Project Title/ID Number Advance Visualization for Seismic Performance—4182003
Start/End Dates 10/1/03—9/30/04
Project Leader Mike Bailey (UCSD/F)
Team Members

F=faculty; GS=graduate student; US=undergraduate student; PD=post-doc; I=industrial collaborator; O=other

Click on images to enlarge in a new window

1. Project Goals/Objectives:

One of the obstacles that have held back widespread integrative visualization has been the Tower-of-Babel problem. That is, everyone communicates their results in a different way, forcing specific specialty programs to be written to perform visualization. Visualization innovations often get applied to one specific problem, and are not available to the rest of the community.

We propose establishing a Structural Visualization Information Interchange (SVII) file format so that many different structural modeling systems can feed a variety of visualization applications. The general flow would look like this:

figure 1

Data-producing structural applications (purple) would make the SVII one of the formats that they can export. Data-consuming applications, such as visualization, would make the SVII one of the formats they can import. This would motivate innovation in the visualization domain because any significant effort could be amortized by use over many datasets. This would also encourage interoperability between simulation models as they can be displayed together and the results compared.

2. Role of this project in supporting PEER’s mission (vision):

 

3. Methodology Employed:

figure 2
For the past year, we have been working with the PEER project at the University of California Berkeley to develop and use a testbed SVII. (A copy of the running specification is included as a separate document.) From that effort we have learned the following:

Defining the format is hard work. It needs to be easy to produce, easy to parse, and covers all needs. The one we have been using has turned out very well for our needs. The following illustrations show two of our applications. The first is a frame vibration model. The second is a coupled bridge-soil model. We started with an ASCII format because:

  1. It is easy to produce,
  2. It is easy to parse,
  3. You can look at the file and see if you got what you think you should get, and
  4. It finesses any machine-dependent binary format problems. The ASCII format consisted of lines of text, consisting of a keyword and zero or more numerical arguments.

But, these files turned out to be huge and took quite a long time to read in to the visualization application. So, we also defined a binary format. The binary format consisted of the same keyword characters, but followed by the arguments in a 32-bit binary format. This made the files somewhat smaller, but the real improvement was in the read time, where we saw speedups of around 5X.

figure 3

We did this application as a standalone C++ program because we wanted to speed benefits of getting right into the OpenGL. We also wanted to experiment with some of the new OpenGL features, such as shaders.

For this OpenGL version, we used GLUT (GL Utility Toolkit) to manage interaction with the window system and GLUI (GL User Interface) for the user interface widgets. This allowed the program to be completely portable among Windows, Linux, UNIX, and Macintosh.

For some of our volume visualization applications, we had already added range sliders to GLUI. A range slider can specify a range of inputs. The user can control just the low end, just the high end, or both together. We found good uses for these in this program.

We realized that a web-based version of this same application would be possible. It could use Java for the program and Java3D or JOGL (Java-based OpenGL) for the graphics. It could also use an XML-based version of the SVII as the input, although this would exacerbate the file size and reading time problems.

We built stereographics into this program. In doing so, we realized that this feature is a must for this type of visualization where understanding spatial relationships is so critical.

4. Brief Description of past year’s accomplishments (Year 6) & more detail on expected Year 7 accomplishments:

 

5. Other Similar Work Being Conducted Within and Outside PEER and How This Project Differs:

 

6. Plans for Year 8 if project is expected to be continued:

figure 4We propose that the NEES community establish some form of SVII to encourage application interoperability and visualization innovation.

Both standalone and web-based visualization applications should be considered. Note that even standalone visualization applications can still use web-based file delivery mechanisms just by opening a socket connection. (We use this fir an SDSC image-delivery system.) So, both versions should use web-based databasing.

Some of the visualization innovations we would like to consider are:

  1. Immersiveness - there will be advantages if the program can also run in a multi-panel projector system, or a CAVE, or a dome.
  2. Volumes - problems such as soil liquefaction are better visualized as 3D volumes. Some of the new graphics hardware now available makes this practical in an interactive environment, as shown here with this underground rock layer model.
  3. Haptics - we would like to see experiments in haptic interfaces, where researchers can feel the vibrations as well as see them.
  4. Experiment with True 3D displays. We have a Perspecta in our lab. We would like to create a version of the Peer Visualization program that runs on it.

7. Describe any actual instances where you are aware your results have been used in industry:

 

8. Expected Milestones & Deliverables:

 

back to top