MATLAB · GRAPH THERMAL MODEL · SYNTHETIC DATA

Cooling fields,
made visible.

An independently implemented spatial electrothermal digital twin for testing how boundary shape, coolant direction, faults, and adaptive cooling influence a multi-cell assembly.

Open the thermal lab View methodology

Demonstration assumptions, not calibrated cell predictions. MATLAB R2026a local evidence.

Peak temperature

—from result manifest

Maximum spread

—across graph edges

Cooling effort

—energy proxy

Energy residual

—normalized accounting

INTERACTIVE RESULT EXPLORER

Follow heat through the assembly.

The animation reads the exported MATLAB result file directly. Choose a scenario, play it, or scrub one frame at a time.

Simulation time0 s

Inspect current frame as data

Cell temperatures for the selected scenario and simulation time.
Cell	Row	Column	Temperature

ONE SCHEMA, FOUR BOUNDARY IDEAS

The boundary is part of the experiment.

MODEL

Transparent enough to interrogate.

01 / ENERGY BALANCE

Cᵢ Ṫᵢ = Q_gen,i − Σ Gᵢⱼ(Tᵢ−Tⱼ) − G_b,i(Tᵢ−T_b,i) − G_c,i(Tᵢ−T_c,i)

Each cell is a graph node. Conductive edges, ambient interfaces, and coolant links all expose their sign convention and SI units.

02 / ELECTRICAL HEAT

Heat follows current, SOC, and temperature.

Joule heating is always available; an optional signed entropic term can be enabled. Parameters remain visibly synthetic unless a licensed validation dataset is added.

03 / COOLANT MARCH

The outlet remembers the inlet path.

A finite-capacity flow balance marches coolant through declared cell segments. Reversing the channel order changes the spatial field and is covered by an automated test.

04 / CONTROL

Bounded, rate-limited, observable.

The baseline zonal controller is deterministic. Optional optimization reports solver status and falls back visibly—never silently—when it cannot produce a command.

VERIFICATION

Evidence before adjectives.

Class-based MATLAB tests cover configuration failure modes, equilibrium, relaxation, energy accounting, scalar/vector equivalence, graph permutation behavior, flow reversal, controller constraints, scenario-specific benefit, faults, deterministic seeds, and export traceability.

Read the verification record →

31/31

local tests passed

MATLAB R2026a Update 4 · 0 failed · 0 incomplete

ARCHITECTURE

One result contract. Multiple consumers.

Validated config→Graph + boundaries→ODE solver→result/v1→Tests · UI · Web

LIMITATIONS

Useful because its boundaries are explicit.

Lumped isothermal cell nodes; no electrochemical diffusion or ageing prediction.
Synthetic parameters; no held-out real-cell accuracy evidence.
Controller demonstrations are scenario-specific and are not safety certification.
Simscape products are installed locally, but structural model generation remains policy-gated and unclaimed.

REPRODUCE

Clone. Run. Inspect every assumption.

startup
result = runDemo
buildtool test
generateArtifacts

Repository path Reproducibility guide

Source and verification evidence are published on GitHub. The 31-test suite passed locally on MATLAB R2026a, and the pinned core workflow completed successfully on R2024a.