work


Time
-
varying, multivariate data compression


Utilize domain knowledge


Visualization
-
specific compression

Thank you!

Extra slides

Visualization
-
specific compression


Regions of interest are around the given
surface


Data precision can thus vary according to
the distance to the surface


Our solution


Non
-
uniform quantization


Space
-
time coherence utilization


Decompression on the fly using graphics
hardware

Compression result


Compression ratio: ~20x


Each voxel only uses less than 2 bits per
variable on the average


Advantages


Reduce data transfer among disk to main
memory, and main memory to video memory


Fast offline compression and online
decompression


Preserve fine details near the surface and
maintain the overall image quality

Compression result

original

compressed (~20x)

Compression result

original

compressed (~20x)