I want to use this class to finish the experimental platform I developed in HTMAA. You can find here the wind tunnel I made (core structure + laminizer chamber + wind flow generator) But there are more things needed. I am aiming to develop a presice experiment carriage that will provide:

A nice mechanical clamping to the wind tunnel section

A presice meassure of Lift and Drag loads

An interface to read the measured data

Room to implement controllers and develop dynamical testing

Extra, some smoke flow visualization

1. Load Cells Disposition on Low Speed Wind Tunnels

When designing your experiments, two main strategies are used on wind tunnel to measure the reaction loads of the body inmersed on it.

1.1 Horizontal Cantilever of Load Cells

1.2 Vertical Cantilever of Load Cells

Something I want to continue working with is manufacturing a 2 Axis load cell. This will allow me to have only onse sensing harware module that I will attach to the different locations so we will be able to change the wind wunnel sensing strategy anytime when needed.

Week 3

I got finally decided with the two architectures I want to build for the two different platforms. For the vertical measurements I am going to build a Steward Platform. I have sketched how I can meassure all the axial loads of the bars and next week I will work towards that. For the Horizontal config, I am going to build two L shape load cells with a rotatory element placed in the laterals of the wind tunnel.

This week I decided to dive in first in have a feeling about if Im able to build my own sensing platforms with arbitrary shape. To do so I am going to build some classic loadcells. Firstly I am going to make a quarter bridge with a simple monolithc bar just to warm up and to see if I can read voltage. Strain gauges are variable resistors but we need to convert that Ohms meassurements into voltage as it is way more handy to be readded by a microcontroller. The circuit that enables you to do so are Wheatstone bridges.

Zach shown me in the lab a lot of different strain gauges. They were very old, ordered (or at least manufactured) in 1986.

So for the first sensing devide I took a monolithic square bar of aluminum, I prepared the surface by sanding it, and applied a Loctite surface activator and Loctite 401. Later, I attached the strain gauge into a bigger cable and protected all with flexible epoxy:

I also build the Wheatstone Bridge and milled with the Roland. I was able to meassure mV when slightly bending the monolithc bar!

Now I was able to jump into a more complex model. From a quarter bridge I will jump into a differential bridge and 2 axis load cells. To do so I cad and waterjet this shape.

This type of shape works in bending. Doing so, the deformed shape of the holes I made made some of them work in compression and others works in tension as the schematics shwon:

Doing a full bridge I will have a differential meassurement for any bending force in the structure. Also, this type of configuration I made is able to read loads in 2 axis. In my case I am thinking on lift and drag.

I treated the surface and glued the strain gauges :

And provided two connection pins for every strain gauge to attach bigger cables. I also solder and routed all the cables into laterals PCBs to be able to read it more comfortable.

Week 4

code here You can check the video here. This week Ive been working in the code above, reading the load cell and plotting real data. Also Ive been doing microstepping raspi code with a NEMA17. Im ready now to controll the angle of attack, read the load and act according.

My ultimate goal is make the wind tunnel talks with the morphing wing.

Week 5 to 11

This last weeks Ive fix the final design for the horizontal balance, manufacture subcomponents, assembled them, failed a bunch of times, set up the teensy to serial read the HX711 at 80Hz and somehow make all work together.

Design

Manufacturing

EDM LOADCELL

I firstly decided to jump from waterjet to EDM the load cell. I decided so to controll way better the tolerance of the thin section that the L structure have. Of course Zach helped me with the EDM process.

Later, I realized that maybe its not needed as you'll need to sand the lateral face to make the superglue work.

Ive also did cut the main base using the waterjet and tapped all the holes to be M5.

WIREPLOTTING WITH THE ZUND

Sam Calisch , PhD student from our lab developed a beautifull hack for the zund as is wireplotting. You can check here the process (last chapter). That allows you to print a continuum (hamiltonian) path over a sticky material that, in my case, I used later to paste it on the loadcell and route all the neede cables.

As in the last loadcell it was a nightmare to route one by one those tiny cables I decided to automate it and even use thinner cable (41 AWG). First I did a dxf of a closed circuit that later will be cutted the perimeter in a way that breaks the continuity of the path, with the Zund in the same job

Check here a bit of the process: And heck here the result

INSTALLING STRAIN GAUGES

After ordering propper strain gauges and faling installing some, I did come with a method that works :

Assembly

There's a reason why the protective kapton layer is more than needed.

But when it works, and one attached, is so satisfying when folds to its place:

Once all is assembled, I milled a PCB to read accordingly the strain gauges on its order:

And the full element looks like this:

Lets make it run

Final Weeks

Singal processing

After reading for a while now the horizontal platform Ive been manufacturing and also, after adding the pitot tube, I noticed the need of implementing signal processing of the readed values.

The system architecture I decided to follow inside this project is:

Teensy 3.2 to read, process and tare all the signals from the loadcells and pitot tube. I want to make the teensy to do the hard work as is running in C++. This will provide a faster data (80Hz) and will generate a serial with all the filtered and translated information ready to read and plot.

Raspberry Pi 4.0

to receive the information through serial USB. As it runs in python I will use the raspi as the main computer of the wind tunnel. That means it will support the GUI, reading and plotting data from the teensy, and moving the Angle of Attack servo.

I started to research which filters could I use and Ive been implementing a Kalman filter, using a library, and it worked. After commenting it with Dave and Chris, they told me that my use was the perfect situation to implement a Low Pass Filter. This one yes, I implemented it by code. Here you can check how it smooth the signal of the pitot tube.

PCB design and manufacturing

It was a meeeesss the way I had hooked up all the sensors using multiple breadboards, impossible to trace back any issue and I was procrastinating the decision of making a PCB to clean that.

Thinking towards the final presentation

I took this class excited to learn more about system architecture, learn more to play around boards sensors, integrations and doing custom sensored platforms, and so far, Im sure ive learnt a bunch.

For the final presentation I want to show a couple of wings and , ideally, a morphing one showing different L/D values (that will require to read and store data on a sd card, its gonna be a bbbunch of work.)

Maybe, the question I am asking is, Can I make a wind tunnel able to be replicated in a Fab Lab? . The intention has been to do everything by myself, not buying elements (only if neccesary) and so far I think I am putting in the table a strong and precise instrument that I will have fun with.