Precise Anemometer From Scratch

This is about making an anemometer from scratch, helplessly overengeneering it and having fun with a ton of boring napkin maths needed for this. 

Along the following jurney i learned a lot about designing and claibrating anemometer. I encountered a a lot of other interesting stuff too, but unfortunatelly this might be far out of scope of this article. So enjoy the incoherent ramblings about making an anemometer from scratch ^^

Although i live in an area with only little wind to harvest, im hopelessly romantic about having a smal wind turbine on my roof. So i need a way to finally tell my self, that i live in an area in which such a turbine wont work, no matter under which aspect. For this reason i want to measure the actual wind speed over a period of time, in order to see for my self, that the wind speed is just to slow in general. For the anemometer to be usefull in the first place, it has to measure down to 2.5m/sec wind speed (lowest speed the turbine i want, starts to harvest). It also has to be weather proof and it has to be designed around one KY-024 hall sensor as pick up.

This is what i came up with. A robust housing with removable cups, internal electronics, one 608Z bearing and removable attachment points. 

How to measure wind speed anyway?
There are many ways in order to measure wind speed, like pitot tubes, cup anemometers, impeller anemometer and many more. The great thing about the cup anemometer is its general robustness, that it is easily to be made (compared to some others) and also quickly set up. Its basically like fire and forget. Concerning a cup anemometer, the wind speed could be assumed with the following equation:

→ V(rotation)=2*PI*r*U/sec
With:
- V(rotation)=Rotational speed at the cups center point [m/sec]
- PI=piece of cake
- r=radius from center of rotation to cups center point [m]
- U/sec=revelutions per seconds [hz]

This equation has the asumption that the rotational speed at the cups center point is equal to the surounding windspeed. This neglects all other aspects of a particular anemometer of course, like hub geometry, used bearings, cup shape/size and also friction. Unfortunatelly, this is going to cause this formula to deviate horribly from the true wind speed in the real world. So what now?

A quick google search is going to reveal some easy general purpose equations for geting a feel for the wind speed. They might be fine, but  also tend to neglect some aspects of the real world too. I tinkered a lot with them and found that they seem to be fine, if you want to get around +-2 m/sec accuracy. They migh also deviate up to 5m/sec though… I tested one equation, which was very close to the measured wind speeds. If you want to design your own anemometer, you are going to find all information/sources at the end of this article[2]

During web research i luckily found a thesis[1] for using and also calibrating cup anemometers. There, all the physics and inner workings got explained in great detail. One aspect hit my eye though. Equiations, like the one above, never got used even once. Within this thesis, a particular anemometer got strapped into a wind tunnel and while different windspeeds got applied, the rotational speed [hz] got plotted over the current wind speed. The different plots always showed a linear relation, and so a easy linear compenstation function was found.
The beauty with this approach is not only its simplicity but also that it takes regard to all the particularities of the tested anemometer, like aerodynamics, friction and shape. This is because the function is directly linked to the anemometer it self. AWESOME! So lets take this approach then ^^ 

First try and a bench sized wind tunnel
The only caviot with the above mentioned compensation function is, that i not only need a wind tunnel, but also an already calibrated anemomenter. Fortunatelly i have a simple one in my drawer, i rescued from sale one year ago. The claibrating setup consists of a big cardboard box, with a fan on one side and a slit shaped outlet on the other side.
 

The fan was controlled with a dimmer switch to different speeds. The windspeeds on the outlet got recorded with the hand held anemometer for reference beforehand. 
Note: It is important to have not to much turbolence in order to test the whole setup. Therefore the wind tunel features a  flap arrangement infront of the fan, in order to cancel the streams spin. I also made sure with the hand held anemometer, that the wind speed is consistant at all points of the outlets cross section.
I also wanted to check if the lenght of the cups had any effect. Therefore I tested for R75mm, R90mm and R140mm (R=Radius from rotational axis to cup center)
 

So, what does all the data show? First of all, i had one arduino configured on an interrupt pin to take the time between interrupt activation, every time the interrupt got triggered (every anemometer rotation). The results got plotted to serial every 0.4sec, until 50 mesurements got taken. In order to rule out the influence of temperature to my setup, i took the whole setup outside, and measured a second time. With all this, i got all data at 18.1°C (Inside) and also 1.3°C (Outside). Indeed, the measurements did not deviate from each other. Better safe than sorry.

We can see that with lower wind speeds, the time for one rotation does increase. We also see that the standard deviation betweend data points (at the same wind speed) does also rise. This is expected, because the measured time rises in lenght, the deviation is going to rise too. The last graph shows that the std dev in % does stay roughly constant though, underlining the previous observation. So in my book these are very nice results, arent they? ^^

Confident about the table top data, i strapped the anemometer to my car in order to collect some data with higher wind speeds. I repeated the test from above but in order to change wind speed, i tried to aim for certain speeds by driving in different gears. It was suprisingly difficult to hold the cars speed constant, have a look on the road, read the anemometer and also start the experiment. who had thought? 

 Although the pole looks like it was leaning, it was perfectly straight in reality. At least thats what he said. The data collected, was consistent with data we saw earlier. Thats a relief! So, what did we get?

Here i plotted the air speed over time for one rotation in herz. The relation seems to be nicely linear, so the earlier mentioned corectional function was also found with no hassle. We also have smal conundrum here. This is because usualy its a good thing to have a big change in the measured bandwith, with only little change in the measured quantity. So in this case, the R75 (0,2hz-9,2hz) shows roughly double the bandwith compared to the R140 (0,2hz-4,2) at same range of wind speeds (2m/sec-12m/sec). In our case, this comes with a little caviot though. Because the R75 also has the bigger deviation of all the sizes, it is not simply the best to choose. To pick the correct size now, is finding the best compromise. 

Concerning the big jump in deviation at 2 m/sec, im not sure what happened there. Since this occured three times and with all cup sizes and also seems to go away with higher speeds, i think this migh be some oddity about my cars aerodynamics a the mounting point. Since this didnt show with the wind tunnel test, im going to ignore this bump further on. Smal note here: I guess this bump might show how a anemometer is influenced if the flow comes from slighty below, above or is turbulent in genereal. This might be consistant with findings in the thesis[1].

Although a lower inertia, aka shorter cups, is better in order to have a faster response to very quick gusts of wind, this design already has a high inertia because of the big hub. Because of this, i came to the conclusion that the R140 size is suited best for all implementations of this anemometer. 

Conclusion
Thats it. 
We found the function for this particular model, of self made anemometer to be:
wind speed [m/sec] = 2.59 * rotation/sec [hz] + 1.31
We also saw that the deviation can be expected to be around +-3.5% in the range from 2m/sec to 13m/sec.
In my book, this constitues to be a success ^^

Im going to release all the files for the anemometer soon, so you can make your own and have it basically “calibrated” out of the box with the above mentioned function. So if you are interested in such a thing, stay tuned :)

Sources:

[1]
German thesis about calibrating cup anemometer.
“Eine Untersuchung im Windkanal zum Einfluss der Turbulenz bei Halbschalen- und Ultraschallanemometern für den Einsatz an Windkraftanlagen”
http://stroemungsakustik.de/old.mv.fh-duesseldorf.de/d_pers/Kameier_Frank/a_abschlussarbeiten/arbeit_deiss_lackmann_kurz.pdf

[2]
As promised, here a linkt to one awesome project by jattie. The anemometer they made, is a different aproach to the same problem. Also, one of the above mentioned “general purpose” equations got used. I found that this equasion worked to be correct to +-1,5m/sec, if used with my design. The nice thing about this is, that it might work with many different geometries and styles. So if you want to build your very own anemometer from scratch, id highly encourage you have a visit here:
https://www.printables.com/model/617911-modular-anemometer-with-roller-scate-bearing-and-d