A physics-informed machine learning surrogate model for predicting lithium-ion battery pack temperature distribution in CubeSats under low-earth-orbit thermal cycling.
CubeSats in LEO experience ~5,500 charge/discharge cycles per year with eclipse-induced cold soaks reaching battery freezing limits. High-fidelity FEM/CFD thermal analysis is too slow for design-space exploration and real-time onboard estimation. This project develops a [PINN-based / POD-RNN / hybrid] surrogate trained against [planned simulation tool] to predict pack temperature distribution at orders of magnitude lower computational cost.
Early stage. Currently in literature review phase.
- Project scoping and literature review
- Baseline thermal model (Python finite-difference TMM)
- CFD validation case (battery pack geometry)
- Surrogate architecture selection and training
- Validation against high-fidelity simulation
- Hardware-in-the-loop test with thermal mockup (3D-printed)
- Writeup / preprint
literature/— annotated reading list and notesmodels/— thermal models and surrogate code (coming soon)data/— simulation outputs and training data (coming soon)notes/— engineering log and decisions
Emilio Barrera, Mechanical Engineering, Escuela Politécnica Nacional, Quito, Ecuador. Volunteer engineer with SpaceLab EC CubeSat program.