- Joined
- May 27, 2017
- Messages
- 14,458
- Reactions
- 19,470
- Location
- Los Alamos, NM
- Site
- www.mavicpilots.com
I recently received a request to try to determine the location a lost Mavic Pro that lost power and shut down in mid-air. The logs were interesting from several perspectives. Firstly - the summary graph, which tells the basic story:
The aircraft ran out of battery and lost flight control and propulsion at just under 5,000 ft AGL, but the logs continued for another 6 seconds or so. The log data showed a consistent and steady wind field (~ 30 mph out of the west), but when I ran the usual numerical simulations of the resulting descent it became apparent that the drag coefficient was wrong - at the end of the log it had already exceeded the predicted terminal velocity by around 2 m/s.
Anyway - averaging a number of simulations, varying wind speed/direction and drag, yielded a likely crash location with an uncertainty circle of around 150 meters radius. The crash site was found around 50 meters from the center of the circle, which is reasonable, but in confirming the log data it also confirmed that the variation of drag with altitude is not negligible. Drag scales with fluid density and the square of the velocity, and at 5,000 ft AMSL the air is significantly less dense than at sea level, making my previous constant drag coefficient significantly wrong at higher altitudes. Adding a density-dependent drag, based on altitude in a standard atmosphere, shows the size of the variation:
That shows a 15% variation over 5,000 ft altitude change. Implementing that directly in the numerical simulation fixed the problem quite well - it yielded a match to both the vertical descent speed at the end of the log and the actual impact point to within a few meters:
While it's only matched to one data set, since we don't often see 5,000 ft free fall events, I'm fairly confident that the altitude-dependent form of the drag is pretty close, and I'll use that for future incidents.
As for the flight itself - don't try that at home.
The aircraft ran out of battery and lost flight control and propulsion at just under 5,000 ft AGL, but the logs continued for another 6 seconds or so. The log data showed a consistent and steady wind field (~ 30 mph out of the west), but when I ran the usual numerical simulations of the resulting descent it became apparent that the drag coefficient was wrong - at the end of the log it had already exceeded the predicted terminal velocity by around 2 m/s.
Anyway - averaging a number of simulations, varying wind speed/direction and drag, yielded a likely crash location with an uncertainty circle of around 150 meters radius. The crash site was found around 50 meters from the center of the circle, which is reasonable, but in confirming the log data it also confirmed that the variation of drag with altitude is not negligible. Drag scales with fluid density and the square of the velocity, and at 5,000 ft AMSL the air is significantly less dense than at sea level, making my previous constant drag coefficient significantly wrong at higher altitudes. Adding a density-dependent drag, based on altitude in a standard atmosphere, shows the size of the variation:
That shows a 15% variation over 5,000 ft altitude change. Implementing that directly in the numerical simulation fixed the problem quite well - it yielded a match to both the vertical descent speed at the end of the log and the actual impact point to within a few meters:
While it's only matched to one data set, since we don't often see 5,000 ft free fall events, I'm fairly confident that the altitude-dependent form of the drag is pretty close, and I'll use that for future incidents.
As for the flight itself - don't try that at home.
