AI helping to create new rugged microstructures

Each journey from point A to B is more than a mere transition; it's an experience made safer and more reliable through sophisticated engineering marvels that underpin our vehicles. At the core of this engineering prowess lies the world of microstructured materials – unsung heroes that bolster our vehicles, ensuring they withstand the rigours of travel with unmatched durability and strength.

In an exciting development from the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory (CSAIL), a group of forward-thinking researchers has ventured beyond the conventional trial-and-error fabrication methods. They have introduced a revolutionary approach to material creation, combining the realms of computational design, physical experiments, and advanced neural networks. This trifecta of technological integration has led to the creation of microstructured composites that redefine the benchmarks for toughness and durability in engineering materials, finding applications in industries as varied as automotive and aerospace.

At the heart of this breakthrough is the ingenuity of Beichen Li, a doctoral candidate in electrical engineering and computer science at MIT, a CSAIL affiliate, and the project's lead researcher. Li and his team have ushered in a new era of composite design and fabrication, one that promises to extend its influence far beyond the confines of solid mechanics. “Our work's implications could reach into diverse fields such as polymer chemistry, fluid dynamics, meteorology, and even robotics,” Li elucidates. Their work stands as a testament to the power of computational design, offering a versatile blueprint adaptable across a myriad of scientific domains.

In the intricate dance of atoms and molecules that constitutes materials science, finding the perfect partnership between elements is crucial. Li's team focused on marrying stiffness with toughness, two paramount material properties. They embarked on an exploration within a vast design space, juxtaposing hard, brittle materials against soft, ductile counterparts, in search of the ideal microstructures through various spatial arrangements.

One of the project's innovations lies in its employment of neural networks as surrogate models for simulations, a strategy that streamlines the material design process. “Utilising an evolutionary algorithm, enhanced by neural networks, we can efficiently pinpoint the highest-performing samples,” Li shares, highlighting the efficiency and precision of their approach.

The team's methodology commenced with the fabrication of 3D printed photopolymers, each no larger than a smartphone but considerably thinner. These were then subjected to a unique ultraviolet light treatment, followed by tensile testing using the Instron 5984 machine, to assess their strength and flexibility. This physical testing, paired with advanced simulations within a high-performance computing framework, allowed for the prediction and refinement of material characteristics with unprecedented accuracy.

A pivotal aspect of their research was the development of a method to coalesce different materials at a microscopic scale. This process, characterised by a complex pattern of minuscule droplets that amalgamated rigid and supple substances, achieved a harmonious balance between strength and flexibility. The close correlation between simulation outcomes and physical testing results underscored the methodology's validity.

The culmination of their efforts is the “Neural-Network Accelerated Multi-Objective Optimisation” (NMO) algorithm. This tool navigates the intricate landscape of microstructure design, uncovering configurations that exhibit near-optimal mechanical properties. This algorithm acts as a self-optimising mechanism, constantly refining its predictions to more closely mirror actual results.

Despite their achievements, the journey has been fraught with challenges, particularly in achieving consistency in 3D printing and melding neural network predictions with simulations and real-world experiments into a cohesive workflow. Looking ahead, Li and his team are dedicated to enhancing the usability and scalability of their process, envisioning fully automated labs that minimise the need for human intervention and maximise efficiency.

Contributing to the research alongside Li are senior author and MIT Professor Wojciech Matusik, Associate Professor Tae-Hyun Oh from the Pohang University of Science and Technology, and MIT CSAIL affiliates including Bolei Deng, now an assistant professor at Georgia Tech; Wan Shou, now an assistant professor at the University of Arkansas; Yuanming Hu, Yiyue Luo, and Liang Shi. Their groundbreaking work has received support from the Baden Aniline and Soda Factory (BASF), underscoring the collaborative effort behind this innovative venture into the future of material science.