Mesicopter Progress Report
November 1999

Summary

See the October report for the previous update.

Rotor Development

Computational Aerodynamics (Kunz, Kroo)

The performance effects of variations in airfoil thickness distribution and edge shape have been further explored. The first set of sections considered were an NACA 0002 airfoil, and two 2% thick flat plates, one with a blunt leading edge and the other with a circular radius leading edge. The second set was an NACA4402 airfoil and a 2% thick plate with the 4402 camber line and a blunt leading edge. All of the plate sections have a nearly blunt trailing edge. It was necessary that the airfoil close to a single point at the trailing edge for gridding purposes. Comparisons were made at Re 6000, Re 2000, and Re 1000.

The INS2d analyses were completed on higher density grids than have previously been used to more accurately capture the flow characteristics in the regions of high surface curvature. For typical airfoil sections 256 by 64 grids have been sufficient, but for the plate analyses the gridding has been increased to 512 by 128. This provides considerably more resolution in the flow field to capture local separation regions and also has the effect of improving the accuracy of the original surface integration method used for the determination of forces. Comparison of surface integrated forces and forces determined from the control volume approaches agree well for the higher density grids up to the point at which the flow tends to become unsteady and the steady-state analysis fails.

From a design methodology standpoint it is interesting to note that the control volume method applied to the 256 by 64 grids provided force values that were quite close to the results from the 512 by 128 grids. Although the lower grid density appears to be insufficient for surface force integration, the flow field would appear to adequately resolved to obtain reasonably accurate forces from the control volume approach. If an iterative design method were implemented, the use of coarser grids would yield considerable computational savings. Finer meshes could be used for validation either at the end of the process or intermittently during the process.

The drag variation between the blunt-edged plate and the NACA 0002 appears as a constant percentage drag penalty for a given Reynolds number, except at high angles of attack where the penalty increases slightly. At Re 1000 the blunt-edged plate exhibits a nearly constant 5% increase in drag up to 4-degrees angle of attack. The penalty then increases up to 6.7% at 7.5-degrees. The Re 2000 results show a 6.2% penalty, increasing to 8.1 at the 5-degrees maximum angle of attack. The penalty at Re 6000 increases to 8.8%, but at this Reynolds number the onset of unsteady flow is very rapid so there is no gradual increase in the drag penalty. The cambered sections exhibit similar behavior.

The effect on drag of rounding the leading edge of the plates appears to be much smaller. This generally resulted in less than 0.4% reduction in drag from the blunt leading edge. The drag reduction was higher near the maximum angle of attack, reaching 1% to 2%.

One area requiring further study is the effect of the leading edge shape on the maximum lift coefficient. The results to date indicate that the blunt edges may reduce the maximum attainable lift coefficient. The steady state drag polars generally fail to converge for the blunt sections 0.5 to 1-degree earlier than the smooth NACA 4-digit sections.

Experimental Aerodynamics (Kunz)

Initial thrust tests with the 5-blade rotor and the 5mm Smoovy motor indicate that there may be some structural deformation occurring at operating speeds. The measured thrust degraded over several separate test runs as shown in the plot below. The input power to the motor controller also decreased for a given RPM, indicating that the rotor was producing less torque and drag for a given RPM. Further work is need to quantify the performance loss and to determine the nature of the deformation.

Prior to the degradation in thrust performance, the initial 5-blade rotor generated 1.4 grams of thrust at 45,000 RPM. The predicted thrust was significantly higher. A second rotor was fabricated with a uniform 2-degrees pitch increase. This rotor generated 1.8 grams of thrust at 45,000 RPM. The discrepancy between the predicted and measured performance may be due to several factors. The as-built geometry has not been accurately determined to match the design specifications. Work is ongoing to resolve this issue. Viscous wake effects are not currently incorporated into the performance model. This may be partially resolved by incorporating the wake velocity deficit into the performance model. There could also be inaccuracies in the two-dimensional section properties. Work is continuing to develop an experiment to validate these results. The current emphasis is on fabricating small gliders using the current airfoil sections. By flying the glider at the characteristic Reynolds number of the rotor and measuring its glide performance, a reasonable estimate of the section properties can be calculated.

Fabrication (Cheng)

In October, two trials were conducted to build rotors with better defined airfoil cross-sections. Based on the problems encountered previously, a third trial made some improvements in the model and the manufacturing processes. The rotor was fabricated up-side-down (-Z direction) as in trial 2. The root of each blade (connecting to the hub) was thickened to an elliptical shape. Both ball and flat end mills were used as in trial 2. When finishing the machining of the top surface, over-cutting the outside edges was applied. Careful checks were done in the CAD model to make sure the over-cutting would not affect the adjacent blades. In this trial, the flash problem was solved and the connection between blade and hub is structurally strong enough to survive machining and handling. However, another problem arose when the flash was removed—the thin trailing edges often chipped during machining. This suggests that the airfoil shape at this small thickness is not practical using the current manufacturing processes. Additional aerodynamic analysis suggested that the very thin trailing edge shape was not critical to performance so subsequent designs employ a new section shape that maintains more constant thickness over the chord. Both epoxy and polyurethane rotors were fabricated. Lift tests were conducted for both materials. Polyurethane rotors tend to flat out during the rotation, resulting in reduced lift. Epoxy rotors did not obviously have this problem. However, the lift (~1.6 g) is less than expected. As described in the previous section, the testing also showed the maximum lift decreased for different test runs. Two degrees of additional incidence were added to the blade design and the new rotors were again tested The lift was increased by about 12% , but was still below the designed value. We speculated that loading during the rotation (lift and pitch moment) might cause appreciable deflection that should be taken into account. Therefore, structural analysis was done for bending case due to the lift distributed all over the blade. Both finite element and analytical solutions are calculated and compared. In this analysis, the blade was assumed to be a cantilever beam subject to various distributed loads along the r direction. The beam was approximated by 19 elements, 20 stations. The moment of inertia, I, at each station was calculated in the CAD software and was assumed to be liner along the element. The force density was assumed constant along the element. The computed deflections associated with bending were very small and more refined analysis that includes torsion due to the pitching moment will be covered in December’s report.

Battery System (Brunet, Fabian)

In order to evaluate the performance of the various batteries a test bench, consisting of Hewlet Packard DC load meter and a PC was assembled. The PC runs Labview. A small Labview application was programmed. This application controls the load meter via a GPIB interface, reads the measured data in, displays them on the screen, and stores in a file. The motor tests have established specific power requirements. With the current controller we will need approximately 9v and 150mA per motor to reach 45,000 RPM with the present rotor. This corresponds to1.35W per motor or 5.4W total. Our mass budget suggests that we have only 5-7 g available for batteries on this intermediate-scale device and since the idea is to fly on commercial batteries to start, a comprehensive review of the options has been underway. The tables below list some possible candidates. We note that because of the limited selection of very small batteries, we are being driven clearly to the use of a voltage multiplier. Some discussions with companies working to develop MAV’s for DARPA indicate that we may be able to achieve the necessary power and voltage levels using a single cell.

Model	Make	Chemistry	Mass (g)	Nom Rat Capacity (Ah)	Rated Voltage (V)	Energy (Wh)	Estim. Energy Density (Wh/g)
N50AAA	Sanyo	NiCd	3.52	0.050	1.2	0.06	0.02
V40H	Varta	NiMH	1.52	0.040	1.2	0.05	0.03
V20HR	Varta	NiMH	0.89	0.020	1.2	0.02	0.03
2L76	Energizer	Li/MnO2	3.28	0.190	3.0	0.57	0.17
CR1220	Energizer	Li/MnO2	0.78	0.035	3.0	0.11	0.13
675 ZA	Rayovac	ZnO2	2.11	0.190	1.6	0.29	0.14
V76PX	Varta	AgO2	2.11	0.190	1.6	0.29	0.14
G13	Panasonic	AgO2	2.18	0.190	1.5	0.29	0.13
675G	Rayovac	AgO2	2.14	0.190	1.5	0.29	0.13

Data on selected batteries of interest for the 10-15g mesicopter.

Model	Make	Chemistry	Mass (g)	Nom Rat Capacity (Ah)	Rated Voltage (V)	Estimated Max Power (W)	Energy (Wh)	Power Density (W/g)	Energy Density (Wh/g)
N50AAA	Sanyo	NiCd	3.52	0.050	1.20	1.961	0.06	0.557	0.02
V20HR	Varta	NiMH	0.89	0.020	1.20	0.004	0.02	0.004	0.03
2L76	Energizer	Li/MnO2	3.28	0.190	3.00	0.787	0.57	0.240	0.17
675 ZA	Rayovac	ZnO2	2.11	0.190	1.55	0.005	0.29	0.002	0.14
G13	Panasonic	AgO2	2.18	0.190	1.55	0.283	0.29	0.130	0.14

Measured battery characteristics.

Since weight is so critical we examined potential weight reduction options with commercial batteries. We opened up two batteries of different chemistries and sizes and weighed their contents. The results suggest that for medium size (2-4g) batteries the casing constitutes about 45% of the weight, while smaller batteries (<2g) have cases that weigh about 65% of the total. Through some machining or custom battery development we may be able to reduce the casing weight considerably.

Structures/Frame Development (Prinz, Cheng, Fabian)

A mesicopter frame for the 5mm smoovy motors was designed and is currently being manufactured. The material used for printed circuit boards (PCB) was selected for the frame. This way the PCB is used also as a structural component. The frame must be large enough to accommodate the electronic circuitry of the mesicopter. For every frame design the whole circuit was routed in order to determine the required frame area. This way the minimum spacing between the electronic components could be found using single layer PCB.

The initial design for the 5mm smoovy-based mesicopter served as an example to become familiar with the Unigraphics CAD system. The second design included modifications based on recent stability analyses to move the c.g.. Batteries were also moved inboard to reduce the moment of inertia and the rotor spacing was reduced This design showed following drawbacks: a lot of unused frame area, still too large and components in the airflow of the propellers. The third and last design addressed these issues, and included the computed rotor tilt angles to maximum damping of the system dynamics. This design is shown in the figure below and includes the high solidity rotors described in the previous report (in holiday colors). We will next be defining shrouds for the mesicopter rotors to improve robustness and efficiency.

Stability and Control (Fay)

The linear analysis of the stability of the mesicopters has been revised and we have computed the effects of vertical c.g. position and rotor cant angle on the dynamics. This resulted in the revised geometry shown above. Next month’s report will include a summary of these results.

Sensors and Communications (Kroo, Holden)

Intel and Stanford have joined forces to explore control and communications for the mesicopter based on emerging wireless LAN technologies.  A layered system for closed loop control over the internet is envisioned, that consists of three components, namely 1) a human graphical user interface GUI for longer time constant interaction, 2) a Java servlet for short time constant interaction, and 3) a wireless actuator sensor unit (WASU) on the mesicopter for real time control.  The GUI uses a web browser or other application via a dynamic web page.  The resulting HTTP requests are sent over the internet (or to the web server running locally on a PC) to invoke a Java servlet which acts as a proxy for the WASU.  (Latency problems for real time interaction are addressed by directly interacting with the servlet on the same computer that is in radio contact with the WASU.)  The servlet uses the Java native interface JNI to access the serial port or USB port which communicates with a microcontroller and a 2.4GHz radio per the IEEE 802.11b wireless LAN standard.  Initial prototypes at Intel use a 900 MHz radio capable of 30-50Kbps.  The WASU on the mesicopter consists of a microcontroller, a radio, a 6 DOF MEMS based accelerometer (Analog Devices), and motor control.  The servlet communicates with the WASU and returns an HTTP response. This architecture will allow the WASU to perform fast motor control in response to accelerometer changes.  The servlet performs dynamic modeling of the 6 DOF rigid body and short term path control.  The human controller issues longer time constant commands.  As controller algorithms develop, more autonomy can be shifted to the servlet or WASU.  This architecture also provides convenient monitoring of the internal state of the mesicoptor via
the internet. In the next few months we will focus more on some of these communication issues and will start to consider the interaction and control of mesicopter groups or swarms.

System and Application Issues (Kroo)

Discussions with JPL continue on Mars applications and an upcoming test in the Mars environmental chamber. We’d like to complete an initial test there in early February. In preparation for this we are doing some low pressure testing of the motors and rotors at Stanford over the next month. We have also talked with researchers at AeroVironment, Inc. regarding power electronics and their Mars aircraft work. We are considering working together on a nearer-term version of a mesicopter for Mars. Prof. Kroo will present a paper on the mesicopter work at the American Helicopter Society meeting in mid January.

---

Edit

Last update: 15-Dec-99 10:15:35 AM
WebEdit servlet by I. Kroo