Academic Projects
The projects were completed all or in part by Ryan Lee during his time at the University of Texas at Austin in Cockrell School of Mechanical Engineering and McCombs School of Business.
Spring Semester 2023
Trade Fairness Scoring System
This project focuses on developing a trade offer scoring system for Tradeblock to improve the success rate of trades. Data exploration and analysis were conducted to identify the relevant variables that influence the acceptance of trades, and to determine their relative “weight” in the model. The trade fairness scoring system will then be used to predict the likelihood of a trade being accepted, helping Tradeblock improve its trade acceptance rate and encouraging users to make more generous offers.
Spring Semester 2023
Initializing New Word Embeddings for Pre-Trained Language Models
This project focuses on the problem of Out-of-Vocabulary words in natural language processing, by testing different methods and models for generating an embedding for these words. The models are developed by leveraging machine learning to produce meaningful embeddings for words not in vocabulary by using the context of the word through a probability distribution of words in the context sequence. This technique of language modeling is analogous to how a person would infer the meaning of a word not in their vocabulary.
Fall Semester 2023
College Football Postseason Predictor
As the 2022 NCAA Football regular season comes to an end, our goal was to develop a machine learning model that can predict the outcome of postseason bowl games. If we are able to develop such a model, we could then place wagers on the outcome of the games and make a net profit. Our initial approach was to train our machine learning model using historical game data from several years and use that to predict the outcome of future postseason games. The model can take in important statistical features such as a team’s strength of schedule, offensive and defensive statistics, and other variables in order to learn patterns and relationships in the data and then come up with an accurate prediction for the outcome of the bowl game.
Spring Semester 2022
Final Report:
Presentation:
Demonstration:
Sensor Deployment Mechanism for Wall-Climbing Robot
The sensor deployment mechanism depicted in the images on the left was the senior design project I worked on with a group of 3 other students.
​
Problem Statement
Due to the risks and costs associated with dry-cask storage container inspection, a sensor deployment mechanism will be designed for a wall-climbing robot and will permanently attach HGUW sensors to the canister surface to reduce inspection costs by reducing the frequency of inspections.
Requirements and Constraints
Our team was required to develop a reloadable mechanism to hold and deploy sensors on dry-cask storage containers. This mechanism was expected to operate in an environment with elevated temperatures and low amounts of low-energy gamma and neutron radiation. We had to ensure protection and prevention of wire entanglements for the wires soldered to each sensor. Our system needed to maintain a rigid connection to the wall-climbing robot at all times and coaxial cables were required to maintain a connection to each sensor.
Our team’s mechanism was constrained in size to being less than 4” in length, 6” in width, and 4” in height and could weigh no more than 2 lbs. We were limited to materials and adhesives that are non-degradable when exposed to elevated temperatures and radiation. We were constrained to designing a mechanism fully actuated by mechanical components.
Solution
The design of our sensor deployment mechanism is entirely mechanical and consists of a sensor pusher assembly, sensor storage assembly, and a stamping mechanism. Prior to entry into a DSC, the top of our storage assembly is removed and Epoxy, Loctite® M-121hp is applied to six HGUW sensors which are then loaded in the storage assembly.
​
The first Bowden cable, connected to the sensor pusher assembly, is actuated to move the sensor and carriage assembly into the stamp assembly. Next, the second Bowden cable pulls the stamp down to attach the sensor to the DSC. Springs are used to return each moving component to its original position.
The stamp component is made out of AISI 4140 steel due to its ability to resist deformation. Additionally, Aluminum 6061-T6 is used for its lower density and strength benefit. The outer housing and the rest of the components are made of polyetherimide plastic due to its high max service temperature, radiation resistance, and low density. The metal components will be machined and the thermoplastic components will be 3D printed. The cost to produce one unit is $3,271.16.
​
Fall Semester 2021
Cookie Cutter Machine
Final Report:
Presentation:
The food sculpting machine depicted in the image on the left was a project for my mechanical engineering design methodology course. The project was done with a group of 5 students and lasted the duration of the Fall semester. The goal of the project was to design and build a machine that could sculpt or cut designs into food in at least 2 axes of motion with a $250 budget. More specific design requirements were compiled from potential-customer interviews and an analysis of their needs. Many design iterations were required before building the final prototype design and extensive testing was conducted on this final design to ensure that our design choices were not flawed. The final prototype was modeled in Solidworks and the individual components of the machine were either bought online or 3D printed on the UT campus. The complete design process is detailed in a final report which can be viewed in the links.
Demonstration:
Spring Semester 2021
Remote-Controlled (RC) Car Project
The RC car in the picture was designed and built with a group of 5 students. The goal was to create a car that would take the least amount of time to travel 3 times around an oval-shaped track that was 16 m long and 4 m wide. Each group was given a torque motor, servo motor, controller, receiver, and battery (depicted in the image on the right) as well as a $50 budget to spend on materials such as wheels, driveshafts, bolts, gear, and other manufactured items. A complete CAD assembly was created using Solidworks and individual part modeling was used to manufacture parts such as the laser-cut plywood chassis and the gray 3D-printed motor mount and steering supports.
Presentation:
Report:
Demonstration:
CAD Assembly: