Leveraging Physical Intelligence for the Self-design of High Performance Engineering Structures

Abstract

The design of complex engineering structures largely relies on computational intelligence in the form of science-based predictive models to support design decisions. This approach requires modeling and manufacturing uncertainties to be accounted for explicitly and leads to an inescapable trade-off of performance for robustness. To remedy this situation, a novel self-design paradigm is proposed that closes the loop between the design and manufacturing processes by leveraging physical intelligence in the form of real-time experimental observations. This allows the real-time product behavior to participate in its own design. The main benefit of the proposed paradigm is that both manufacturing variability and difficult-to-model physics are accounted for implicitly via in situ measurements thus circumventing the performance-robustness trade-off and guaranteeing enhanced performance with respect to standardized designs. This paradigm shift leads to tailored design realizations which could benefit a wide range of high performance engineering applications. The proposed paradigm is applied to the design of a simply-supported plate with a beam-like tuned mass damper (TMD) introduced to reduce vibrations based on an equal peaks performance criteria. The experimental setup includes a low-cost 3D printer driven by a simple decision algorithm and equipped with an online vibration testing system. The performances of a small population of self-designed plates are compared to their standardized counterparts in order to highlight the advantages and limitations of the new self-design manufacturing paradigm.