On Design Mining: Coevolution and Surrogate Models.

Area of Science:

• Computational intelligence
• Engineering design
• Evolutionary computation

Background:

• Design mining leverages computational intelligence for iterative optimization of physical object attributes.
• It facilitates exploration of novel materials and processes, bypassing traditional models and simulations.
• Complex design tasks are often decomposed into concurrent sub-design threads to manage complexity.

Purpose of the Study:

• To investigate the coevolutionary dynamics of decomposed design processes.
• To develop strategies for sampling sub-design threads for system-wide evaluation.
• To optimize the construction and utilization of surrogate models within coevolutionary frameworks.

Main Methods:

• An abstract, tunable model of coevolution was employed.
• Strategies for sampling sub-design threads were analyzed.
• The construction and use of surrogate models in coevolutionary scenarios were examined.

Main Results:

• The study identified effective strategies for managing concurrent sub-design threads.
• Optimal methods for integrating surrogate models into coevolutionary design were determined.
• The findings were applied to the design of a heterogeneous vertical-axis wind turbine array.

Conclusions:

• Coevolutionary approaches, coupled with effective surrogate modeling, enhance the design of complex systems.
• The proposed methods offer a robust framework for design mining.
• Successful application to wind turbine array design demonstrates practical utility.