I've been spending a lot of time over the last few weeks trying to figure out the optimum battery, motor and prop configurations in my skyfun to provide the maximum range and endurance options. Operating on a limited budget, I've been trying to avoid trial and error and make the right choices based on specs.
This is not easy but I've settled on a lower kv motor than many have chosen, running a larger prop with a 3300Mah 3s battery at only 20C (lower C batteries weigh less). The motor I've chosen is a Scorpion II SII-2208, 1100kv, 130W, 45grams, 12Amps, 10 x 4.5 prop.
I'm still mulling over ESC choice.
The final choice was also based somewhat on easy availability in New Zealand (Turnigy is not widely stocked).
There are a lot of suggestions on this and other forums about motor and prop choice which is helpful but what is missing is a semi technical discussion about endurance, range and speed and how to maximize these on any given platform. In the commercial UAS and military worlds, range, endurance and speed can mean the difference between a useful platform and something that is useless. In general, many of the systems people are building here do not seem to be optimized and are achieving quite poor endurance and range.
In commercial aviation, the optimum cruise speed is not based on endurance or range but rather to minimise time. Airliners need to get to their destination in the minimum time while using the least amount of fuel possible so the optimum cruise speed is the speed that flies the fastest possible speed using the least possible fuel per hour.
UAS's are totally different, in general our missions are to fly to a destination, loiter over a target for as long as possible and then return. In this scenario, we need to maximise range for the first part of the journey and get there as efficiently as possible (minimum watts per mile). Once at the target we need to switch to an endurance mode and fly at the minimum speed possible to maximise the amount of time over the target. We then need to return home at a speed that maximises range again (this might alter if a payload is dropped).
Alternatively we might need three modes, in an emergency you might need to get to a target quickly (optimal cruise speed as per airliner), loiter for as long as possible (maximum endurance) and then return as efficiently as possible (maximum range).
I'm not much of a maths guru but using logic I figure the first and easiest step is determining the maximum endurance of an airframe which would be done by finding the minimum amount of thrust required to maintain level flight. At this point, power consumption is minimized and the endurance is maximised (while speed and range are sacrificed), this would have to assume that motor and ESC efficiency are linear (from what I have read brushless motors are more efficient in the medium power bands anyway).
At the other end of the spectrum is the maximum range which I'm a little less sure on how we calculate on a model. For this I figure we need to know about the drag characteristics of each airframe across the full range of speeds.
I also figure that to calculate the optimal cruise speed (max speed with minimal power as per airliner) we would also need to know about drag at different speeds.
Is there anyway the APM logs could be used to figure out useful drag measurements to calculate these speeds for our airframes?
Or, are these speeds the APM could actually calculate itself by performing an 'in flight calibration'?
You could define in the flight planner which portions of the journey should be for calibration and which for maximimum range, endurance or efficient speed.
The APM could then slowly adjust the crafts speed downward in the calibration phase to determine the minimum amount of power required to maintain altitude and store this as the loiter speed. It could then slowly increase the power and measure which power setting delivers the maximum range for input power and then the maximum speed that uses the least fuel per hour.
I realise that there is also windspeed which could affect these numbers, which might mean that the APM might always need to be in 'calibration mode' and once put into one of these modes, constantly optimize the power so as to deliver maximum range, endurance or minimum time depending on what you want.
Once my new motor arrives (my stock motor now barely maintains flight) I will do some testing and add to this post, but would be interested in others thoughts on this topic as there seems to be a severe lack of discussion on such a relevant topic to UAS.
Comments
I think it's a limitation on Mavlink
I wonder why the current APM XBee data transmission is not configured to report parameters like, motor current, engine temperature, RPM, outside temperature, sonar height? Of basic interest, in my opinion.
How can that be??
I suggest that the developers add this into the telemetry reporting.
One would need to know the amp draw in real time, otherwise you'll have to fly, land and check drain using a charger, but that would include take-off draw, so could be wildly inaccurate.
The voltage is available over the XBee connection, but not amp draw, so one would need to fit a full FPV kit for this test.
Interesting! Maybe if you elaborate this there could be a APM standard procedure to have the APM "map" the performance factors of the vessel. This data might provide means to the MissionPlanner to calculate pretty accurate energy requirements for a planned trajectory at every phase of the flight (pre-flight, or in-flight). So you will have fuelplanning in you flightplan that caters for vessel performance and prognose/ actual wind, pressure, temperature, climb, cruise, decend. Turns etc... Now isn´t that pretty cool?
(Or is this already implemented...? I am kind of new in the ArduPilot realm...)
@ken, this is a useful article:
http://www.eaa1000.av.org/technicl/perfspds/perfspds.htm
It is where I stole the above graph from.
There are two parts to all this and I think we need to keep them quite seperate.
The first part is choosing the parts for your airframe to make it as efficient as possible.
For this you can use things like motorcalc and the general information that is available on the web.
The second part is giving the ability to the APM to find the optimum settings for any platform you have chosen on its own with as little configuration and input as possible. The APM should not need to know what sort of motor it has and what sort of prop its running, it should be able to figure out the optimum power settings for endurance, range and time to target on its own.
I've also done a lot of reading on motors and props over the last few weeks and in trying to understand the science and math I've also tried to think of a simple way of communicating it so that a relative newcomer can make good decisions and understand what it all means. Perhaps we can add this to the documentation somewhere on selecting an airframe (it might need fleshing out and technical correction if I have something wrong).
Here is what I have so far and I'd be interested to know if anyone has any further input (or would like to correct my assumptions).
Speed - to get the fastest possible airframe we need to create a thin stream of air moving as quickly as possible. The best way to do this is to spin a small propellor as fast as possible. This is where the 'screamers' come in and there is tonnes of info on various forums on how to achieve max speed.
The key to speed is to find the biggest prop you can run that does no significantly reduce the rpm of the higest RPM motor you can incorporate (the bigger the prop the wider the diameter of the stream of air). Two things affect RPM, the mass of the propellor itself and the friction of the air on the turning propellor. As you increase propellor size both the mass of the propellor and the friction of the air increase exponentially (assuming the same pitch props). Therefore a good way of finding the best prop for a given motor is to start with a very small prop and listen to the pitch of the motor RPM at full throttle. Gradually the increase the size of the propellor until you notice at full throttle the RPM starts to drop off, go for the biggest prop just before this RPM dropoff which should give you maximum speed.
Efficiency - There is far less info on the web about flying a model aircraft efficiently, in order to this we don't want a thin stream of fast moving air, we need a wide stream of slower moving air. We need to spin the biggest prop our motor can cope with before the motor is using more energy on simply rotating the mass of the prop than what it is on moving air. This would tend to lead us to a lower RPM motor which is capable of rotating a greater mass. This is probably a much bigger prop moving much more slowly than might be practical as the most efficient motor / prop combination would probably not give enough speed to create a useful system (unless you want to stay in the air a long time and don't care about speed). As a result you might need to think about what maximum speed you need and then work backwards to find the largest prop spinning at the slowest speed possible you need to achieve a given speed.
Also, consider using a gearbox. It will help you keep motor RPM higher for a larger prop.
I love both the mathematical modeling and dynamic calibration. Both would be extremely useful.
The model would be very useful to plan a design and to benchmark your build. If your model isn't performing where you would expect, it could help you troubleshoot some implementation error. Also the model could help determine the most important features when buying an airframe, motor or other part. Could you provide the equations that you used to generate the graphs. Also any directions on reading material from anyone would be great.
A famous quote goes "all models are wrong, but some are more useful than others." I think the dynamic calibration would be great, as it would provide the ability to overcome many simplifications in the model and adapt to our dynamic world (whether changes in current mission goal or changes in wind etc).
It seems to me that a model of prop performance would be necessary for decent accuracy of the overall model. The thrust is not only dependent on power input but also wind speed. Also typically with our models, we have a fixed pitch on a prop (not only fixed pitch, but also not many options on the choice of pitch).
Commercial airline holding patterns usually resemble long straights with reasonably sharp turns as this uses the least fuel. A very very large circle might be a little more efficient, but the circle would have to be so big you would not be close enough to your target to make it useful.
However, while this is the most efficient, I still think a hexagonal holding pattern where you cross to opposite corners of the hexagon would give the best balance between efficiency and crossing directly over the target the most number of times from multiple directions while using the least amount of energy possible.
I'm also not sure you need an airspeed sensor. For example, if you perform the above mentioned hexagonal holding pattern but maintained constant engine power then your groundspeed would alter based on the wind direction. This would allow the APM to work out the windspeed and direction. Once you calibrate and can eliminate windspeed then the APM would always know what its speed should be for a given power level. If it knows that airspeed should be 10m/s at 50% power and the groundspeed is 8m/s for this power level then that tells you the windspeed is slowing you down by 2m/s it could then decrease power accordingly to increase efficiency. If it sees it is actually travelling at 12m/s ground speed it might need to increase power slightly in order to maintain lift. If you again perform a hexagonal hold pattern then you will get different speeds going in different directions which should give you enough information to calculate the windspeed and direction and ultimately figure out what the groundspeed should be at a given power level. It just needs to know the airspeed constant for each power level.
However, I wouldn't be too concerned about wind direction as this would require a calibration for each flight, the main thing we need to know is actual airspeed at a power level which will be more or less depending on the angle of attack into the wind. If you performed a single 'calibration flight' which consisted of multiple passes in multiple directions at different power settings, then the APM would learn its airspeeds without the need for an airspeed sensor. This would also eliminate the need to know windspeed and direction and would keep everything much simpler.
Perhaps I'm missing something and oversimplifying it all though.
-
1
-
2
of 2 Next