R. L. West
The concept of experimental spatial dynamics modeling (ESDM) embodies the experimental sampling of both surface shape and surface velocity of a structure under test by scanning lasers interfaced to an engineering workstation for data acquisition, modeling and scientific visualization. The goal is to derive statistically qualified spatial dynamics models of structures from in-the-field measurements. The resulting shape and dynamic response are experimentally derived spatial models that allow analysis, post-processing and visualization of the dynamics over the structure. These models are the experimental counterparts to the structure's existing finite element model.
The objective of the ESDM scalability project is to demonstrate the utilization of the distributed scalable computing resource through the use of supercomputer resources on campus for simulation and visualization in experimental structural dynamics. This research would particulary benefit from "realtime" simulation-visualization of 3D model predictions in a CAVE environment. The proposed demonstration plan follows in three phases:
* Research in ESDM is currently conducted by prototyping problem solving approaches and model formulations in Mathematica and graphics simulation using SGI's Inventor on laboratory SGI and HP workstations. The laboratory workstation computing level is predominantly used for model exploration, validation, data acquisition and signal processing and visualization.
* The process of reconstructing both the shape and dynamic response models from several thousand samples of shape and velocity is computationally intensive. Upon validation of a solution concept, critical computational elements in the problem formulation are coded as C++ objects and applied to the acquired data. The opportunity of utilizing a distributed scalable computing resource is directly compatible with the execution of prototype codes to model and post-process the large datasets.
* Reconciling the experimental data with computational models and model updating under a probabilistic framework requires computational resource at the supercomputer level. At this level the computational model must accommodate several thousand samples taken over a range of frequencies. Additionally, model order studies must also be carried out to obtain estimates on the "goodness" of the updated model. With CAVE resources much of these parametric studies could be done more efficiently in realtime with specially design 3D visualization CAVE tools.
This scalable approach to research oriented problem solving has worked well and is directly applicable to the concept of leveraging distributed scalable computing resources. Several elements of the overall ESDM concept have been realized and interfaced to "virtual" and working instruments on existing SGI and HP workstations and, in turn, used on industry problems. This project also targets the practical issues of utilizing the simulation-visualization computing resources for solving industry problems on a scale from laboratory workstations through university compute resources to the supercomputing level at NCSA.