A team of researchers at MIT has made a breakthrough in creating materials that are tougher and more durable than traditional designs. By integrating physical experiments, physics-based simulations, and neural networks, they were able to discover microstructured composites that excel in stiffness and toughness.
The new system uses a combination of computational design, simulation, and machine learning to navigate the discrepancies between theoretical models and practical results. This approach allowed the researchers to find optimal materials by exploring various spatial arrangements of base materials with different properties.
Lead researcher Beichen Li stated that their methodology provides a blueprint for adapting this design approach to diverse fields such as polymer chemistry, fluid dynamics, and robotics. The team used neural networks as surrogate models for simulations, reducing time and resources needed for material design.
The researchers started by creating 3D-printed photopolymers and testing them using a standard machine. They then simulated the material characteristics before even creating them, using high-performance computing to predict and refine the results. The biggest breakthrough was in binding different materials at a microscopic scale, using an intricate pattern of minuscule droplets.
The new system achieves near-optimal mechanical attributes by navigating complex design landscapes. The workflow operates as a self-correcting mechanism, continually refining predictions to align closer with reality.
For the next steps, the team aims to make the process more usable and scalable, aiming for fully automated labs that minimize human supervision and maximize efficiency.
Source: https://www.techbriefs.com/component/content/article/52479-ai-discovers-stiff-and-tough-microstructures