Arc Flash Motor Contribution

[WDeanN]

Here at our facility we have several large synchronous motors. The question of motor contribution to our arc flash hazard has recently come up, especially with a software upgrade to SKM Power Tools for Windows, version 6.0.2.1, which allows the user to specify synchronous motor contributions over a time period. On one bus in particular, because there was no protective device on the motor that would clear the fault quickly, (we are using a two second maximum for our calculations, unless other conditions warrant otherwise) the motor contribution to the arc flash hazard was very significant. The calculations revealed 17 cal/cm2 with all 6 synchronous machines running, but only 2 cal/cm2 with no motors running.

While trying to research our options, I came across an article by J. C. Das, "Design Aspects of Industrial Distribution Systems to Limit Arc Flash Hazard" in the Nov/Dec 2005 issue of IEEE Transactions on Industry Applications. Mr. Das states that "The synchronous motor short-circuit contribution can be neglected in six to eight cycles."
Although he doesn't reference it directly, I found support for this in the 1997 edition of IEEE 399, the IEEE Brown Book. Chapter seven addresses short circuit studies.

Section 7.4.1 states:
Time delayed currents are the short-circuit currents that exist beyond 6 cycles (and up to 30 cycles) from the fault initiation. They are useful in determining currents sensed by time delayed relays and in assessing the sensitivity of overcurrent relays. These currents are assumed to contain no dc offset. Induction and synchronous motor contributions are neglected, and the contributing generators are assumed to have attained transient or higher value reactances (see Table 7-1).

This can also be found in the 1993 edition of IEEE 141, the IEEE Red Book. Chapter four addresses short circuit studies. Section 4.5.4.3 states:
For an application of time delay relays beyond six cycles, the equivalent system network representation will include only generators and passive elements, such as transformers and cables between the generators and the point of short circuit. The generators are represented by transient impedance or a larger impedance related to the magnitude of decaying generator short-circuit current at the specified calculation time. All motor contributions are omitted.

This is for breaker duty calculations, so the question remains: Can this be applied to arc flash studies? Otherwise, what is a reasonable time to assume for the motor contribution to an arc flash event at the upstream bus?

 

[dpc]

There is no guidance on this in IEEE 1584. The contribution will decay over time, but while it is there, it will certainly contribute to the arcing current at the fault. I don't see how totally ignoring it can be justified.

If SKM software allows you to specify time period, I would use a conservative value of 10 or 15 cycles. Otherwise, the conservative approach is to include it for the duration of the fault.

 

[WDeanN]

I should also add that I've done some calculations for motor short circuit contributions for RMS Current, assuming no load, constant input voltage (the motors are self exciting) and trying to account for the resistance of the arc. In these cases, the motor contribution decays in about 5 cycles to 100% of rated current.

I think this still gives me a conservative estimate of the motor contribution to the arc, but it still does not agree totally with the references above. It also does not account for the motor load. How do I incorporate the motor load, Wk2, into this calculation, and is this necessary?

 

[dpc]

I'm not sure there is any benefit in trying to do such an "exact" determination of motor contribution since the arc-flash equations themselves are certainly not terribly accurate. The calculation is a very crude approximation of expected maximum arc energy.

I would do the calculations assuming the motor fault contribution is present for the entire fault. If that answer seems unreasonably high, try it with no motor contribution and compare the results. Remember that the maximum arc-flash energy will not always occur at the maximum current scenario.

 

[JensenDrive]

The 5 cycles seems a tad fast, in my limited knowledge. That is the decay time I would expect out of an induction machine; I did not hear before now that synchronous motor contribution would decay so fast. Do you have any published Td'?

 

[WDeanN]

The six to eight cycle decay time quoted by Das troubled me, which is why I began doing my own calculations.

I do not have published Td'. I have been using the default values in SKM for bolted fault values of Td' and Td'' for my calculation. With input from Bob Wilkins, I lowered these values from 26ms for Td'' and 420ms for Td' to 2.6ms and 42.0ms, respectively, to account for the resistance of the arc, in terms of L/R. I am still trying to verify this assumption.

I also have been unable to verify what effect the motor load (in terms of Wk2) will have on these values, if any. I’m not a motor guy, so I’m still trying to piece most of this together. Any input I receive will be much appreciated.

 

[TurbineGen]

From what I understand, motor contribution can supply it's locked rotor startup current for only a few cycles. Once the excitation field of the rotor falls, motor contribution is minimal. Perhaps this is where the 6 to 8 cycles comes from?

I agree with dpc. Arc flash is in its infancy and to try to get an exact number might be more trouble than it is worth. When calculating I assume the motor contribution to be the locked rotor start-up current for eight cycles. Since the purpose of arcflash is to establish a boundary and proper PPE, I would lean on the conservative side of things.

There are some references you might try for this:

http://www.brainfiller.com

has some free information regarding arc flash.
Also, there is arcflashforum.com . It is small, but has much information.

-TurbineGen

------------------------------------------------------------------------
If it is broken, fix it. If it isn't broken, I'll soon fix that.

 

[JensenDrive]

I have nothing more to add, but am a bit curios on the reasoning to use a Td' of 42ms; I am not sure what resistance of the arc does to the matter. Here is my thought process:

From what I understand, Td' and Xd' is something inherent in how the rotor field windings respond to rotor flux that is generated by the fault currents. If you have an increase in positive sequence current (e.g., any fault, but especially a three phase fault) you get a net increase in rotor flux, coming into the rotor from the stator. First the damper bars fight the increase in flux, and hence a voltage and current is induced in the rotor bars. As these currents die out, a voltage is induced in the rotor field winding. The induced rotor field currents die out over Td' and the field current returns to whatever current the exciter voltage is trying to drive.

Are you thinking that resistivity in the fault (arc) current affects the time constant of the rotor field? If so, I am not seeing that this is appropriate; I mean I am not seeing how arcing fault has a role in how fast the rotor field currents return to normal. Maybe there is some connection but I am not sure how much, and my technical grasp of the matter is stretched to the limit by what I have written thus far.

 

[Modula2]

WDeanN,when you stated that the S/W used by you, allowed specifying synchronous motor contribution over a time period, did you mean that it produces arc flash results at a number of fixed contributions, probably two. Wouldn't actual incident energy be a summation from initial to final, so that a calculation with motors decayed to off, would not be representative. That could however be a case during maintenance or testing.

 

[WDeanN]

Modula2,
The software, as I understand it, performs a step calculation, figuring the arc flash contribution assuming maximum motor contribution for a time period, and then will allow you to assume a smaller motor contribution for the rest of the time.

The inputs are:
Reduce Generator / Synchronous Motor Fault Contribution To
___ % of Rated Current after ___ cycles

If Das, and the IEEE Red book are correct, I could change this to:
Reduce Generator / Synchronous Motor Fault Contribution To
_0_ % of Rated Current after _6_ cycles

In reality, I understand that the motor contribution will decay with time, and I am trying to figure out the correct formulas and assumptions to come to reasonable, yet conservative, values for this. The 0% after 6 cycles values seem too low, and don't agree with the few calculations I have performed.

 

[WDeanN]

JensonDrive, I’m not sure myself on the effect of the resistance of the arc to Td’ or Td’’, but if I assume the following equation:

Iac (t) = E*[((1/Xd'')-(1/Xd'))*?^(-t/Td'')+((1/Xd')-(1/Xd))*?^(-t/Td')+(1/Xd)]

The resistance of the arc will have an effect on the total impedence of the motor, with respect to the resistance seen at the fault. Inputting this resistance into Td may have a similar effect?

 

[countingcrow]

Hi, We have a similar study going on to WDeanN with the difference that we have alot of smaller Induction machines grouped onto a few busses. Can i assume from JensenDrive's post ("The 5 cycles seems a tad fast, ... That is the decay time I would expect out of an induction machine") that the decay time for our machines is much smaller than for your sync machines both because they're generally smaller and the fact they're induction machines? does anyone have any rules of thumb for induction motor decay time?

 

[dpc]

See ANSI C37 or IEEE Red Book. For small induction motors, defined as less than 50 hp, contribution can be ignored after 2 cycles.

 

[Morquea]

Hi, I used SKM for arc-flash studies less than two years ago, before it gives the option to specified the duration of motor contribution. I remember producing four set of results :

1- Every motor online and contribution as per multiplification factor specified in the IEEE Red book with the 2 seconds limits. We used "Interrupting Network" factors of table 4-1 and 4-2.

2- Every motor offline with the 2 second limits.

3 & 4 - Same as 1 & 2 but without the 2 seconds limit.

The "every motor offline" scenarios were to simulate "shut-down operation maintenance". The purpose of not using the 2 seconds limit was to simulate working postions don't allow getting out of fault zone within 2 seconds and to find mis-coordination of protecting device for arc-flash fault.

About motot contribution :
Since the arc-flash level are cumulative energy (cal/m) over time, applying a cycle limit of the motor contribution would give me a model that is between my scenatios 1 & 2.

Whould I use that new SKM option :
I think that if I had this option less than 2 years ago, I wouldn't had used it. This cause I made many case to find the worst one according to possible operations planning.
You seem to conduct your study on only one scenario, so in that case I would apply it. This cause lower current fault might give you higher cumulative arc-flash energy. There is a chance that you end with lower level cause you are using the 2 second limit. I don't like using that limit without study of working procedures.

Simply cause, I remember that the scenario 4 was the worst one cause the arc-flash fault current fall under clearing curve or had over 10 sec clearing time.

That was my two cents

Good day

 

[AusPowEng]

Hi,

I have also been thinking about the effect that motor contribution makes to incident energy calculation. The various methods of motor contribution calculation are, as I understand them, based on the premise of a bolted-fault. This represents worst case as the system voltage in 0V at the point of fault and the maximum motor contribution is therefore considered. Of course, it decays over a few cycles in the case of asych machines and for longer with Synch machines.

However, in arcing faults the situation is completely different. The arc is a resistance and surely the arc voltage must be taken into account when determining motor contribution. For arc faults, it seems that the motor contribution will be less than for bolted faults.

Any ideas?

 

[Modula2]

AusPowEng, I expect the arc current, generally less than the bolted fault, would then include a reduction in the motor contribution. Stated another way, all contributions are accordingly reduced in going from the bolted to the arc current.

 

[AusPowEng]

Modula2,
Agreed. This is my thinking also. The question is, how is this determined by calculation. The affect on incident energy outcomes can be huge is the reduced motor contribution is overlooked, especially in LV motor control centers.

 

[WDeanN]

Morquea,
I have performed a couple of scenarios with different motors running. As stated in my original post, the energy goes from 2 cal/cm2 with no motors running, to 17 cal/cm2 with all 6 motors running. We have adopted the two second rule, and in this case, it would probably be supported since there is ample clearance to the sides and in front of the switchgear.
Using a worse case energy of 17 cal/cm is possible, but just doesn't feel right intuitively to me. There has not been an event on this particular bus in the past that I can find, but on other buses with large motors, the worst case does not hold up to observed events in similar line-ups.

As stated previously, the main breaker trips on an instantaneous setting with or without the motors running (it is not affected by motor contribution). If I continue to use the 2 second rule, this assumes that the motors will continue to contribute their full fault current to the event for the entire 2 seconds.

 

[WDeanN]

AusPowEng, Modula2
Thank you for your input. I have been seeking some aid on this problem for some time. Other then the paper by Das, above, I have not seen anything written on this.

My thoughts on the calculation are that if you could reasonably calculate the arcing resistance then you could input this back into the motor bolted fault equations as a contribution to Xd'', Xd', and Xd, effectively coming up with an arcing Zd'', Zd', and Zd.

The approach taken by Bob Wilkins, (see

http://us.ferrazshawmut.com/arcflash/getting_help/downloads/ICPS-wilkins011904_web.pdf

) however, is to input the arc resistance in place of the motor L/R, greatly reducing the Td’’, Td’, and Td. I am trying to confirm as well whether or not this approach is valid. The effect here is to greatly reduce the time period that the motor contributes to the fault, with the arc resistance acting effectively as a dynamic brake on the machine.

Do you have any thoughts on either of these two appoaches?

 

[dpc]

Regarding machine time constants - an external impedance between the machine terminals and the fault will change the time constant for T'd and T"d, I believe. T'd0 and T'd are the limits and the exact value depends on the external impedance so will generally be somewhere in between these two values.

But of course, the time constants will generally be different for each motor.

If you have the Edith Clarke texts, I think it is covered there, and probably in Anderson's "Analysis of Faulted Power Systems".

ANSI C37 has some discussion of source contributions in the standard for short circuit calculations related to breaker duties.