Real ECLSS (Environmental Control & Life Support System) integration test campaigns move through levels: each subsystem alone ("unit"), then pairs, then subsystem groups, then the full system. The point of climbing through those levels — rather than jumping straight to a full-system test — is that some faults only exist at an interface between two subsystems and are physically invisible to a unit test.
This 2D network view lays the four subsystems out as the four corners of a graph with all six possible interface edges drawn between them (the same complete graph K₄ the 3D tetrahedron uses — a flattened square with both diagonals shows every pair exactly once). Each subsystem has a fixed base electrical/thermal load Pi, and every pair gets a coupling load that only switches on when both are active simultaneously (e.g. the water-recovery distillation heater raising the load seen by the power/thermal bus):
Combined load L = Σ Pᵢ·aᵢ + Σ(i<j) κᵢⱼ · S · aᵢ · aⱼ
Usable bus cap C' = C_bus · (1 − margin)
Result: FAIL if L > C', else PASS
ai ∈ {0,1} is whether subsystem i is switched on for this test, κij is a fixed interface-coupling constant per pair, S is the coupling-strength slider, and margin is the reserved capacity fraction. Because the coupling term vanishes whenever fewer than two subsystems are active, a single subsystem tested alone can never trigger a fault — the same reason a unit test passes while the full-system test fails once every subsystem is finally combined.
- Subsystem checkboxes — choose which mix is under test right now: one box = unit test, two = pair test, three = subsystem-group test, four = full-system test.
- Coupling strength — how tightly the subsystems' resource demands actually interact; raising it makes interface faults easier to trigger at higher integration levels.
- Design margin — capacity deliberately held in reserve during the test, exactly as spacecraft power/thermal budgets do; a stricter (higher) margin catches marginal interface faults that a looser budget would let slip to flight.
- Run Integration Test — animates the resource-flow check along every active interface edge and evaluates the load equation above; passing interfaces glow green, a faulting configuration glows red.
- Drag the network diagram to pan it and use the scroll wheel (or pinch) to zoom; the "Reset View" button snaps the camera back.
The coverage matrix below the diagram is the classic test-engineering artifact this whole workflow produces: a cell per subsystem pair, green once that interface has been exercised without a fault, red the moment it has faulted at least once, and grey while still untested — exactly the sign-off record a qualification review checks before flight.
Real-world relevance: this staged unit → pair → subsystem → full-system approach is exactly how NASA and ESA qualify closed-loop life-support hardware (e.g. ISS ECLSS) before flight — interface faults caught on the ground here are the ones that don't become in-flight anomalies.