What’s the right next question to ask?
That mantra has been drumming in my head throughout the holidays of 2016 and until now. By asking that question, and designing the next right question to ask, and then answering it, I can feel the pace of this project growing and growing.
Progress, is a byproduct of pace.
Big progress was made last month in 2 big areas:
- Motor, Inverter, and Battery research
- Studying Tesla Motors patents
- Understanding the state of the art
- Understanding analysis techniques
- Project Planning
- Defining the critical path of the project
- Laying out the critical path throughout the year
- Defining milestones and objectives to hit
Just under 1 week into the new year, and big advances have been made! The first three milestones have been completed, right on schedule.
Milestone 1: Verify the PROP_DESIGN program’s real-world accuracy or make corrections
To accomplish this, I pulled out the good ol’ motor test stand from before!
Like that one friend everyone has that just can’t get enough of going to the gym, this thing is outrageously beefy! It can withstand about 2000N of down force, and measure that up to 1000N.
It took some time to rig up an angular speed (RPM) sensor (“tachometer), and get it to work accurately… Like three days… Too much time honestly. Turns out the response rate of an LED is not so good as a phototransistor… Nevertheless the system was calibrated (with an off-the-shelf tachometer) and worked accurately up to about 8000RPM! The biggest drawback of this setup is that it must be done in total darkness, otherwise ambient light shifts the signal of the LED receiver to the wrong threshold!
I also rigged up an off-the-shelf anemometer (wind speed sensor) above the test stand to take axial wake velocity measurements.
To measure electrical power, I bought a beefy 50A DC shunt from All Electronics in Van Nuys, and hooked it to my multimeter, manually taking measurements.
The motor? The SunnySky X3508S-580KV. The ESC? Just a normal 30A ESC… The Battery? I used both a Lithium-Ion 4S 3400mAh battery rated at 30-40A draw and a Lithium-Polymer 2200mAh rated at well over 30A. I switched to the Li-Po permanently after some tests indicated that individual cell voltages were going too low on the Li-Ion.
The results? Well, if graphs make you giggle with excitement (like they do for me!) :
If graphs aren’t your thing, no harm no foul; I will explain!
The initial prognosis was not so good… PROP_DESIGN seems to have an error up to 50% at the high end and low end of the RPMs. Ouch!
With that said, the axial wake velocity measurement was very good. On average, 12% error in that department. That’s a good thing since all of the cascaded configuration measurements were based on this measurement of flow propagation.
Once I had seen that amount of error, I went hunting for the cause. The largest cause I could come up with is an error in the way PROP_DESIGN was generating the geometry. In other words, was PROP_DESIGN thinking the prop looked different than it actually does?
I decided to test the results via a 3rd source: Computational Fluid Dynamics. I set up a simple run in the Autodesk CFD 2016 software, modeling angular velocity by a simple linear distribution of velocity along the x-axis, with the flow in the +y-direction. Pulling a periodic all-nighter, I ran a number of tests with the varying RPMs I had tested. It took a total of 24hrs to do all the runs!
The results, are just as interesting as I could have ever imagined! The PROP_DESIGN results agree very well with the CFD results, in all measurements except axial wake velocity! CFD did a bad job predicting those measurements… Very bad. Like 1000% bad.
So, most likely, my test stand thrust & torque measurements were off (load cells may need recalibration). The one sensor which worked very well was the off-the-shelf wake velocity anemometer, and that agreed with PROP_DESIGN!
So the accuracy of PROP_DESIGN seems to be within the margin specified in the CFD agreement plots. The motor test stand needs some software tweaks to get better data, and account for the hysteresis of the constant load on the load cells.
Additionally PROP_DESIGN is good enough at axial wake velocity measurements to continue without significant corrections.
Milestone 2: Calculate adjusted cascaded configuration for Generation 1 Revision 2
“Generation 1” is the 500N system, designated Revision 2 as this is a revision on a previous initial design which was made.
Since there wasn’t much need for PROP_DESIGN adjustment, for this milestone I primarily went through and checked all the G1.2 calculations. I went through and verified that I wasn’t going to attempt to break any laws of physics (generally a bad idea). I checked that conservation of mass, energy and momentum were all being satisfied. They are!
Milestone 3: Systems Engineering for Generation 1.2 (G1.2)
This is where the fun begins!
Systems Engineering & Design is a great deal of fun and creativity.
The “visual” results of my labor are the following system architectures. Each fan stage will incorporate this “big picture” system:
The geometric assembly of the system is as shown above! M1, M2 etc… designates motors, and I1, I2, etc… designates inverters.
The system is liquid cooled, for all components excluding the battery (I think… Still deciding on that one).
The real hard work here was not the visualizations, but the physics!
The detailed work is no-where near being done. In fact this is like… Step zero in about 50 billion. Even though that is the case, initial systems engineering does not delve into the details.
With that said, some initial attributes of the system are expected to be as follows:
Total Thrust = 551.0 N
Total Power = 17.39 kW
Total Effectiveness= 32 N/kW
Supply Voltage = 300V
Power Range = [1.65 , 3.65] kW
RPM Range = [2500.00 , 7600.00] RPM
Torque Range= [4.59 , 13.94] Nm
Power Range = [0.82 , 5.47] kW
RPM Range = [1250.00 , 11400.00] RPM
Torque Range= [2.29 , 20.91] Nm
Freq Range = [41.67 , 380.00] Hz
Amps Range = [2.75 , 18.25] A
Amps Per Phase= 6.08 A
DC Power Range= [0.82 , 5.47] kW
—-Battery—- (based on this cell)
Cells Per Battery = 82.00
Batteries Per Pack= 2.00
Cells Per Pack = 164.00
Cost Per Pack = 1374.16 $
Weight Per Pack = 7.71 kg
kWh Per Pack = 2.12 kWh
Voltage Output = 303.40 V
Max Amps Output = 20.00 A
Max Power Output = 6.07 kW
Time Range = [23.01 , 152.73] mins
These were calculated using design equations coded inside a Python program I wrote to go through all of these factors. I will figure out some way of sharing this program in the near future.
What’s the right next question to ask?
Objective 4: Detailed Engineering of Power System
As of now, I am beginning the road of motor, inverter, and battery detailed engineering cycles. This is a significantly more time consuming task! But, I can feel the pace picking up, and I can see the potential growing!
Ultimately, I plan to share as much of the detailed work I do as possible. I’m considering the best way to involve more people in this endeavor in a positive way. Open-source software engineering projects like Linux or Pixhawk/Ardupilot have been wildly successful in the past, but hardware is a different animal.
The goal here is to increase the speed of innovation in aviation, and ultimately to move air transportation to fully electric & sustainable.
To that end, it makes very little sense to keep anything proprietary or closed-source.
The potential gain in this project is not measured by dollar bills put into my pocket, or any one individual’s.
It is measured in the number of sustainable aircraft in the skies every day.
Today, that number is very close to zero.
Let’s bring that number up.
Stay tuned for more details on how I will open-source this project as much as possible!
Ideas? Post them in the comments!