Abstract
A major shortcoming to interval uncertainty approaches in computational mechanics is the lack of an interval field to represent uncertainty. This paper introduces the supervised interval field (SIF), a model to quantify spatially and temporally dependent uncertainty using machine learning. We introduce deep learning model architectures that are used to develop the SIF for domains of any dimension using deep recurrent neural networks. We demonstrate how to unify the SIF with the Interval Finite Element Method (IFEM) and show an experiment using soil layer data.
Keywords: Interval field, Spatially dependent uncertainty, Machine learning, Finite elements, Computational mechanics, Materials