Non-intrusive system state reconstruction from indirect measurements: A novel approach based on Hybrid Data Assimilation methods

The problem of estimating in real-time the state of a system by combining experimental data and models has been extensively addressed in literature. In particular, there have been a lot of developments in reduced order modelling techniques integrated in a data assimilation framework. This coupling allows to reconstruct the variable of interest considering a priori knowledge (i.e. the mathematical model) and some measurements of it. However, in some engineering systems not all the variables of interest may be measurable and hence the typical framework cannot be applied. However, these methods can be extended to reconstruct the full state of a system, by means of partial observations only. In this work, a novel approach will be introduced, based on a two step method: first the measurements are used to determine the parameters describing the system; then, the full state is estimated by means of reduced order modelling techniques. This new approach will be compared with the state-of-the-art both on a simple thermal hydraulics case and on an innovative nuclear reactor design, the Molten Salt Fast Reactor. •Application of the GEIM vectorial strategy to thermal–hydraulic problems.•Indirect Reconstruction algorithm for reconstruction of non-observable fields.•Numerical benchmark (differentially heated square cavity): reconstruction errors.•Application to the Molten Salt Fast Reactor (MSFR) core (Loss of Heat Sink transient).