High School Summer Research Program

About

As part of its strategic plan, mission, and vision, the LSU College of Engineering is dedicated to developing students into the next generation of transformative problem solvers for the local region, the state, and beyond. The High School Summer Research (HSSR) Program is an outreach initiative aimed at engaging high-achieving high school students in real research in the fields of engineering, computer science, and construction management. HSSR interns have opportunities to learn from faculty research groups and understand how they work, what inspires them, and how to continue in fields related to engineering in college and beyond.

In addition to their work on a research team/project, HSSR interns also attend workshops that include trainings on safety, research best practices, ethics in research, and communication. The program culminates in a poster presentation for students to present their research projects, which is mandatory for the completion of the internship.

Program Goals

To give high-achieving, highly motivated, and hard-working local high school students meaningful experiences in engineering, computer science, and construction management research during summer.
To develop students’ curiosity, research methods, and intellectual abilities before they have completed high school or made decisions about which college to attend and what bachelor’s degree to pursue.
To develop students’ abilities to communicate technical content using both written and oral modes of communication.
To teach students about the engineering design process and how it can be applied to both fundamental and applied research.
To introduce these students to the opportunities available at the LSU College of Engineering and showcase the impressive faculty and research available to them at the state flagship university.
To support faculty in their research projects and outreach efforts.

Information Sessions

Three in-person information sessions will be held, as well as one virtual session:

October 30, 2024 at 6:00 p.m. (virtual)
November 20, 2024 at 6:00 p.m. (virtual)
December 11, 2024 at 6:00 p.m. (in-person)
January 22, 2025 at 6:00 p.m. (in-person)

Registration is required to attend an information session. Meeting details will be provided via email after registration.

Program Details

Students selected as HSSR interns are matched with a College of Engineering faculty member's research team. They will receive guidance from the professor, as well as graduate-student and undergraduate-student mentors, as they work on a project related to the research team's ongoing research. Here are examples of past research projects and an article about a student/project from summer 2020.

HSSR Interns will not be paid for their work.

HSSR Interns will be held accountable for their work responsibilities by the College of Engineering and will be expected to compete in regional science fair competitions. The program will provide information and training regarding science fair participation.

HSSR Interns will have to complete detailed safety paperwork and training through the course of the spring 2025 semester in order to begin work on a project in summer 2025 and beyond.

HSSR Interns will be required to work about 15-20 hours per week during summer 2025 for a total of about 120-140 hours. Weekly schedules can be flexible depending on summer travel/activity schedules, however students should not miss more than 4 working days.

The HSSR Intern application and selection process will be highly competitive due to high interest and a limited number of available positions.

Apply

Application Deadline: February 7, 2025

The LSU College of Engineering is seeking qualified local high school students to apply for a limited number of High School Summer Research (HSSR) Intern positions available in summer 2024. Please see the program details and eligibility before applying. If you have questions about eligibility and program details, please contact Raynesha Ducksworth at [email protected].

To be eligible for this program, you must:

Be at least 15 years of age.
Be currently enrolled as a 9th, 10th, or 11th grader.
Have a 3.5 (or equivalent) high school GPA (as listed on a current high school transcript).
Complete and submit this application by end of day on February 2, 2024.

To apply for this program, you must:

Complete the LSU College of Engineering 2024 High School Summer Research Intern Application.
Include/upload a 500-word personal statement focused on how you think this opportunity will impact your future as you pursue an education/career in engineering and what skills and experience you expect to gain.
Include/upload a current unofficial high school transcript.
Include/upload a signed Parent/Guardian Statement of Support.
Include/upload a current resume of academic, volunteer, community service, leadership, and extracurricular activities.
A letter of support from a current science teacher must be emailed to Raynesha Ducksworth at [email protected].

Current Projects

2024

Project Title	Abstract	Program	Mentor
AI Text Completion	Artificial intelligence (AI), largely in the form of Large Language Models (LLMs) used for text completion, has become one of the most rapidly growing and widely used technologies for business and personal applications. With this rapid growth comes issues as there have not been enough studies to test the ethics and reliability of these LLMs used widely. Users could blindly trust incorrect information, which is especially dangerous in a business setting as incorrect answers can lead to costly errors which affect everyone involved. LLMs tend to struggle to correctly answer more complex prompts that are more open-ended or require longer answers. The goal of this research is to discover under what circumstances LLMs respond with incorrect answers and the reasons for these errors. To do this, the GPT2 LLM was used alongside input saliency, a method where the LLM highlights the words used in the prompt that led it to its response. Using this, it can be determined what the LLM thinks it is answering and what the reason for the correct or incorrect answer is. Two different classifications of prompts were used in this study, simple prompts which consist of geography, history, sports, and general knowledge requiring one or two-word responses, and complex prompts consisting of various topics requiring longer and more complex answers. The LLM correctly answered 169 of 227, or 74.4%, of the simple prompts and 18 of 64, or 28.1%, of the complex prompts. In all trials, there was a clear bias towards the United States, specifically San Francisco, where GPT was created. Geography questions, when incorrect, were almost always answered with San Francisco, and history and sports questions were answered correctly at a much higher rate when they were related to the US. In addition, complex prompts were correctly answered at a very low rate because the answers were not as straightforward and the complexity meant that the bias towards American topics affected these answers even more than the answers to the simple prompts. There is a clear issue in the training of GPT2 not being widespread enough. The amount of bias the LLM has towards the US is a clear indicator that the training was focused mainly on things related to the US. The lack of ability to answer more complex prompts also points to training issues, mainly that GPT2 has limited knowledge, and more obscure topics cause issues because the training was too surface-level.	Computer Science	Umar Farooq
Using Metal-Based Zeolite Catalysts to Recycle Plastics	Only 14% of all plastic waste is collected for recycling, while only 5% of this waste is recycled after sorting. The disposal of single-use plastic materials has caused 40% of all plastic waste to be discarded and landfilled, posing a significant environmental challenge. Most of this plastic comprises polyethylene (-CH2CH2-)n, a simple polymer commonly found in plastic bags and packaging. Using catalysts, especially metal-based zeolites, has proved to be a crucial tool in chemical recycling to depolymerize polyethylene. Pt/Ni-based zeolite catalysts were found to chemically recycle low-density polyethylene (LDPE) and produce fuel such as gasoline and diesel with coke (carbon) as a byproduct. This project aims to examine the properties and products of different catalysts to determine which is the most successful at depolymerizing LDPE, creating the most fuel products, and minimizing the amount of coke produced. Coke formation can block the active sites of the catalysts, therefore hindering it from continuing to react with the polymer. After synthesis by ion exchange (K-Zeo + Metal salt à Metal-Zeo), the reaction with LDPE was done by melting the plastic into the catalyst bed (Fe3O4 : catalyst, 1:1) and exposing it to radio frequency to get a bed temperature of 375˚C in a fixed bed continuous reactor. Following the reaction, the catalyst products were analyzed with temperature-programmed oxidation (TPO) and x-ray diffraction (XRD); the former was used to show fuel products, unreacted polymer, and coke, and the latter was used to determine different phases of materials on the used catalyst. Upon analyzing results, it was found that the Ni/K-BEA catalyst was more stable than the Ni/K-MFI catalyst, and the MFI catalyst produced more gas products. The results from XRD showed that there was a formation of iron carbide (FeC) and carbon in the used catalyst. Deactivation of the catalysts were caused by the buildup of coke.	Chemical Engineering	Kerry Dooley
Optimizing Ultra High-Performance Concrete using Python	Although commercial concrete is a cheap and reliable material for building roads, bridges, and buildings, its 50-100 year life span makes concrete a significant source of construction waste. Ultra-High-Performance-Concrete (UHPC) reduces this issue with its more than 200 year lifespan. The material itself is also longer-lasting, more ductile, and has a compression strength of 17,500 psi (3-4 times that of standard concrete). Because the quality of the concrete depends on its particle size distribution, the ratio of the dry components (cement, sand, silica fume, and slag) must be optimal for making the densest material possible. The goal of this project was to create a Python tool that would calculate those ratios using minimization. The optimal particle size distribution (PSD) had already been calculated mathematically. By applying the Root Mean Squared Error equation, the ratios of the ingredients would change based on what would create the smallest deviation in PSD from the optimal values. Three conditions were tested. The “no boundaries” set had no conditions, only what would result in the smallest RMSE value. The “balanced” set had it so that silica fume could be no more than 20% of the material, and that slag had to be at least 9.1% of the material. The “economical” set had the same conditions as the “balanced” set, including that the amount of sand should be at least the same as cement. It was especially important to use slag because it is a waste product that could now have a potential use. After the ratios were printed, new batches of concrete cubes were created according to the given values. The cubes were cured in water for 7 and 28 days until being compression tested to see if they indeed resulted in a stronger material. The 28 day “balanced” set was the only one to reach 17.5 kpsi. With that, we have created a more sustainable UHPC with the incorporation of slag	Civil & Environmental Engineering	Yen Fang Su
Traffic Light Detection and Recognition on Autonomous Driving Systems	Traffic accidents are the cause of death of many people in the United States and over 25% of traffic fatalities occur at or near an intersection. At intersections, drivers are surrounded by cars on three sides and pedestrians are everywhere, all of which can tempt the drivers' focus. Additionally, non-traffic-related circumstances can corrupt the decision-making abilities of drivers. This probable loss of focus and poor judgment can cause injury to drivers, passengers, and pedestrians. However, implementing traffic light recognition systems to assist drivers can prevent fatalities and improve road safety for all. This work utilizes the You Only Look Once(YOLO) model in a physical simulation to mirror the detection and interpretation of traffic signals in real time. Through the use of a Turtlebot3 Burger equipped with a Raspberry Pi camera, a dataset of over 500 images of traffic lights with varying arrangements, lighting, and signals was captured. Each image's traffic light and traffic signal of interest were annotated using Roboflow, a computer vision tool. In the custom YOLOv8 model, 70% of the images were used for training, 20% were used for validation, and 10% were used for testing. The model was trained through the command line interface for 78 epochs with a batch size of 8. With a precision of 96.3%, mAP50 of 92.1%, and an exponential decrease in class loss, the model was highly effective in the detection of traffic lights and signals. The results indicate that traffic light detection systems are a viable option for improving road safety and decreasing fatalities by ensuring the correct decisions are made at traffic lights.	Electrical & Computer Engineering	Xiangyu Meng
Efficiency of Photovoltaic Panels Under Baton Rouge Weather Conditions	For my research, three distinct simulations were conducted using the TRNSYS software to evaluate the performance of photovoltaic (PV) panels with varying surface areas. Each simulation used the online plotter built into TRNSYS and used TMY2 Baton Rouge weather data, type15-2, for the month of January. The simulations used Type 103b photovoltaic panels; these panels are a basic model with MPPT (Maximum Power Point Tracker). The only variable input was the surface area of the panels, every other input was kept constant to ensure only the change in surface area would affect the outputs. The surface area of these panels were increased from 2000 ft2 to 413,820ft2 (10% of LSU campus), and finally to 120,323,174.4 ft2 (5% of Baton Rouge). Throughout these three simulations, four different types of graphs were generated from each simulation: power output over time (Graph-1), power output and temperature over time (Graph-2), power output and wind velocity over time (Graph-3), and voltage and current over time (Graph-4). Graph-4 demonstrates the I-V curve, which can be used to find out the power output, and also used to find the Maximum Power Point (MPP). MPP is the point on a current-voltage (I-V) curve of a solar PV panel that corresponds to the maximum power output. Operating a PV panel at its MPP will make sure it is at optimal efficiency and maximum energy output. Graph-1 is demonstrating the power output over time which leads to finding the efficiency. The formula to find efficiency of the panels is power output over irradiance multiplied by the surface area, η= (Power Output/ Irradiance×Surface area) ×100%. Graph-2 and Graph-3 provide insight on how different weather conditions affect the panel’s power output. By comparing the simulations, a notable observation is the increase in surface area influenced the panels' power output. Also throughout the simulations, correlations between the power output and the temperature as well as the power output and wind velocity is shown, thus providing valuable insights into the scalability and performance of large PV arrays.	Construction Management	Arup Bhattacharya
Soft, Liquid Metal Embedded Fluoroelastomers for Space Equipment and Seals	Lunar dust build-up has become a serious problem in space exploration technology. When the dust builds up on space seals, many qualities of the space mission are altered including negative impacts on the mechanics of the shuttle, the electronics of the equipment, and the health of the astronauts and passengers on board. Designing and using appropriate space gasket seals is beneficial and crucial for the reliability and longevity of the seals. Past studies have shown that the conductors, liquid metal and silver flakes, achieve high conductivity and reach high temperatures before degrading, making them the perfect candidate for an electric current to run through while in space. For this to be possible, we designed an O-ring mold in SolidWorks, a platform used to design all sorts of products, to have a base model for what we wanted the seals to look like. We then mixed different concentrations of Viton and MIBK, the solvent that melts Viton and mixes to become an elastomer, liquid metal, and silver flakes, testing to see which composition worked best for space conditions and eventually embedding sensors. We tested each composition’s resistance to temperature changes using a hot plate. After finding the best ratio between each material, we started to embed the mixture into a ring of a different elastomer. From our methodology, we found that liquid metal and silver flakes embedded into the space seals allow for them to be far more conducive and can reach high temperatures before fully breaking down and becoming unable to work. We also found that the samples containing 30% Viton, as compared to the samples containing 40% Viton, were far more conductive due to the amount of the nonconductive material Viton. From our findings, we can conclude that for the sensors embedded in the O-rings to work to the best of their ability, less Viton needs to be used. For the initial problem to be properly dealt with, advances in space-sealing technology need to be approached.	Mechanical & Industrial Engineering	Robert Hebert
Developing the Usage of Hydrophones in Underwater Robots	Using hydrophones with underwater robots would allow users to understand through hearing the state of his/her underwater robot. The main method used to understand the state and location of an underwater robot is sonar, however, sonar does not give users specific information. Hydrophones are used in fields or grids for surveying certain areas but do not travel with the robots themselves. The research and work have been dedicated to understanding how a hydrophone could operate onboard an underwater robot. Before attaching the hydrophone to the robot, sound recordings were conducted with the hydrophone in a small tank (1ft x 1ft x 8in). The recordings included various elements such as, tapping on the tank, tapping on the hydrophone, moving the hydrophone through the water, etc. These recordings were analyzed and processed using the Fourier Transform so that ideal frequencies could be prioritized, and unideal frequencies could be removed. Therefore, when hydrophones were used with underwater robots, the user could filter out signals in order to listen to only the operating systems of the hydrophone to understand its state. This process could also be used in reverse order, filtering out the noises created by the robot to understand the environment it is placed in. Additional testing is being conducted to see how the hydrophone responds to the sounds made by the thrusters of the underwater robot.	Mechanical & Industrial Engineering	Corina Barbalata
Machine Learning for Polymer Topologies	Physical tests to determine the properties of polymers take months and often years. This process has been sped up, as one can virtually test the polymers in a computer simulation. In the last few years, this process has been sped up even more, with the rise of machine learning. Using ML models and previous data about the properties of polymers, we can accurately predict the properties of an untested polymer. In this project, we investigated bottle brush polymers' Glass Transition Temperatures and Diffusion constants. To simplify the calculation, the polymer is not modeled on the actual molecules, but each bead of the polymer becomes a vertex of a graph. Then, certain descriptors of this polymer, including relatively simple ones like the number of vertices/edges, average degree, and network density, and more complex ones. We first looked at the descriptors vs. the Tg and D values and discovered some patterns in correlation, as well as stronger correlations of descriptors with D than with Tg . We then created a neural network that takes such topological descriptors of a polymer graph, and uses them to train/predict gas transition temperature and diffusion constant values. While the accuracy of both was reasonably strong (Tg prediction had an R 2 of 0.87 and D had an R 2 of 0.98), diffusion constant predictions were much stringer due to a wider range of values, as well as heavier correlation to some of the topological descriptors.	Chemical Engineering	Yaxin An
Lane Detection Through Use of Mobile Robots	The turtlebot3 is an ROS-based mobile robot used in education and product prototyping. This mobile robot has the capability to perform many functions such as teleoperation, SLAM mapping, gazebo simulations, lane detection, and traffic sign detection. In the modern age, vehicles with the capacity to operate autonomously have grown into a large and expanding industry, emphasizing a need to understand the processes that drive these systems on a deeper level. Autonomous driving is usually achieved using two methods: Machine learning, or in the case of this project, lane detection. The turtlebot 3 works by using the camera and computer programming to achieve the desired lane detection. On the front side of the robot there is a Raspberry Pi camera connected to the SBC (Single board Computer). After calibration of the camera, it can differentiate different hues to detect where the lanes are through a preprogrammed computer program. By placing the robot in the middle of a model lane, it will be able to detect where the yellow hue is in the camera and use that information to detect where the left lane is. This process is then repeated to the other side until the turtlebot at any point in autonomous driving mode can know where both lanes are. From this point, the turtlebot3 then creates a line through the middle of both the yellow and white lanes to follow a path to drive through the model track. This was successfully completed on the earlier computer simulation gazebo, a program that has its own model track, however not fully realized on the actual turtlebot3 within the timespan of this research project. Future steps of this project would include finishing the lane detection on the actual turtlebot3 and creating a function that also allows for traffic sign detection.	Electrical & Computer Engineering	Xiangyu Meng
Hurricane Performance on Storage Tanks	Storage tanks have historically been subject to major natural disasters which have necessitated spending millions of dollars in repair over the recent decades. The Murphy Oil spill, an incident occurring during Hurricane Katrina, exemplifies this issue, in which intensive flooding caused projectiles to impact the storage tank surfaces at such high speeds that resulted in leakages of over one million gallons of crude oil. The oil contaminated almost 2,000 houses in the nearby vicinity and required billions of dollars in its restoration process. This project aims to model the damages projectiles can induce on storage tanks due to hurricanes so that we can find effective solutions in a quicker and more effective manner. Specifically, the study investigates the effects of projectile impacts at two extreme angles: 0 degrees (perpendicular) and 90 degrees (parallel). The underlying hypothesis is that by examining these two extremes, the obtained data can be extrapolated to predict damage across intermediate angles. Additionally, cases have been considered depending on the location of which the projectile hits the storage tank and how that correlates to the maximum amount of stress the projectile brings about. There is a starting point of impact and four additional cases considered for each angle with distances shifted 0.5, 1, 2, 3, and 4 meters respectively in the z direction. In both cases for 0 and 90 degrees, we see similar patterns of stress levels where the middle section showed larger deformation which absorbed more energy. Thus, it generated smaller amounts of stress relative to the ends, where stress was shown to be the highest. This work is significant because it addresses a critical issue in the design management of storage tanks when subject to extreme natural disasters like hurricanes. By modeling the impact of projectiles on storage tanks, this research gives us valuable insights into how these structures can be better protected from such failures. These failures have historically led to large-scale environmental damage and huge financial losses.	Civil & Environmental Engineering	Sabarethinam Kameshwar
Algorithm Design for Autonomous Vehicles: from Cyber System to Physical System	Self-driving technology is becoming more prevalent and has the potential to revolutionize our transportation system. Self-driving cars are expected to be operated by a computer reducing accidents caused by human error. However, self-driving cars are still not trusted enough for real-life use due to concerns about their safety. Self-driving vehicles need to make intelligent decisions in as little time as possible. Yet, the algorithm for autonomous vehicles is not completely reliable and accurate causing safety issues for the passengers. The objective of this study is to create an algorithm for self-driving vehicles that would follow a reference trajectory and would run smoothly with minimum errors. To achieve this objective, a Proportional– Integral–Derivative (PID) controller algorithm was developed to be used in an autonomous vehicle simulation. To enhance the results of the PID controller, we created a Model Predictive Controller (MPC) that may help the model car follow a predesigned trajectory. In this study, five research tasks were conducted. First, I learned how PID and MPC controls work and how they can be developed. Second, I learned how to develop PID and MPC using python. Third, I found the derivative/integral/proportional values for the PID control algorithm. Fourth, I integrated and tested the PID and MPC control algorithms in a model car simulation. Finally, I refined the algorithms repetitively under accurate results were obtained. Results of the simulation drive showed that the PID algorithm followed the reference trajectory but with some errors around the horizontal curves. On the other hand, results of the simulation for the MPC algorithm showed that it was very accurate on the directed course. This study concluded that the PID controller algorithm was useful and easy to develop when following a track but had limitations around horizontal curves. In addition, the MPC controller is a better alternative than the PID controller/trajectory following algorithm because it integrates the speed and steering control systems.	Electrical & Computer Engineering	Xiangyu Meng
Analyzing Influence of Lighting Parameters on Biological Surface Imaging	Image analysis can help aid in the visual categorization of biological surfaces. Visual indicators of damage can be found on the surfaces of soybeans, making image analysis suitable for identifying damage types. Such damage types can be categorized as smooth, cracked, and shriveled. Acquiring high quality images will make the imaging process more efficient. Different lighting and background conditions can be used to make imaging easier by creating more contrast between the regions of interest and the background. Currently, the USDA has an approved official background for manual inspection of soybeans. This study investigates using various lighting, black, USDA yellow, white, and blue backgrounds to determine which combination provides the most contrast for further image analysis. Using ImageJ, an image processing app, the mean color values of the backgrounds and soybeans were measured and compared to find which parameters created the most contrast. On average, it was found that the red channel of soybeans that were imaged on a blue background had the most contrast. Specifically, the red channel of smooth and cracked soybeans also had the most contrast on a blue background and the blue channel of shriveled soybeans had the most contrast when imaged on a white background. These insights into optimal imaging conditions can lead to the production of better, higher-quality images, improving the accuracy and efficiency of image analysis in future applications. By using background colors and lighting to optimize contrast, researchers can enhance their ability to categorize and assess soybean surface damage, ultimately contributing to better quality control of soybeans.	Biological and Agricultural Engineering	Kevin Hoffseth
Using Machine Learning to Predict Cross-Site Scripting Vulnerabilities in JavaScript	Cross-site scripting (XSS) attacks are one of the most common web vulnerabilities, but difficult to track and prevent. Websites have a very large XSS attack surface area, with every user input on a website leaving potential spaces for a vulnerability. Developers can easily mishandle user inputs, leading to an XSS vulnerability. Currently, the most effective method for identification is taint tracking, a form of dynamic analysis. Its high precision in detection comes at the cost of resource consumption and time constraints that make it impractical for many applications. Using a database of Javascript functions scraped from the web and labeled as vulnerable or safe, I compared two machine learning models, a Deep Neural Network (DNN) and a Random Forest Decision Tree, trained on data that I processed using two different hash functions. I found that a DNN trained on a Term-Frequency Inverse Document Frequency (TFIDF) hash function maintained the highest accuracy at a high recall. Using the DNN with taint tracking would reduce the number of functions that taint tracking needs to analyze by 242.2x, at a recall of 95%. This would vastly reduce resource overhead and speed up analysis, making it possible to apply in-browser while maintaining a reasonably high detection rate.	Computer Science	Phani Vadrevu
Fabrication of Non-biofouling Nanochannel UV-Curable Devices	Nano/microchannel devices are a promising technological advancement that develops research in selective biomolecule transport or DNA detection. This makes way for single molecule analysis as a nanoparticle’s presence is indicated via electric current as it passes through a nanopore. The study of Non-biofouling devices observes electrically charged fluid flow within a channel under the nanoscale (less than 100nm). The fabrication process is unique in a way that allows us to optimally reproduce several samples in an affordable, convenient, and portable manner. To create these devices, a technique using Ultraviolet nanoimprint lithography (NIL) is implemented to generate nanoimprint patterns at a high throughput and low cost. Using a plastic substrate is more favorable under lab conditions due to its disposable nature, plastics possess diverse surface properties which potentially reduce the need for anti-fouling (non-sticking) treatment. The goal of this project is to test and compare the types of UV-curable resins to observe which PEG hydrogel solution optimizes the stability of the nanochannel devices over an extended period. As a hydrophilic compound, Polyethylene glycol is a network of polymers resistant to protein adhesion and biodegradation. It is ideal to create Non-biofouling devices to avoid blockage within the channels due to dust particles or unwanted biomolecule agents. There are three types of Polyethylene glycol that were tested in this experiment. GDD (Glycerol 1,3-diglycerolate diacrylate), GDM (Glycerol 1,3-Dimethacrylate), and PEGDA (Poly(ethylene glycol) diacrylate). Each resin type differs in mechanical and chemical properties which can affect or alter the consistency of stability values. The Hydroxyl-enriched polymers exhibit a high level of hydrophilicity and can facilitate solution fillings without any surface modifications, unlike thermoplastics (PMMA, COC).	Mechanical & Industrial Engineering	Sunggook Park
Renewable Bio-Oil from the Pyrolysis of Sugar Cane Bagasse	As part of the global effort to identify renewable sources of energy to replace fossil fuels, several methods have shown promise in utilizing the abundance of organic biomass. Pyrolysis, the thermal decomposition of biomass in the absence of oxygen, is an existing method that converts biomass into sustainable bioproducts, namely bio-oil, biochar, and syngas. This project aims to further understand and improve upon the pyrolysis process on a laboratory scale continuous induction heater (CIH). The biomass used, sugar cane bagasse (SCB), has high fiber which has all sucrose removed as part of the sugar production process, and sugarcane yields a high content of biomass (Fennell, et al., 2014). Prior to operation, SCB undergoes milling into smaller particles (2 mm sieve) and heating to remove excess moisture. Then, the SCB is loaded into a chamber, nitrogen purges the system (15-20 min), and SCB is pushed through the CIH at an average rate of 124 g/hr and with a residence time of 20 min in the main reactor. The CIH is manually set to the desired experimental temperatures (500, 600, 700 ◦C), and the effect of these temperatures on bio-oil quantity and quality after 5 ½ hours of run time is analyzed. To maximize liquid yields, bio-oil is first collected in a 2L round-bottom flask in a condensing ice bath and is followed by a second stage in a 0.5L flask in dry ice and connected to an electrostatic precipitator. In between the reactor/induction coils and first stage collection, heating tape is used on a pipe leading to the first flask to maintain a stable temperature of 275 ◦C. Solid black char is collected at the end of the CIH. Collection of bio-oil includes separation into its two distinct phases: a light brown translucent phase and a dark heavy tar phase (using acetone). Across the 6 total trials (2 trials per temperature), the highest average percent liquid yield, 41.56%, was reached at 600 ◦C. Total Carbon and Total Nitrogen analysis (in mg/L) of light phase oil samples showed that as temperature increased, carbon content gradually decreased while nitrogen content increased. Further analysis (primarily GC-MS) is required for a more comprehensive conclusion on oil/char quality and their potential in industries such as aviation/ground transportation fuels. Maintaining a continuous flow of biomass was an operational challenge of this CIH, so future scaled-up continuous pyrolysis systems should utilize a dry mixer or secondary auger for maximum efficiency.	Biological and Agricultural Engineering	Dorin Bolder
Sustainable Energy from Photovoltaic Panels	The state of Louisiana produces an average of 216 sunny days a year. As of today, Louisiana’s energy is only about 3% renewable energy. Historical weather data from Baton Rouge shows promising production of renewable energy across the city. Using the TRNSYS weather platform, our research found that in the month of January, 5% of Baton Rouge’s area covered in photovoltaic (PV) panels produces a maximum of about 75 Watts in a day and a minimum of about 39 Watts. The average household in the United States uses around 30 Watts. Louisiana has the capacity to produce energy to run a city using majority sustainable energy. When hurricanes and other bad weather passes through the state, PV panels contain a battery to preserve energy. The PV panels can produce the preserved energy on cloudy days or during the night when sun rays are not available. However, a setback with the panels is that when temperature rises too high, PV panels stop working and cannot produce energy. When the temperature peaks in Baton Rouge, the power produced by the PV panels drops significantly. As global temperatures are only rising, PV panel production is threatened. However, PV panels can also compensate for rising temperatures across the world. For example, in an average parking lot, 1 in 7 spots are left empty. The sun rays are reflected off of the black cement and only further heat the area. One possible solution is to shield these lots with roofs covered in solar panels. While there are notable variables to study with solar panels, like how temperature may affect its productivity, solar panels have the potential to replace non-renewable energy resources as resources like oil, natural gas, and coal depletes.	Construction Management	Arup Bhattacharya
Enhancing User Experience for d/Deaf Community Using Machine Learning	With over 300 different types of sign languages and over 72 million users, sign language is an integral part of millions of people’s lives. Deaf and hard-of-hearing people primarily rely on sign language to communicate with their friends and family. This study aims to document the available sign language resources and create a machine learning algorithm that can accurately identify a sign and convert it into text or audio, and vice versa. First we carried out a need-finding study with the d/Deaf community and conducted a thematic analysis to pinpoint current challenges and future expectations. In a survey taken of both signers and non-signers, signers generally felt that communicating with people not fluent in sign language was hard. Meanwhile, beginners in sign language felt that signers were signing too quickly. Then we started gathering as much data as we could find on the different types of sign languages and made a website to store the data in one place. Using all the data gathered, we plan to create a Convolutional Neural Network capable of identifying a sign and converting it to text and vice versa. Current research focuses on sign language translation using image processing; however, the need for implementing a real time detection and translation platform is still necessary. There is still a need for a simple medium that can be used by both signers and non-signers to communicate with each other. The implications of this study are great, as deaf people will now be able to communicate effectively with people who do not know sign language well. Additionally, our tool can be used in international meetings to convert audio into sign language, tailored to the participant's native sign language.	Computer Science	Mahmood Jasim
Multi-Agent Reinforcement Learning for Gaming	This project explores the application of multi-agent reinforcement learning (MARL) within the context of a hide-and-seek game simulation, leveraging advanced tools and frameworks to model and train autonomous agents. The hide-and-seek game is implemented using MuJoCo, a high-performance physics engine, and OpenAI's environment generation tools, facilitated by the MuJoCo-Py and MuJoCo-Worldgen libraries. The project involves setting up a virtual machine with an Ubuntu Linux distribution, followed by the installation of MuJoCo, MuJoCo-Py, MuJoCo-Worldgen, and OpenAI's environment generation framework. Once configured, an example hide-and-seek scenario is executed to demonstrate the interaction between agents trained to master hiding and seeking strategies through reinforcement learning. This setup not only highlights the capabilities of MARL in complex, dynamic environments but also showcases the integration of simulation tools to advance research in autonomous agent behavior and emergent strategies.	Electrical & Computer Engineering	Hao Wang
Ion Surface and Bulk Diffusion Coefficient Measurements in Microchannels of Microfluidic Devices Utilizing Current-Time Monitoring	Diffusion is a crucial process for ion migration in microchannels due to high surface-to-volume ratios, impacting ionic transport's efficiency. In microchannels, diffusion can be separated into surface and bulk diffusion, with surface diffusion occurring on the walls of the microchannel and bulk diffusion moving through the interior of the microchannel. Fick’s Second Law of Nonsteady Diffusion ( ) can be modified to represent ∂φ ∂𝑡 = 𝐷 ∂ 2φ ∂𝑥 2 = 𝐷∆φ ⇔ ∂𝐶(𝑥,𝑡) ∂𝑡 = ∂ ∂𝑥 𝐷 ∂𝐶(𝑥,𝑡) ∂𝑥 diffusion in 3 dimensions. Due to a large volume ratio, we can neglect in the original ∂ ∂𝑥 expression, allowing the concentration of the KCl solution in the microchannel to be integrated over length 𝐿. Prior experimentation shows that current is directly proportional to KCl bulk concentrations, which converts the concentration equation into a simplified current-time equation. To find the diffusion coefficient, current-time data can be curve-fit into the exponential rise to maximum model 𝑓 = 𝑦 , and with the previously integrated current-time 0 + 𝑎𝑒𝑥𝑝(− 𝑏𝑥) equation, form the relationship of 𝐷 = , where is the diffusion coefficient. To form the 𝑏𝐿 2 π 2 𝐷 microfluidic devices, a UV resin mold was created through replication of a Si master mold, which was then nanoimprinted (using an Obducat machine) into a COP1020R substrate after being covered in polyimide thermoplastic and polycarbonate. After drilling into the reservoirs, a 3-minute UV-Ozone treatment was applied, followed by bonding with COC8007 (102 µ𝑚) in an Obducat NIL machine. Microchannels of the microfluidic device were filled with .1 M KCl (via air pump) and held with a pressure jig before implanting electrodes over enclosed reservoirs. .1 M KCl was replaced with 1 M KCl, .5x Tris EDTA solution, and voltage was applied every 20 seconds for surface diffusion and every 2 minutes for bulk diffusion. Diffusion coefficients were calculated for 10 mV, 15 mV, and 20 mV, before using linear regression to extrapolate the diffusion coefficient at 0 mV, which is . 3709 × 10 ) for surface diffusion and −8 ( 𝑚 2 𝑠 8. 4028 × 10 for bulk diffusion.	Chemical Engineering	Yaxin An

Past Projects

2023

Project Title	Program	Mentor
The Use of Glass Sand to Prevent Erosion in Coastal Louisiana	Civil and Environmental Engineering	Clint Willson
Quantifying the Benefits of Freeboard Policy on Louisiana Parishes’ Class in the Community Rating System	Biological and Agricultural Engineering	Carol Friedland
Studying the Effects of Distortion on the Hydrodynamics in the Lower Mississippi River Physical Model	Civil and Environmental Engineering	Clint Willson
Coordination of Fully Automated Vehicle Platoons for Crossing Non-stop Intersections	Electrical and Computer Engineering	Xiangyu Meng
Optimal Path Planning with Applications to Automatic Parking	Electrical and Computer Engineering	Xiangyu Meng
Utilizing Machine Learning Classification Algorithms to Detect Pancreatic Cancer with Fluorescence Spectrum Data	Electrical and Computer Engineering	Jian Xu
Weighted Gene Co-expression Network Analysis (WGCNA) Analysis of the Genotype-Tissue Expression (GTEx) data from the left ventricle	Biological Engineering	Jangwook P. Jung
Pancreatic Cancer Detection by Artificial-Intelligence-Assisted Raman Spectroscop	Electrical and Computer Engineering	Jian Xu
Development of Earthen Building Materials Inspired by the Nest Construction Techniques of Mud Dauber Wasps	Civil and Environmental Engineering	Hai Li
Machine Learning-Based Colloidal Self-Assembly Phase Identification	Chemical Engineering	Andres Lizano, Xun Tang
Unraveling the potential of ChatGPT and AI in optimizing the Average High Schooler’s Daily schedule	Computer Science and Engineering	Hao Wang
Examining Drivers’ Behaviors to Connected and Autonomous Vehicles	Civil & Environmental Engineering	Hany Hassan
Comparison of Bone’s Natural Microstructure to Applied Speckle Patterns	Biological and Agricultural Engineering	Kevin Hoffseth
A Machine Learning Approach to Analyze Energy Burden in U.S. Low-Income Households	Construction Management	Amirhosein Jafari
Toluene Production Capacity of a Microbial Community Derived from Colorado River Sediment	Civil and Environmental Engineering	William M. Moe
The Photobleaching Effect of Fluorescent Proteins for Cell-Free Biosensor Development	Biological Engineering	Yongchan Kwon
3D Printed Co-culture Platform to Study Bacteria Induced Endocrine Resistance in Breast Cancer	Chemical Engineering	Adam Melvin
Tracking Augmented Reality/Virtual Reality (AR/VR) Users' App Usage Duration through Push Notifications	Computer Science and Engineering	Chen Wang, Ruxin Wang
Improving Low Temperature DRM by Deposition of CeO2 Overlayers on Ni/Al2O3 Catalysts	Chemical Engineering	Kerry Dooley
Microfluidic 3D Co-culture of Estrogen Receptor Positive (ER+) Breast Cancer and Stromal cells Study Endocrine Resistance	Chemical Engineering	Braulio Andres Ortega Quesada, Adam Melvin
Backdooring AI Models with Data Poisoning	Computer Science and Engineering	Hao Wang
Improving Superwood by Optimizing the Delignification Process	Civil and Environmental Engineering	Hussein Alqrinawi, Hai Lin

2022

Project Title	Program	Mentor
Optimization of Cell-Free Protein Synthesis	Biological Engineering	Yongchan Kwon
Evaluation of Staining Method for Analysis of Cortical Bone Geometry	Biological Engineering, Civil and Environmental Engineering	Akshay Basireddy, Simone Muir, Beatriz Garcia, Alexander Lee, Kevin Hoffseth
3D-Printed Co-Culture Platform to Study Bacteria-Induced Chemotherapeutic Resistance in Breast Cancer	Chemical Engineering	Rocio Larenas Bustos, Stephanie Price, Emmaline Miller, and Adam T. Melvin
Solute Movement in Surface Water With Different Stream and River Geometries	Civil and Environmental Engineering	Emily Chen, Clint Willson
Keystroke Privacy Leakage From Zoom Meetings	Computer Science	Collin Clement, Long Huang, Chen Wang
Artificial-Intelligence-Aided Laryngeal Cancer Identification	Electrical Engineering	Mariana Cuadra, Zheng Li, Huaizhi Wang, Jian Xu
3D-Printed Soil Bricks Inspired By Mud Dauber Nest	Civil and Environmental Engineering	Josephine Day, Joon S Park, Hai Lin
Single Cell Analysis of Deubiquitinating Enzyme (DUB) Activity Using a Droplet Microfluidic Trapping Array	Chemical Engineering	Veda Devireddy, Alireza Rahnama, Adam Melvin
Accelerating Reinforcement Learning	Computer Science	Ryan Ding, Hao Wang
The Role of the Genus Azospira in Transforming Arsenic-Containing Compounds	Civil and Environmental Engineering	Andi Hayes, Kali Martin, Bill Moe
Designing RNA Gene Circuits With Coherent Feedforward Loops	Chemical Engineering	Benjamin Hogg, Xun Tang
Demonstrating UAV Propulsion Using an Aircraft and Flight Model With Hardware in Loop Approach	Mechanical Engineering	Nicole Lin, Shyam K. Menon
An Investigation into the Role of Fluid Shear Stress on Enhanced Cancer Extravasation during Metastasis	Chemical Engineering	Josie Ostrowe, Braulio Ortega Quesada, Adam T. Melvin
Nanoengineering Balsa Wood for Resilient Superwood	Civil and Environmental Engineering	Addison Schempf, Hussein Alqrinawi, Hai Lin
Reinforcement Learning in Flappy Bird	Computer Science	Kaitlyn Smith, Hao Wang
Steel Fiber Reinforcement in 3D Construction Printed Concrete	Construction Management	Kaiser Stentiford, Ilerioluwa Giwa, Hassan Ahmed, Ali Kazemian
Detecting Hidden Security Threats With a Thermal Camera	Computer Science	Kenzie Stentiford, Ruxin Wang, Chen Wang
The Impact of an Integrated Local Fan in a Central Cooling System on Occupant Thermal Comfort in Working Environments	Construction Management	Sarah Thomasa, Seddigheh (Tala) Norouziaslb, Amirhosein Jafari
Designing RNA Gene Circuits With Incoherent Type-1 Feedforward Loop	Chemical Engineering	Ahan Zaman, Xun Tang
Multimodal Label-Free Monitoring of Stem Cell Differentiation: Confocal Microscopy	Mechanical Engineering	Laura Zapata, Sreyashi Das, Ram Devireddy
Geotechnical Analysis and Comparison of Recycled Glass Sediment for Coastal Restoration	Environmental Engineering	Louisa Zhu, Julia Mudd, Clint Willson

2021

Project Title	Program	Mentor
Application of PCR to Detect Aromatic Hydrocarbon Producing Bacterial Populations in Sediment Samples from South Louisiana	Civil and Environmental Engineering	Bill Moe
Role of the Genus Azospira in Biological Nutrient Removal	Civil and Environmental Engineering	Tamara K. Martin, Bill Moe
Investigation of Physical and Mechanical Properties of a Mud Dauber Wasp Nest	Civil and Environmental Engineering	Joon S. Park, Hai Lin
Hurricanes vs. Oil Storage Tanks	Civil and Environmental Engineering	Sabarethinam Kameshwar
Effect of Sand Content on Metakaolin Based Geopolymers	Construction Management	Ruwa AbuFarsakh, Gabriel Arce
A Data-Driven Approach to Improving Energy Efficiency in Buildings	Construction Management	Amirhosein Jafari
Crystal Phases of Metal Oxide Materials	Chemical Engineering	Yuming Wang, James Dorman
Optimization of Hydrogel Identity and Composition in an Open-Air 3D Printed Microfluidic Device to Study 3D Cell Migration	Chemical Engineering	Kalena Nichol, Adam Melvin
Development of a Modular Microfluidic Device to Study the Effects of Fluid Shear Stress on ER+ Breast Cancer	Chemical Engineering	Blake Nassar, Adam Melvin
3D Bio-Printing of Tumor Phantom in the Larynges for Tumor Resection Training Applications	Biological and Agricultural Engineering	Kaushik Sunder, Michael E. Dunham, Jangwook P. Jung
The Effects of Bone Dye Techniques on Numerical Microstructural Analysis	Biological and Agricultural Engineering	Kevin Hoffseth
Droplet Interaction with Propagating Shockwaves	Mechanical Engineering	Shyam Menon
Colorimetric and Spectroscopic Sensing of Biomarker for Cystic Fibrosis Using a Smartphone	Mechanical Engineering	Elnaz Sheik, Manas Ranjan Gartia
Preventing Handheld Device Distraction for Drivers Using Smartphone Motion Sensors	Computer Science	Chen Wang
Preventing Driver Distractions Via Acoustic Sensing	Computer Science	Long Huang, Chen Wang
Machine Learning Methods on Raman Spectroscopic Cancer Data for Early Diagnosis	Electrical Engineering	Zheng Li, Jian Xu

2020

Project Title	Program	Mentor
Simulating Cortical Bone Structure in Large Vertebrates	Biological Engineering	Kevin Hoffseth
Microstructural Geometry and Damage Detection in Cortical Bone Images	Biological Engineering	Kevin Hoffseth
Characterization of Fluorescent Proteins Produced in the E. coli Cell-Free Protein Synthesis System	Biological Engineering	Yongchan Kwon
Meta-Analysis of Cardiac Extracellular Matrix Proteins: Information Extraction for 3D Bio-printing	Biological Engineering	Philip Jung
Dynamic Photoluminescence Response of Dipole-Modulated Rare Earth Doped Core-Shell Nanoparticles to Local Changes in Temperature and Solution pH	Chemical Engineering	James Dorman
Machine Learning-Based Feature Analysis and Classification for ICG-Assisted Vibrational Spectroscopic Data of Pancreatic Carcinoma	Electrical Engineering	Jian Xu
3D Tumor Spheroid Generation Using a Droplet Microfluidic Device	Chemical Engineering	Adam Melvin
Circulating Microfluidic Co-Culture Device for the Dynamic Analysis of the Tumor Secretome	Chemical Engineering	Adam Melvin
Development of a Modular Microfluidic Platform to Investigate the Role of Fluid Shear Stress on Cancer Cell Phenotype	Chemical Engineering	Adam Melvin
Using Pulsed UV Light for Enhancing Advanced Oxidation Water Treatment	Environmental Engineering	Samuel Snow
Using Pulsed UV Light for Enhanced Water Disinfection	Environmental Engineering	Samuel Snow
Shockwave Induced Droplet Breakup	Mechanical Engineering	Shyam Menon
Characterization of Animal Nest-Building Geomaterials	Civil Engineering	Hai Lin
Breath Monitoring: Analyzing Breathing with Wireless Bluetooth Earbuds	Computer Science	Chen Wang
Evaluation of the Field Performance of Stabilized and Non-Stabilized Asphalt Overlays in Louisiana	Construction Management	Momen Mousa
The Use of RAP and WMA Mixtures in South-Central States: Challenges & Limitations	Construction Management	Husam Sadek
Variability and Uncertainty of Overlay Tester Testing Data, Analysis, and Results	Construction Management	Husam Sadek

Photos






	Finalists: Josie Ostrowe, St. Joseph’s Academy Kaitlyn Smith, Baton Rouge Magnet Veda Devireddy, Baton Rouge Magnet

Contact

The program administrators are responsible for the facilitation of the program from start to finish by creating the policy/structure, providing regular communication to all stakeholders, serving as the key liaisons between all stakeholders, and generally supporting/directing the program throughout each cycle.

Program Administrator Contact Info:

Raynesha Ducksworth
Assistant Manager
225-578-5335
[email protected]

Corina Barbalata, PhD
Assistant Professor of Mechanical Engineering