Recently a customer asked whether the groundspeed recorded by the flight data recorder was right. They had (allegedly) exceeded a taxi speed limit by a large margin and suggested, very politely, that we had got this wrong. This prompted me to do some checks and as the results were interesting I thought I would share them with you.
On the grounds of de-identification I am not going to name the customer, aircraft type, location or equipment installed in the aircraft, but this is the actual data.
The simplest way to check the groundspeed scaling is to compare the groundspeed to the true airspeed, ideally on a day with no wind.
If the two agree, it is fair to say that they are measuring the same speed. As the data source for the groundspeed is from an ARINC 429 digital signal, any data conversion error would either half or double the correct value, and with a speed of 464kt in the cruise, that means that either this parameter is correct or the groundspeed was in fact 232kt with a 232kt wind. Somewhat implausible!
So in flight the data looks fine.
OK, so we’re feeling confident about the speed through the air, but let’s check out the flight across the ground. This flight lasted 50 minutes and covered 281 nautical miles from origin to destination as a great circle route. (Where would we be without Google Earth?). This gives an average groundspeed of 337kts. However, this does not allow for the fact that part of the flight will be in the wrong direction because of turns in the climbout and approach. If we are a bit more sophisticated and use the Distance Travelled of 320 Nm, we get an average speed of 384kts which is very similar to the mean level of the graph above.
[A quick terminology reminder for users of POLARIS; Distance Travelled is across the ground, Distance Flown is through the air].
We can do the same test using just the travelling down a taxiway and if we are selective we can get a very accurate measurement of the distance. From the heading trace we found that the aircraft turned off at the rapid exit taxiway at the left of the Google Earth image, then continued straight down the yellow ruler line to turn right towards the terminal buildings. This gives us a nice straight, clearly defined, section of the taxi to measure.
Here is the recorded groundspeed for this section of the flight, and the ground distance computed by integrating the speed for the 112 seconds of this part of the flight.
The distance computed by integrating the recorded groundspeed is 1.64 nautical miles. The distance measured using Google Earth is 0.78 nautical miles.
So why does the aircraft record the correct groundspeed in flight, but almost exactly twice the correct groundspeed during the taxi in?
My late father-in-law had a phrase “looking here, looking there” for searching for an answer to a mystery and this is what we did. Here are the questions we asked, and the answers.
Was taxi out affected as well as taxi in on this flight?
This was inconclusive, as the taxi out did not have a nice straight section to measure. However, there was a clear spell lasting four minutes where the aircraft recorded a groundspeed of 23kts (with a little variation to make it appear “real”). This happened before pushback and before engine start.
Were taxis affected on other flights?
Yes. The subsequent flight recorded excessive groundspeed during the taxi out but measured the correct groundspeed during the taxi in on that flight.
Was this specific to this aircraft does it affect all of this type?
Looking at the groundspeed exceedances across many operators we can see occasional exceedances, but they were valid. If you taxi one nautical mile down the length of a taxiway in one minute, the aircraft really is taxiing at 60kts. The rate of events over 45kts for aircraft of this type excluding this specific aircraft was 1.33 per thousand flights. The rate for this specific aircraft was 22.7 per thousand flights. Ignoring the many aircraft which never triggered an event, here is the rate per aircraft for those which exceeded 45kts:
So we can say this is probably an issue affecting this single aircraft, and not common across all of this fleet.
Does this happen on aircraft of different types?
There are high speed events during taxiing for other types, but only at a low rate and the speed measurements have never been questioned before.
Does it happen in specific places?
Speed events for the global fleet happen at locations where there are long straight taxiways (perhaps we should consider chicanes?) but there is no indication that the event rate is high in one airport more than another.
The operator has been advised and may consider maintenance action to correct this signal. In the meantime a change is being introduced to provide a check against the measured position during ground operations only. This will come into operation in a future release of the POLRAIS suite and can be seen at https://github.com/FlightDataServices/FlightDataAnalyzer/blob/master/analysis_engine/derived_parameters.py. Search for “GroundspeedSigned” to see the affected code.