ADVANCES IN ARTIFICIAL NEURAL NETWORKS NESTED IN FINITE ELEMENT ANALYSIS

To address the difficulties in accurately describing complex material behaviours using conventional empirical constitutive models, the feasibility and engineering significance of nesting artificial neural networks (ANNs) within finite element analysis (FEA) are investigated. The concept of machine learning nested in FEA is reviewed, focusing on ANN-based constitutive modelling, the challenges associated with nested training and backpropagation, and the related contributions of the authors. The reviewed studies demonstrate that nesting ANNs within FEA enables continuous model updating under physical constraints using structural- or process-level measurements, leading to improved agreement between numerical predictions and experimental observations. The nested ANN–FEA framework provides an engineering-feasible approach for data-driven constitutive modelling and digital twin development, showing strong potential for complex materials and multiphysics applications.