So where are we in terms of a simple risk analysis, say using parameters like likelihood, severity, duration?

Difficult. Statistical methods tend to make some assumtions or others - e.g. all starting conditions being equally likely - which may or may not be relaistic in practice.

For instance, if we had a situation where L1 and L2 had similar loads on them and L3 was unloaded (perhaps not unreasonable if the break affected say a pair of semis) then the 400V between L1 and L2 would be shared roughly equally between the two sets of loads, so each seeing something near 200V. Being lower than normal L-N voltage there would perhaps be little chance of loads frying and so "automatically disconnecting" themseleves as it were, so the situation might persist for some time. On the diagram N would be about halfway along the straight-line between the L1 and L2 points - so could well fall into the red zone (depending on the exact figures involved).

As Graham says, in practice things may well not be static though - even if nothing fries it's likely that some appliances will have change their consumtion from time to time - fridges will switch compressors off or on, thermostats on heaters will click. So things may swing into and out of the red areas. There's then the question of how the devices respond in that situation - do they latch off as soon as a voltage out-of-range event occurs or do they only remain switched off while the event persists and then re-close automatically when the voltage appears to be within range again? BS 7671 doesn't seem very specific on that point (it's not permitted to re-close while the voltage is still out-of-range, but doesn't specify whether than needs to be manual or can be automatic), and I can see arguments both ways - I can see a system that results in an entire neighbourhood finding their cars flat in a morning just because someone's vacuum clearer fused and the old BS 3036 fuse was a bit slow to clear the fault might not be entirely popular. C.p.c. current monitoring is certainly an extra safeguard, but not one that any standard currently insists on for these devices - so is that something we can rely on?

    - Andy.

So where are we in terms of a simple risk analysis, say using parameters like likelihood, severity, duration?

Difficult. Statistical methods tend to make some assumtions or others - e.g. all starting conditions being equally likely - which may or may not be relaistic in practice.

For instance, if we had a situation where L1 and L2 had similar loads on them and L3 was unloaded (perhaps not unreasonable if the break affected say a pair of semis) then the 400V between L1 and L2 would be shared roughly equally between the two sets of loads, so each seeing something near 200V. Being lower than normal L-N voltage there would perhaps be little chance of loads frying and so "automatically disconnecting" themseleves as it were, so the situation might persist for some time. On the diagram N would be about halfway along the straight-line between the L1 and L2 points - so could well fall into the red zone (depending on the exact figures involved).

As Graham says, in practice things may well not be static though - even if nothing fries it's likely that some appliances will have change their consumtion from time to time - fridges will switch compressors off or on, thermostats on heaters will click. So things may swing into and out of the red areas. There's then the question of how the devices respond in that situation - do they latch off as soon as a voltage out-of-range event occurs or do they only remain switched off while the event persists and then re-close automatically when the voltage appears to be within range again? BS 7671 doesn't seem very specific on that point (it's not permitted to re-close while the voltage is still out-of-range, but doesn't specify whether than needs to be manual or can be automatic), and I can see arguments both ways - I can see a system that results in an entire neighbourhood finding their cars flat in a morning just because someone's vacuum clearer fused and the old BS 3036 fuse was a bit slow to clear the fault might not be entirely popular. C.p.c. current monitoring is certainly an extra safeguard, but not one that any standard currently insists on for these devices - so is that something we can rely on?

    - Andy.