Can anyone think of an situation where it would be necessary to install a cable which was inadequate for its own fault current - as in John's question.
Good question. SUNRAY has mentioned one case in which we would install a cable whose (upstream)
overload protection was theoretically 'inadequate' (a 2.5mm² spur from a 32A ring final which originated at the CU) - so the question remains as to how adequate (I would suspect very adequate!) the fault protection was ....
... I think it's probably true to say that most of us (certainly myself) do not usually give any appreciable thought to the adequacy of fault protection of a cable, our assumption being (almost certainly correctly) that if the overload protection is adequate (easy to determine), then fault protection will also be adequate.
If one does want to confirm adequacy of fault protection, then one has to undertake an adiabatic calculation and I think one of the main reasons why most of us don't do that (even when the 'confirmation' would be nice to have) relates to uncertainty as to what value of disconnection time ('t') to use for the OPD convened, since such figures don't really exist in the sources we usually look to (BS7671, OSG etc.).
If one considers a radial circuit with a Zdb of 0.35Ω (hence 'worst case' PFC, at the CU, of 657A at 230V or 723A at 253V) wired in 2.5mm² T+E and protected by, say, a B32, the adiabatic calculation (using 'k'=115) appears to indicate that the maximum acceptable disconnection time would be 0.191s at 230V/657A or 0.158s at 253V/723A - which I imagine is within the capabilities of a B32.
Looked at the other way around, if we consider PFC of 723A at 253V, then, if we assumed a disconnection time of 0.1s (the lowest plotted in the BS7671 graphs), then that would be adequate to protect a conductor down to about 1.99mm², whilst if we looked at 0.01s (some graphs go down that far), it would protect a conductor down to about 0.63mm².
However, interpretation of all that relies on knowledge of what 't' (disconnection time) we should be considering, which probably requires access to "I²t graphs".
Kind Regards, John