Deep active learning for constitutive modelling of granular materials: From representative volume elements to implicit finite element modelling

Constitutive relation remains one of the most important, yet fundamental challenges in study granular materials. Instead using closed-form phenomenological models or numerical multiscale modelling, machine learning has emerged as an alternative paradigm to revolutionise constitutive modelling However, deep neural networks (DNNs) require massive training data and often fail make credible extrapolations. This aims develop a active strategy (i) identify unreliable forecasts without knowing ground truth; (ii) continuously improve verify data-driven model until desired generalisation is satisfied. The role instantiated through three scenarios: off-line strain-stress pool materials; interactive labelling; (iii) finite element (FEM) driven by learning-based models. results confirm capability advancing materials toward developing faithful surrogate with less data. same can also be applied other data-centric applications across various science engineering fields.