Design Optimization with Uncertain Application Knowledge

Ravi Kapadia, Markus P.J. Fromherz

Abstract

Designing modern electro-mechanical devices is a non-trivial task, due to their versatility, their increasing performance requirements, and the integration of software and hardware. Competition in the market place motivates additional requirements, such as high performance and low cost, and an important part of design is often to optimize conflicting objectives.

In this paper, we present model-based techniques for design optimization. We develop a parameterized machine model in the form of design, scheduling and cost constraints and present a tool that determines values for design variables that optimize the performance and cost of the machine.

Furthermore, we present a method for analyzing design solutions in the absence of precise workload knowledge. A qualitative method for classifying optimal sets of machine parameter values with respect to workload distribution, helps the designer to understand the interaction between potential application contexts and optimal solutions, and may be used to generate configurable designs.

Postscript file