Alessandro Chiodi

This work describes the experimental and numerical validation of a Multi-Fidelity (MF) model for the efficient simulation of Heat Exchangers (HX) based on Triply Periodic Minimal Surface (TPMS) lattices. The MF model is based on the porous media approximation developed in our previous work, where the porous lattice is modelled as a homogeneous medium characterised by variable permeability, Forchheimer, and heat transfer coefficients. Lattice coefficients are assumed to be dependent on the local lattice geometry and flow conditions, and are computed by a Machine Learning (ML) model. The ML model is trained on a database of LF simulations performed on individual lattice cells and a handful of High-Fidelity (HF) runs data. The resulting framework allows to run simulations at a fraction of the computational cost of a full-resolution CFD simulation. The validation of the proposed model is performed against both HF Conjugate Heat Transfer (CHT) simulations and experimental data. Experiments are performed on a series of Additively Manufactured (AM) HXs, tested under controlled thermal Boundary Conditions (BCs), at different inlet mass flow rates. Comparison with experimental data shows an excellent agreement for temperature distribution. A qualitative comparison with experimental measurements indicates that also the pressure drop is predicted correctly by the MF model. In particular, the MF model is able to capture the order of magnitude of the pressure drop and the overall trend with respect to the inlet mass flow rate. These results confirm the accuracy of the MF surrogate model and demonstrate that the proposed approach provides a reliable and computationally efficient tool for the simulation and optimisation of TPMS-based lattice HXs.