17 hours ago •
After the first successful (success == not flying apart) rotor test last week, the design was tweaked to change the orientation of the servos so that the centrifugal force held the internal gears in place instead of tearing them out. The change required moving the axle gear, and a new set of hub plates were cut with the gear pockets moved. Below is a shot of the reassembled hub with the reversed servo orientation.
The rebuild of the rotor allowed the opportunity to weigh the EVPR by itself (versus attached to the hub). The complete assembly weighs 2.7 lbs (note: in the future I still need to add the alternator, but the batteries give about 10 mins of run time right now). The fixed pitch rotors weighed 1.25 lbs, so adding variable pitch only adds. 1.45 lbs, better than I'd anticipated (The 7075 aluminum axles reduced the weight considerably). For a full set of 4 rotors the total increase in weight will only be 5.8 lbs.
In contrast, the grid fins control concept is anticipated to weigh 6 lbs a piece (not including actuators and mounting hardware. A set of 4 would add 24 lbs of weight. If the variable pitch rotors perform as hoped, that will mean upwards to 18 lbs of weight savings.
With the rotor rebuilt, two more tests were conducted. Both tests were 60 seconds long and the rotor held together. Both blades have survived 3 start-up/shut-down cycles and 3 minutes of run time. It's looking like the mechanical bugs have mostly been worked out.
With the hardware performing well, the prize money from the Wings, Wheels and Walkers round was invested in buying enough parts to build a full set of variable pitch rotors. The majority of the hardware is from Servocity.com and the parts arrived on Friday.
I've been really happy with their hardware so far and if you send them pictures of your project, you can get a discount. It's also pretty cool to get a handwritten note with your order!
More testing needs to be done on the first rotor to build up the time and cycles on the hardware to make sure things will hold up in the long term. The other big priority is to to mature the firmware and demonstrate controlling the servos.
As time allows, the buildup of the second rotor will continue. The hub plates have already been cut and the rotor blades are almost complete. Work on the build instructions have started now that the hardware has reached a somewhat stable state.
08/12/2017 at 17:21 •
Since the last project log, the design has undergone a few iterations. The main problem had been keeping the rotor blades attached to the hub. One failure mode was that the blade came detached from the axle and the second was that the axle detached from the hub.
The axle was fixed by added a second retaining ring so that both pillow blocks keep the axle from flying out. It also became apparent that the pillow block bearings were not adequate to take the axially loads. A thrust bearing was added at each pillow block to improve the design.
Better bonding techniques were used to attach the blades to the axles. Another test was conducted (see the video later in the log) and for the first time everything survived past the engine starting up. Then after running for about 90 seconds the blades detached from the axles.
It was decided that only using epoxy to keep the blades attached wasn't going to work and that rivets should be added. However the 1045 carbon steel axles are difficult to machine (at least with the equipment I have) due to their hardness. After some research it was decided to switch the steel axles out for 7075 aluminum which has nearly the same tensile strength, but is easier to machine and much lighter.
The latest iteration of the axle is shown above. A knurled surface was added to further improve the adhesive bonding to the axle. After the axle was placed into the blade, holes were drilled though the assembly and six solid aluminum rivets were added.
The new axles and rivets did the trick. For the first time the whole assembly survived startup, running for about 60 seconds and engine shutdown.
A few other issues have popped up as well. The screws holding the servo in place allowed some movement of the servo which wasn't desired. Torquing the screws down further wasn't an option with the plastic flanges. The solution was to use precision shoulder screws to mount the servos.
The screws are shown in the image above next to the red and black wires. They work great, but at $2.24 a piece they are pricey.
Losing rotor blades caused vibrations that damaged the quadcopter frame. Three of the gussets on the rotor arm were cracked. Below is a shot of one of the worst cracks.
Part of the problem was the concave corner that I had cut too sharply. The gussets were replaced, but a large radius was added to the corner to prevent a crack from starting.
I'd love to say that the hardware design is complete, but one last issue popped up during the last test. Inspecting the hardware after the test, one of the servo was not holding position and would make strange noises when the blade was rotated out of position. Taking about the servo it was found that one of the gears which just has a friction fit had come loose.
The culprit gear is the brass gear in the middle of the assembly above. Why did it come loose? While the gear case holds the gear in place from top, the centrifugal forces are likely pulling the gear out of the friction fit. To keep this from happening the servo should have the top pointing inwards instead of outwards so that the centrifugal force helps the gears stay in place instead of pulling them apart.
Looking at the design it should be straight forward to make the changes. It'll have the added benefit off keeping the main servo gear attached to the spline as well. The only downside will that it will more difficult to adjust the blade pitch with the gears inside of the hub, but it'll still be doable.
07/24/2017 at 01:26 •
While it's impossible to know the exact sequence of failures during the previous test, it's clear that losing power was a contributing factor. Using small gauge solid wires with soldered connections and standard servo connectors was a poor choice considering the high vibration environments.
The power supply board was re-worked afterwards to come up with a better design. The connectors for the servos and the batteries are all screw terminals now and all of the solid core wires have been eliminated. The image below shows the power supply board with the screw terminals.
Another test was conducted, this time with just the electronics powered and the blades and axle removed (The replacement parts aren't ready quite yet, but I wouldn't have put them on for this test anyways).
The test was a success. The start-up and shut-down sequence was conducted a total of four times and the electronics never lost power. I should have done a test similar to test before attaching the blades in the first place, which would have eliminated some of the re-work I'm having to do now.
The new blades are nearly complete and a spare set is in-work to ensure that if there is another failure, the project won't get slowed down. Below is the two epoxied blades and the cores for two more space blades.
One other design change in the works is the addition of thrust bearings. The hardware for that will get here tomorrow and then the blades will be ready to be installed with the thrust bearings. If everything goes smoothly, the full prototype will be assembled again and ready for testing by the end of the week.
07/16/2017 at 16:27 •
The first full prototype had been assembled and some basic firmware was written to do some simply tests. The first was a self test where the on-board ESP32 command the servos to sweep through the servo travel range. Below is the video of the self-test.
Note that the servos have way more range than is actually required. More gearing could be added, but that would mean a loss in response time. It was decided to keep the current travel range, but limit the range in software.
For the next test, the firmware was setup simply hold the blades at a specified angle of attack. The blade angle was set to a near zero lift condition (Due to the blade twist, the isn't really a true zero lift condition). The position hold functionality was tested by attempting to twist the blades by hand and the blades maintained their specified position. With the firmware functionality confirmed, the vehicle was turned on to test out the full prototype similar to the hub spin up test done a few weeks ago.
The test didn't go well. The vehicle had an unusually hard start-up, causing the vehicle to flex more than usual. Shortly afterwards the electronics on the rotor lost power. Afterwards, it's hard to determine exactly what happened next. Even at 60 fps, the blades are spinning too fast to see what's going on. However, without power, the servos can't maintain the blades. It's possible that they rotated into the belts and were subsequently ripped off.
Here's a photo of the rotor after the test. One of the blades was simply ripped from the axle. It appears that the bonding between the shaft and the blade wasn't that great in places. The black wire is the battery wire that became disconnected during the test.
The other blade (the one that bounces page into the frame in the video) took the axle with it. To do this, one of the retaining rings and the steel gear had to be stripped off (I still haven't found either of those pieces in the shop). It appears that when the blade impacted the wall, the axle was kept going and popped out through the blade.
Repairs are in-work and shouldn't take too long. More robust power connections are needed. I'll start with slightly bigger wires and screw-down style terminal blocks. Mechanical limit stops will also be added to prevent the blades from rotating too far. It'd probably also make sense to add some additional clearance between the rotor and the belts.
UPDATE (7/16/17 10PM PDT)
I finally found the missing gear today! The gear was embedded in the foam enclosure around the CNC router. The 2" thick foam was the best possible material for the gear to impact and the steel gear is completely unharmed.
07/11/2017 at 01:20 •
The previous project log showed the assembled prototype complete with rotor blades. However, the construction of the blades for EVPR hadn't been documented. The rotor blades use a similar process as what's documented in #Inexpensive Composite Propellers/Rotors, but with a few tweaks.The blades are built using a composite layup process. The image above shows all of the materials that go into making the rotor blade. The process starts with the pink object, which is the foam core. The core is cut into the shape of the rotor blade using a CNC router. The core as shown above is after the supports have been trimmed off.
To start, epoxy is mixed and a thin layer is applied to the core. Next, the two strips of uni-directional carbon fiber tape (The two black strips on the far right) are applied over the quarter chord on the top and bottom of the blade. Next, 3 layers of fiberglass (white woven fabric in the image) are applied individually and wetted out with epoxy as they are applied. The fiberglass is then covered with Aeroveil (gauze looking material below the fiberglass fabric in the image) which gives the composite a smooth finish when complete.
The next step is to wrap the blade with peel ply (green plastic) which keeps the bagging material from sticking to the epoxy and then with breather cloth (white felt) and then everything is placed it n the vacuum bag (yellowish plastic).
When vacuum is applied to the bag, the air is removed and the excess resin is sucked through tiny holes in the peel ply and into the breather cloth. This makes the part as light as possible. Below is the part after the epoxy has cured and the item has been removed from the vacuum bag.
Finally the ends are trimmed off, and the axle is attached. This is done by drilling a hole into the root side and using epoxy to adhere the axle to the blade. With the axle attached, the blades can then be installed onto the hub.
07/05/2017 at 03:56 •
The rotor blades and the electronics are now complete and have been added to the hub.
For now, the prototype will just be operated on battery power. Once the rotor has been tested, and shown to work, then the primary power source will be added.
Two 3.7V LiPo batteries provide the power (and will be the backup source for the final version). The batteries are tied together to provide the 7.4V needed to drive the servos. A voltage regulator is mounted on one half of the hub and provides power for the ESP32 dev board on the other side of the hub.
Small proto-board with battery connections and voltage regulator.
ESP32 dev board.
The hardware is now ready to test all of the capabilities, except for the primary power source. The next step is to develop the Firmware so that more complex testing can begin
06/21/2017 at 04:46 •
The prototype hub was assembled with just the mechanical components. (No batteries, controller or blades). It was decided that it'd be good to do a quick test to ensure that the assembly would hold together before proceeding further with the rest of the project. Therefore the hub was mounted onto the vehicle and a quick spin up test was performed.
The hub held together, so the next step will be adding the batteries and the controller board. I've also started making the first set of rotor blades, which I'll document in a future build log.
06/09/2017 at 02:31 •
While the primary power source for the rotor will eventually be an axial flux generator, batteries will be present to provide startup power and serve as a backup. The latest servos, HS-5685MH, run at 7.4 V max and draw a maximum current of 2700 mA.
Packaging was the primary driver for the battery selection. It was desired to have batteries that would fit inside of the hub. The battery chosen was the UBP053450/PCM. This battery will tuck into the gap between the servo and the upper plate and the main mounting bolts. The batteries are highlighted in blue in the figure below.
Each of the batteries have 900 mAh, so the total capacity is 1800 mAh. This is sufficient to provide 10 min of operation time at the maximum stall current for both servos, and should be adequate for controlling the vehicle if the axial flux generators fail.
The batteries should arrive next week. In the meantime a design for the battery clip needs to be created. After that, the ESP-32 dev board can be mounted inside the hub and the electronics testing can begin in earnest.
06/03/2017 at 01:48 •
The mechanical elements of the second prototype was assembled today and the gear drive was tested. Below is a short video of the gear drive in action (Note that the top plate is removed to share the gear movement. For this test, a temporary power supply was wired up with two 3.7V LiPo batteries to get a 7.4V power supply and a spare RC reciever was used to provide the signals.
The pillow blocks, servo mounts, and the servo gear are all off the shelf components. The shaft gear was custom made from gear stock. Below is a picture of the gear, which also includes a set screw for mounting to the shaft.
Here's an overhead view of the gear and servo on one half of the rotor.
The next step is to put the hub on the quadcopter to check the fit and the clearances. Then the quadcopter will be run with the hub in place to make sure that the assembly holds together. If that works then it'll be onto wiring up the prototype electronics and making the first set of blades.
05/27/2017 at 05:12 •
A few weeks back I assembled the major mechanical components for the first EVPR prototype. Shown below
This prototype used a simple Acetal bushing. This was done to keep the design simple and the weight down. After assembly, it became apparent that this wouldn't work as well as hoped. The dimensional stability is less than desired. When originally assembled, the shaft and bushing fit was acceptable, but with variations in temperature and the torque on the screws, the fit is quite variable.
The second issue was the lever arm concept for rotating the shaft with the servo. The mechanical movement wasn't working smoothly and the movement range on the shaft was less than desired.
After researching different configurations, it was decided to change the bushings to bearings and use a gear drive instead of a lever arm to rotate the shafts. Even better, most of the hardware needed could be source for a reasonable price from Servo City. A new branch was created for the repository and after a quite a bit of CAD work, playing around with various configurations, a gear drive re-design was settled upon. A screenshot of the geared setup is shown below:
The servo has a 24 tooth brass gear that fits directly onto the servo spline. On the shaft is a 40 tooth steal gear giving a gear ratio of 5:3. The shaft is supported by two pillow blocks. The only piece that needs to be custom made is the steel gear. That will be machined out of gear stock from McMaster Carr. The servo, servo gear, servo mount and pillow blocks are all off the shelf hardware from Servo City.
The hardware was ordered earlier this week and most of it's already been delivered, so if things go smoothly, the goal is to have the second prototype ready by end of Memorial Day weekend. Below are the hardware bits that came from Servo City yesterday.
I also plan to have more details of the servo sizing and selection in the near future.