# ESM4714 Scientific Visual Data Analysis and Multimedia Sources of Scientific Data Sets

In this section we introduce you to a variety of data types. The sources of the data types come from numerical supercomputer simulations, experimental measurements, and complex analytic functions. We will show that in all cases if each data type can be organized in a similar fashion then the same three-dimensional graphic tools can be used to explore and discover new relationships. Because some of these data sets are used by the homework assignments we also use this section to define information about the data sets needed to complete the homework.

CONTENTS:
1. Three-Dimensional (3-D) Coordinate Space
2. Supercomputer Numerical Simulations
3. Parameter Space
4. Visual Insights

## 1. Three-Dimensional (3-D) Coordinate Space

The most common source of data comes from numerical simulations where we predict gradients in physical properties both in space and time. If properly organized the viewer can begin to interpret not only the large volume of data but also to understand how property relationships change in space and time.

Perhaps the best known 3-D experimental data source is the X-ray computed tomography (CT) scan where we observe gradients in density. Instead of the more common biomedical application, here we introduce a material application where the source of data is a CT scan of a turbine fan blade. The objective is visual detection of the existence, location and 3-D distribution of cracks, voids, or other inhomogeneities which are physically significant. Below we show how these data are organized.

### EXPERIMENTAL (coordinate space): One property: Density distribution.

Figure 1. Schematic of Density Distribution Measured by X-rays with a Computer Tomography (CT) Scan System.

You can get more information about how this data set is organized on your optical disk, along with PV-Wave procedure files that will be used to visualize the data.

You can also get some background information, Problem Statement and Visual Insight, about this data set from the researcher who worked this data set.

Typical CT scanners can measure 512 slices parallel to the x-y plane where each slice contains 512 x 512 individual measurements points. The final result is a volume of 134,217,000 points. If each point (a volume of 20 x 20 x 20 um) is represented by 256 shades of gray or 256 colors then we require 8 bits of memory for each data point which will total 1 gigabits of memory for a complete scan. Obviously 3-D graphical tools can help us sort through this large data set.

## 2. Supercomputer Numerical Simulations

Another source of 3-D distributed data has come from supercomputer numerical simulations. The most popular has been in fluid mechanics. A variety of complex interrelated physical properties can be studied in 3-D simply because of the recent advances in massive parallel CPU architecture. On some computers each CPU corresponds to four finite difference mesh points where there can be up to 262140 mesh points or a 3-D cube with 64 mesh points on an edge. The idea is to approximate the fluid with a continuum mesh; the finer the mesh the more accurate the approximation. Although the volume of data is not as large as the CT scan data sets, fluid mechanic simulations predict how properties change with time. If we allow a calculation with 512 time-steps then we again accumulate a volume of 134,217,000 data points but now in space and time. Below is a numerical simulation of a chaotic mixed structure of jet fuel. Two data sets are studied at the 250th and 750th time step. Here we need 3-D graphic tools to study how the mixed structure changes with time.

### NUMERICAL (coordinate space): One property: Fuel concentration

Figure 2. Parameters Defined for Gas-Air Elliptic Jet Mixing Simulation

Example of how this ASCII data set was created and organized into 3-D format.

```       program name
integer image(64,64,44)
open(6,file='brown.ascii.start',status='unknown',err=888)

--  numerical calculations for fuel concentrations  --

--  Scale concentration into integers  ( 0 min.  to  255 max )  --

write(6,10)(((image(i,j,k),i=1,64),j=1,64),k=1,44)
10 format(64(1x,i3))
88 stop
end

```
You can view the FORTRAN program used to create a SDS-HDF file of this data set.

You can get more information about how this data set is organized on your optical disk, along with PV-Wave procedure files that will be used to visualize the data.

You can also get some background information, Problem Statement and Visual Insight, about this data set from the researcher who worked this data set.

Three dimensional graphical tools can also be used to study how distributed properties are interrelated in space and change with time. Below is a numerical simulation of two volumes mixing along a common interface where the relationships of density, temperature, and the u,v,w velocity components can be observed to change with time.

### NUMERICAL (coordinate space): Five properties: Density, temperature, and u,v, and w velocities

Figure 3. Parameters Defined for a Compressible Mixing Layer.

You can get more information about how this data set is organized on your optical disk, along with PV-Wave procedure files that will be used to visualize the data.

You can also get some background information, Problem Statement and Visual Insight, about this data set from the researcher who worked this data set.

## 3. Parameter Space

The advantages of using these 3-D graphic tools in coordinate space can be extended to parameter space. Again the source of data sets for parameter space can be from experimental, analytical, or numerical origins. The idea is the same, that is, to use the same 3-D visual tools we used in coordinate space to help in our visual thinking (analysis) of how properties are related to other parameters where these other parameters are substituted for the x,y,z coordinate axes, hence the name "parameter space". Parameter space visualization is only for those who are intimately involved with their data sets and require yet another dimension for interpretation and evaluation.

An excellent example is given below where we can observe the relationship of strain concentration at a fiber/matrix interface as a function of an approaching crack, surrounding fiber volume fraction, and the ratio of the fiber/matrix moduli. The objective here is to use the visual tools to better understand how these four properties (parameters) are related.

ANALYTICAL (parameter space): One property: strain concentration.

Figure 4. Parameters Defined for Composite Material Fracture Model where the Axial Strain in the Unbroken Fibers are Plotted as a Function of Crack Length, Fiber Volume Fraction, and Ratio of Fiber to Matrix Modulus.

You can get more information about how this data set is organized on your optical disk, along with PV-Wave procedure files that will be used to visualize the data.

You can also get some background information, Problem Statement and Visual Insight, about this data set from the researcher who worked this data set.

Because parameter space is often unique to the observer-researcher we will not elaborate on this example although 3-D parameter viewing of data has the most opportunity for visual discovery.

Next, let us see how researchers have used visual tools with these different data sources to gain an understanding of the physically significant relationships embedded in their data.

End of Information on Sources of Scientific Data Sets

Send comments to: rkriz@vt.edu
Ronald D. Kriz
Virginia Tech
College of Engineering
Revised April 4, 1995

http://www.sv.vt.edu/classes/ESM4714/data_sets/data_sets.html