Peridynamic Neural Operator

Peridynamic Neural Operators (PNOs) are a novel class of data-driven models that leverage the nonlocal nature of peridynamics to learn complex material behaviors from experimental or simulation data. Current research focuses on developing PNO architectures for diverse applications, including constitutive modeling of heterogeneous materials like biological tissues and predicting brittle damage and crack propagation, often incorporating techniques like convolutional neural networks and long short-term memory networks. This approach offers advantages in handling complex material responses and noisy data while ensuring the preservation of fundamental physical laws, leading to improved accuracy and efficiency compared to traditional methods. The resulting models have significant potential for accelerating materials discovery and design, particularly in scenarios involving complex microstructures and failure mechanisms.