Physics-Aware Soft Sensors for Embedded Digital Twins

In this paper we show some paradigmatic examples of physics-aware soft sensors for embedded digital twins, i.e. algorithms for edge computing that are formulated with a scientific machine learning approach, where scientific computing methods are blended with machine learning and, in particular, neural computing. A physics-aware soft sensor is a numerical algorithm that performs an indirect measurement by exploiting a physico-mathematical model plus a possible data-driven extension, both used within an estimation algorithm.