Lowdown LabsFELA PDE
Fourier neural operator that predicts the steady-state 2D temperature field of a battery pack from its layout, heat load and cooling, standing in for a finite-volume solve on a plain CPU.
Model details
- Architecture
- Fourier neural operator (892,545 parameters)
- Input
- 8-channel 96×96 field (pack mask, heat source, conductivity, convection, ambient temperature, coordinates, domain size)
- Output
- 96×96 temperature field (°C)
- Training data
- Synthetic
- Trained on
- Self-generated battery thermal simulations
- Uncertainty
- Not reported
- Physics constraints
- None
- License
- other
Limitations
- A screening and design aid, not a certified thermal safety tool; validate against a full solver before relying on it.
- Trained and evaluated only on its own generated thermal data.