I'm a mechanical engineer who loves fabrication and hands on learning.
Check out all of the cool stuff I've made so far!
As an undergrad and masters student at UC Berkeley, I found a passion for mechanical design. I love to take advantage of all the resources I have to create and bring to life my ideas. Rapid prototyping and additive manufacturing are what I enjoy most because I can turn my designs into reality so quickly.
May 2021 - Present
I am working to create and run a brand new makerspace at Chapman University.
May 2019 - May 2021
I had the opportunity to work in a cutting edge startup pioneering the use of robotics, artificial intelligence and machine vision to create an entirely new category in the gigantic and lucrative beauty industry. I worked on numerous projects including designing and prototyping a safe end effector and designing a holder for parts that are .15mm in diameter. I gained extensive experience with the Form 2 SLA printer including experimentation with various materials, learning how to optimize the repeatability and resolution of the printer, and performing maintanence.
May 2020 - August 2020
In my time working remotely for this civil engineering start up, I worked on CAD designs using Fusion 360 with a team to design modular tracks for transportation pods associated with Hyperloop technology. Additionally, these tracks were designed to be 3D printed for both a small and large scale model.
May 2018 - May 2019
The Hybrid Robotics Lab is a UC Berkeley Mechanical Engineering controls lab. I worked on designing a 3D printed cycloidal gearbox for our Thigh-Actuated Bipedal Robot for Extreme Agility. I was fortunate enough to work with the Markforged 3D printer which has carbon fiber infused materials. Both the concepts of 3D printed planetary and cycloidal gearboxes were weighed to determine which would be preferred for this application. I performed tests and analysis on different materials and manufacturing methods to conclude that a planetary gearbox would not meet our requirements and therefore designed and 3D printed a cycloidal gearbox for testing.
May 2018 - May 2019
Pioneers in Engineering is a student run non-profit at UC Berkeley that provides an extremely discounted robotics competition to 300+ students from underserved high schools in the bay area. As Director of Engineering, I managed and advised the hardware related engineering teams as well as took the lead on sourcing and acquiring the cheapest parts from manufacturers to cut costs. I also created the drawings for our custom parts to outsource in large quantities from manufacturers.
August 2017 - May 2018
I led the 10 person mechanical team in ensuring there would be enough mechanical parts for 23 teams of high school students to run the competition. Additionally, I explored the most cost efficent methods of acquiring and producing parts, both to make certain that we would be within budget but also to ensure that we would not spend too much of our own labor. We redesigned the modular kit metal used by the students for assembling their robots as well as carefully documented all of our processes to apply for grants.
August 2016 - May 2018
The UC Berkeley student chapter of Engineers Without Borders, Peru team, worked to implement rainwater catchment systems that provide clean water to replace the arsenic filled well water in rural Peru communities. In August 2017, I had the opportunity to travel to Carancas, Peru to work with communities members to construct a rainwater catchment system at the local primary school.
These projects were completed for my internships, research, classes, clubs or for fun.
The original aim of this project was to create a 3D printed planetary gearbox for a 3D printed legged robot. The size constraints on the gearbox led the teeth of the spur gears very small. After some static testing on the 3D printed gears, it was determined to be too weak for the gearbox's purpose. With more research done, it was concluded that a cycloidal gearbox would be stronger because of its larger and therefore stronger teeth. Because it is not as commonly used as spur gears, I had to read several papers to understand how to design it.
I love additive manufacturing and have always wanted my own 3D printer, however I never wanted to spend money on a nice one or even a cheap one. As a result, I learned to 3D print using the printers at my university or in my internships however COVID-19 significantly reduced my access to fabrication tools. As a result, I decided that I could make my own cost-effective 3D printer while learning a lot in the process. I then realized there may be other students in the same situation as I am and could perhaps share my design with them. I recieved a grant from the Jacobs Institute at UC Berkeley in Fall 2020 to design, fabricate, and assemble a low cost "Maker Machine" that can potentially be replicated by other students. My original pitch was to build a hybrid 3D printer and CNC mill however I only had time to build the 3D printer.
For my Human Centered Design class, my team and I participated in a semester long project to develop a product to address the challenges of cooking for wheelchair users. Please click the more info button to find out more!
As a continuation of the 3D printer Maker Machine that I recieved a grant for in Fall 2020, I built a separate, desktop sized CNC mill with a grant from the Jacobs Institute at UC Berkeley for Spring 2021. To make this design more accessible, an Instructables was created so that others could use the design I made and learn from my process.
As part of a group project to create a planting robot, I designed the system to water the plants. A gravity fed watering system was implemented with a 2L bottle used for storage. A bottle cap was 3D printed out of PLA, and in order to prevent leaking from the cap, a gasket was 3D printed out of the compressible material TPU. To route the water, a rigid polyethylene tubing was used at the entrance and exit of the solenoid valve and at the exit of the bottle cap. A flexible, long silicone tubing was used to easily route the tubing and allow the bottle to flip over for refills. In order to balance the pressure and ensure flow out of the solenoid valve, a check valve was installed on the bottom of the bottle for one directional flow of air into the bottle. Additionally, a laser cut structure was designed and manufactured to hold the bottle upside down. The solenoid valve was actuated using a transistor as as switch controlled by a microcontroller to apply a voltage across the solenoid valve.
As part of a semesster long project where we explored the entire design process, from How Might We statements to user interviews to low and high fidelity prototypes, The Amazing Cane was the outcome. Designed to ensure the safety of visually impaired people, The Amazing Cane is the result of many iterations exploring other ideas like robot guide dogs and using feedback from visually impaired individuals. This product is an attachment to the existing white cane that notifies the users of objects above head and chest level, a pain point among users. The attachment slides onto the handle of the cane and requires a screwdriver to clamp onto the cane, similar to a shaft collar. Inside the attachment is an Adafruit feather, which takes input from an ultrasonic distance sensor and relays feedback to the user through a vibration motor and speaker. Flashing neopixel lights are used alert others to the user. In order to reduce the noise in the sensor, a median filter was employed. Additionally, a boolean flag feature was used to ensure that the audio feedback would not be running continuously. The body was outsourced and SLS printed in Nylon and the attachment parts were printed using PLA.
For a fun personal project, I wanted to see if I could make a multistage planetary gearbox using only wood. There were a lot of manufacturing challenges involved making it difficult to produce greater than a two stage gearbox, though I tried for three. I used dowels for the shafts and had to test different sizes of holes to get the appropriate clearance and interference fits for the shafts. One challenge was constraining the gears properly so they had enough play to move smoothly but without too much that could cause them to fall out of place. Another issue was the manufacturing tolerances, the gears were so small that the kerf of the laser significantly impacted the geometry of the gear, causing a lot of backlash. Overall, I was pretty pleased with how it turned out.
As part of a personal project to understand how motor controllers work better, I researched and assembled an H-Bridge using four NPN BJT transistors. By connecting this to a microcontroller I can control both the direction and the speed of a motor with this circuit. In the future, I would like to use this circuit to control a small robot car.
As part of a final projct for a class to learn Printed Circuit Board design, I designed a PCB for a microcontrollerless remote control car. I designed the remote PCB with four momentary buttons and an on/off switch meant to be held in one hand. The circuit uses buttons and an encoder to transmit signals using an antenna to a board on the car which will recieve and decode the signals. Those signals will then be sent to the motor controller. The PCBs were assembled but the functionality had some issues which were not able to be resolved due to COVID-19.
For my control of unmanned aerial vehicles class I learned how to model and control an autonomous quadcopter. The code implemented predictor-corrector estimators to estimate the location and orientation of the vehicle using the onboard sensors. Using the state estimation and the complex quadcopter dynamics, a cascaded controller was implemented using single integrator controllers to control the attitude and horizontal and vertical positions of the quadcopter. The estimators and controllers were used to autonomously lift the quadcopter off the ground, fly 1 meter in both orthogonal horizontal directions and land relatively smoothly. Due to COVID, the class was unable to fly real drones as in past semesters so we used a simulator to model how the vehicle would behave in real life.
As an assignment to design an experiment, my group chose to characterize the elastic modulus of Russian birch plywood, typically used for applications like laser cutting. Our experimental setup consisted of a vice with two points of contact supporting our specimen. A load cell was attached to the moving plate of the vice and would measure the load applied at the center of the beam. A linear encoder was attached to the side of the vice, measuring its displacment and thus the maximum deflection of the specimen. Using the model for a simply supported beam and known properties about the specimens, the measured data gave us many quantities for the elastic modulus. These many quantities were used to find a general elastic modulus using a linear fit of the modeled equation. The linear fit gave us the uncertainty bounds for a 95% confidence interval. To take into account the uncertainties in the measuring equipment, we used Gauss' method for quantifying uncertainty at each data point and found the maximum value. Additionally, a full report in ASME format was created outlining the entire experiment and results.
As a project for the Planar Machinery class, my group chose to go with an external, mechanical deadbolt unlocker. It is much cheaper than on the market smart locks and does not require replacing the entire doorknob, making it useful for renters. I took the lead on the mechanical design, choosing a 3:1 gear ratio to supply more torque from the servo motor and to provide the correct degrees of rotation due to the servo motor's 270 degree range. I took care to choose cheaper parts and to ensure the stack up would be appropriate in the CAD before prototyping. The box and gears were laser cut from acrylic to display what was happening on the inside with bearing press fit into them to ensure the shaft would turn smoothly. As a part of the project, the CAD model was animated and a motion analysis was performed on it using Solidworks.
As an individual class project, I optimized a control law for Unmanned Aerial Vehicles (UAVs) to effeciently map (or collect pictures of) all targets while not crashing into each other or any obstacles. This control law was optimized using a genetic algorithm through machine learning to find appropriate weights that would direct the agents (UAVs) in the optimal direction at each time step. To do this, a simulation was constructed to model the behavior of the agents for specific weights. In this simulation, the agents were modeled as point masses experiencing only a propulsion force and drag. If agents got too close to each other or to obstacles they would crash. For a given set of weights, the simulation would output a cost function to score them on their performance. A lower cost indicates a better performance. For the genetic algorithm, 50 sets of weights were randomly generated, run through the simulation and ranked by their cost. The best sets of weights were bred to have offspring which would be paired with more randomly generated sets of weights in the next generation. This was repeated for 100 generations and the resulting simulation is shown in the animation.
As an individual class project, I modeled and optimized a free form robotic 3D printer. The printer was simulated to be a robot arm with three links and an extruder at the end. Each link had a fixed length and a constant angular velocity. At the extuder, drops were dispensed every time step with a given extrusion velocity. Once in the air, the drops were subjected to gravity, drag, and electrical forces from the charged print bed. Using Newton's second law, the forces were calculated and the position and velocity of each droplet was updated using Forward Euler numerical integration. Again, a genetic algorithm was used to optimize the constant angular velocities of each arm and the extrusion velocity to produce a desired pattern. The further this pattern was from desired, the cost function would output a higher score.
To determine if a 3D printed gearbox could be feasible, it was necessary to test the 3D printed material in comparison to traditional gear materials. Due to size constraints on the desired planetary gearbox, the diametral pitch and number of teeth were optimized to create the largest teeth possible. The Markforged 3D printer was used for its carbon fiber infused nylon material and compared to gears of the same pitch and diameter that were made of acetal and aluminum. The setup was created such that one gear was bolted down and the other was free turning. The two gears meshed and the free turning gear was attached to a torque wrench that measured the torque needed to break the teeth. The result was that the 3D printed gear, even with the carbon fiber infused, failed at a far smaller torque than the other gears.
Continuing the work to produce cheaper gears from resin casting, we looked to injection molding. Using 3D printed molds made from Digital ABS printed with the Objet polyjet printer, we were able to design detailed molds with gear teeth without expensive tooling. Multiple revisions were made to reduce warping and flash by creating a more consistent wall thickness but keeping a hub for mounting purposes.
As part of a semester long project for a Design Innovations class, I created a low fidelity prototype of a robot dog meant to guide visually impaired people. This laser cut dog could turn its head with a servo motor and "bark" using a piezo buzzer when a button was pressed. The dog was controlled using an Adafruit Feather microcontroller and breadboardless electronics soldered together and placed in a 3D printed enclosure. This low fidelity prototype was used in user interviews to obtain UX data.
In a local makerspace I had spent some time learning to make things on a CNC router. As one of my first interactions setting up toolpaths in CAM, I learned a lot. I have used a few softwares including VCarve and Fusion360 to generate the gcode. I've worked on a couple of materials including MDF and plywood and learned a lot about the necessary speeds and feeds for a particular tool. I hope to learn more complex roughing and finishing techniques in order to create some relief carvings in the future.
As part of a never ending goal to cut costs, the mechanical team, under the direction of an older advisor, learned to resin cast gears. This new inhouse manufacturing method cut costs from $5 per gear to less than a dollar per gear. I helped with designing the 3D printed postive of the gears in Solidworks and then helped make the mold using silicone as well as cast the gears using polyurethane. Fortunately, our prototyping turned out well and were able to mass produce the gears rather inexpensively. The gear molding process was extensively detailed in a grant application for funding.
This was the final project for the Intro to Robotics class. Our group had Baxter solve a puzzle based on the computer game, tetris, and place the tetris pieces into a frame. The robot would use the AR tags placed on the pieces to determine its location and orientation and then place them in its final location an orientation. My part involved laser cutting the tetris pieces with AR tags as well as helping with determining the AR tracking orientations.
This project, through the organization Pioneers in Engineering, is meant to develop a simple robotics learning tool for high school students. I did nearly all of the designing of the mechanical system, from creating the CAD to sourcing the cheapest parts that would still enable high school students a robust structure with design versatility. The object of this kit was to be used in workshops taken to schools where the students will learn about gear ratios and elementary circuits. This year we were able to host 4 workshops using these kits and the students enjoyed racing their new cars that they got to keep.
As part of the mechanical team, we design the parts that the students build their robots with. The parts have to be versatile so students can use them for many different structures or mechanisms. We made many adjustments to the kit metal parts from previous years including decreasing the hole spacing, changing the bearing blocks from square tubing to C-Channel for easier access, and modifying the base kit design. There was a fair amount of prototyping in making these adjustments. The adjustments to the custom made parts required new engineering drawings so I created the new drawings as well as reached out to many local manufacturers to compare prices and find the cheapest one.
Orange, CA
92866 US
miyukiweldon99@gmail.com
m.weldon@berkeley.edu
(909) 633 4298
