Session: S03: SYMPOSIUM IN MEMORY OF PROFESSOR RICHARD J. GOLDSTEIN III

Paper Number: 134321

134321 - Advanced Thermal Management Using a Generative Multi-Layer Approach 

Abstract: 

As transistors approach the nanometre scale, mirroring Moore's Law, a surge in power density challenges effective heat flux management in densely packed electronic systems. To address these challenges, we present an AI generative design framework that fuses topology optimization, constructal theory, and bio-inspiration. This tool is then used to streamline the development of passive heat transfer solutions with thermal metamaterials, aligning with recent advances in multi-layer manufacturing.

We first demonstrate the use of this synergistic framework to mitigate the challenges of navigating vast combinatorial design spaces, such as local optima convergence and initial conditions sensitivity. Our approach improves Pareto front hypervolume, enhancing heat extraction capacity while promoting temperature uniformity in conductive heat sinks for the area-to-point problem. We then boost modelling resolution using a hierarchical design strategy for complex metamaterial structures to handle multiple heat flux and temperature objectives.

The modularity of the framework allows versatile design exploration, generating continuous spatial shapes through latent space interpolation, tailoring samples for additive manufacturing. This enables high-throughput production of optimized multi-material structures with tuneable thermal properties.

This research initiative establishes AI as an essential component in a comprehensive workflow for optimizing, designing, and manufacturing thermal metamaterials The generative approach shows promise in improving thermal passive cooling across applications like printed circuit boards, solar collectors, and power electronics.

Presenting Author: Matei-Cristian Ignuta-Ciuncanu Imperial College London

Presenting Author Biography: Matei is a driven graduate student with a keen interest in thermal management through the use of innovative computational tools like generative design. Having completed an MEng degree at Imperial College London in 2022, he distinguished as a Dean’s List student and developed an acute interest in academic research.

With a passion for applied mathematics and science, Matei specialised in areas like thermodynamics, fluid mechanics, and computer science. His academic journey has been marked by curiosity and commitment to pushing the boundaries of technology, particularly in the context of sustainable energy solutions.

Authors:

Matei-Cristian Ignuta-Ciuncanu Imperial College London
Ricardo Martinez-Botas Imperial College London