ThermoPINN Cell Lab

Set up a cell on its own, a phone battery inside a phone, or a row of cells along an EV pack's coolant channel. Physics-informed networks compute the temperature field in your browser in under a second, and every result says whether it sits inside the range the networks were trained and checked on.

What to simulate

Cell

Geometry
Parts and materials

Material properties

Values and sources come from data/materials.json. Edit any value and the cell is rebuilt and rerun.

Thermal properties
Cooling
Load

Loading the trained networks…

Temperature field

Drag to rotate · scroll to zoom
–
–
–

Can you trust this run?

    Temperatures

    Heat generated W per cell

    Cell current A

    Terminal voltage V

    State of charge %

    How the results were checked

    What runs here

    Two networks were trained on the heat equation alone, with no simulation data: one for radial cooling of a long cylinder and one for a plane wall. A finite cylinder is the product of the two; a pouch cell is the product of three plane walls. A Thevenin circuit model supplies the heat, and Duhamel superposition turns the unit responses into temperatures for any load.

    What it was checked against

    An independent finite-element model (scikit-fem, quadratic elements) solving the full 3D problem with no symmetry assumption, run through the same electro-thermal loop. The table lists the largest temperature difference over each run. This page's engine reproduces the Python engine to within 0.02 mK.

    What it is not

    A screening tool for uniform-heat, homogeneous cells with constant convection. It does not replace detailed FEM or CFD, does not predict thermal runaway or abuse, and is not a certification method. Preset properties and circuit parameters are illustrative placeholders, not measured data.