The Eclipse flight project was established to demonstrate a reusable-launch-vehicle concept developed by Kelly Space and Technology, Inc. An F-106 delta-wing aircraft was chosen as the towed vehicle, and a C-141A transport-type airplane was selected for the towing vehicle. These airplanes are shown in Figure 1. Dryden Flight Research Center was the test organization with responsibility for safety of flight on the Eclipse project.
To enhance safety of flight, simulations of the two airplanes were implemented along with a simple mathematical model of a tow rope. A computational simulation of an F-106 airplane had been implemented at Langley Research Center to support some vortex-flow flight experiments, and this simulation was revived at Dryden. The C-141 simulation was adapted from an existing B-720 simulation at Dryden by replacing the mathematical model of the aerodynamics of the B-720 airplane with linear aerodynamic coefficients based on the performance of the C-141 airplane. The mathematical model of the B-720 engine was modified with a thrust multiplier to match the C-141 static sea-level thrust. In addition, the simulation was updated with C-141 weight, inertia, and center-of-gravity data. Existing simulation cockpits were used without modification.
The tow-rope model assumes that the tow rope lies on straight line between the two airplanes. On the basis of results from laboratory tests, the rope tension was modeled as quadratic in elongation and linear in elongation rate. This tow-rope model was verified initially by implementing it in a glider simulation and having a glider pilot subjectively evaluate the performance.
Initial studies were performed with the F-106 simulation alone. In these studies, it was assumed that the C-141 airplane was a point mass that would be unaffected by the forces on the tow rope. C-141 takeoff trajectories were generated and recorded in the C-141 simulation. These trajectories were played back in the F-106 simulation to study the takeoff performance of the towed F-106. This first cut showed some interesting results. The F-106 performance on tow was quite different from that of a sailplane. There appeared to be a lower and an upper bound on the tow angle between the two airplanes. Flight beyond these bounds would cause divergent pitch and sometimes roll oscillations. Fortunately, the oscillation amplitude would increase slowly enough that the pilot was able to recognize the problem and correct for it by flying back within the bounds. The simulation was already providing important information to the flight-test team.
To make the simulation study more realistic, it was decided that simulations of both airplanes should be performed simultaneously. To do this, it was necessary to link two independent six-degree-of-freedom (6-DOF) simulations - essentially creating a 12-DOF simulation. Although this seemed challenging at first, it turned out to be quite simple. The two simulation computers were linked with a fiber-optic reflective memory interface; this linkage enabled the sharing of airplane positions, velocities, and tow-rope forces between the two simulations.
To obtain consistent results, it was decided to synchronize the two simulations. The frame rates of both simulations were increased to 100 Hz, and flags in shared memory were created to enable the simulations to synchronize by polling. The interrupt driver in the F-106 simulation was used to generate the 100-Hz frame pulse, and the C-141 simulation simply waited for the F-106 simulation to indicate that a new frame should be started. The synchronization scheme is shown in Figure 2.
The results of the linked simulations confirmed the results of the F-106 simulation. The assumption that the C-141 airplane could be treated as a point mass turned out to be a good one. The C-141 pilot could not feel the effects of the F-106 doing normal small-amplitude maneuvers on tow.
The availability of two independent simulations also afforded a capability to achieve quicker, more productive, simulation sessions. Instead of generating a C-141 trajectory and then preparing and transferring the resulting data for playback in the F-106 simulation, the C-141 pilot could simply hit a "simulation reset" button and immediately try a different takeoff profile. This enabled the F-106 test pilot to quickly get the feel of the towed operation, and soon this pilot's task became easy. This setup also proved valuable for evaluating various failure scenarios during full mission simulation with the control room being fed by a stream of data generated by the simulator and transmitted by pulse-code modulation.
Six towed flights were performed in a demonstration program that was completed on February 6, 1998. Extensive instrumentation was used so that flight results could be compared with simulation results. It turns out that the simulation tow model was good at predicting rope tension, but a little conservative in predicting stability. The F-106 pilot was able to fly to more extreme tow angles before encountering the divergent oscillations. Part of this difference between the simulation and the flight tests may be due to the assumption of a straight tow rope in the simulation. During the flight tests, the tow rope would "sail" and develop significant curvature. In later flights, the tow rope was marked at regular intervals and video images were recorded so that this phenomenon could be studied in more detail. With the flight-instrumentation data and video images, it should be possible to develop a more realistic tow-rope model that can be incorporated into the simulation.
This work was done by Ken Norlin and Jim Murray of Dryden Flight Research Center and Joe Gera of Analytical Services and Materials, Inc. DRC-98-33