Graduate Student Research

Driving with Touch
The Industrial and Systems Engineering Department supports a wide range of student projects in data science, manufacturing, automation, HSE (Human Factors, Safety and Ergonomics), Port Management, Operations Research, Supply Chain Management, robotics, and process improvement. Many projects have a significant software component where our students use advanced modeling techniques to solve problems and contribute in many areas.

The department has grown to 30 doctoral students. The following are examples of current student research projects:

Solving a train assignment problem: An optimization model with a graphic user interface

Student: Masood Jafari Kang - mjafarikang @ lamar.edu

Advisor:  Dr. Maryam Hamidi

This project solves the train assignment problem (TAP) in classification yards, known as shunting or marshaling yards. The project focuses on local rail yard as case study. After multiple in-site observations and meetings with their managers, the project developed a computerized system for railyard terminal operation including a user interace and optimization.

Optimization of Emergency Shutdown System (ESD) to reduce waterhammering in the piping network of the LNG cargo terminal

Student:  Ravinder Singh (rsingh2   @   lamar.edu)

Advisor:  Dr. Xinyu Liu

The sudden opening or closing of a valve during the process operation can create a pressure surge or hydraulic shock, which may exceed the design pressure. The high surge pressure can then damage the pipeline, its supports and equipment such as pumps and sensors. This phenomenon is commonly known as water hammer. The pressure surge could be quite severe for a long pipeline with large diameter.

 

A case study in using synthetic image data for ear plug detection

Student: Samuel E Casco - secasco @ lamar.edu

Advisor:  Dr. James Curry

The biggest bottleneck in the development of computer vision-based solutions is data, since it can be expensive to collect and annotate. An alternative such as the use of synthetic data can provide significant savings of both cost and time as well as accelerate the use of this technology in fields such as industrial safety. In this research we will train a computer vision model with a small synthetic dataset to detect earplug use. We will use software including Blender, GIMP, Python and Detection2.

Resilient ports and waterways with a focus on port infrastructure

Student:  Nader Madkour (nmadkour  @  lamar.edu)

Advisor:  Dr. Berna Tokgoz

Natural disasters can adversely impact port infrastructures including soil erosion and debris in the waterway and ports. This project explores was to monitoring and inspecting those problems to support preventive maintenance and emergent repair.

Simulating blind spot monitoring feedback systems

Student: Acyut Kaneria  akaneria @ lamar.edu

Advisor:  Dr. Yueqing Li

Electronic automotive crash safety systems are considered to be important for driver and passenger’s safety and are widely used by manufacturers in all newer model cars in the United States. Automated blind spot monitoring system is one of those systems which plays a vital role in warning the driver during lane changes on a multilane road. My research focuses on identifying the effectiveness of blind spot monitoring system on the driver during clear weather versus wet rainy conditions. This system uses a radar to detect the presence of an object in the driver’s blind spot and shows a signal to the driver in the form of a warning light on outer rear-view mirrors. My research includes using a STISIM driving simulator to first program the road/traffic conditions and then perform experiment on subjects. This research will find any limitations and/or advantages of using the blind spot monitoring system under varied road conditions. This research may open windows of improvement to the currently used radar based blind spot monitoring system.

Detecting defects of railway tracks

Student:  Premkumar Ravishankar  (pravishankar  @  lamar.edu)

Advisor:  Dr. Berna Tokgoz

This project use computer vision methods to identify common problems including rail damages, missing fasteners, and deteriorated crossties. 

 

Identifying key skills in the US labor market

Student: Yimei Long - ylong @ lamar.edu

Advisor:  Dr. Alberto Marquez

This project examines key employment skills on the US-based job postings and their relation to different education attainment levels (EDL) and job levels (JL). We analyzed the relative importance of employability skills (RIES) based on US-based job postings using Quantitative Analysis of Textual Data (QUANTEDA) and cross-tabulation analysis methods. A dictionary was created by reference to the US Department of Education's ESF (employability skills framework) to identify the required skills in each job posting. Both statistical models and visualizations techniques are used to explore the relationships between education and job skills in the labor market.

Deep learning-based anomaly detection for compressors using audio data

Student:  POOYAN MOBTAHEJ - pmobtahej @ lamar.edu

Advisor:  Dr. Maryam Hamidi

This research project has the goal to develop an automated system that classify normal audio signals and anomaly audio signals accurately for midstream compressor systems. This research has proposed deep learning-based method by using ResNet50 for feature extraction which can classify the normal and anomaly audio signals collected from a compressor system.  Additional to that result showed that the combination of multiple-resource features (2D MFCC features and 1D SC features) can improve the performance for normal and anomalies classification. Additional models are being developed and test. The project will also develop real time software with user friendly interface that can show and explain anomalies in real time.

 

Simulating driver inattention in automated vehicles

Student: Elahe Abbasi - eabbasi at lamar.edu

Advisor:  Dr. Yueqing Li

Drivers’ inattention is one of the principal reasons for most crashes. With the increasing tempo of technology, its interaction with drivers’ attention has become a serious challenge in the safety area. Automated vehicles and In-Vehicle Information Systems (IVIS) are two samples of technologies that may affect the drivers’ attention and at the same time help them fulfill their comfort and safety demands. The aims of this study are to (a) explore various aspects of IVIS and automated vehicles which may affect drivers’ attention, and (b) investigate the trade-off between distraction and the IVIS used in conventional and automated vehicles by using the eye tracker, EEG, and fNIRS. The results of this research can be employed to design the best features and functions of IVIS in automated vehicles to help motorists drive with the maximum level of attention.

Modeling human factors in driving simulations

Student: Yi Liu - yliu8 @ lamar.edu

Advisor:  Dr. Yueqing Li

The project aims to explore the relationship between human factors and driving. The driving simulator lab can use the STISIM software to simulate different kinds of scenarios to investigate the drivers’ behavior and driving performance under different situations. Multiple sensors ar used to record the driver including eye-tracking, EEG, and fNIRS system to study drivers’ eye movement and estimate cognitive workload.

VR training for lifeboats

Student: Mahmood Nawaz Baig - mbaig3 @ lamar.edu

Advisor:  Dr. James Curry

Lifeboat drills and emergency events have significant risks from lifeboat use. This project develops a VR environment to train crews with the goal of improving lifeboat safety and reducing the need to training via drill events on cargo vessels.

Driver workload in automated vehicles and conventional vehicles under different traffic densities

Student:  Ruobing Zhao - rzhao1 @ lamar.edu

Advisor:  Dr. Yueqing Li

The project I am currently working on aims to assess driver’s workload between automated vehicles and conventional vehicles under different traffic densities. In SAE level 2 automation, lateral and longitudinal vehicle motion control is transferred from the driver to the vehicle, but the driver still needs to monitor the driving environment and be able to intervene when necessary. Common ways to assess driver’s workload include subjective measures such as the NASA-TLX questionnaire, task-related performance, and physiological activities. With the help of our driving simulator, eye-tracking glasses and fNIRS system, a more accurate and comprehensive method for evaluating driver’s workload can be achieved.  

Effectiveness of lifeboat testing procedures

Student:  George Kochuparampil 

Advisor:  Dr. James Curry

Are the current testing frequency and procedures of lifeboats on merchant vessels is optimal for crew safety is a key question for improving cargo vessel safety. This project examines multiple data sources to determine lifeboat safety and use patterns. The project examines alteratives and identifies key barriers to improving lifeboat safety.

Exploring the impact of music on task performance

Student:  Fatih Baha Omeroglu - fomeroglu @ lamar.edu

Advisor:  Dr. Yueqing Li

Music has always been a very big part of mankind and the way we live. It affects our mood, cognition, and even our thought process; it induces various emotions and feelings.This study aimed to identify how the human brain and short-term memory are affected by various background music conditions. Electroencephalogram (EEG) brain signals were collected from individuals during the music listening and the Corsi block tapping task (CBT) was performed under 3 different conditions; the relaxing music participants prefer, the music piece they dislike, and no music phase. Results showed that participants achieved significantly better CBT scores while listening to relaxing music. Additionally, relaxing music had a significant effect on beta power spectral density. This study can serve as a valuable guideline to understand the effects of sonic stimulations on cognitive processes such as memory, learning, problem-solving, and decision making.

What do Likert scale surveys really tell us? Predicting average response order using natural language models

Student: Saikrishna S Gubbala   sgubbala at lamar.edu

Advisor:  Dr. James Curry

Likert scale surveys are often used to identify areas of strength and weakness based on ranking of question responses. This project uses deep learning models such as BERT to determine if survey responses can be predicted based on prior surveys in the literature. Using a leave one survey out validation approach, the project tries to predict survey responses order for safety culture surveys. If the model can achieve a high correlation for the rank order of questions (above 0.7 Spearman correlation), then the response order is primarily driven by the wording of questions as opposed to organizational factors.

 

Simulating different touch screen locations in cars

Student: Saumil Patel - spatel74 @ lamar.edu

Advisor:  Dr. Yueqing Li

There has been a rapid increase in ‘In-vehicle touchscreens’ usage over the last decade, particularly for information and entertainment functions. And because of the dynamic interface functionality, the touchscreen becomes a major attraction for the automobile industry. In the research study, we are evaluating 8 different possible designs. (2 HDD location for touch screen display- Top & Middle, 2 different main menu layout- Horizontal & Vertical, 2 screen mount positions- Fixed horizontal screen vs Tilted screen towards driver). The study focused on assessing possibilities for different HDD locations within close proximity of the driver. This experiment aims to find the most effective and efficient touchscreen position with the least possible discomfort and distraction. In the experiment, participants were instructed by the facilitator to perform a series of different Music, A/C, Seat control tasks on the developed In-vehicle information prototype. The experiment tasks were focused on controlled and real-life open task-based scenarios.

Supply chain for 3D printing

Student: Mudassir Ali - mali11 @ lamar.edu

Advisor:  Dr. Robert Kelley Bradley

The project explores how 3D printing impacts the supply chain with focus on business processes and supplier relationships.

 

Analyzing vessel traffic in waterways using AIS data for design and operational optimization models

Student: Sepideh Zohoori - szohoori @ lamar.edu

Advisor:  Dr. Maryam Hamidi

A spatial-temporal analytical approach is developed to analyze vessel traffic in waterways using AIS data. For traffic performance analysis, a novel algorithm is expanded to prepare AIS data for quantifying practical traffic measures. This algorithm's competitive advantage is to separate vessel trips into three categories: outbound, inbound, and stop state, covering all changing movements between origin and destination. Our proposed algorithm has high computational speed and could handle large size of data in a short time. This analysis shows waterway traffic features, channel capacity, waterway strengths or weaknesses, and potential congestion points, and is useful for traffic performance analysis. 

The Houston Ship Channel is considering expansion at different sections of the waterway. The results of this research can determine which section and restriction is causing a more severe delay for vessels and should be the target of expansion. An optimization model based on the traffic pattered is developed to schedule vessel to optimize the channel transportation efficiency.

Past Dissertations

  • Rail track geometry degradation and maintenance modeling
  • Developing a real time online scheduling system for a manufacturing service company : achieving visibility
  • An advanced capable-to-promise system for hybrid production strategy
  • An analysis of workers' attitudes toward safety culture in maritime industry
  • Study on chemical vessels scheduling problem in a port using MIP, CP, and priority job scheduling heuristics
  • A multi-level modeling and simulation framework for port management
  • Maritime crane near misses and injuries
  • Determining inventories of medical supplies during disaster response operations with a system dynamic approach
  • A multi-level modeling and simulation framework for port management
  • Efficient scheduling for energy saving for food processing industry - a case for an ice cream processing facility
  • Multi-objective optimization for natural gas pipeline network operation
  • Measuring efficiencies and economic impact of air transportation sector in the U.S. economy using data envelopment analysis and Leontief analysis
  • Analyzing brand loyalty in automotive sector using the Hidden Markov Model and Support Vector Machine
  • Cleaning large data sets with a coordinated machine learning and manual approach
  • Developing effective cross tabs to visualize data sets
  • Refinery production planning considering nervousness
  • Analysis and classification of near miss data in the maritime industry, including the use of text mining software
  • A data to text framework for describing of regression models : an optimization approach for content determination
  • Using natural language generation to document portfolio performance : an optimization approach for content determination
  • Tool wear and tool life in micro-milling under dry and MQL conditions
  • Inventory metrics for lead time focused manufacturing