View Single Post
Old 3rd Sep 2016, 14:06
  #95 (permalink)  
BRDuBois
 
Join Date: Nov 2015
Location: Seattle area
Posts: 213
I've uploaded a proof-of-concept video to my weTransfer page. This is intended to show the kind of data I'll capture using the X-Plane simulator. This is a 15 mb MP4 video. The frame rate is low, but the goal here is to capture data, not to make a smooth video experience. https://we.tl/GU7wQaiDOZ

I envisioned three levels of testing with the simulator. The first was to test basic handling characteristics of an Electra. One of the objections raised to my analysis has been the proposal that Flight 706 not in a shallow bank but was knife-edge flying when it hit, and this explains the slow descent. Hand-flying the simulator shows that this is impossible because Flight 706 was going about 100 knots too slow for knife-edge flying. Hand-flying is sufficient for this proof, but that's about my limit as a pilot.

The second is to determine the characteristics of an Electra in a banked turn. Turn radius is critical to understanding how Flight 706 got to where it impacted on the airport grounds, and the angle of attack it could have had. So the next set of tests involves putting the simulated plane into turns of varying angles and different rudder deflections and measuring the turn radius of each. A pre-requisite was to write a simulator-controlling harness, which now is showing basic functionality. The first job in this test series is to calibrate the simulator and ensure that it accurately reflects what we could expect of an Electra. The link shows part of such a calibration step.

Non-Techies skip this part: The internal control is via a Lua plugin running inside Gizmo. Gizmo is a Lua/X-Plane interface plugin. This Lua script is a dumb plugin; it passes simulator data (eztracted by Gizmo) to the external program and the external program passes data values for the plugin to stuff into the sim through Gizmo. The plugin makes no decisions. The external program is a Delphi 10 (Pascal) program that has rudimentary flying abilities. Delphi 5 was my last installed version, and it's the best IDE I ever used, but it's unable to deal with W10. Embarcadero is giving away an intro-level Delphi 10 for free for the next four or five days, so it you ever want it this is a great opportunity. The intro level has plenty horsepower for this job.

The uploaded video shows the level of control needed to establish Electra flight characteristics. The harness can hold the sim plane within about a half a degree of bank angle, and within very few meters of height. There is a ringing effect, so the longer it runs the smoother it gets. With this harness I can put the plane into a stable state and then mess with rudder inputs and the like. The large readout shows the time it takes to make a circle, control surface deflections, headings, etc. It's not all wired up yet. The goal at this point is to calibrate the simulator and refine the Delphi flight controller so it smoothes more quickly. The video shows the Electra (in 1960's livery) circling over Elliot Bay and Bainbridge Island.

Not shown in the readout are the longitude and latitude values that determine exact coordinates. These are captured in a log file with other data, and I use those to compare to the speed and timing, which determines circumference, to check on the circle dimensions. There is a distortion in these measurements due to the way the simulator handles the 'flat earth problem', so I'll have to do the real measurement flights in Jakarta, not Seattle. Circles flown near Seattle turn out to be north-south-elongated ellipses, buit circles flown near the equator are circles.

The third set of tests will be to replicate possible flight paths for Flight 706. I won't be able to work on this until I have more data, a feel for how the sim handles, and a better Delphi flight controller. If the sim cannot be shown to be a good representation of Electra handling, then more work would be wasted. Even if the sim is only a marginal flight model, there should be some useful information about how rudder deflections alter the turn radius. If the sim turns out to have a dependable flight model, then there may be more useful data forthcoming. At this point I can't be sure the sim will generate anything useful beyond disproving the knife-edge flying conjecture.
BRDuBois is offline