The SURROG-GAIT project is developing a new AI-supported methodology for the simulation of composite structures at the functional level. Taking modern elastic prosthetic feet made of composite materials as an example, the project is researching surrogate models that can accurately and efficiently map complex deformation behaviour. This appraoch combines finite element method (FEM) simulation, biomimetic robot testing and multimodal measurement technology to generate high-quality training data for AI models. Based on this, AI surrogate models can be trained and integrated into multibody simulations, enabling precise, real-time simulation of complex structural behaviour.