Embedded Files
Single-Axis Cold Gas Reaction Control System
Project Overview
NOTE: This is not a finished page, and will be updated in the future. If you want any extra information or to hear more, feel free to contact me and ask away! I have additional media and a whole lot more explanation that will be coming to this page soon, but I'm more focused on the project itself at this point. Thanks for understanding!
***This is out of date, scroll down to see more current progress***
A personal project that I'm currently working on, with the goal of designing a single-axis cold gas reaction control system, with the ultimate goal of creating a system that can be used to stabilize a weather balloon payload. The project is ongoing, but has so far involved designing and modeling a frame in CAD, 3D printing components, calculating dimensions of and designing a rocket nozzle in CAD, performing CFD simulations of the rocket nozzle, testing the rocket nozzle, assembling the system, integrating an IMU, writing code for and tuning a PID controller, maintaining a budget sheet, tracking progress and recording issues, and more. A lot of the project has been helped along by referencing similar projects, especially from papers published by an ongoing senior design project at the University of Huntsville Alabama. The system currently runs on 100psi compressed air stepped down from a 3000psi air tank, but I'm looking to increase the working pressure and reduce weight by changing to a carbon over-wrapped pressure vessel from the current aluminum pressure vessel.
The software is written for Arduino, and uses a 9DOF BNO055 IMU for positional data. The hardware is mostly 3D printed, but uses steel threaded rods for the main supports, as well as aluminum extrusions for the thruster arms. The air tank is a 40 cubic inch, 3000psi paintball tank which is internally regulated to ~800psi, further stepped down to 100psi by a secondary regulator.
Rocket Nozzle Design
Mostly relying on NASA's Glenn Research Center's rocket thrust equations page (I started this project prior to starting thermodynamics), I was able to come up with an initial design for a rocket nozzle that would maximize thrust from 100PSI air. After calculating the throat diameter and expansion ratio, I modeled the nozzle in Ansys Fluent to validate my calculations with CFD. My calculated theoretical maximum thrust was 0.63N. Because I was planning to 3D print the nozzles due to cost concerns, I estimated a slight decrease in actual thrust, and expected ~0.4N for my actual thrust. I then designed a test stand for the rocket nozzle, and 3D printed a number of nozzle prototypes, working to integrate an NPT thread for connecting airlines. After a few iterations, I was able to set up and test the nozzle for the first time. The first test did not give me thrust values near what I was expecting, and I realized that the 3D printer's layering was causing issues with turbulent flow. I then cleaned and cut down the nozzle by hand to reduce imperfections, and I was able to record a maximum thrust of 0.5N at 100psi. This gave me an efficiency of ~78%, which I was content with. Though I could likely eek out a slightly better efficiency, I believe this is a result of the small nozzle and low resolution of FDM 3D printing.
PID Controller and Software
My initial design called for the use of an OLED screen to display live data on the system for easy debugging (and because it would look really cool), so I integrated an I2C screen which would take the data from a BNO055 IMU and then display rotational velocity, a gravity vector, and some other relevant information. This worked out well on the test stand, but when I tried to add the solenoids to activate the RCS thrusters, I ran into issues of the system overshooting and acting as if it had a stuck solenoid. After confirming that the solenoids were working appropriately, I tested a few other things, and realized that the screen was designed to load the display into memory in every loop, then transfer the data to the screen all at once. This unfortunately was leading to a ~40ms delay after each reading from the IMU, and was nearly halving the sensor's refresh rate. To combat this, I attempted to integrate a second Arduino, which would offload the screen's load from the sensor/solenoid Arduino. I connected the two Arduinos through UART, but even with this attempted solution, I was unable to keep the sensor refreshing any faster than ~50ms. So, I decided that the design would have to forgo the screen, in order to prioritize the functionality of the system.
Frame and Integration
Containing over 30 individual components and parts, Almost all of this project has been designed and verified in Fusion360, allowing for easy fitment and design validation. This currently does not include any of the air lines, as the first version of the design only used one rocket nozzle per arm. Now that I've been able to test functionality of the software and validate the design so far, a transition to two nozzles per side will reduce the currently present translation caused by asymmetric thrust.
The frame is based on threaded rods, and is designed to keep the center of mass as close to the center of the vertical axis as possible, and to extend the nozzles outwards to allow for more efficient use of the limited air pressure. The dimensions are also laid out to allow for the theoretical use of stabilizing a weather balloon payload, but I currently do not have plans to pursue a weather balloon launch on my own (I can't afford that just yet!).
Version One Done!
"Pizza Box Prototype"
No, seriously, it's a pizza box! Before putting everything together, my first test after integrating all the components together was based on a pizza box. This is before I've done any tuning to the PID controller, so the oscillation is to be expected.
Most Recent Test
This is the first version of the semi-completed design. Though it only had 1 thruster per side (leading to some translation), and the PID controller hadn't been tuned for this frame at the time, it was pretty great to have this working as well as it did. From here, the plan was to integrate a second set of thrusters, and to tune the PID controller. Then, work on optimizing the thrust and cut down the weight.
PID Tuning
This is effectively the finished first version/prototype. The PID controller has been tuned, the system is assembled, and everything is working as anticipated. This doesn't mean this project is over, however. In fact, this is just the proof-of-concept. Now the real fun begins, where this design is refined to meet weight and performance goals that V1 was not intended to reach.
Upgrades and Changes
Version 2 Overview
One of the biggest concerns in moving towards the second version is designing the payload to be flyable. Easily, flyable, that is. Under FAA part 101, a high altitude balloon under 4 lb doesn't need to be registered to fly, and it's generally much easier to get one in the air without any paperwork. So, reducing the weight is a top priority. Here's a chart of the proof-of-concept components and their weights. As expected, the big heavy aluminum air tank is big and heavy. Hence, carbon fiber tank for V2.
Going down the list, most of the components are swapped for higher-performance lighter versions. The threaded rods are replaced with carbon, the regulator is a smaller, lighter version, and the frame size is reduced from ~4in square to 2in square. Another major change is the size of the moment arms. Compared to the first version, the arms are tiny!
But that's not without reason. After looking into some videos of other balloon flights and consulting with my boss (who has lots of balloon flight experience), I was able to come up with an estimated angular acceleration that the payload would experience. Then, by modeling the whole system on a 2D plane, I was able to calculate the moment of inertia, and was able to come up with this graph. The Y axis is the amount of force require to counteract the expected acceleration, and the X axis is the arm length. As I assumed in starting these calculations, there's an ideal arm length where shorter arms cause a smaller moment on the center of mass (thus needing higher force and using more fuel), and longer arms increase the moment of inertia so far that any losses are nullified. With this, an ideal arm length of ~13cm was calculated.
Flight Computer
This is the new flight computer I've been spending a significant portion of my recent free time working on. Entirely rebuilt from the ground up, and with a host of improvements, it's a big step up from the Arduino that preceded it. It's now built around an Adafruit Feather V2 ESP32, which makes it possible to run a variant of freeRTOS. That means true parallel processing, baby! With two cores to use, and the ability to create multiple tasks to run at the same time on the same core, this flight computer can do things that the previous version couldn't dream of doing. It's able to process the BNO-055 IMU data at the same 100Hz refresh rate as before, but it can use the second core to poll other sensors at the same time. This means that it's happy to have 8 (or more!) separate tasks going on at the same time. One for each sensor, a few for controlling the thrusters and PID loop, and another that's able to log all of the sensor data in real time to the newly integrated SD card. All of that, packed into a nice box with a redundant battery to keep logging data in the event of a main battery failure. Amongst those sensors, there's also a GPS unit (which is also recorded). Best of all, it's a mere 92 grams! With the IMU separated to be more easily separated from potential interference from the solenoids, there's still room for a few more sensors to be added in the future. Another great new feature is the OLED screen. With it, data flow can be validated, and sensor status can be checked (forget to plug in the SD card? No problem, have a big "X" to show you the error of your ways).
Custom Valve Manifold
An attempt at one, at least.
After cutting the weight of the heaviest components, the next up was the solenoids. The amazon solenoids are big and chunky and don't really need to be. So I looked around at a few different ways of making valves, and decided to try to design a custom manifold that would work with some MAC bullet valves I purchased on eBay.
Cut on my CNC router over a few hours, it's made of PMMA/acrylic sheets that are bonded chemically. Of course, as all things go, the first version has some issues. From critical failure to minor inconveniences, they are as follows: 1) I permanently affixed the final layer on backwards. Oops. 2) The eBay valves were listed incorrectly and are only rated of 30psi. 3) Tapping is difficult (or I'm just bad at it), especially without a vice or drill press.
So needless to say this is still a work in progress. Perhaps there's a way that I'm able to get this to work and I can reap the benefits (of which there are many), but I think this will limited to a cool learning experience, rather than something that will end up on the final version. I'll be able to stay under 4 lb with the older solenoids, and I'm searching for lighter ones as well.
Experience and Responsibilities
Derived design specifications by analyzing research papers on balloon flights to determine expected payload behavior and describing a standard flight profile
Used thermodynamic principles and rocket nozzle equations to design a rocket nozzle based on pressure and expected force requirements, size limitations, and manufacturing processes
Modeled and conducted CFD simulations using Ansys Fluent on rocket nozzle designs to verify prior calculations and validate expected theoretical thrusts
Built and utilized a test stand with 0.1-gram accuracy to measure thrust and validate the rocket nozzle design, determining an isentropic efficiency of 78%
Optimized parts and components selection by leveraging additive manufacturing, staying under 86% of the budget for a proof-of-concept design
Developed custom PWM code for solenoid activation, protecting components and maintaining system accuracy within 10 degrees of the target
Integrated a commercial camera gimbal to improve stabilization to remain within ±0.005° of the target
Modeled and tested over 50 components in Fusion360, reducing assembly weight from ~6lb to less than 4lb, ensuring compliance with FAA regulations for high-altitude balloons
Developed, tested, debugged, and optimized over 1300 lines of code using FreeRTOS to manage eight concurrent tasks across two cores, enabling a 100Hz PID control loop update rate while maintaining separate data logging to an SD card
Incorporated a GPS, BME688 environmental sensor, MCP9808 temperature sensor, LTR390 light sensor, and a 128x64px OLED screen into a 2-inch square flight computer enclosure
Designed and soldered a power delivery board with an integrated backup battery, ensuring data collection for up to 150% of the expected flight time in case of primary system failure.
Google Sites
Report abuse