Physics assisted knowledge representation in process systems engineering

Advances in Machine Learning (ML) algorithms and automatic differentiation (AD) have facilitated knowledge representation of various cause–effect relationships, that are conventionally represented by PDEs (higher order), to be captured in nonlinear mapping frameworks such as Neural Networks (NN).  Using such knowledge representation frameworks, physics laws/mechanistic models are modeled using data-driven learning frameworks to give rise to a new class of deep learning approach known as Physics-Informed Neural Networks (PINN)s.  Research in my group has focused on incorporating physics in knowledge representation systems to develop robust representations of complex behaviour for the purposes of advanced decision making.