My 787 battery fire theory

[ka9q]

I've looked over the NTSB's latest data dump, and I'm convinced I know what happened.

There was a ground fault between one of the li-ion cell cases and the metal battery box.

Nothing else can explain the evidence. And I see no fuses or circuit breakers that would have interrupted the fault current.

If anyone wants my detailed reasoning, let me know. But you heard it here first...! :-)

[cjameshuff]

On cell 5, producing the protrusion on side 3 described here? They found signs of arc damage, battery case material on the outside of the cell case, etc. It seems clear that a short occurred, but it might have happened as a result of failures elsewhere.

It'd be nice if they hadn't made the quick disconnect lever out of something that melted in the fire. They beat the thing up pretty badly first trying to unbolt it or pry it loose and then cutting it free.

[ka9q]

Yup, that's part of the evidence. But the real smoking gun (!) is the overcurrent damage to the battery box ground jumper and the shield of the signal cable. Neither can carry battery current under any conditions other than a ground fault within the box, because neither side of the battery string is connected to the box. Nor would I expect it to be, as that would create a ground loop.

I am not familiar with aircraft wiring practices, but I presume their 28V DC systems are similar to our 120V AC house wiring systems in having a single-point ground where all the ground and DC return wires ("neutrals") come together. So a short between a cell case and the box would have created a current path that proceeded out the ground jumper and signal shield in parallel, to the aircraft ground point, and back to the battery through its negative lead.

I see no fuses or circuit breakers anywhere that would have interrupted this fault current. Maybe it was an oversight, or maybe Boeing made a conscious decision that losing the battery to a spurious fuse blow was a bigger risk than a short circuit.

But it does rekindle an old, old argument about whether we should ground the neutrals in our electrical power systems. Had the (-) 28V DC rail not been tied to ground, no current would have flowed through this ground fault and, if it was the cause of the fire, no fire would have occurred.

I had this argument years ago when I installed my PV system. I believed then, and I believe now, that it's a mistake to ground DC system returns and quite possibly also AC system neutrals for precisely this reason. But it's entrenched in a century of practice and is probably impossible to change now.

At least modern EVs like the EV1, Prius and Leaf isolate their HVDC systems from vehicle ground.

[ka9q]

Some additional evidence for a ground fault:

The only cell not found shorted was cell #8, on the (+) end of the series string. This is the only cell that could not have contributed current to a ground fault (i.e., if its case were the one to short to the box, though I'm not saying it did.)

The aluminum current collectors in several cells were found melted. An internal short in a single cell could not have done this. The fault must have involved several cells.

The question now is how the ground fault developed. Several possibilities:

1. A faulty insulator or improper installation.
2. Repeated ambient pressure cycling caused the sealed battery cells to swell and contract slightly on each flight,  causing the vents to abrade through the insulators.
3. A cell vented for some other reason, and the petals of the vent cut through the insulator and touched the battery case.

[RitchyRev]

Hi ka9q,

I have been looking at the 787 battery problem myself. You said the following

The only cell not found shorted was cell #8, on the (+) end of the series string. This is the only cell that could not have contributed current to a ground fault

Why do you say that? I have been trying to work out if the individual cells cases are + or -. Do you know?

But it does rekindle an old, old argument about whether we should ground the neutrals in our electrical power systems. Had the (-) 28V DC rail not been tied to ground, no current would have flowed through this ground fault and, if it was the cause of the fire, no fire would have occurred.

I had this argument years ago when I installed my PV system. I believed then, and I believe now, that it's a mistake to ground DC system returns and quite possibly also AC system neutrals for precisely this reason. But it's entrenched in a century of practice and is probably impossible to change now.

Curious as to why you think having grounded neutrals (m.e.n. system), or grounded DC is a mistake?

[ka9q]

This accident is a perfect example of the drawbacks of grounded electrical power neutrals. When the ground fault formed, current flowed out of the battery case, into the battery box, out along the ground bonding wire and control cable shield (in parallel), through the airplane's ground bus to the common grounding point, and back to the negative battery terminal through its negative supply conductor.

Had the negative side of the 28V DC power system not been connected to the airplane ground system, no current would have flowed even with the ground fault from the cell case to the battery box. The two DC terminals would have assumed voltages with respect to ground depending on which cell shorted to the battery box and those voltages could easily be detected to trigger an alarm that a fault had occurred, but it would not create an emergency situation.

Even with the grounded neutral all DC-powered equipment still have to isolate their power terminals from chassis ground to prevent ground loops. These are currents that flow through the aircraft ground system under normal conditions that can result in different potentials at various points in the grounding system that cause unwanted currents to flow in the shields and grounds of signal cables.

Our AC outlets have separate ground and neutral wires for the very same reason; in a legal installation there is only one connection between ground and neutral, and it's in the main service box.

The one valid argument for neutral-ground bonding in AC power systems is to provide a low impedance ground path in case lightning strikes a power line or the insulation in a transformer fails. I personally think these situations could be handled with spark gaps and MOVs. In a DC power system, however, there is no connection to the outside world. That leaves no need that I can see for a low impedance neutral-ground bond.

[ka9q]

Why do you say that? I have been trying to work out if the individual cells cases are + or -. Do you know?

According to the NTSB, the current collectors on the + side are aluminum and copper on the - side.

I had thought the - terminals were bonded to the case based on some CT images in the NTSB report, but I found another report that says the cases were nominally floating with respect to the cell. However it also says there's 1.0-1.4 V between the + terminal and the case and 2.6-3.0V between the - terminal and case, which suggests they're not completely isolated either.

So if there was a large ground fault current from cell #5 to the box, it does seem probable that something happened to connect at least one of cell #5's electrodes to the cell case to provide a low impedance path. I.e., the ground fault, though it occurred, was not the first event in the failure sequence. That's the point I'd been wondering about.

[RitchyRev]

One of the biggest problems I see with the MEN system is if the neutral conection is compromised.
If there is a high impedance joint at the utillitty pole, or in your main box, you can get a voltage increase and greater current flow through the earth ground loop.
If the neutral is broken, and the ground stakes are high R, then you can have full mains potentiall on the earth.
An R.C.D. won't trigger in this case. You would need an ELCB to monitor the earth current as well.

I think the seperate earth and neutral distribution returns are the safest, TN-S I think, but the extra conductor adds extra cost.

I agree with you regards the PV system DC grounding. I don't see why it should be.
We quite often use mains isolated units, via transformers, that have no direct connection to the mains circuit.
However, I would assume that the reasoning for grounding the PV DC system would be in case of a fault in the interface to the mains.
 

[RitchyRev]

I see they are modifying the battery box. It is going to be a sealed stainless steel unit with an outsde aircraft vented titanium tube.
Also better insulators. They are also modifying the BCU. Making it a softer charge. They still have the ground bonding terminals in the pictures I've seen.

What I have found as I have been following this, is more questions than I can find answers for. Without these answers, I can only make guesses.

... The individual cells have voltage measurement. Is this data recorded and availlable?
... How does the individual cell monitoring affect the charging and fault parameters?
... There are temp sensors. Is this data recorded? Is it availlable? How does this affect charging and fault parameters?
... The battery is 32V connected to one of the 28V busses. What are the buss voltage parameters and does the equipment connected to it have voltage regulation?
... Is the battery always connected to the bus wnen the APU unit is running?
... Is the BCU connected to the bus, and charging the battery by bus voltage monitoring?
... And many more.

Like you, I have no doubt there was a breach in cell no.5 to the case. This then tracked through the ground bonding wire and signal returns, and then back through to the battery negative through some common ground point.

I wonder what started the chain of events.

I was looking at the 4 sec 44 amp increase in charge after the Bat dropped from 32 to 31 V.

I am still trying to work out if it is possible for a cell to have shorted internally, thereby inreasing the voltage over the other cells due to the charger.
Hence my question on the individual cell voltage monitoring which should have shown a fault and disconected the battery.

With the cells over charged they possibly went into thermal runaway. This led to cell 5 rupturing. Maybe.

[ka9q]

I'll comment on your 787 questions later, but I wanted to remark on the grounded neutral issue because it's been a pet peeve of mine for some time.

When I installed my PV system in 1999, the NEC then required a "PV ground fault protection" device on all roof-mounted PV arrays. The only unit that qualified for this purpose was sold by Trace (since bought by Xantrex) and cost several hundred dollars. It consisted of two mechanically ganged DC circuit breakers: one at 100A and another at a fraction of an amp. The latter breaker was shunted by a power resistor, the size of which I can't remember at the moment.

The 100A breaker was installed in the positive array lead. The low current breaker formed the only connection between the DC return and system ground. If a ground fault developed in the PV wiring (or anywhere else in the DC system) that permitted current to flow, then the fractional amp breaker would trip and open the 100A breaker with it.

This was absolutely useless. The 100A breaker cut off array current, but since the most likely position for a ground fault was somewhere in a PV panel this did nothing to interrupt the fault. The fractional amp breaker did interrupt the fault current, but this could have been accomplished by simply doing without the breaker, and the system would function just fine without it.

The neutral-ground resistor should be there to drain any static charges on the DC circuit and to provide a monitoring point. Any voltage drop across this resistor would indicate a ground fault somewhere in the system which could be investigated without shutting the system down first.

I had a long email discussion with the NEC columnist for Home Power who wrote that NEC rule. He never could give me a scenario in which the PVGFP, as designed and used as intended, actually provided protection better than simply leaving the DC (-) bus grounded through a resistor rather than hard-grounded. He just kept saying that the authors of the NEC are firemen who don't like fighting residential attic fires. This may well be true, but it had nothing to do with the issue at hand. In fact, I could think of several realistic failure scenarios that  the PVGFP would do nothing to stop, particularly a cable staple causing a direct, medium impedance short between array + and array - somewhere inside an attic. Fuses are useless here because PV panels are constant current devices that normally operate very close to their short-circuit currents.  The only protection is a careful installation.

It turns out he had developed the Trace PVGFP, which I thought was a serious conflict of interest. But apparently that sort of thing is both accepted and common in code-writing and standards-writing groups. It shouldn't be.
 

[ka9q]

Regarding the 787 battery, my understanding is that the individual cell voltages and temperatures were only available to the battery management unit inside the battery box; only the total battery voltage was logged in the flight data recorder.

I presume the DC bus voltage is allowed to vary with the battery state of charge, just as it is in other DC systems with batteries (like a car). That means all the DC loads have to be designed to operate over a voltage range centered on the nominal 28V.

There is a separate battery charger that ensures the battery voltage doesn't exceed safe limits, and if an individual cell temperature or voltage gets too high it will get that information from the battery management unit and shut down. I think the NTSB said that one purpose for the DC power contactor inside the battery box was to disconnect the battery from the DC bus if the bus voltage got too high. This contactor is a normally closed type, and I can't see any reason to disconnect the battery from the bus during routine operation, just as a car battery remains on a car's 12V DC bus.