It is a pleasure to share another very recent paper by Nina Morozova V.Vanovskiy, Carles Oliet, E.Burnaev and myself:
N.Morozova, F.X.Trias, V.Vanovskiy, C.Oliet, and E.Burnaev. “A CFD-based multi-fidelity surrogate model for predicting indoor airflow parameters using sensor readings”. Building and Environment, 270:112533, 2025.
In this study, we introduce a multi-fidelity machine learning surrogate model that predicts comfort-related flow parameters in a benchmark scenario of a ventilated room with a heated floor. The model leverages both coarse- and fine-grid CFD simulations employing an LES turbulence model. To make the model practical, we mimicked real sensor placements by using temperature and velocity magnitude readings from just two specific locations as inputs. Parameters like the room’s width aspect ratio, inlet flow velocity, and heated floor temperature were varied to build a robust dataset.
All multi-fidelity methods significantly reduce computational costs while maintaining accuracy. Among these, co-kriging stood out for achieving the best balance between cost and precision. This work highlights the potential of multi-fidelity modeling in cutting computational expenses for applications in indoor climate optimization and beyond.
https://linkinghub.elsevier.com/retrieve/pii/S0360132325000150