A fatigue crack in the aluminium skin grows a little on every pressurization/turbulence cycle, following Paris' Law:
da/dN = C_eff · (ΔK)^m
ΔK = Y · Δσ · √(π·a)
K_IC = Y · σ_max · √(π·a_c)
Here a is crack length, ΔK the cyclic stress-intensity range, Y≈1.12 a standard edge-crack geometry factor, and m≈3, C≈6.9×10⁻¹² m/cycle are typical Paris constants for 2024-T3 aluminium. The panel fails once a reaches the critical length a_c set by the material's fracture toughness K_IC.
The nanorobot swarm represents an autonomous inspection-and-repair layer: each bot random-walks the skin sharing a simple gradient field (a stand-in for mesh communication) that biases it toward the crack front once damage is sensed. Bots that arrive at the tip deposit self-healing resin, which is modelled as reducing the effective Paris constant:
C_eff = C · max(0, 1 − bots_at_tip × repair_rate)
More bots on station, or a higher repair rate per bot, slow crack growth — enough of either can drive C_eff toward zero and effectively arrest propagation, exactly the "preventive, no-teardown" maintenance case the source article describes. Turning the swarm off removes this term entirely and the crack grows at the bare material rate.