Automating scientific and engineering computing

Abstract

This paper describes research on automating scientific and engineering computing by combining domain knowledge,mathematicaltheory, artificial intelligence techniques, and numerical software. Wepresent two case studies: kinematic analysis of mechanisms and phase space analysis of dynamical systems. The case studies illustrate our general strategy of identifying an important task domain, formalizing the domainknowledge and analysis tools of experts, and incorporating the formalization into a useful problem solver. The problem solvers could not workwithout a robust, high-level interface to conventional numerical software. We describe an initial interface that managesa root finder, a continuation package, an ordinary differential equation integrator, and a Lyapunov exponent calculator.