Exploring Various Sequential Learning Methods for Deformation History Modeling

Conference : Engineering Applications of Neural Networks Title : Exploring Various Sequential Learning Methods for Deformation History Modeling Abstract : Current neural network (NN) models can learn patterns fromdata points with historical dependence. Specifically, in natural languageprocessing (NLP), sequential learning has transitioned from recurrence-based architectures to transformer-based architectures. However, it is un-known which NN architectures will perform the best on datasets contain-ing deformation history due to mechanical loading. Thus, this study as-certains the appropriateness of 1D-convolutional, recurrent, and transfor-mer-based architectures for predicting deformation localization based onthe earlier states in the form of deformation history. Following this in-vestigation, the crucial incompatibility issues between the mathematicalcomputation of the prediction process in the best-performing NN archi-tectures and the actual values derived from the natural physical proper-ties of the deformation paths are examined in detail. Not : These are inital codes this repisotory will be updated completely.