Active-Learning Combined with Topology Optimization for Top-Down Design of Multi-Component Systems

DS 116: Proceedings of the DESIGN2022 17th International Design Conference

Year: 2022
Editor: Mario Štorga, Stanko Škec, Tomislav Martinec, Dorian Marjanović
Author: Lukas Krischer, Anand Vazhapilli Sureshbabu, Markus Zimmermann
Series: DESIGN
Institution: Technical University of Munich, Germany
Section: Artificial Intelligence and Data-Driven Design
Page(s): 1629-1638
DOI number:
ISSN: 2732-527X (Online)


In top-down design, optimal component requirements are difficult to derive, as the feasible components that satisfy these requirements are yet to be designed and hence unknown. Meta models that provide feasibility and mass estimates for component performance are used for optimal requirement decomposition in an existing approach. This paper (1) extends its applicability adapting it to varying design domains, and (2) increases its efficiency by active-learning. Applying it to the design of a robot arm produces a result that is 1% heavier than the reference obtained by monolithic optimization.

Keywords: topological optimisation, artificial intelligence (AI), data-driven design, systems engineering (SE)

Please sign in to your account

This site uses cookies and other tracking technologies to assist with navigation and your ability to provide feedback, analyse your use of our products and services, assist with our promotional and marketing efforts, and provide content from third parties. Privacy Policy.