Chronological Entries - 2022

Year Summary

I did a lot of bouncing back and fourth this year. It was a pretty stressful year for me though with 11th grade and college so I don't feel too bad about it, but It did show me that I tend to wander sometimes. Next year, I'll work on a better system to keep myself aligned with deadlines and goals. I did, however, get some pretty important data about Rev02. No glaring issues in the design expect for sourcing and assembly. Software is also progressing well, MATLAB was a good solution to arm control. I do also want to look into a camera tracking method that keeps the cameras mounted to the arm itself to keep setup quick, also I think it would be a cool feat. My skills in software and design have also gotten significantly better.

1/09/22 - 3D Printing the updated parts!

I will spend the next couple days printing the new and improved components using my new 3D printer!

When I redesigned these parts I really made sure to design with 3D printing in mind, so no crazy overhangs, so small structural parts, and overall not to complicated.

*Here is my build log of this DIY 3D printer -

sites.google.com/wiufsd.org/3dprinter-buildlog/home

1/14/22 - Assembling Main Base Stand

I just got a chance to cute the 1010 and mount the angle brackets. I plan on for a prototype, cutting the aluminum parts out of laser cut wood. I will see how strong it is and I might just leave it as wood.

I also plan on putting some clear acyclic panels on the side since this is where I will be storing all of the electronics.

1/20/22 - Assembling the base of the of the arm

I have laser cut the plates and finished 3D printing the main stand for the motors and arm. I will need to start cutting shaft to test belt lengths since I am quite concerned about those.

Other than that everything seems to be fitting well including the motors on the sides. The bevel gears mesh nicely.

1/26/22 - Assembling J1 Electronics

After just wiring up the base actuator, I noticed that I going to need a couple more "layers" of electronics. I'm thinking of some sort of shelfs to put the components on.

2/02/22 - Belt length adjustments for J2

After fitting the belt on J2, it seems to short. I had to re-measure the proper length and re print the main body since the bearing would not fit properly. I did calculate what the center distances should have been be, but it seemed to be a tad to short.

1/29/22 - Prototyping Connector Board

Here I was prototyping how I wanted all the connectors to sit. I decided on have a board with each connector organized for each joint. Having printed this test print, I adjusted dimensions for a better fit. I will be printing the full size one next.

2/09/22 - Final Main Connector Board

Here is the full sized connector board. I will be mounting this to the inside of stand, hopefully this keeps the electronics much more manageable and helps me troubleshoot in the future.

Each Joint has a main motor connector and an ethernet cable for the encoder. Excluding the last two joints since I will be using lightweight, high power servos for them which doesn't require an encoder.

2/12/22 - Current Control System

I'm running the oDrives with a Pi currently but I'm skeptical if this will have enough power for the vision system. Ideally, I would like to have this arm run completely without a computer. This means running both the arm software and vision processing both on an internal computer.

After dong more testing and I realize if the PI doesn't have enough GPU power for the computer vision stuff, then I will swap to some other, more powerful micro computer.

A80AD5B1-96F0-40B2-AD9E-8103F1485617.MP4

2/13/22 - Troubleshooting J1 with the ODrive

After about 2 days of trying to figure out why the ODrive would spit out so many errors when I tried to run the motor, I finally got it to work! Here, I just ran it to some endpoints, I plan on re-assembling this back into the arm since I have worked out the problems.

D97F01A0-69AE-4875-82B1-91E6302CE619.MP4

2/19/22 - Troubleshooting J2 Motor

After wiring up the J2 encoder and motor identically like J1, I got these aggressive vibrations when the motor was position holding. After reading some forums it seemed the issue was encoder noise and feedback. So after properly re-calibrating the encoder, these vibrations went away.

IMG_0211.MOV

2/21/22 - Re-assembling J1 to see how it looks!

Super exciting, finally get to see the arm move! Running the motor at 75% max velocity, It seems super quick. There seems to be a torque issue. The motor has plenty of torque to resist a push on the arm, but seems to throw an error every so-often. I think I need to adjust the motor amps.

2/28/22 - Continuing Assembly!

Here is the MidBody Section. I am working on machining the slot for the pulley still, but for now assembling the as much as I can!

2/28/22 - Finishing body tubes

I have cut the tubes to length, added servos to bracket at the end, and routed extension wires! Also finished soldering all connectors for remaining actuators so I can start putting it all together!

3/01/22 - Electrical board

Finished designing and cutting a test electrical board! After a failed attempt or two, I have a board that fits all the components where I want them while also having spots to zip-tie cables down to keep everything neat!

3/01/22 - Finished up wiring cables!

All connectors and power cables are wired from the boards to their corresponding connector on connector board. I tried to keep everything as neat as possible so it's not one big rats nest.

3/08/22 - Reassembling Arm

With everything wired, I can start to fully assembly the remaining parts of the arm!

Assembly of this arm has been a breeze! I credit that to the thought I put into how assembly will work when designing in CAD!

3/08/22 - Continuing Assembly!

Almost all wires are routed along with all the tubes fastened down. I still need to add the side actuators along with their bevel gears.

3/12/22 - Major Assembly Done!

Complete arm assembly is almost done, still need to:

Adjust Bevel Gears for better meshing
Shorten cables
Attach Paddle
Finish Slotted Shaft

Other than these things, this arm is looking complete!

3/14/22 - Fixing Immediate Issues

I decided to redesign the mid-arm joint slightly to give us a little more range of motion.

I also threw on the paddle to get a look at the arm as a whole. I'm noticing how light the entire assembly is, I may have to add weights or sandbags to the bottom to keep the frame sturdy.

I will also need to finish machining the shafts for the pulleys and bevel gears.

3/16/22 - Machining Shafts

I need to machine one of the internal shafts to put a slot in them so that they can mesh to the 3D printed pulley. Expectantly, the set screws on the main Joint 3 Pulley cant dig into the tool-steel axle.

I drew up the parts and learned a nice way to cut the slot, using a woodruff cutter. I will also be milling a flat on the shaft for the set screws of the bevel gears to grab onto.

3/18/22 - Jetson Nano!

Since I'm only able to work on the shaft in school, I thought I'd test the vision system running on the raspberry pi. After experimenting with the program, the main problem I see is the pi has trouble processing both videos streams and running tracking on top of that. I tried down-res-ing the video streams and while this did improve performance, the tracking system had much more trouble finding the ball.

My understanding of memory and processing power is very limited, but I can say that my code isn't the most optimized code in the world, and neither is python the most optimal choice for my application.

I decided to try running my program on Nvidia Jetson Nano. I could take advantage of the extra GPU power and hopefully run my vision system at significantly high frame rates to keep up. I found some cheap ones on eBay and thought I'd give it a try.

3/19/22 - First Boot Up

When I first got the nano, the power supply I found turned the display black. Apparently you cant power the board and monitor from the micro usb port. After sourcing a 5v 4amp barrel jack power source, it booted up like charm. Currently installing all the necessary software to get this working!

3/20/22 - Re-working electrical board

Decided to take new measurements to add the new necessary components to the wood board. I'm going to need to bring the voltage down from power supply to 5v for the jetson and the servos.

I will be using DC-DC Buck convertors and wiring up the proper power source for the nano to the supply. I will be cutting the new board soon so I can test all the boards and motors to see how they run.

3/21/22 - Arm Rendering!

While I'm waiting the cut the board, I quickly started to render a scale model of the arm in the program. I will be splitting up the code into two separate project for simplicity and going back to integrate them. One for vision and tracking and the other for robot arm control. I don't quite know exactly how I will be controlling the rotations of each joint with respect to one-another, but I have a good feeling my linear algebra knowledge will help me out!

3/21/22 - Rotation Matrices!

After looking over my notes on linear-algebra, I knew how I was going to control the arm joints. I will be figuring out each little detail one step at a time so I will be starting only with one joint. I know how to rotate a body some degrees around some point, but I don't know how I'm going to get the other joints to respond accordingly, like some sort of hierarchy.

\/ Link Below \/

Linear algebra for game developers ~ part 4 - Wolfire Games Blog

3/25/22 - Finished Bottom and Arm 1 Axis!

After teaching myself how to use opengl, I started to get the hang of how transformations work and coordinates. There is no way to reference your current position, you have to keep track, no absolute cords, all transformations are relative to the last and rotations work around the origin. I can see this becoming a big dysfunctional mess of code, so I am REALLY disciplining myself into keeping this neat.

2022-03-27 20-53.pdf

TrigArmControlDemo.mp4

3/27/22 - Math for following Arm2!

Once I started moving the second arm I knew that I would have to translate arm2 along with rotating it. To get the new positions I sketched out a model and started reasoning with how I will get the new positions based on the angle and lengths of the arms below it.

2022-03-28 10-34.pdf

3/28/22 - Math for following Arm3!

After the successful reasoning behind getting arm2 to move, I started similarly with arm3. Starting with a sketch and trying to get equations for the x y and z translations with respect to angles and lengths of arms that I will have to code in.

3/28/22 - Semi-Working Arm Render

The Full Arm Render works pretty well, just some bugs to work out.

Next up is Inverse Kinematics! Since the ball, table, and arm are all on the same coordinate system, I will start with trying to get the paddle to follow the ball. But even before that, I need to understand IK a little better. I will try to find a super simple algorithm that I can mock up and try out to get the arm moving based on (x, y, z) cords rather than joint angles!

2022-03-28 22-31.pdf

3/29/22 - FABRIK!

Forward And Backward Reaching Inverse Kinematics (FABRIK), is a method of Inverse Kinematics that avoids the use of rotational angles or matrices, and instead finds each joint position via locating a point on a line. Basically put, its super easy to understand and implement. I understand this strongly in 2D space so I would like to find a way to keep it in 2D. My arm is only 5DOF and pretty simple, so I think I have an idea to keep the IK in 2D. I will be doing more research and trying to implement this in my code!

2022-03-29 08-01.pdf

2022-03-29 20-11.pdf

3/30/22 - More Conceptual IK in 2D

After some more time familiarizing myself with FABRIK, I starting trying to write down my idea for keeping the IK for this arm in 2D space. I am not claiming this was my idea, but I did come up with this on my own.

Since I can only rotate on the XY place from the base actuator, put simply, I can calculate the radius that any tracked point would follow around the base point (Bottom Picture). Then map this to be in line with the arm, thus removing an axis to keep in 2D. Then do IK on the arm on the ZX plane to line up the arm with the point in this plane. Then rotate the base to become in line with the point. BadumBing BadumBoom, the points now line up, theoretically.

I'm going to try this out as I implement FABRIK in my code, but I feel quite solid about it.

2022-03-29 20-12.pdf

Working3DIK.mp4

3/31/22 - 3D Inverse Kinematics

Having all this thought behind something, then implementing it in code and it just working is a sight to behold. There are a few things that I need to tune but other than that, I am extremely happy with how this looks.

Next up is working on arm3 and the paddle. I think keeping arm3 level with the table for now is a good way to keep things simple because I will have to return the ball so I can free rotate the paddle when I need to. Maybe my logic will change but for now it will remain level, just need to find a way to tilt the paddle level along with arm3.

4/2/22 - Finishing Touches on Arm

As I prepare to start running this thing for real, I want to make sure the arm is fully built before I break anything. I have spent today using some spare cable sleeves from when I built my computer and neatened up the cables. I also took the time to shorten the cables to lengths that allowed for a decent range of motion. I still have the second shaft to turn but in the mean time, I will work on connecting the Jetson Nano up to the ODrives and servo motors.

4/4/22 - Finishing up IK

I'm almost done working on the arm IK placement with incoming ping-pong balls. I've been working on setting up a rebound target for the arm to aim the incoming ball back too. For now, this target point is stationary, but maybe I could implement a facial / hand tracking to aim the ball away from players current position. However, as this is all untested, I will keep it as simple as possible. Still have some work to do to get this thing running.

4/7/22 - Return Algorithm

For now, I've added the class structure to allow for the target point. Now, I can set a new target point and the final arm will rotate the paddle to be perpendicular to the ray from the end-effector to the target point. I have yet to add the paddle-swing-thing, but this will require some physics.

I'm going to need some more info about the ball's motion along with some equations to calculate how fast I should swing the paddle, which I don't know how to do yet, however I'm getting ahead of myself! I need to tweak the TODO's, such as constrains on the rotation distance for final arm, and generally get the whole system working solid as its not working to well right now.

Here's a summary of the change I will making:

I have spend at least 2 weeks getting a model of the arm and rendering that arm. The end result was a janky, sometimes would break arm, that I had complete control over. I'm proud that I made it completely on my own, however, this jankiness will get out of hand really fast when I add more stuff.
This said, I will be moving to a program that will not only take care of the all of the joint behaviors and motion profile for the arm, but I can use a model of my robot to have an even better visual of the assembly as a whole.
It will also better optimize the IK computations speed wise and accuracy wise.
This is MATLAB's Libraries and Simulink!

5/13/22 - It's time for a Pivot

AP's are finally done! Had to take a break and prioritize school, but its time for more robots stuff!

The break was nice since it allowed me to step back and realize some simple solution to some big problems regrading my robot control software. I've been talking to a bunch of people about returning the ball and the model of the arm as a whole, and I think its best to pivot. There's an easier way to do this.

While yes, it's great that I've learned so much for making my own rendering and model of the arm, its time for an upgrade.

5/13/22 - Learning MATLAB Simulink

This idea to swap to MATLAB came from a mentor of mine. He sent me the video to the right, briefly going over what Simulink does.

Over the next couple days I will be learning the basics of Simulink, what its for, and how I can use it for my situation. Ideally, I can use Simulink to visually model my arm and take care of inverse kinematics and dynamics. I then would have to integrate this into my current python vision scripts.

5/14/22 - Preparing the CAD

Simulink uses a modified CAD assembly of the arm to build its "Simscape Multibody". Initially, I started copying the Inventor Files into a separate folder to then export the xml file required for Simulink. However, Inventor 2022 is not supported.

Since Inventor is painfully not backwards compatible, I had to export different subassemblies into step files and rebuilt the assembly in Solidworks since I had it installed already.

After fully constraining the model as it would be assembled, each joint then needs its own labeled Origin and Coordinate System, shown to the left.

5/15/22 - Installing Simscape Multibody Add-Ins

Now that the model is ready for export, I then had to install MATLAB to add the Simscape add-in to Solidworks. There is documentation to detail this process more in-depth, so I will briefly go over the process I followed.

First, I had to obtain a MATLAB licenses. Luckily being apart of FRC, the offer discounted licenses, I just had to purchase the Simulink add-on. Then, after running the MATLAB server, the Solidworks plugin can then be installed. Finally, I can then export the Multibody as an xml file.

Tomorrow, I will be uploading these files and playing around the model and Simulink!

SimulinkStart.mp4

5/16/22 - Playing with Simulink and Simscape!

After first trying to debug why the files would not upload, I found out that a model cannot contain more than 1000 models. After going to Solidworks, simplifying the model, and re-exporting, I could then upload the model to Simulink.

5/17/22 - Some Changes

So today I spoke to a MATLAB developer and he brought me through some design recommendations along with an alternate animation and robot control methods. Long story short, it seems that I misunderstood some things. Simscape is meant for testing simulations. So you run the sim, collect data, run the sim again, collect data, ect. This is not what I need.

What I need, is arm control that works in real-time. To do this, I will continue to use MATLAB Simulink since it simplifies a lot of the arm mechanics, however I will have to switch out Simscape for the "Robotics Toolkit". This will allow me to model the environment and run the physical arm in real-time along side my vision software. Thankfully it very similar process , I just have to export a different file type, a ".urdf" file.

I don't yet know how I will end up integrating the separate systems into one as I dive deeper into MATLAB, but as of now, I fell that I will end up creating one conglomerate MATLAB project with all systems working together. I'm loving how MATLAB works, and see a lot of its features useful just because of how much math I do in my program. However, I am getting ahead of myself!

5/19/22 - Issues with URDF

I have been in contact with a few Solidworks devs since I am having a very rough time trying to get my Solidworks edition to work with the URDF Plugin. We seems to have found the problem being that I only have a Student license to Solidworks and not a full one. This is not allowing me to install the plugin-in needed to export the URDF file.

This has been an extremely frustrating process since I have wanted the file exported quickly so I can work on MATLAB but I think I am unable. I don't have a solution as of right now, so I have to find some work around because I cannot purchase a full license. Updates to come...

5/20/22 - Downtime Experimenting

Since I'm stuck on getting the actual model transferred, I think it would be best to better learn MATLAB and Simulink and how I can use them for what I need.

I will be following tutorials as well as recreating a majority of the vision system I had in python to run in MATLAB. I feel that If can have everything run in MATLAB, it would be ideal. However, no more python. I'll start with getting the tracking system to work on ping-pong balls, then move to reconstruction geometry if I still haven't found a solution by then.

5/27/22 - Solidworks Licenses!

One of the benefits of documenting your struggles is that others can come to help when you need it! Through interning at SDP/SI, they were thoughtful enough to gift me a full Solidworks license to solve this problem. As soon as I receive the details to the license I will install the Add-Ins necessary and continue with the exporting process!

In the mean-time, I will continue to work on the vision system to better learn how MATLAB and Simulink work.

5/29/22 - Code Updates

Today I have been practicing writing super clean and readable code to save me some time later. I also started a new GitHub Repo which is where I will be storing all of the MATLAB code for the project. I have a feeling I will be moving all to MATLAB as I am growing much more familiar to the style plus all of the awesome libraries I will be making use of in the future.

I will be getting the licenses soon, so as soon as I get word, I will continue uploading my files to work on the robot control aspect of the project!

5/11/22 - Difficulties

School has ended! Sadly though I feel like I'm going through a lack of motivation. MATLAB has presented some problems and overall difficulties since I'm not experienced enough to do some of the tasks I want to do. Learning each individual step is taking forever. Part of me wants to just give up. But I WILL not! I have been taking a little break to recoup myself from school since I think its just a change in routine that's making me feel this way. I know this is kind-of a personal entry, but I feel it's important to detail this so I have record of this so I can hopefully improve on this as I grow!

6/16/22 - Updates

Sorry for being a bit radio silent. I should be full focused on this in a little bit! Back in a routine and have been having good progress on my other projects, enough to warrant a swap of focus back to the robot. I have gotten a couple things done on the MATLAB code, but just need a little but more time before I can switch gears back to the robot!

7/01/22 - Picking Up Where I Left Off

After being busy with other projects and work, I decided to refocus back onto this project! Today I was able machine the shafts with flats, holes, ect. from the drawings I pulled from the CAD model in order to mesh with the bevel gears and pulleys! Still waiting on some hardware to come in the mail, but mechanical assembly of the arm version 0 is reaching its conclusion. I took a break from MATLAB to finish wiring the servos to the Jetson, as well as implementing the new electrical board that's holds all of the new components.

With the version 0 coming to an end I don't see any immediate mechanical problems with the arm. All joints have sufficient travel room, and all belts have the perfect tension. Obviously this doesn't mean that it wont fail, just initially good signs!

7/05/22 - Software Coming Together

Software is really hard to document well on this site alone. I have been hard a work implementing all of the algorithms I have in python to MATLAB. There are subtle differences that add up and make it an interesting change. After figuring out how to use the GUI editor in MATLAB, it was pretty straight forward. As of now, I'm finishing up the ball tracking aspect of the program. The vision seems to be preforming with the same consistency and maybe even a bit faster. I don't expect the 3D reconstruction part to take too long as its just a few algorithms. I'm exciting to finish this, since I can then work on the reason I swapped to MATLAB, robotic arm control! More to come!

7/09/22 - Finished MATLAB Tracking

After learning the GUI mechanics and a few other quirks about MATLAB like functions and scripts, putting together the detection algorithm was pretty simple. I added a few other features like an initial HSV tuning stage as well as a better way of making sure the working settings are saved. Here's a clip of the tracking system working. I detailed what steps are happening to find the ball in a previous log (10/24/21).

7/10/22 - Rewriting the Analysis Section (1/3)

Since I was able to finish the tuning and custom detection algorithm, its on to the image geometry section!

I'll be starting with the a frame that has the user click on the corners of the table in the frames of the camera. This is mainly a GUI thing which I learned from the tuning GUI previously. Only difference is clicking on the images. Using the plot item in MATLAB, each button click on the image has a registered coordinate. After some time I was able to get the section of the analysis section complete.

(image of analysis section code)

7/13/22 - Math, Math, and More Math (2/3)

Next is the math programing. MATLAB makes this part super easy and a lot more intuitive in my opinion. Honestly, I just copy and pasted the previous code file over and MATLAB-ifed it. Main thing to work out now is the 3D rendering of the whole program.

If you're curious what math I'll be specially rewriting, take a look at entry 8/29/21 and on. Of course my code is saved on GitHub if you would like to look there as well.

(image renderings)

7/15/22 - Learning 3D Rendering in MATLAB (3/3)

Learning this was a little more tricky since I needed to integrate both the robot and table into the same figure. However, starting bit by bit I was able to create a 3D figure of the table and ball just like in the python script.

Before I move on too much, I wanted to test the fully completed system out on a real table. Theoretically I didn't change anything, so it should work, right?

(vision demo)

7/24/22 - Vision Testing Complete!

Looks pretty good all things considered. The ball is tracked throughout in somewhat controlled conditions. I see this getting a little out of hand maybe outside where shadows are changing.

I definitely would love to have some sort of off-table detection so it knows when to stop tracking and when someone has won a point or lost. Because of this, and since I don't have access to the mechanical robot right now, I'll work on something like this implementing this in my next entry.

Serve Detection Method

As we know that objects traveling through Earth's atmosphere are effected by Earth's acceleration due to gravity, we can use this to our advantage in this situation.

1) We are given the ball's 3D position over time from the cameras. From this, we can get the ball's position, velocity vectors, and acceleration vectors since math is our friend.

2) Just before a ball is served, it is held by someone. Meaning, the ball's acceleration does not match Earth's acceleration due to gravity. Thus, we can simply say when the ball's acceleration is close to 9.81m/s/s down (we can tune with drag if need be) we can assume the ball is in free fall. Any other value, and the ball is being held (Ignoring bounces for now).

3) Armed with this knowledge, as the ball is being held (not -9.81m/s/s accel), we can assume that the ball is about to be served. As soon as the acceleration of the ball matches, the game has begun and we should attempt to keep score now.

7/25/22 - Serve and Point Detection!

What does it mean to serve the ball? Better question, how can we detect when the ball has been served. Thinking over this, I came up a method to detect this (left).

I'll start by placing an indicator light that shows if the computer thinks the ball is in freefall or not.

Next, using the Ping-Pong table in the 3D render as a reference, I'll detect if the ball had fallen of table during a live match and end the game.

The main issue I think I may have is placing the cameras in a way so it can see if the ball has fallen since it would go out of frame pretty quickly. I probably can use a predication algorithm to see where the ball would end up, but if its solved by moving the cameras back a little, I'll go with that.

8/01/22 - MATLAB Robotics Toolkit

Today I started learning about the robotics toolkit in MATLAB. I'm excited since this is why I wanted to use MATLAB in the first place! I'll start with their general guide. It brings one through the documentation, examples, and some videos. Hopefully, this will give me a good sense of what I can do using the toolbox.

I also organized the workspace to keep each aspect of the project in their own directories. Reading through it seems this toolkit is a lot to swallow as one pill. So it may take me a bit to try to understand everything.

I'll be documenting all of the learning projects they run through to learn the toolkit before I tackle my own problem. Not all of them are applicable to what I'll be doing, so I will just glace over them and move on. Seems super exciting!

8/12/22 - Project on Pause

I think I'm going to put the project on pause for a short duration. Don't worry, I have intentions of continuing! College applications, test prep, school starting, ect have been taking a lot of my time. I thought it would be best to focus on those things and continue when I have the necessary time. More updates to follow...

12/19/22 - Back At It!

After a nice long break, I'm ready to attack the project again! I want to get back to the mechanical side since I have a few things to update and too much software gets me impatient ;).

Here's the plan of attack:

1) I think I broke one of the actuator 3D prints since I've been moving it all over the place, so I'll start by reprinting and installing a new one.

2) The bevel gears on the first few joints need flats machined on the shafts, so I'll get those done in the shop before winter break.

3) The back pulleys need new keys since I lost the ones I made. So I'll 3D print them out of solid ABS and it should be well off (at least until I find time to machine new ones).

4) I need to fix electronics plate with new t-nuts, along with the rest of the cross braces.

5) Re-tighten and add missing bolts across the whole arm.

6) I'm going to leave out the c-clips I left for shaft retention since it seems to be doing fine. If this comes back to haunt me, it wasn't my fault...

This should be enough to hold me over for a week-ish, so next week we will work backwards from where we left off on moving the actuators and fixing any electrical problems. Updates to come...

12/20/22 - Encountering Shaft Issues

So It seems that after leaving the the arm for a while, the bearings have seized to the shaft. I think it's because of temperature changes as the room can fluctuate during winter and summer, I could be totally wrong though.

In any event, I had to take the shaft out to put the new bevel gears on and could not. I tried everything, from cooling and heating the bearings and shaft to get enough wiggle, then the hammer, then pliers, but nothing worked. I could not get the shaft out.

I ended up prying the parts apart to isolate the shaft. I still have to find a way to tap the shaft out of the bearings but It's much easier to reach isolated.

It kinda felt nice tearing these old supports down as they were out of PLA and poorly printed. So, I took the opportunity to redesign the supports with the new knowledge of the issues the previous ones had and also use a stronger material, ABS.

I also think I'm going to swap materials for the shaft after having such trouble machining the tool-steel ones (in hindsight, completely overkill) and switching to an aluminum. I did end up gnarring up the ends trying to pry the shaft out, oops.

12/21/22 - Problem Identified!

After some "careful dismantling" I was able to remove the problem shaft and really take a close look at it. It seems the bearings have rusted to the shaft. They spun freely but would not budge at all for anything except for one which I was able to shift a little.

Here's that bearing. I honestly have no clue why this rust formed or how to prevent it. I'm going to bring it into school tomorrow and see if I can get any info about it. I'm also, obviously, going to need to find a material that won't do this again, so I'll also ask that.

After all this I have a lot to redesign, but honestly there was a lot I didn't like about the previous design. Everything from too many bolts to no cable paths.

I think I'm going to take a break for before the new year and come back with a new perspective. Happy Holidays!

Page updated

Report abuse