Surrogate Assisted Version

Surrogate-assisted modeling aims to replace computationally expensive simulations or experiments with faster, more accessible approximations, thereby accelerating scientific discovery and engineering design. Current research focuses on developing and refining these surrogates using diverse machine learning techniques, including neural networks (e.g., convolutional, encoder-decoder, and Sobolev neural networks), Gaussian processes, and tree-based models, often incorporating physics-informed features to improve accuracy and generalization. This approach significantly impacts various fields, from optimizing complex engineering systems (e.g., electric machines, pressure vessels) to analyzing high-dimensional data in applications like structural health monitoring and wildfire prediction, enabling more efficient exploration of design spaces and improved understanding of complex systems.