Department of Computer Science

10/26/1995

3:45pm-5:00pm

This seminar is an overview of a research project that is exploring two nontraditional approaches to engineering design. Each of these two threads is composed of several synergistic subprojects; these subprojects combine analytical mathematics, numerical simulation, artificial intelligence techniques, control theory, and physical experiments.

The first approach involves the active use of chaos to improve the analysis and design of various electrical, mechanical and chemical engineering systems. For example, the dynamics of the differential equation that is normally used to model the driven pendulum bear only a very general resemblance to the dynamics of real devices -- and nonlinear regression on time domain data is not a good enough tool to elucidate the differences. Nonlinear dynamics techniques like bifurcation analysis provide a much more powerful comparison. An example of how chaos can be exploited in design is the broadening of the capture range of the common phase-locked loop circuit, which was recently demonstrated in a physical circuit. Another potential application, which we are just beginning to explore, is to cause gases to burn more effectively by inducing chaotic mixing in a combustion chamber.

The second approach to improving engineering design is through computer automation. The ultimate goal of this endeavor is a suite of programs that explore different high-level analysis and design paradigms. The first step along this path, currently under development, is a program called PRET, which autonomously constructs mathematical models from information about a system, presented in a variety of forms and formats: measurements made directly on the physical device; a user's observations described to the program, in varying degrees of precision, symbolically or graphically; partial models and hypotheses suggested by a user; and general mathematical theory that is encoded in the program's knowledge base.

*Refreshments will be served prior to the talk.*

Department of Computer Science

University of Colorado Boulder

Boulder, CO 80309-0430 USA

webmaster@cs.colorado.edu

University of Colorado Boulder

Boulder, CO 80309-0430 USA

webmaster@cs.colorado.edu