Model-Based Support for Mutable Parametric Design Optimization

Proceedings of the AAAI Conference on Artificial Intelligence, 16

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 16

Model-Based Reasoning

Traditional methods for parametric design optimization assume that the relationships between performance criteria and design variables are known algebraic functions with fixed coefficients (that correspond to user preferences). However, performance criteria may be dynamic, i.e., the functions and/or coefficients may depend on input tasks and run time behavior. We present a framework to support parametric, dynamic, design optimization using model-based reasoning techniques that derive event models to represent the effects of the system’s parameters on the material that flows through it. Next, we use these models to automatically discover relations between the system’s design variables and its optimization criteria dynamically. We then present an algorithm that searches for "optimal" designs by employing sensitivity analysis techniques on the derived relations.