Northrop Grumman X-47B Design & Build

Hey everyone,

This is my first article on Flite Test. I generally post over on RCGroups, but figured that this community might be interested in seeing my latest project - an EDF powered model of the X-47B UCAS. 

Some background about myself - I'm a recent graduate of the aeronautical and astronautical engineering program at Purdue University and work at a major aerospace company. I used many of the design and optimization tools I had learned in my studies to determine the important aspects of the aircraft, so sizing, battery weight fraction, wing and power loading, airfoil, and wing twist distribution were all found mathematically. In addition, I perfomed some wind tunnel testing after the build in order to assess the performance and control. I've done my best to document the entire process, and will update this post as flight testing continues. Enjoy!

Weight Estimation

I needed to figure out how big the airplane would be before I could start CAD design and pick a power system. The first step to sizing the model was estimating the total weight and empty weight. I first put together a flight profile in order to determine how much total energy the batteries need to contain. It's pretty much a big list of guesses and estimations, and it's usually repeated a few times over the course of the preliminary design. The profile I came up with is:

- A 10 second takeoff run to a speed of about 60 ft/s
- A 30 second climb at a speed of 75 ft/s
- 3 minutes of cruise flight at about 90 ft/s
- About a minute of turning flight at 90 ft/s (this needs to be accounted for separately than cruise, since the airplane produces more lift & drag while turning)

There are some performance approximations as well, such as:

- A lift-to-drag ratio of 16
- Speed controller efficiency of 98%
- Average motor efficiency of 60%
- Propulsive (fan) efficiency of 40%

My code then calculated the energy required for each phase of the flight profile and then added them all up. The result is a fraction that represents how much of the airplane's weight should be made up of battery. It comes out at 16% for the values listed above, which seems reasonable as a first estimate.

Next, I made a plot in Excel that's just a straight line of payload weight plus battery weight versus total aircraft weight. In this case, "payload weight" is the Ardupilot along with all its sensors and wiring. I plotted another "historical data" line that represents average battery weight versus airplane weight for a list of similar R/C models. The point where the two lines intersect indicates the AUW that I should design for. The result is between 5 and 5.5 lb. You'll see later that the actual AUW ended up around 6 lb, which fortunately was close enough to the design value to meet all of the performance constraints.

Constraint Diagram

The next step in the process was generating the constraint diagram. This takes advantage of the fact that pretty much every performance parameter depends on wing loading and power loading. It essentially produces a range of feasible wing and power loading combinations for my model. The inputs are:

General Assumptions:
- Aspect ratio of 3.99
- Zero-lift drag coefficient of 0.03
- 40% propulsive efficiency
- 60% motor efficiency
- Oswald (span) efficiency factor of 0.60

Takeoff Performance:
- Zero elevation (for air density)
- Maximum lift coefficient of 1.3
- 75 foot takeoff run

Landing Performance:
- Zero elevation (for air density)
- Maximum lift coefficient of 1.3
- 75 foot takeoff run
- Rolling friction coefficient of 0.2

Ceiling Constraint:
- 1000 ft ceiling (this constraint is almost never relevant for R/C models)

Rate of Climb Performance:
- 1000 ft/min

Maximum Speed Performance:
- 300 ft altitude
- 90 ft/s maximum speed

Turn Performance:
- 300 ft altitude
- 70 ft/s airspeed
- Load factor of 2 (sustained 60 degree bank)

The program solves the equations for each constraint in terms of power loading over a range of wing loading values. The design space is the area where all of the constraints are satisfied, and it's marked out on the plot. From here, it seems like a 25 oz/sq.ft wing loading and about 150W/lb will fit comfortably in the design space. The active constraints are landing, rate of climb, and maximum speed, while the rest of the constraints are easily satisfied. However, a good rule of thumb is to multiply wing loading for a flying wing by 0.7 since the twist causes the tips to be very lightly loaded. So, I'll be designing the airplane for a wing loading of about 17.5 oz/sq.ft. For this planform, it'll end up around a 55" span.

Weight & Balance

I planned to use XFLR5 to analyze aerodynamic performance and dynamic stability, but I needed inertia properties first. This is kinda like the "chicken and the egg" paradox, since I couldn't build the detailed CAD model until after the analysis is done but I needed the detailed inertia properties in order to run the analysis. The best I could do to start is to let XFLR5 calculate the inertia for the wing structure itself, then approximate everything else as point masses. I used a flying wing CG calculator to find a baseline center of gravity location, and then I found weights for all of the components and distributed them around the model. I tweaked some things at the end so that the actual CG would be close to where it's supposed to be. This should be accurate enough for the analysis, and it also helped me to locate components when I start the CAD process.

Aerodynamic Analysis

To choose the airfoil for this model, I started with the NACA 64(1)-212 and added twist and refex to control the lift distribution, which is important for tailless models. I elected to go for a bell-shaped lift distribution that's essentially a sine curve to the power of 2.5. I believe this approach was pioneered by the Horten brothers, and it not only creates longitudinal stability but some lateral stability as well. The design ended up with a range of twist from zero degrees at the root to -3.6 degrees at the tip. The resulting lift distribution in level flight matches the "ideal" distribution quite nicely, and the airplane is both statically stable and trimmed in pitch. 

CAD Design

After I knew the wing span and airfoil, I could get started with the CAD design in CATIA V5. I imported a 3-view of the full scale aircraft, and used it to outline the model. This process took several months, but it's pretty well summarized from start to finish in the set of renderings below.

Mold Fabrication

I started manufacturing molds right after the CAD design was completed. They're made from MDF, which was cut with a 5-axis CNC mill and sealed with Feather Fill primer. This method is durable enough to produce a few airframes, but not too expensive. 

Skin Layups

The composite skins are a sandwitch of fiberglass cloth with a thin foam core. It took some experimenting to determine that this method was best for stiffness and weight. Everything was done using a wet layup and a vacuum bag.

Internal Structure

The internal structural parts were cut by laser from some 1/8" plywood. They were assembled first, then glued inside the skins during the joining process. Much clamping ensured a tight fit.

Removal from Molds

De-molding composite parts for the first time always feels like Christmas morning. Here are the X-47B parts fresh from their molds.

Gear & Doors

It was easiest to install the landing gear and doors next, while the model could still be rested inside the molds. All of the hinges are laser cut plywood, and no two are identical. This is because the many complex curves mean that the hinge geometry must be precisly controlled. Each gear door has it's own small servo, and an Arduino controls the sequence of motions.

Ducting & Fan Mounting

The ChangeSun 70mm fan was mounted next. The inlet and exhaust ducting are made from fiberglass, shaped using a positive plug. A hatch in the electronics mounting plate allows access to the fan after the model is assembled.

Control Surfaces & Wing Mounting

The control surfaces, servos, and remaining electronics were the last to add. Pin hinges were used for elevators and ailerons, while tape hinges were used for drag spoilers. The outer wings slide on with a carbon joiner tube and are retained by a thumb nut inside each landing gear bay.

Wind Tunnel Testing

When assembly was complete, the entire airplane was mounted into a large subsonic wind tunnel. This let me gather lift, drag, and pitching moment data and allso allowed me to trim the model before it ever left the ground.

First Flight

Finally, the model was ready to fly! I installed some vertical tails for initial flights that will stay until the ArduPilot's yaw controller is tuned. The plane flew great, and looks awesome in the air. No trimming was required.

That's it for now, thanks for reading! I'll be doing some more flight testing over the next couple of weeks, and will also be painting and detailing the model. Stay tuned!

Update: 6/7/2014: 

Got the model painted and finished! Paint is Testors Neutral Gray, and decals were custom made by Callie Graphics. I'm very happy with the result! Hopefully the solid color won't cause any orientation troubles while flying. 

Update: Here is some additional flight video from the 4th and 5th flights.

Update 2: I've only flown the X-47B a few times recently, as I'm not making much progress on tailless flight. Debating a switch from APM to AS3X, as I only need stabilization and I think it might be an easier approach. Here are some more flight photos.