Technologies for Enabling System Architecture Optimization

Abstract

Optimization of complex system architectures can support the non-biased search for novel architectures in the early design phase. Four aspects needed to enable architecture optimization and the author's views on how to solve them are discussed: formalization of the architecture design space, systematic exploration of the design space, conversion from architecture model to simulation model, and flexible simulation of architecture performance. Modeling the design space is done using the Architecture Design Space Graph (ADSG) implemented in ADORE. Systematic exploration can be done using evolutionary or surrogate-based optimization algorithms. Architecture to simulation model conversion can be done using an object-oriented approach using class factories, or using theMultiLinQ tool to synchronize a central data repository. Finally, simulation environments should expose a flexible and modular interface to be used in architecture optimization. A jet engine architecting problem is presented that demonstrates various aspects of system architecture optimization.