The MAXIMA consortium has released Deliverable D4.2, which defines the electromagnetic, thermal and mechanical low‑order models that underpin the project’s Digital Twin for an axial‑flux permanent‑magnet machine. This preliminary version documents the modelling methodology and validation work; a more complete document will follow once all results are officially published.
Low‑order electromagnetic model from 3D FEA
The first part of the deliverable presents a methodology to derive a real‑time capable electromagnetic low‑order model from a 3D finite‑element (high‑fidelity) model of the axial‑flux machine. The reduced model preserves access to quantities such as flux, torque, iron losses, permanent‑magnet losses and winding losses, enabling its use both for control and for loss‑driven optimisation in the Digital Twin.
Data‑driven thermal model based on CFD and ML
The second part focuses on a low‑order thermal model built using a machine‑learning workflow trained on computational fluid dynamics (CFD) simulations. This approach provides fast predictions of temperature distributions inside the machine while maintaining sufficient accuracy to assess risks such as permanent‑magnet demagnetisation and insulation ageing.
Mechanical model and prognostics for bearings
The third part develops a mechanical model for the machine’s roller bearings, combining physics‑based modelling with data‑driven prognostics to estimate remaining useful life (RUL). The thermo‑mechanically coupled model supports real‑time prognostics of bearing health within the axial‑flux machine, enabling condition‑based maintenance strategies.
Foundation of the MAXIMA Digital Twin
Together, the three low‑order models form the core of MAXIMA’s multiphysics Digital Twin, which will be used to monitor internal temperatures, predict losses and assess component degradation during operation. By extracting reduced models from high‑fidelity simulations, the consortium ensures that the Digital Twin retains high accuracy while remaining suitable for real‑time automotive applications.