Failing Exhaust Valve Analysis

This is very interesting. Can one get access to this analysis without signing up to their full management service? It doesn’t sound like it:

FEVA™ is included with a paid subscription to SavvyAnalysis Pro, which costs $129 per year for singles and $199 for twins

Also, how can one tell an exhaust valve is about to fail? Surely it is a single piece of metal which either remains in one piece or not. I could see one could look for indications of a valve getting stuck in the guide.

One failure mode of the exhaust valve seems to be that if it doesn’t make contact with its seat over the full circumference, hot spots may develop and the valve may subsequently fail. Apparently you can usually see this in the EGT curve, as oscillations due to the rotator cap (which rotates the valve)

Mike Busch has expained this here, I would expect that FEVA is just an automated tool to look for these oscillations.

Would that show up on a compression test?

I have seen EGTs going “walkabout” for a long time, but all my compressions are 78-79/80. At one stage I pinpointed the walkabout to the wingtip strobes affecting the EDM700 probe wiring, but it is still seen sometimes. The variation is about 1% of the EGT so while alarming to watch on the instrument, in “engineering terms” it is tiny, and I think all the cylinders do it. I put it down to the fuel servo.

Also the EDM700 displays to 1F resolution but records the data to 5F resolution (see my other GAMI test thread) which is a bit stupid because it makes it much harder to make sense of the recorded EGT data if looking for small variations, or if trying to do the GAMi test flight. I strongly suspect this is a recent firmware change.

Would that show up on a compression test?

Unlikely, unless the valve has already failed (i.e. bits have broken off)

A minor case of a poorly seated exhaust valve may show up in a compression test if you listen for hissing at the exhaust pipe and breather pipe to identify where the air flow is going.

Real time diagnostics are interesting but obviously you do tend to see some ‘solutions in search of a problem’. The exhaust valve monitoring thing seems like it would be valuable if you could figure out the best way to tell the pilot without undue complication. Dropped valves seem like one of the few kinds of engine component failures that don’t announce themselves much ahead of time.

At work we have a small group that works on this stuff and they occasionally come and talk with me to understand the hardware and failure mechanisms. One time a young guy started describing gearbox/clutch problems, and that they’d been encouraged to come up with diagnostics that could extend the TBR for the end users. I realized he was taking about Thielert engines, but nobody had bothered to tell him much about the hardware.

From that link:

As for

I would say that only if you have had too many Ouzos – I’d say half a bottle. A 30F-60F EGT variation over a minute or so is going to be blindingly obvious. That’s halfway between peak EGT and 120F ROP (best power)…

Also, anybody who actually looks at their EDM downloaded data will see this from a mile away.

A 30F-60F EGT variation over a minute or so is going to be blindingly obvious

Yes. But still, if I were Savvy I’d want such a software badly, because it saves them having lots of A&P’s to look through user submitted engine monitor traces. And if I had such a tool I’d advertise this too.

The fact that it’s blindingly obvious probably makes it possible to have a low enough false positive rate while having essentially zero false negatives…

A quick and dirty FFT should make this sort of thing obvious, though the daft 5F quantisation of the recorded EDM data won’t help with picking it up early.

I took this video today – FL100 in cruise

It does show a periodic variation, mostly about 10F peak to peak, but when you watch the EGT bars carefully you can see that they tend to vary at the same time.

I think think this variation is caused by the fuel servo doing its “servo bit”, not by the rotation of an improperly sealing exhaust valve, because the latter would hardly be synchronised across multiple cylinders.