Pelican's Perch #67:
Analysis of an In-Flight Engine Failure
GA engine failure captured digitally in full color! AVweb's John Deakin shows us engine-monitor data from an aircraft that lost power on takeoff just after an annual inspection. As you might expect, John disagrees with the engine manufacturer's post-mortem.
The EventAccording to the recorded JPI data, flights prior to the incident were normal (see chart), and then the airplane went in for an annual at the Danbury, Conn., airport (DXR). The failure incident occurred on the first flight after the annual. He reports that he did a very thorough preflight and runup, and everything appeared to be normal. He departed Runway 17, which requires flight down a valley with "good-sized hills" on either side. At about 500 feet he noted a slight odor, which he thought at the time might have been from some sort of solvent used to clean out the engine compartment. He now thinks it was the #5 cylinder, overheating. I have no idea what the smell might have been, but his thoughts sound reasonable. Shortly after that, he noted a slight abnormal vibration. At about 1,200 feet, the vibration increased. He noted that the JPI instrument was showing an alarm (a blinking display), and that the #5 cylinder was showing a problem. He very wisely did no further troubleshooting, informed the tower of his problem, turned around at about 2,000 ft AGL, and landed straight in. Nicely done. VERY nicely done. No "knee-jerk" panic, just thoughtful, timely, and decisive action. They immediately opened the cowl, and found the #5 cylinder was cold to the touch. A compression check showed zero. The data was downloaded from the JPI EDM-700, and he was kind enough to send it to me, along with data from two previous flights.
Previous FlightA glance at the previous flights is sufficient to see that there was no malfunction showing on the engine monitor. Here's one:
The "Failure Flight"Here's the chart. Take a look at this, and see if you can puzzle it out.
Update from John Deakin, Feb. 2012:
I'd like to correct what I wrote above, while keeping it for historical accuracy.
In 2003, when I wrote this article, we were just learning about the display on an engine monitor of detonation and preignition, and I recall a good deal of heated (you should pardon the pun) discussion on it. Today, I would say, "Typical preignition is indicated by an initial drop in EGT, while light and medium detonation shows little or no change in EGT. Preignition causes a very rapid -- often catastrophic -- rise in CHT, unless stopped by going rich on the mixture, or pulling off power to "break" the event. Light and moderate detonation may cause some rise in CHT, but often that may not be noticed unless it progresses on to prolonged medium to heavy detonation or preignition. Other factors may, as in this case, modify the typical initial presentation of a preignition event, as seen on the engine monitor and in the downloaded data. In this case, the spark plug ceramic had failed and it is likely that debris from the ceramic was causing the exhaust valve not to fully close. This resulted in the unusual and variable pattern in the #5 EGT seen in this data."
I've asked George Braly to elaborate on this new research.
Addition from George Braly of GAMI, Feb. 2012:
The #5 EGT started "wandering" for this flight before takeoff. Further, it is true that the #5 EGT started up slightly before the preignition event. This is significant because, by its very physical nature, a preignition type of event will exhibit a falling EGT, not a rising EGT as seen here. Since this article was first written in 2003, we have subsequently seen and collected data on dozens of preignition events. Over the last year (2011 through early 2012), we have been able to correlate the engine monitor downloads with actual preignition as documented by in-flight pressure transducers. There is no doubt that the "typical" preignition type combustion event is accompanied by a falling EGT on the affected cylinder. In fact, the physics of the preignition combustion event dictate that the EGT must decrease as part of the event ... unless there is some other confounding factor.
So what is the confounding factor? The evidence points to the spark plug. If you click and expand the picture of the spark plugs (below), you will notice that part of the ceramic nose core is missing on the upper spark plug on the left side of the photograph. The most likely analysis of what took place is this: 1) During taxi, the spark plug began to shed a piece of ceramic. At some point just after 15:44:46, a piece (possibly even just ground ceramic "grit") of that broken ceramic fragment from the spark plug became lodged around the seat of the exhaust valve and that small leak in the exhaust gasses caused the otherwise unexplained rise in #5 EGT during taxi.
Immediately after power up, either the spark plug (which had lost its thermal protection due to the missing cracked insulator piece) or even a piece of the ceramic floating around in the combustion chamber turned red hot and promptly started the preignition sequence associated with the very rapid rise in the #5 CHT.
While this was going on -- and on an intermittent basis -- some of the hard ceramic pieces continued to randomly interfere with the complete closing of the exhaust valve, which resulted in the erratic #5 EGT pattern.
An alert pilot might notice the unusually high #5 EGT during taxi. Maybe. But the pilot has to be routinely comfortable with the appearance of the "normal" EGT display in order to appreciate something like this event and do so in time to take effective action and abort the takeoff.
|(Click photos for larger versions)|
|Damaged Cylinder Walls|
|Damaged Cylinder Top|
|The upper spark plug is on the left and the lower plug is on the right.|