Hello guys,

Even though I start talking about SKM, its more a general question.

Initially I had a question about SKM, I contacted our supplier who forwarded my question to SKM. This was the reaction:

Quote:

You can inform the user that unfortunately he would not be able to simulate the dynamic generator specific fault current contributions in arc flash. In the Arc Flash Study Options, it can only be reduced to a specified percentage of the FLA at a certain time (cycles).

But why..

I know SKM is an American based company, and that in America they are a lot further with arc flash hazard/risk assessments then here. I also understand that I am doing the assessment for a ship and that means the fault location is close to the generators but on mainland the dynamic characteristics don’t play a big role. But what about the American classification societies like the American bureau of shipping?

I suspect that I might be misinterpreting the effect that the dynamic short circuit current has when you use it as input for an arc flash calculation. I would say the short circuit current is dynamic so the arc flash is dynamic to. Simple example if the short circuit current is 1000% of the FLA at 2 cycles then that is the current to use to determine the arc flash current at 2 cycles, if the short circuit current is 400% of the FLA after 30 cycles then that is the current to use to determine the arc flash current at 30 cycles.

Is this correct or am I really lost?

IEEE 1584 provides a means to calculate the incident energy given a BOLTED FAULT current. The current edition does not speak to subtransient or transient fault currents. Current practice by some software packages (SKM and ETAP) is to model it in a piece-wise linear fashion. That is, calculate the arc flash E1 from say time 0 to t1 at current I1, then E1 is from time t1 to time t2 at current I2, and so forth. Then the resulting incident energy is repeorted as E = E1+E2+...

The complication with any arc flash modelling is that the arcing current is variable and somewhat random. It is also voltage dependent. The difference between the predicted arcing current and bolted fault current disappears as voltage increases and is largely irrelevant above 1 kV. This is the reason that the current IEEE 1584 standard recommends calculating the arc flash energy at bolted fault current and at 85% of the bolted fault current, along with using the overcurrent protection device opening time calculated at both points.

Theoretically of course the arcing current could be at ANY current between the bolted fault level and 0. However as a practical matter IEEE 1584 looked at a large database of actual laboratory measurements and found that the above procedure captures the actual measured incident energy with 95% confidence. They have also performed a numerical study coupling the IEEE 1584 prediction with predictions from arc flash PPE chosen following the ASTM 1959 standard and found that the equation successfully protects against injury 95% of the time when subjected to the data in the testing database and using the incident energy value as a threshold, while in the real world, we would almost never have an arc flash that worked out with that precision. In other words although say 8 cal/cm^2 PPE may be chosen, the actual calculated value will hardly ever be an exact 8.0 value. So it is likely to be decades before we can even gather enough data from actual arc flash incidents to be able to validate the IEEE 1584 equation. So far there have been ZERO injuries with over 10 years of actual performance data.

Hey PaulEngr, thanks for your reply. I do still have a few questions though.
Because I misinterpreted and it doesn’t matter or because of various reasons and it is coming in the next update?

Quote:

That is, calculate the arc flash E1 from say time 0 to t1 at current I1, then E1 is from time t1 to time t2 at current I2, and so forth. Then the resulting incident energy is reported as E = E1+E2+...

For SKM this is not true. E1 based on the initial value and the time off the arc flash. For E2 this is different: I2 and t2 are 2 values that can be set to be anything but they have to be set manually if you want to use this 2nd step in the current flow. There is no +E3+E4+…

After reading my first message I see that I have been a bit vague. I hope the next part is a bit more clear: The short circuit current is a big influence in the arcing current, the short circuit current is dynamic therefore the arcing current is dynamic. True or false?

In my mind it is true but I can’t really find how to determine the resistance off the arc and if this has a linear characteristic. I thought I knew a fair bit about what I was talking about but after this, my conclusion is that I definitely do not know.

Both. The actual tests for the current edition of 1584 used fixed X/R conditions because these were laboratory tests done at high voltage/current labs. As I understand it in the upcoming edition it will incorporate a more complex model that includes time series data to improve accuracy. All that the model expects is a "bolted fault" current and TCC curves for protection devices.

Quote:

For SKM this is not true. E1 based on the initial value and the time off the arc flash. For E2 this is different: I2 and t2 are 2 values that can be set to be anything but they have to be set manually if you want to use this 2nd step in the current flow. There is no +E3+E4+…

SKM and ETAP I know do try to incorporate some form of transient modeling. SKM is a little more adjustable and a little more individualized whereas ETAP takes a more global and brute force approach. Neither one has been "validated" against anything except a theoretical concern. I haven't used Easypower or any other competitors so I can't comment on these. Arcpro definitely doesn't do any of this kind of modeling.

Quote:

The short circuit current is a big influence in the arcing current, the short circuit current is dynamic therefore the arcing current is dynamic. True or false?

Quote:

Partly true. In terms of physics at the current zero crossing (not voltage), the arc extinguishes. At the next half cycle once the voltage exceeds a threshold, the arc restrikes. Once established the arc resistance rapidly decreases until it reaches a point where power transfer is maximized at roughly 1/2 of the available fault current (bolted fault current). Then at zero crossing the cycle repeats itself. The restriking voltage is dependent strongly on the arc gap and the air temperature, which is obviously a function of previous arcing cycles, as well as enclosure size, air density (which changes...arc blast), and even the electrode materials (copper, aluminum, etc.). Any good book on circuit breaker design or high intensity discharge lighting will describe the myriad of variables involved. Furthermore, all these factors plus again the enclosure shape/size, shape/geometry of the electrodes, and distance play a factor into the incident energy at the point of interest (worker's body). Needless to say, PREDICTING this is troublesome at best. There are some theoretically based models. ArcPro and Duke Heat Flux are two commercial programs that do so. I think at this point the heat flux software has been shown to be inadequate but at medium voltages in particular, ArcPro seems to work just as good as IEEE 1584 and more importantly, it doesn't suffer from the problems of the equations producing nonsense at the limits as voltage increases over 15 kV. This makes it the best choice for predictions for utilities.

In contrast, IEEE 1584 took a pure empirical approach rather than trying to derive anything theoretically. The result is fairly accurate and has stood the test of time despite the fact that it is obviously glossing over an incredibly large and diverse set of details. We're talking curve fits here. Although a lot more is known about arcs in general over time, there are still some things that elude understanding in the data. A recent model proposed by someone at Mersen or Eaton I believe (can't remember for sure) which was a time series based model discusses some of the key parameters and shows that although the model fits the data much better than IEEE 1584, there is still something nonsensical with regards to a couple of the parameters.

So the real answer is that its a curve fit, nothing more, nothing less, to a set of about 300 measurements. Buying and reading the actual IEEE 1584 standard is very enlightening in this regard but I can also understand someone not wanting to pay the rather exhorbitant price for it.

Keep in mind...we are also talking about one model within the IEEE 1584 standard. There are actually roughly 3 sets of models. There is a theoretical model which is nothing more than the Lee theoretical equation, which has a lot of issues outside of a fairly limited range of currents and voltages. There is the empirical model we are referring to which also has problems at its limits but the application range is larger than Lee in terms of producing results that are close to actual test data. And there is a set of empirical curve fit models for fuses because the empirical model does not do so well modeling current limiting fuses. These models are based on fuses from one manufacturer only but that is also not clear from the standard, and these models don't generally make it into the available power analysis software. No one has spoken up about validity with fuses from other manufacturers so at this point I suspect they perform reasonably well. Fortunately they are so simple to use relative to the empirical equation that it is easy to do it by hand.

I don't know if this is the answer you are looking for but it appears that you are attempting to make some sort of correlation or somehow figure out the theoretical underpinnings relating current to incident energy. Don't try. There is no theoretical underpinning in the empirical equation...it's empirical, and that's the reason that there is no theoretical relationship. Current plays a major role in the incident energy obviously, but the relationship is extremely complex due to the fact that the arc extinguishes and reignites 120 times per second and that the arcing time in each half cycle is a factor of several other variables. The last item that I did not mention is that to a certain degree, it is also random and not 100% predictable. There are graphs within the IEEE 1584 standard showing this very clearly but I have seen plenty of data out there studying especially low voltage conditions where arcing can actually cease on one phase for several cycles and then restart again as if nothing ever happened, and arcs are significantly different from theoretical predictions, and can be different on every run. IEEE 1584 attempts to stay away from this and only considers data in the "stable" region of arcing but a better understanding of unstable arcs (120/208/240 V) is clearly needed in a variety of situations. Similarly modelling from a theoretical basis gets a lot easier at medium voltages where the restriking phenomena have a relatively minor role to play. Hence the reason ArcPro works so well when it is a theoretical model.