AI Optimizes Mycelium Composites with Graded Microstructures

A new research paper introduces an advanced method for optimizing sustainable materials, specifically mycelium-woodchip composites, by leveraging AI-driven surrogate models. These models are 'conditioned' on microstructural variables using a hypernetwork, allowing for accurate predictions of multiscale mechanical behavior even with limited training data. This approach addresses the challenge of efficiently simulating and optimizing materials with engineered hierarchies and microstructures, which typically require extensive data. The conditioned surrogate model makes it feasible to perform multiscale simulations for functionally graded structures, which was previously computationally intensive. As a demonstration, the researchers optimized a graded multiscale disk, achieving a 42% reduction in peak stress compared to a randomly structured counterpart. Furthermore, the network can be conditioned directly on manufacturing variables, offering a practical pathway to engineer microscale properties for desired macroscale performance. This work highlights the potential of microarchitectured structures and how conditioned surrogate models can accelerate the development and design of future sustainable materials.