Marcelo wrote:
I believe, that over the years I certainly have seen pictures of electrical faults that claerly arced on both sides of an overcurrent device such as a switch or circuit breaker. Whether it happened within the span of time you would consider an arc flash event (milliseconds to 2 seconds) or over a longer period of time I could not tell. I have seen arcing faults also transfer from one set of conductors to another in laboratory equipment tests when the barriers that should have been put in place were ommitted.
I've seen similar photos. But here's the inherent problem here. Let's assume that the fault starts on one phase and then propagates to the other phases. IEEE 1584 is based on test data that includes 3 buses that are all arcing with no phase barriers. Although the single phase arc may occur and damage is likely to be considerably less, this is not what the model covers. The only weakness here is that IEEE 1584 currently does not model single phase arcs so it will produce higher incident energies than may actually occur in most cases, but will adequately model an arc that propagates from L-G to L-G-L to L-L-L.
Second with the case of an arc propagating from load to line side of a breaker. This SHOULD be covered in the methodology used by most arc flash practitioners. Although I am particularly interested in the incident energy on the load side if for instance I'm doing voltage testing directly at a motor termination, within the MCC bucket itself I'd be much more concerned about the potential incident energy of a bus bar fault which is the line side of the breaker. I don't particularly care if the fault originates on the load side and propagates to the line side or not. Either there is a significant possibility of an arcing fault within the MCC bucket or not.
The two cases that I'm aware of that keep being brought up is propagation from downstream equipment across some sort of "main" breaker. For instance in a panelboard it is easily conceivable that an arc initiates anywhere within the panelboard and then subsequently propagates across the main breaker so the recommendation would be to analyze the panelboard as if it is a main-lug-only panelboard. Where this practice gets more questionable is with switchgear and MCC's. In this case there is not much if any barrier within an MCC section so it is conceivable that buckets placed in the same vertical section as a main breaker could propagate to the line side of the main breaker. Where this gets far more questionable is the idea of propagating from one MCC SECTION to another, and similar arguments occur with some switchgear designs. Metalclad switchgear clearly has very little chance of propagating by any mechanism other than magnetically propelled arcs. Some metal enclosed switchgear is little more than a fancy panelboard and thus deserves similar scrutiny while other designs are clearly isolated between breakers. Portable skid mounted switchgear in particular is sometimes little more than an enclosed box with various switches, fuses, and breakers mounted inside it. However around 95% of such gear is also custom built on a case-by-case basis and there are no standards so you can't categorically state anything about the design. Gear with "underground style" sealed modular equipment has no chance of propagation, while some skids designed for remote pumping applications take advantage of the lower cost of "open" gear and have little or no internal isolation whatsoever.
For the record I buy a lot of this kind of equipment, usually 5-10 units a year. Some substations get relocated every 60 days, and some starter skids are relocated every 1-2 weeks. I specify it with isolation between compartments of breakers and also provide isolation between the incoming/outgoing compartments and the remaining gear in order to facilitate connecting and disconnecting equipment while some compartments remain energized. Due to the relatively small transformer sizes (500-3750 kVA) relative to the voltages (2.3, 4.16, or 7.2 kV), arc flash incident energy tends to be 20 cals/cm^2 or less. We use resistance grounding on almost everything from 480 to 22.9 kV. Failures tend to get people excited but we see very little damage other than typically destroying the equipment in a given compartment with little major damage to the compartment itself. Major changes that we've made in recent years due to equipment damage are:
1. Stopped using ABB MV disconnects. There is something about their particular design (more compact) where we've had up to 5 switches fail in one year. All cases were contamination leading to partial discharge and eventually L-G flashovers as moisture accumulated overnight during warm steamy Southern U.S. days. This equipment is particularly susceptible to this problem. We've revised our specs to require Powercon or Federal Pacific switches because those seem to be more robust to this problem.
2. Stopped using 23 kV enclosed fused disconnects. Same kind of problem. We've gone back to using outdoor overhead switches and/or fused cutouts. The cost of a gang operated unitized overhead switch and/or cutouts is so much less than an enclosed fused switch with the same performance that it makes no sense to bother with enclosed gear and the incident energy is less because of the increased working distance. The cost actually decreased when doing this, and contamination is no longer a factor since nature provides a wash down periodically on the equipment. Heavy fog/mist or sometimes freezing conditions can still potentially cause a flashover but these are rare here except around cooling ponds where I work.
The next move that I'm planning and testing now is to replace the unitized switch plus fuse combination (where both are used) with an overhead recloser. Some of them (Tavrida, T&B) are low cost enough that they are the same installed price as a combination fuse/disconnect configuration and I can provide full breaker protection for the transformer and I can even add secondary bushings to the transformer secondaries and virtually eliminate the vulnerability of a failure between the transformer and downstream protection. In fact I can simply treat this arrangement as a "virtual breaker" and lower the overall cost. However we don't have much "run time" with reclosers here so the first step is to install a few and see how it goes before redesigning the portable skids.
So I hate to say it but in my experience I've never seen arc propagation "upstream" which crossed from one compartment to another except when everything is all inside a single enclosure such as with a panelboard. I don't doubt that it can happen but the circumstances for it to happen make it seem extremely unlikely. I could easily see with 5+ MVA, 480 V secondary transformers and 3000 A metal enclosed breakers where this sort of thing could happen. But I wouldn't even sign my name on that equipment. At that size, I'd split it into 2 transformers and add one more breaker and either a third breaker, fused disconnect, or perhaps even just a wiring compartment to provide a main-tie-main arrangement for some redundancy, and this would also lower arc flash incident energies from crazy high down to reasonable values. This would be the prudent thing to do even without concerns for arc flash because the breakers go for $50K at that size, the copper gets very expensive, and the expense and risk of a fault is so high that the long term cost of operating such equipment vastly exceeds the short term benefit. That's even if it doesn't make sense to distribute at 4160 with further downstream subs which with those dimensions might make more sense in the first place.