Research Associates and Assistants: S. Crabill, S. Weerawarana
Sponsor: ARPA, NSF
One of the next "Grand Challenges" for computer applications is creating a system for the design and analysis of physical objects. Example applications include mechanical design, biomedical design, and so on. These systems will provide accurate computer simulations of physical objects coupled with powerful design optimization tools to allow prototyping and final design of a broad range of items. When such systems become a reality, they will have an even greater impact than systems for electronic design, one of the great achievements of computing technology of the past decade.
It is recognized that the design of physical systems is an iterative process where typically the geometry of the frames is modified, perhaps often. This is the result of some design optimization with respect to some static or dynamic criteria. Creating modeling systems whose geometry changes with time is a challenging problem that is impossible to solve without some "intelligent" support from the software components of the system.
It has been observed that the design and simulation of complex physical systems is a very computationally intensive process that requires massive computational power. It is known that realistic 3-D problems cannot be accurately modeled by existing computers. Thus, it is necessary to implement Electronic Prototyping for Physical Object Design (EPPOD) systems on massively parallel hardware platforms.
This project has the following two main thrusts:
Development of "Smart" Problem Solving Environments for Electronic Prototyping for Physical Object Design (EPPOD)
Future design environments of physical systems must operate in an automatic mode, support users with different computational objectives (students, engineers/scientists, experts) and carry out a distributed (parallel) analysis of the physical system transparently from the user on multilevel software/hardware platforms. A fully automatic system is the one where the user specifies the design/simulation problem at a high level along with some design (accuracy) tolerances and computational objectives. An interactive automatic system is one that allows some form of dynamic communication between the user and the computational engine. It is clear that the most natural forms of machine-user interaction are visual and audio.
One of the main research issues here is to identify the most efficient computer representation of the "continuous/discrete" atomic components of the physical systems that permits the implementation of such design/simulation environments. In this study we are primarily interested in the design and simulation of physical systems defined over "flexible or adaptive" geometric regions. The atomic components of these systems tend to correspond to the well defined substructures (atomic components) of the geometry. Thus, for the implementation of problem solving environments for such physical problems, it is natural to include a geometry-based platform that includes high-level geometry-based problem specification and control, powerful data structures, and the geometric functionality that is necessary to support automation, interaction, parallelization, simulation, debugging, diagnosis, and visualization of numerical simulation.
Development of High Performance Scalable Algorithmic Infrastructure for Solving PDES
The trends in computer technology for large scientific computation are characterized by the diffusion of multiple processors cooperating synchronously or asynchronously towards the execution of an application program.