I am also not aware of any special hazards from this. A ground fault on one phase that blows a fuse would probably blow the other fuses. Single phasing will certainly increase the load on the other two phase conections and in my experience is usually due to loose conections or contact faults. 

As an aside this was a common way to get a small amount of three phase from a single phase supply. Find a suitable size three phase motor and arrange a single phase supply to two phases. Wrap some string around the motor shaft and give a good tug to get the motor spinning then turn on the single phase power. Three phase power was then available at the motor terminals. Don't try this at home unless you have a small industrial machine you want to use in your home workshop.

The protection for the other two phases will operate due to overload, but that's a lot slower than fault protection.

The problem is that the prospective fault current may be orders of magnitude lower than that from the grid, and fault protection may not therefore operate or if it does it will be delayed. In that time conductors overheat and insulation begins to fail. 

This used to be a very common way to get 3 phase from a single phase supply, these days of course we just use a drive. A pull string start is one way, but you can just use a motor start capacitor as you would with a single phase induction motor. If this condition occurs when the motor's already running then it will just continue on. 

it will be delayed. In that time conductors overheat and insulation begins to fail. 

If that can happen then surely your overload protection is flawed?

I do that the point that ADS might be somewhat compromised though if the voltage at the point of the fault can be fed from downstream of the protective device (maybe an issue even with some single phase motors spinning down, or grid-connected generators/inverters). I guess the simple answer is to use a protective device that opens all poles, including N (if any) so the remaining voltage is at least separated from Earth.

   - Andy.

unlike the grid,  the energy available from the spinning motor is limited - indeed as a 3 phase creator it is distinctly floppy -that's why we only do it if there is no easy way to get a proper 3 phase supply.

  There may be enough current to hurt someone on the created phase, at least from a big motor, but probably not to worry breakers with asymmetric arc traps or burn out wiring or anything like that.

Mike.

AJJewsbury said:

I guess the simple answer is to use a protective device that opens all poles, including N (if any) so the remaining voltage is at least separated from Earth.

Ideally yes. But imagine a fault upstream of the DB protecting the motor - that's more likely to be protected by e.g. BS88 fuses which by their nature don't provide all pole disconnection.

It would be very difficult or impossible with convention devices to meet ADS going backwards from the motor because Zs from the motor is likely to be very high.

I suspect the voltage won't hold up very well into a fault, so maybe not a shock risk. But it's still going to be feeding current which could overheat conductors.

And the basic principle of ADS doesn't work very well if a secondary source of supply isn't also disconnected.

Maybe I'm completely overthinking this. But there seems to be an absence of literature on this failure mode - either that or I'm looking in completely the wrong places.

30 years ago rotary converters were something you could buy. Now we are so dependent on drives that you get a weird fault and misbehaving motor running amok and it feels like I need more of a historian than an engineer.

AJJewsbury said:

I guess the simple answer is to use a protective device that opens all poles, including N (if any) so the remaining voltage is at least separated from Earth.

Ideally yes. But imagine a fault upstream of the DB protecting the motor - that's more likely to be protected by e.g. BS88 fuses which by their nature don't provide all pole disconnection.

It would be very difficult or impossible with convention devices to meet ADS going backwards from the motor because Zs from the motor is likely to be very high.

I suspect the voltage won't hold up very well into a fault, so maybe not a shock risk. But it's still going to be feeding current which could overheat conductors.

And the basic principle of ADS doesn't work very well if a secondary source of supply isn't also disconnected.

Maybe I'm completely overthinking this. But there seems to be an absence of literature on this failure mode - either that or I'm looking in completely the wrong places.

30 years ago rotary converters were something you could buy. Now we are so dependent on drives that you get a weird fault and misbehaving motor running amok and it feels like I need more of a historian than an engineer.

Ideally yes. But imagine a fault upstream of the DB protecting the motor - that's more likely to be protected by e.g. BS88 fuses which by their nature don't provide all pole disconnection.

It would be very difficult or impossible with convention devices to meet ADS going backwards from the motor because Zs from the motor is likely to be very high.

:

And the basic principle of ADS doesn't work very well if a secondary source of supply isn't also disconnected.

That's the point of all-poles disconnection (e.g. 4-pole MCB or even fuses with a phase failure relay) - the grid supplies the fault current on the initial fault to cause disconnection, then with all poles including N disconnected, it doesn't matter if there's voltage generate from the motor, since it's disconnected from the system's N-PE link so it's effectively separated from earth - so no shock hazard (for conventional shock situations at least) even if the fault itself is to PE.

I suspect the voltage won't hold up very well into a fault, so maybe not a shock risk. But it's still going to be feeding current which could overheat conductors.

But that current should still be flowing though the motor's (usually local) overload protection (usually built into the starter) - so that should provide more than adequate protection to the supply cables if the fault is between the DB and starter.

   - Andy.

AJJewsbury said:

That's the point of all-poles disconnection (e.g. 4-pole MCB or even fuses with a phase failure relay)

Phase failure relays detect undervoltage conditions and operate when a phase's voltage has dropped. Maybe they would work because the voltage will be all over the place, they probably will but maybe they won't. 

Using a starter with an integrated overload would help, but I think it's only going to tickle the thermal and not the magnetic trip. One of my favorite graphs here from The IET design guide nicely illustrates how a low fault current is very bad news, these curves are fuses but they're not that different to the thermal trip on a breaker. If your prospective fault current is low then you'll still trip the protection but you'll do it after the adiabatic equation tells you that you may have melted some insulation somewhere.

The beauty of ADS is that if I go and do something stupid like hammer a nail into a cable, the ADS should operate quick enough that the damage is limited to where my nail went in. You should be able to get an electrician to cut out a few cm of damaged cable, resplice and the installation should have suffered no secondary damage. If an induction motor is nearby then that might not hold. And I'm not sure we can bank on the fault being between the motor and the source of supply - it's far more likely to be on a parallel branch.

Keir.

Missing from those curves however is the fact that the blue line is only true for faults that are so fast  there is no time for any of the heat generated to leave the copper and start spreading into the insulation.

Consider adding some additional lines  at the steady state current rating of each of the core sizes, you can then  join them to the blue adiabatic lines for  the same core size, and sketch in a bit of a curve at the intercept.

If you do this you can also see that the 'it is always 5 seconds'  assumption is also a pretty  poor rule of thumb for sizes that are at the extremes - looking at the intercepts - for 16mm2 it is more like a minute and a half and for 1mm2 perhaps 30 seconds. Perhaps half to a quarter of this time is the "safely adiabatic" limit.

All of these things (the adiabatic lines and the steady state current ratings) are just comfortable working assumptions, and if you know exactly what you are doing, for short/modest duration loads you can dissipate a lot more without endangering the cable,  compared to what the simple rules suggest.

Actually this is something which we know from experience - for example  2,5mm flex will supply a short shower at 10kw , so long as the person stays under the shower for only  a few mins, and you have to let it cool for at least 4 times that time before doing it again.
Not that anyone in their right mind would design that of course, but it is the sort of silly thing that you find that has been working for years when installations have been extended without too much thought.  In the same way motor inrush etc does not normally do anything to the cable even on motors that take more than 5 seconds to spin up or down, as these tend to be on the heavier cables. The combined steady state rating and adiabatic curve overlay shows why.

Mike