The MALEAF project targets increase of the efficiency for the energy production and propulsion devices using rotating machines, in particular for marine or aeronautical applications. The objective will be obtained via the development of efficient and inexpensive analysis tools for hydrodynamic or aerodynamic problems. More precisely, advanced linear and nonlinear predictive reduced order models will be trained and calibrated by the use of data-driven methodologies. A distinctive feature of the project is that the models will be obtained via smart data instead of big data. Databases for data-driven training of the models will be obtained via data expansion techniques based on data assimilation, starting from high-fidelity dedicated numerical and experimental database obtained on two test-benches, namely an axial compressor and a hydrofoil in a water tunnel. A case of intermediate complexity (a compressor cascade) will also be studied numerically. The development of this project will also provide valuable information about open questions in data-driven applications in fluid mechanics such as efficient data usage and storage, systematic manipulation of sparse data and robustness of the training procedures.

A novel first principle approach will be proposed for dealing with multi-scale laminar, transitional and turbulent flows across weakly porous grids. Taking advantage of the separation of scales due to the small size of the grid pores, MultiMatchGrid will match the large far-grid to the small near-grid scales by means of numerical simulations and artificial intelligence. This will allow to perform high-fidelity simulations for regimes currently inaccessible to DNS and asymptotic theory. The new approach proposed by MultiMatchGrid will be tested against experiments and its potentials for prediction of challenging engineering flows will be assessed by comparing with bench top experiments. Potential applications of our scale matching approach include flow control in pipes, mixers and turbomachinery.