A team of researchers at the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) has developed a revolutionary computational design system that enhances the performance of microstructured materials. These materials, ubiquitous in everything from cars to airplanes, offer essential durability and strength. This project, led by Beichen Li, a CSAIL affiliate and an MIT PhD student in electrical engineering and computer science, implements physical experiments, physics-based simulations, and neural networks to identify the optimal microstructures that are tougher and more enduring, with a perfect blend of stiffness and toughness.
The team utilised two kinds of base materials—one hard and brittle, and the other soft and ductile—and explored various spatial arrangements to find the best microstructures. This research, which is fundamental to engineering, could have a broad impact on numerous fields, including polymer chemistry, fluid dynamics, meteorology, and robotics, according to Li.
A vital innovation of their approach was employing neural networks as surrogate models for simulations, leading to a reduced need for time and resources in designing materials. As part of this process, they created 3D printed photopolymers and evaluated their strength and flexibility using the Instron 5984 standard testing machine. Alongside physical evaluations, the team used a high-performance computing framework to predict and tweak the characteristics of the materials, ensuring the perfect balance between strength and flexibility was achieved.
Further enhancing the effectiveness of their methodology was their “Neural-Network Accelerated Multi-Objective Optimization” (NMO) algorithm. This algorithm navigates the complex design landscape of microstructures and uncovers configurations that present near-optimal mechanical attributes. The workflow operates like a self-correcting mechanism, continually refining predictions to better align with reality.
Despite facing challenges in maintaining consistency in 3D printing and effectively integrating neural network predictions, simulations, and real-world experiments into a streamlined pipeline, the team is dedicated to making the process more accessible and scalable. Li envisions a future where labs are entirely automated to boost efficiency. This research was supported in part by Baden Aniline and Soda Factory (BASF).
