Self-Updating Digital Twin of a Semi-Industrial Furnace Via Data Assimilation Approach

Abstract

Data assimilation, i.e., upgrading a numerical model by using experimental observations, is applied to improve the performances of the digital twin of a semi-industrial combustion furnace. More specifically, we implement the technique through the Kalman filter, which allows us to adjust the prediction of our model by accounting for the underlying uncertainties. The digital twin comes from the reduction  (through Proper Orthogonal Decomposition, POD) and interpolation (Kriging) of several Reynolds-averaged Navier-Stokes simulations of the furnace. The experimental campaign concerns the measurement of the axial and radial temperature profile inside the chamber and the NO concentrations at the outlet of the furnace. This study is, perhaps, a one-of-a-kind early study on applying data assimilation to combustion. The results of this new data assimilation framework demonstrate its validity in the combustion field, encouraging its further development.

Keywords: Data assimilation, Kalman filter, Uncertainties, Combustion systems, Data-driven modelling, digital twins