This paper is a wide-ranging survey of how artificial intelligence (AI) methods are being applied to materials plasticity — the permanent shape change that materials undergo under stress. The authors bring together work from materials science and AI to build a clear taxonomy of methods and to explain how data-driven tools are being used to discover, model, and emulate plastic behavior.

From the materials side, the review covers the physical causes and effects of plastic deformation. That includes how microstructure (the tiny internal features of a material) links to macroscopic response, and how these responses are described by plasticity constitutive models (mathematical rules that predict how a material yields and flows). From the AI side, the authors examine a broad range of approaches: classical machine learning, deep learning, physics-informed models (which embed physical laws into AI), and probabilistic or generative methods that treat uncertainty explicitly.

The paper organizes existing work around practical concerns. It reviews methods that find new relationships in data, build surrogate models (faster, approximate models that stand in for costly simulations or experiments), and emulate material behavior. The authors place special emphasis on comparing model architectures, on the amount and type of data each approach needs, and on how well methods perform at prediction within the specific domain of plasticity.

This survey matters because it offers a roadmap for researchers and engineers who want to apply AI to problems in materials plasticity. By grouping techniques and spelling out their data and performance trade-offs, the paper aims to make it easier to choose appropriate tools and to gain physical intuition from data-driven models. The review also highlights approaches that explicitly include uncertainty, which is important when models will guide experiments or design decisions.