See the August report for the previous update.
Thrust testing has been completed on two-commercially available props for use on the 150g to 200g vehicle currently under development. Both props are 2-bladed carbon-fiber designs produced by the WES-Technik of Germany. The results of this testing are provided in the figures below. The 10-inch diameter rotor produces sufficient thrust for the proposed vehicle while operating at a voltage low enough (8.3V) to be provided by three or four of the 12g Tadrian Li-Ion batteries. Since this prop is only available for right-handed rotation the geometry must be ascertained and suitable left-hand rotation rotors manufactured. The experimental performance measurements of this prop/rotor combination will also provide another test point for development and validation of the current design tools.
Based on the use of these rotors with the Firefly motor/gearbox, the preliminary weight budget for the new vehicle shows significant payload capability beyond the primary flight systems:
Surface grids and near-body volume grids have been completed for the 3-D Navier-Stokes calculations on the 4-bladed, 2.5cm diameter rotor. By the end of October it is hoped to have all gridding completed and to begin computations for this first case. The performance effects of unsteady 2-D motions are also being explored. Time-accurate viscous calculations modeling an impulsive start exhibit an interesting trait in lift. The lift coefficient has been seen to rise and maintain a level significantly higher than the steady state value for a distance of several chord lengths. The lift then decreases towards the steady state value. This effect is apparently not easily explained by inviscid thin airfoil results, as in the Wagner effect. This behavior is being explored with several different tools including INS2D, inviscid Euler analysis, and unsteady potential flow.
Laser scanning results of 4-blade rotors have been further analyzed. According to the coordinates of measured points at leading and trailing edges, incidences at different cross-sections can be calculated. The following plots illustrate the incidence distributions along radius of two different rotors. The results show the inconsistency of the incidence variation in different blades within the same rotor. This indicates the epoxy rotors are not as stiff as we expected to retain the geometry within tolerance during handling. Different materials and possible approaches were evaluated. Machining rotors out of aluminum is a feasible option, while adding ceramic powders into epoxy may increase the stiffness.
Efforts were made to fabricate rotors out of aluminum. The strategy is to machine both surfaces of the rotor from bulk aluminum alloy. Considering strength and stiffness, 7075-T6 aluminum alloy is selected. Since 3-axis CNC machine can only access the part from one surface, the part needs to be removed from the fixture after the first surface has been machined and to be flipped over for machining the second surface. As a result, alignment becomes a very important issue. The practical alignment setup is to build a mold that fits the first surface. This mold provides a reference and solid support for the part. There are two possible solutions to attach the part to the mold-glue and vacuum. The initial trial took the glue approach. Two blades of the resulting rotor failed during the machining. This may due to the uneven glue layer that tilt the first machined surface relative to the mold. Consequently, thickness of some areas becomes so small that cannot survive during the machining. Next, the vacuum approach will be utilized to ensure perfect attachment.
Efforts were made to read the angular speed information from a Murata gyro integrated circuit IC by a PIC micro-controller. This efforts include setting up the development environment for the micro-controller, building a simple test circuit with the micro-controller, conditioning the gyro signal (filtering and amplification) and implementing a simple routine to read out the gyro signal and convert it into a pulse width modulated (PWM) signal at one of the micro-controller pins.
In continuation of the ongoing analysis of the dynamic behavior of the mesicopter, the three-dimensional model of the mesicopter simulation has been improved to include aerodynamic effects. The rigid body dynamics remain the same, but the rotor forces and torques have been modified. These forces are now a function of the inflow at each rotor. The equations are the same as those used in the two dimensional simulation. The velocity of each rotor is calculated independently to determine the inflow at each rotor. Simulations of the mesicopter have been run for various initial conditions. The plots below show two such simulations. The first plots show the open loop motion of the mesicopter with a small initial horizontal velocity. The second plots show the open loop motion with a slightly larger initial velocity. In the first case the motion is grows, and in the second case the motion decays. Allowed to run long enough, both simulations will settle into a steady limit cycle with an amplitude that lies somewhere between the two initial values. The fact that limit cycles occur with just these simple aerodynamics demonstrates the highly nonlinear behavior of the mesicopter. The hub forces which dampen these oscillations have not yet been included, and should improve some of the nonlinearities. The motion in the two plots takes place only in the X-Z plane and matches the results of the two dimensional simulation with similar aerodynamics. Note that for each run, plots of the position and orientation as well as the linear velocities and angular rates are provided.
Plot of 3D Simulation with Small Disturbance along X-axis.
Plot of 3D Simulation with Large Disturbance along X-axis.
The last plot below simulates a run with motion in all six degrees of freedom. The mesicopter is initially yawing about its vertical axis with a small velocity in the X-direction. The motion soon couples with the y-dimension and the complex motion shown below is the result.
Plot of 3D Simulation with Motion in all 6 D.O.F.
Currently, the two dimensional linear model is being used to design an LQR pitch controller that will use only angular rates as feedback. Then assuming that the pitch and roll motion are identical and lightly coupled, these controllers will be incorporated into the three dimensional simulation to control roll and pitch. Also, the aerodynamics for the three dimensional model are being improved as well.
The dynamic simulator for the PCB flyer has been improved. The dynamics model computes the spatial acceleration from the forces on the flyer. This acceleration is integrated to obtain the flyer velocity. A stabilization option which uses this "true" velocity ( 3 angular and 3 linear velocity components) has been implemented. The velocity is used in a PID feedback control algorithm to modify the input forces. This form of stabilization works extremely well even when force additive noise is very significant. The flyer bounces around and remains stable even with no additional user control input.
The optical flow algorithm is improving. It can stabilize vertical motion and yaw. It has some noise problems that are being worked out. Smoothed environment textures and more optical flow sample points are being explored to reduce noise in the optical flow measurements. Angular motion is estimated by summing the curl of the optical flow field projected on a sphere centered on the camera center of projection. Linear motion is estimated by summing the flow vectors projected on this sphere. These estimates suffer from the limited sampling of the camera image. Improved motion estimation algorithms are being developed.
The following image shows a snapshot of the simulator showing the physical model performing stabilized flight (using true velocity) within its texture mapped environment.
Work on our feedback control system continues. We are integrating a new vision feedback system into the wire frame demonstration we created this summer. Instead of using the VME point tracking system we used previously (see previous progress reports), common framegrabbers, cameras, and a Linux subsystem will be used to track LEDs placed on the mesicopter.
We have integrated and tested a Linux device driver capable of acquiring frames from a Matrox Meteor framegrabber. Using a modified version of the XVision package from Yale, we were able to display and threshold frames in real time.
We are taking various segments of code (new and old) and massaging these into a ControlShell vision repository (see previous progress reports). We are nearing completion of the initial version of this repository which will consist of an acquisition, a processing (thresholding), and display component for now.
The first iteration of the vision repository is being linked into our code from the original demonstration. We hope to have the demonstration functional in a couple of weeks.
Once the redesigned demonstration is functional, we will modify the code to control both yaw and height of the vehicle. The vision system will track multiple LEDs and we will pass state information to a MIMO controller. The controller will be based on dynamic models developed through the course of this work. We will perform system identification on data from this demonstration. This data will be used to study the aerodynamic properties of the rotors and to model the dampening in the system.
Last update: 16-Oct-00 12:15:17 PM
Click the Reload or Refresh button in the browser to be sure you are viewing the updated page.
Click the Submit button to enter changes and view the formatted file.
WebEdit servlet by I. Kroo, Oct. 1999.