The web says that big data is a catch-all phrase for data sets that are so large that traditional processing applications are inadequate. Even though Savvy Analysis is just approaching its 3rd birthday, let’s call it a traditional processing application. On my last cross-country flight, my passenger was a web developer, and as we scanned for traffic, we brainstormed how the coming ADS-B rule might change engine data analysis. In the 1990s I enjoyed hearing Scott McNealy, one of the founders of Sun Microsystems, talk about smart appliances, and how a refrigerator equipped with wi-fi could monitor its systems, recognize the signs of an impending compressor failure, log onto the net, order the part, then schedule the service appointment – all without human intervention.
So my friend and I tried to apply that to flying. Imagine scooting along an airway, and along with the real-time weather and real-time traffic, you have real-time engine data available. Obviously you have it already – a whole panel full – but in addition to each gauge displaying its own information, you would have a real-time graphic depiction of EGTs, CHTs, FF – all the information from your data monitor, on one screen. So when you perform a lean-of-peak mag check and discover the bottom plug in cylinder 4 is marginal, you’re a couple of button pushes away from taxiing up to the shop at your destination airport – who have been alerted that you’re coming – and swapping out that plug, or cleaning a dirty injector, or replacing a failed probe, or whatever you were able to diagnose enroute. It sounds a little futuristic, but many of us learned to fly when moving maps and terrain awareness seemed like a far-fetched idea.
While we wait for big data, in this Puzzler we’ll look at some of the data we have now. This is data from a 2:22 flight in an SR-22. As Vin Scully says, deuces wild. I set the screen to show me four screens. You can do that on any browser using the settings icon – the golden gear at the bottom ofthe icon list on the right side of the page. Our default is two screens, but here I’m using the four setting and ignoring the lower one, because it helps to be able to see the anomalous data along with good data. Showing just MAP, CHT1 and Oil Pressure on one screen would have looked like a 3D seismic event.
Right away you have to be suspicious of the data. The purple MAP trace shows dramatic fluctuations from the high 20s to around 10 and sometimes to zero. If MAP were really doing that, we’d expect to see it corroborated by EGT and CHT, which are very steady. So if it’s not actual MAP, is it a bad probe or a bad connection? Let’s table that question and come back to it.
On the next graph, the red CHT 1 trace looks erratic, but the others are steady, and EGT 1 (also red but masked by other colors) does not corroborate unusual combustion. Plus, the spikes are not in the nature of the heat dissipation of a CHT. If it’s not real heat, then is it a bad probe or a bad connection? Unlike MAP, which is erratic through the whole flight, CHT 1 eventually calms down. Experience tells us that connections that are wobbly in one phase of flight (climb, cruise or descent) can become better – or worse – with a different angle of attack. Probes rarely do that. So let’s say we’re leaning towards connection but let’s continue.
PV = nRT
On the next graph it’s just Oil Pressure and Oil Temp – pressure is green. Lycoming drivers are probably thinking “gee, that’s low for oil pressure”. This SR-22 has a Continental IO-550-N so 50 psi or just below is ok. Experience also tells us that – despite what we learned in high-school chemistry – a change in oil pressure does not always cause a change in temperature, because the oil cooler steps up and corrects for it.
Now let’s try and connect the dots. There’s really no time when all three of these parameters are reliable. There appears to be a descent at about 1:30, then another half hour or so of cruise. In that descent, MAP is about as good as it got all day, and CHT 1 looks normal, but OP is wobbly. Earlier in the flight, OP and CHT 1 are ok, but MAP is jumpy. We’re leaning towards connections, but just to think it out, is there a scenario that would cause MAP, OP and CHT 1 – but only 1 – to be erratic? Can’t think of one, plus if those MAP fluctuations were real, we would have expected the pilot to land well before the 2:22 mark.
Holy Molex, Batman!
In this case, it helps to know something that isn’t in the data. Cirrus uses plastic connectors on both sides of the firewall to pass data from the engine to the panel. Experience tells us that we don’t always see the same data on the same connectors, and that having MAP, OP and CHT 1 data on the same connector is not unusual. In this case we recommended replacing the connector and the data became normal again.
Let’s finish up with data from a different SR-22, also a Continental IO-550-N. This is the takeoff and initial climb. Max FF is 28.8 GPH. TCM recommends 27.3 GPH and we like 0.5 GPH above that, or 27.8 GPH.
So this FF setting is a little higher than our recommendation. But that’s not the problem that causes EGTs to flip out and three of the six CHTs to show less power than the others. The clue was looking at the lower FF after the 00:09 mark and how that setting was able to achieve normal combustion and let the flight continue. So if less gas means better combustion – and this isn’t a turbo – then maybe we’re not at sea level, which is what that 27.8 GPH recommendation is based on. This data set did not contain GPS or altitude data, but big data came to the rescue.
With the N number and the flight date, Flight Aware confirmed that this takeoff was from KFLG Flagstaff, Arizona – elevation 7,014 MSL.