simulating the Asiana 214 crash in FlightGear


FlightGear is a free / open source flight simulator. Eerily, SFO is the default airport for FlightGear. In simulation, I've flown into SFO runway 28R many times, and 28L a few. The crash was just short of / on runway 28L. Also, the Boeing 777-200ER seems to be a popular plane in FG; it's what I'd been flying for months before the crash. So it was spooky to see a crash in a place and plane that I'm very familiar with. (I haven't read the FG forums, but I'm sure I'm not the only one who was spooked.)

I read the crash reports for the first two weeks (or so) after the crash, although I haven't looked since then. From my armchair perspective, I think I know exactly what happened, and how to demonstrate it in FG.

The script for this, that I mention just below, is derived from my earlier scripts.

The Simulation / Setup

I am using FG in Windows 7, but I'm sure a wide variety of versions and platforms will work.

Before you set up the location in FlightGear, you'll need my script to set up the various plane controls for landing and other conditions that you can't set without a script.

I had a pretty good idea before the diagram, but this diagram was very helpful.

If you are already in FG (the program proper, not the Wizard), go to Location -> Position Aircraft in Air, then:

Make sure the top radio button is lit (white)--the one to the left of "Airport"
Airport: KSFO
runway 28L
the next 4 rows in FG are not relevant because we selected "Airport"
Distance (nm)1.4
Altitude (ft)620
Airspeed (kt)162
Heading 297.91
Leave azimuth and glidepath blank

If you are setting up from the FlightGear Wizard, then set Advanced -> Initial Position as:

Altitude 620
Heading 298
Leave everything else blank

I set the fuel level in my script, but two more parameters that you have to set "inside" the simulator under Equipment -> Fuel and Payload are "Pax/Baggage" and (optionally) "Crew." I have "Pax/Baggage" maxed out at 131,000 pounds, and Crew maxed out to 500 pounds (like anyone cares). My gross weight is 466,000 pounds. I don't know how close this is relative to reality; that probably has been published by now. If the real numbers are too far off, just change the throttle level to get the same results.

Note that each time you run FlightGear, the first time you will crash almost instantly--many seconds before the Asiana 214 crash. This is an artifact of the simulator--the first time the program runs, the physics engine is way ahead of the gui. After the first time, you can do Shift-ESC (reinit) and simulate properly as many times as you want.


As of my Asiana 214 v1.0 script, I have the internal throttle property set to 0.08. This translates to an "N1" throttle percentage of about 25.5--you can see this on the right side of the simulator screen / cockpit. With all of these conditions, if you don't move the throttles, you'll crash in about 30 seconds in a manner almost identical to A214. You might barely make the greenspace (which does not exist in the real layout) and land, but you probably won't make the threshold of the runway, let alone the runway itself.

The real pilot did move the throttles with a few seconds to go--if you do that, you'll probably still crash.

Other Physics Engine - GUI Delay Artifacts

Sometimes you'll get a warning to the effect of speed above maximum for flap extension. This is another simulator artifact--not relevant to the crash. That warning goes away after a few seconds. You might also get a terrain warning--in the first few seconds, that's an artifact. Later, it's real.

Differences between the Simulation and the Real Crash

One difference is the "A/T Arm" switches are turned off by my script. In the real world, those switches were on. However, autothrottle was not on--that's an additional button. (Last I checked, "they" weren't sure why A/T was off.) I turn A/T off because I found that sometimes it seems to come on by itself, which would mess up the simulation.

Slight Difference Between Target Conditions and What You Set in FG

You should set your initial speed about 14 knots above your target and your initial altitude about 20 feet above, due to the lag between the physics engine and gui.

Slight Differences in the SFO layout

The position of the seawall relative to the threshold of the runway is a bit different in FG from reality, but it's plenty close enough.

Commentary--How the Crash Happened

There were three pilots in the cockpit. They said they assumed auto throttle (A/T) was on--speed control, or cruise control. A/T is not auto-pilot like a Google driverless car; it's simply cruise (speed) control. So the engines were way too low (low power, not enough thrust), speed dropped dramatically, and they didn't notice until it was too late.

To some degree, not noticing the speed problem is more understandable than I first thought. 30 seconds before the crash, they are too high for a pretty approach. This isn't a safety problem in itself--they could have made a safe landing--but it was one of the causes of the illusion that they were ok. ~22 seconds into the simulation, the glide slope (vertical angle of approach) still looks good. A moment later, it's more much more obvious you're in trouble, but it's too late to gain speed and the plane is way under aerodynamic stall speed.

I said "to some degree," and I'm tempted to rant, but I'll save that for later or never. I'm sure there will be enough ranting--by parents who lost children, for example.

Page History

Updated Aug 7 (08/07) with weight and commentary.

Written 2013/08/05 3:11am EDT.

Valid XHTML 1.0 Strict