Graduate Theses and Dissertations
Permanent URI for this collectionhttps://hdl.handle.net/10877/135
Browse
Browsing Graduate Theses and Dissertations by Department "Engineering"
Results Per Page
Sort Options
Item A call recognition approach for endangered or threatened chorusing amphibian species using deep learning architectures(2020-12) Islam, Shafinaz; Valles, Damian; Forstner, Michael; Stern, HaroldAudio signal analysis has become prominent in biological domains for detecting endangered or threatened species like Houston toad and Crawfish frog. Researchers at Texas State University and Texas A&M University are working on a project to steward these species and understanding the causes of their decline. The researchers are currently using an Automated Recording Device (ARD), the Toadphone 1, which is an embedded solution. The hardware platform can perform detection tasks without human interruption and can provide near real-time notification. However, this device’s predictive model for the software solution has limited success to serve the primary purpose for which it was developed, which is to provide proper identification of Houston toad calls. Also, the current predictive model for Toadphone 1 was only designed for the Houston toad calls. There is another near-threatened chorusing amphibian, the Crawfish frog, which has become a concern of the researchers working to protect this species.
This thesis research experimented with a modified predictive model for the existing Toadphone 1 software solution, predicting a Houston toad call with decreased false-positive rates. The model can also perform the call recognition task for Crawfish frog calls. This work used the audio data for Houston toad and Crawfish frog collected by the Department of Biology to train the predictive model. Before training, the audio data spectrum was studied to find the frequency range of Houston toad and Crawfish frog call. Next, the audio data have been iteratively preprocessed using digital filters and then applying framing, the Hamming window function to each frame. Mel-frequency Cepstral Coefficients (MFCCs) with their first and second derivatives or Spectral Sub-band Centroids (SSCs) or Mel-spectrograms audio features have been extracted for each frame. These features were used to train the predictive or classification model for Houston toad or Crawfish frog call prediction. Advanced Recurrent Neural Network (RNN) algorithms such as Long Short-Term Memory unit (LSTM) or Gated Recurrent Unit (GRU) and Convolutional Neural Network (CNN) were utilized, which are sub-fields of deep learning network architectures. Several model architectures were experimented with using different combinations of classifiers and audio features with tuned hyperparameters to build the best predictive model. The voting mechanism of ensemble learning was developed to make the final prediction from the three-best models. Lastly, the predictive model was evaluated on a near real-time prediction system.
Item A Comprehensive Economic Analysis of Vertical Farming: Quantitative Models and a Decision Support System for the Competitive Marketplace(2020-08) Moghimi, Faraz; Asiabanpour, Bahram; Ghoddusi, Hamed; Stern, HaroldThere are various problems associated with our conventional practice of farming. In the past few years alone, agriculture has been responsible for a million square kilometers of deforestation. The world is facing a water crisis, and farming is responsible for using 80% of its freshwater. Also, the prospect of global climate change is projecting a much riskier future to practices of conventional farming due to more pesticide incidences, weather uncertainties, changing rain patterns, and more frequent climate extremes. One could argue that these alarming problems might someday be treated as more imminent as the population grows, less fertile land becomes available, and the effects of global climate change become more apparent. Vertical farming solves a lot of the discussed issues associated with traditional farming by using considerably less water, requiring less land, and not relying on the environmental conditions whatsoever. These are all excellent features, but vertical farming is also energy and labor intensive and can be quite expensive in some cases. This study worked to quantitatively model and evaluate the economic prospect of pursuing vertical farming as a business venture in a competitive marketplace under different circumstances. This effort is deeply needed by both the scholarly field and the young industry. This project initially develops a comprehensive stochastic theoretical model to evaluate vertical farming with respect to traditional farming in various conditions. This comprehensive theoretical model is then revised to match the real-world data. The revised model is then utilized to develop a Decision Support System that could help identify the best competitive location alternative to pursue vertical farming as a business practice. The system would account for both profit and risk decision factors as well as stakeholder preferences. Moreover, the developed Decision Support System is employed for a case study to locate the best location alternatives for implementing vertical farming in the US by considering the relative profit potentials and risks in each region. The results from the system identify the most suitable locations to pursue vertical farming as a business venture. These results also contribute towards forming a better understanding of current and future states of the vertical farming industry.Item A Comprehensive Solar Powered Remote Monitoring and Identification of Houston Toad Call Automatic Recognizing Device System Design(2019-08) Bashit, Abdullah Al; Valles, Damian; Forstner, Michael; Aslan, SemihThe Houston Toad is an endangered amphibian living in the edge of extinction. To save from annihilation, localization of their mating calls needs to be determined in order to protect the eggs from being hunted by predators. The current method of monitoring their vocalization lacks real-time and onboard voice recognition capability. In this research, a solar-based battery powered Raspberry Pi is designed along with a microphone that records environmental sound at prescribed intervals triggered by a Witty Pi module. This thesis proposes a naive approach to build a predictor model to detect the Houston Toad mating call signature through recorded audio files from the embedded design. These prerecorded several audio files have been analyzed to determine the unique characteristics of Houston toad call for their identification. The audio file is bandpass filtered, and then preprocessed by multiplying every frame with the hamming window into segments. Next, the Mel-Filterbank and Mel-Frequency Spectral Coefficient (MFCC) are used for feature extraction, and the Support Vector Machine (SVM) and Multi-layer Perceptron (MLP) neural networks are utilized as classifiers to determine the best fit. This experimental result reflects the higher accuracy of the MLP neural network over SVM showing the best potential of classification. The trained neural network predictor model is deployed in the Raspberry Pi to identify Houston Toad, and if there exists trailing in the call, it sends notification of time stamp via Email and SMS over the GSM and GPRS modules to the researchers.Item A Data-Driven Framework for Planning the Expansion of the Trauma Healthcare Network in Texas(2023-08) Saha, Sabhasachi; Perez, Eduardo; Méndez Mediavilla, Francis A.; Valles Molina, DamianTrauma care services are a vital part of all healthcare-based networks as timely accessibility is important for citizens. Trauma care access is even more relevant when unexpected events overload the capacity of the hospitals. Research literature has highlighted that access to trauma care is not even for all populations, especially when comparing rural and urban groups. Historically, the configuration of a trauma system was often not considered but instead hinged on the designation and verification of individual hospitals as trauma care centers. Recognition of the benefits of an inclusive trauma system has precipitated a more holistic approach. The optimal geographic configuration of trauma care centers is key to maximizing accessibility while promoting the efficient use of resources. This research proposes the development of a two-stage stochastic optimization model for geospatial expansion of a trauma network in the state of Texas. The stochastic optimization model recommends the siting of new trauma care centers according to the geographic distribution of the injured population. Data analytics are used to represent the demand for services in different regions. The model has the potential to benefit both patients and institutions, by facilitating prompt access and promoting the efficient use of resources.Item A Digital Twin Framework of a Material Handling Operator in Industry 4.0 Environments(2020-11) Sharotry, Abhimanyu; Jimenez, Jesus A.; Mediavilla, Francis A. Méndez; Rolfe, Rachel M. Koldenhoven; Valles, DamianThe manufacturing and construction industries around the globe have poor occupational health and safety records. Slip & fall, manual material handling (MMH) moves, and forklift accidents are the top three causes for warehouse injuries. Statistics from the U.S. Department of Labor, Bureau of Labor Statistics show that in manufacturing industries, musculoskeletal disorders accounted for 34% of the "Days Away from Work" cases in 2017. Sprains, strains, and tears accounted for the leading type of injury in the manufacturing industry. This research presents a digital twin (DT) approach to assess fatigue in human operators in the material handling industry. DT is an advanced simulation tool that is an exact representation of a physical object. For data collection and analysis, a simulation-based framework is presented. The proposed methodology consists of three modules: Data Collection, Operator Analysis & Feedback, and Digital Twin Development. An optical motion capture system helps develop the DT, which captures simulated material handling activities similar to those in an actual environment. For a pilot study, participants were selected from the university population to perform a series of 'lifting' MMH activities. The participants' physical attributes, body kinematics, and their rating of perceived exertion were measured throughout the experiment. Fatigue was measured as a factor in the subjects' joint angles and analyzed via a dynamic time warping algorithm. To identify the accumulation of biomechanical fatigue, we use an exponentially weighted moving average control chart.
This research aims to conceptualize a DT of an operator and propose a tool that enables the understanding and analysis of the factors that influence human variability and error while performing MMH tasks. The proposed methodology was able to detect biomechanical fatigue in subjects performing MMH tasks and justify the need for a true DT of an operator for fatigue evaluation in the Industry 4.0 era.
Item A Facial Expression Recognition Application Development Using Deep Convolutional Neural Network For Children With Autism Spectrum Disorder To Help Identify Human Emotions(2019-08) Haque, Md Inzamam Ul; Valles, Damian; Resendiz, Maria; Koutitas, GeorgeIn this thesis, a novel idea is presented, which is to teach young children with Autism Spectrum Disorder (ASD) to recognize human facial expressions through the help of computer vision and image processing. Universally, there are seven facial expressions categories: angry, disgust, happy, sad, fear, surprised and neutral. To recognize all these facial expressions and to predict the current mood of a person is a difficult task for a child. For a child with ASD, this problem presents itself in a more complex manner due to the nature of the disorder. The main goal of the thesis was to develop a deep Convolutional Neural Network (DCNN) for facial expression recognition, which can help young children with ASD to recognize facial expressions, using mobile devices. Previously, different neural network models and classifiers have been presented to achieve state of the art accuracy in this sector. Separately, different studies have been performed in studying the ability and performance of children with ASD for recognizing facial expressions. In this thesis, additional features have been added to the DCNN model such that it can correctly classify facial expressions in different lighting conditions and from different viewpoints as the model is trained to do so. Upon developing the DCNN model, an iOS app has been developed implementing this deep learning model as a byproduct and as a medium to use this model in clinical trials for children with autism as a medium of enhancing their communication abilities. The implementation of this proposed idea started with finding datasets containing images of faces with different expressions from different angles.
Further datasets were produced from the original dataset with images of different contrast and brightness with the help of image processing. The performance of the DCNN model was evaluated using these datasets. Once an optimal accuracy is achieved with good generalizability, an app suitable for iOS platform was developed for running both the DCNN model and image processing algorithms. The function of the app is to open the camera of the device, detect a face, classify the facial expression, and show the expression with an emoticon on the screen. As a product of this work, the app can be used by speech-language pathologies, teacher, care-takers, and parents as a technological tool when working with children with ASD. The design of the model and application is targeted to children with ASD to recognize and identify facial expressions in real-time to practice social skills during everyday social interaction.
Item A Game Theoretic Framework to Secure Cyber Physical Systems (CPS) against Cyber Attacks(2018-12) Siddique, Khan; Novoa, Clara; Guirguis, Mina; Perez, EduardoCyber-Physical Systems (CPS) is a term describing a broad range of complex, multi-disciplinary, physically-aware next generation engineered systems that integrate embedded computing technologies (cyber part) into the physical world. CPS are engineered systems that are built from, and depend upon, the seamless integration of computational algorithms and physical components [2]. Generally speaking, they are sensor-based communication-enabled autonomous systems. Wireless sensor network for environmental control, smart grid system and industrial robotics systems can be a good example of CPS. With the exponential growth of CPS, new security challenges have emerged. Various vulnerabilities, threats, and attacks have been detected for the new generation of CPS. Additionally, the heterogeneity of CPS components and the diversity of CPS systems have made it very difficult to study the security problem with one generalized model. This thesis focuses on the development of effective deterministic and stochastic mathematical programming approaches to protect the CPS against a wide range of cyber attacks. The primary goal of this work is to orchestrate an optimization methodology based on a game theoretic framework to protect the CPS and evaluate its results using a simulation model and a real world testbed. To assert that the game theoretic framework yields to an optimized performance, three other heuristic approaches (i.e. Greedy, Greedy-LP, Random) are formulated and their results are compared to the outcome from the game theory approach. The game theoretic model was further extended to include stochastic number of signals and stochastic effectiveness. A two-stage stochastic model was formulated and the results were compared. Further investigations included simulation of a real world system. The simulation model was coded in MatLab Simulink to emulate a real world CPS. As a final step in this thesis, a real life CPS testbed was constructed with functioning cyber and physical components and the results from the different approaches studied are tested and compared. It has been found that the two-stage stochastic programming (two-SSP) model gives most optimized result to protect CPS.Item A Machine Learning Based Victim's Scream Detection System for Burning Sites Using an Autonomous Embedded System Vehicle(2021-11) Saeed, Fairuz Samiha; Valles, Damian; Viswanathan, Vishu; Stapleton, WilliamFire incidents are responsible for severe damage and thousands of deaths every year all over the world. Extreme temperatures, low visibility, toxic gases, and unknown locations of victims create difficulties and delays in rescue operations, escalating the risk of injury or death. It is time-critical to detect the victims trapped inside the burning sites for facilitating the rescue operations. Since human beings tend to scream for help in emergency situations, this type of audio events can play a crucial role to detect victims trapped inside burning sites. The information regarding victim’s presence can help the firefighters to make a faster and a safer rescue plan. This research work presents an audio-based automated system for victim’s scream detection in fire emergencies, through the investigation of three machine learning (ML) approaches: Support Vector Machines (SVM), Long Short-Term Memory (LSTM) and transfer learning with Yet Another Mobile Network (YAMNet). The performance of these three techniques has been evaluated based on a variety of performance metrics. The models with top performance on scream detection were implemented in an Autonomous Embedded System Vehicle (AESV). This research work also presents the performance analysis of these models in field testing. The main objective of this thesis is to develop an autonomous victim detection system in burning sites from scream sounds of victim(s) for effective rescue operation.Item A multi-stage stochastic model for production planning using onsite renewable generation with prosumer approach(2020-12) Ayuwu, Atamgbo; Clara, Novoa; Jin, Tongdan; Zhu, EmilyThis thesis researches on finding a production plan that minimizes the cost of a manufacturing system facing uncertainties on the demand of its final products over a horizon of multiple periods and considering adoption of renewable power as an energy prosumer (i.e. consumer and seller). Researched energy sources are wind turbines and solar photovoltaics coupled with energy storage systems (i.e. batteries). Renewable generation varies because of daily changes in wind speed and weather conditions. To account for the uncertainty on products demand and power supply, a multi-stage stochastic programing model is proposed. First-stage decision variables are the size of the renewable generation technologies, capacity of the batteries, and amount of production for the first set of periods. Second-stage recourse actions to cope with the uncertainty include: (1) storing final products in inventory or purchasing from vendors, as needed, (2) using battery to discharge or store energy and (3) purchasing/selling energy to/from the grid. In the second-stage, a new production decision for the second set of periods is also determined considering the inventory levels, production and purchasing costs. The third-stage includes deciding again on the best recourse actions to the second-stage decision. The model is implemented using the scenario-tree approach, and it is solved under two operation strategies: (1) factory and warehouse consolidated in Amarillo and (2) factory in Amarillo and warehouse in Phoenix. Numerical experiments show that a prosumer microgrid model is cost-effective (annual cost $7,052,410, levelized cost of electricity (LCOE) $37/MWh) if compared to an island microgrid model (annual cost $15,150,000, LCOE $70/MWh). Due to high battery costs, the prosumer option reduces amount of battery capacity adopted and purchases some energy to the grid to save cost.Item A Pilot Study to Formulate Data-Driven Worker Fatigue Models of an Order Picking Operation(2022-05) Suresh, Venkataramanan; Jimenez, Jesus A.; Mendez Médiavilla, Fancis A.; Farrell, John; Russi-Vigoya, M. NataliaFatigue is one of the significant issues manual workers face while performing highly demanding physical tasks. An industrial environment where the workers must perform repetitive manual material handling tasks leads to fatigue. Human fatigue can be physical or mental and leads to musculoskeletal disorders (MSDs) and injuries and a reduction in productivity at the workplace. This research presents an approach to formulate data-driven models to predict worker’s fatigue levels while performing manual tasks using a digital twin (DT) framework. DT is a simulation tool used to represent a physical entity virtually. An order-picking activity involving a combination of manual tasks like picking, carrying, and placing was designed and simulated using human subjects in a bio-motion capture environment. The subject's motion and physiological data were collected using motion capture (MoCap) technology and Hexoskin suit. Cognitive Stroop tests were conducted during the task, and the subject's reaction time for the tests was recorded. The study used Borg's scale to indicate the subject's self-reported exertion levels while performing the task. Using physiological factors like heart rate, breathing rate, minute ventilation, biomechanical factors like body joint angles, and cognitive reaction time, three individual and one combined multiple linear regression model was derived to predict the self-reported exertion level of the order picker. The models were compared. The results show the statistical significance and residual errors of all the models. The proposed methodology using a DT framework was able to predict the self-reported exertion levels of the order picker. This research has the potential to contribute to the field of ergonomics and manual material handling industries to help schedule and assign work to the industry workers efficiently.Item A Robust and Power Efficient Software Encryption Method for IOT Framework Communication Using Zigbee Protocol via XBees(2023-08) Obisakin, Inioluwa; Aslan, Semih; Droopad, Ravi; Valles, DamianIn this study, our focus is on Zigbee, a wireless communication standard that uses IEEE 802.15.4, a prevalent short-range wireless communication standard, for both indoor and outdoor applications. The effectiveness of Zigbee relies on various networking parameters, including transmission distances, deployment environment, hopping, baud rates, and transmission power. Zigbee security and data encryption is based on security defined in the 802.15.4 protocol. The encryption algorithm used in Zigbee is a network-level symmetric AES (Advanced Encryption Standard) with a 128-bit key length. However, newer versions of AES (196 & 256-bit) are more substantial, and asymmetric encryption methods are better suited for systems with a sender and receiver. As such, this study aims to determine a robust application-level encryption method for the Zigbee protocol by examining standard IoT transmission parameters, such as Received Signal Strength, latency, and packet delivery ratio, as well as its power efficiency by examining power consumption during various stages of the Zigbee transmit and receive cycle at different transmit power levels. To achieve this goal, we have implemented AES-256-bit(symmetric) and Public Key Cryptography (asymmetric) encryption methods and compared their performance to the existing AES 128-bit encryption method. Our findings reveal that AES-256-bit encryption leads to higher power consumption, which can be mitigated by adjusting the transmit power levels. On the other hand, PKC encryption provides a better solution in terms of power efficiency, although it is slower in terms of encryption speed and higher latency. This study's contribution lies in its attempt to understand the performance tradeoffs involved in the security of Zigbee networks by proposing application-level encryption methods and benchmarking it against the default encryption. The findings of this study can assist developers and researchers in making more informed decisions when selecting encryption methods for Zigbee networks.Item A simulation-based data-driven analysis for improving collaboration between food bank facilities before and after natural disasters(2021-12) Kothamasu, Monica; Pérez, Eduardo; Méndez Mediavilla, Francis A.; Jiménez, JesusFood banks are non-profit organizations that gather and distribute supplies to agencies serving people in need. Food banks obtain donations (i.e., supplies) from the public and public and private organizations. Catastrophic events like hurricane Harvey leads the way to complexities for serving people in need. Food banks act as food storage and distribution depots for smaller front-line agencies and usually do not give out food directly to people struggling with hunger. The demand for supplies needed by supporting agencies is very dynamic and challenging to predict. The demand for supplies required to collaborating agencies is very dynamic, especially after natural disasters when food banks become significant players in disaster relief efforts. Therefore, planning for the operation of food banks under both normal and disaster relief conditions is a challenging problem. This research aims to develop data-driven models for improving the operation of a network of food bank facilities in charge of providing relief after natural disasters. This research provides a methodology for achieving the defined objective by developing simulation and optimization models for decision-making purposes.
The first part of this thesis is to develop a discrete-event simulation model of a network of food banks to investigate the impact of multiple disaster relief operational policies (i.e., supply prepositioning, distribution center assignment). The model simulates the flow of donations at three food bank facilities and the demand for supplies of 55 demand locations before and after a natural disaster. The simulation model is validated with the real-time data collected at the food bank facility. Discrete-event simulation model experiments are conducted from the results of the stochastic model developed in the second part of the thesis. The value of the simulation model is demonstrated through the analysis of 21 scenarios, five optimization-based demand distribution policies, and six performance measures: unmet demand, total demand met, daily number of trips, delivery cycle time, demand fulfilment rate and the order fulfilment rate. The results of the simulation model show that there is a 20% increase in the overall demand fulfillment rate if food banks operate as an integrated network with supply prepositioning and demand splitting between operating facilities.
The second part of this thesis is to develop optimization-based decision-making policies for pre-positioning disaster relief supplies considering the transportation limitations of network food bank facilities and the uncertainty in demand. The stochastic programming model determines the best distribution decisions considering supply chain disruptions after hurricane events. The first stage models the pre-positioning of supplies between food banks, while the second stage provides recursive actions for supplies prepositioning and models supplies distribution to the demand nodes under different scenarios. The developed stochastic model will identify the least-cost strategy associated with pre-positioning existing supplies that will satisfy the demand needs after a natural disaster by considering various parameters like storage cost, transportation cost, truck availability, and docks availability.
Item A Smart Circular Economy for Integrated Organic Hydroponic-Aquaponic Farming(2023-05) Chowdhury, Hribhu; Asiabanpour, Bahram; Chen, Heping; Huertas Pau, MarWith the global problem of population growth and climate change, current challenges to traditional agriculture include a long-term decline in the available arable land, soil depletion, water scarcity, and food security. To mitigate these challenges, agricultural methods must undergo a significant transformation to become more efficient and environmentally sustainable. Soilless agriculture particularly, vertical farming provides the best possible solution to these growing problems. With the growth of vertical farming, it has been evident that many investors have gone out of business because of high investment costs, high maintenance costs, high energy costs and high labor costs, and high waste generation. This research expanded the integrated organic hydroponic-aquaponic farming system and investigated the feasibility of a zero-waste circular economy to find parameters for the profitability of the system through zero waste and reduced operational costs, byproduct utilization, balancing the capacities, and product selection. The research utilized technology, technical knowledge, and the proposed integrated hydroponic-aquaponic circular model to convert the traditional vertical farming method to a zero-waste circular economy opportunity by using hydroponic plant waste, aquaponic waste, vermicompost tea, and aerobic compost tea/quick compost on the growth of halophyte plants. This study proposes a circular economy model for vertical farming that can contribute to the development of a sustainable agricultural system. The simulation study justified the research objectives and demonstrated the feasibility of the proposed model. The model has the potential to revolutionize the agricultural sector by providing a more efficient and sustainable approach to food production, resource conservation, and economic growth through minimizing waste and maximizing the use of resources. Overall, the results of this study provide valuable insights into the development of a more environmentally friendly and economically viable vertical farming process.Item A Transactive Energy Approach for Design and Operation of Battery Swap and Supercharging Infrastructure(2021-05) Saha, Moumita; Jin, Tongdan; Dong, Sasha; Rosas-Vega, RosarioTransactive energy refers to the planning and control of the two-way energy flow between distributed generation and main grid in regards to the realization of economic benefits. This study addresses two research questions related to transactive energy operations: first, how to allocate renewable microgrid system to energize the battery swap and supercharging stations under demand and supply uncertainty? Second, is it economically feasible for wind turbines (WT), solar photovoltaics (PV), and energy storage system (ESS) to participate in day-ahead transactive energy market as virtual power plants? An optimization framework for sizing and siting WT, PV and ESS in a battery swap and supercharging network considering both island and grid-tied operations is proposed. Mixed integer linear programming models are formulated to minimize the annualized battery service infrastructure cost considering facility setup, spare batteries, and supercharger installs. For island microgrid, reducing the cost of ESS does not significantly stimulate its adoption because renewable generation largely depends on capacity factor of PV and WT is shown. In grid-tied microgrid operation, reducing the PV cost by 50% makes the system to install more panels in both sunny cities and windy cities is shown. For network model, the work shows that by reducing the PV capacity cost by 75% from the benchmark cost makes the system choose more PV for Texas cities and reduces the annual network cost by 29%. The system opts to behave as “prosumer” who fulfills the charging demand of vehicle fleet as well as enhancing grid reliability and security by participating in transactive energy market.Item A Victims Detection Approach for Burning Building Sites Using Convolutional Neural Networks(2019-12) Jaradat, Farah Bilal; Valles, Damian; Stapleton, William; Koutitas, GeorgeIn this work, an approach is proposed to detect people and pets trapped in burning sites. This will be done in a manner that will ensure the safety of firefighters and accelerate the rescue process of victims. The proposed work suggests detecting victims in fire situations autonomously, through a deep learning technique using the Convolutional Neural Network (CNN) model. Research work in the field of firefighters’ assistance suggest using infrared (IR) sensors, which are widely used in this field due to their functionality in fire situations, since such cases are associated with high levels of smoke, that limit the vision as well as the feature of white light cameras since the smoke particles, reflect the light. Thus, thermal sensors found to be a convenient choice for gathering data for detecting victims in the burning site without distortion of smoke. Despite the huge progress is the field of firefighting assistance due to the huge technological advancement, however, deciding if victims exist inside the burning structure without human interference has remained a challenge. Here deep learning model is introduced to identify victims’ appearances in IR images. What distinguishes this work is leveraging IR imaging combined with deep learning models, which were proved to be the state-of-the-art techniques in the field of people detection. The proposed method is extended to detect pets in the burning sites. The thesis work provides a deep learning approach that speeds up and facilitates the process of rescuing victims in vague burning sites layouts; hence, more lives would be saved. The objective of the CNN development is to classify input images sent from the burning site into one of three outcomes classes: “people,” “pets,” and “no victims.” First responders can leverage this information to define their priorities regarding the locations they should target first. Deep learning is a popular method for human detection in the IR image, and it is applied extensively to the field of pedestrian detection and surveillance. However, the literature scarce for applications in fire assistance fields. One reason behind this, is the unavailability of public IR datasets, including infrared images of people and pets. Therefore, a dataset was developed to accomplish the task of human detection in IR images. Furthermore, the dataset was processed to mimic high-temperature environments as in building on fire situations. In this thesis, a cascaded CNN architecture approach was implemented and benchmarked, and the cascaded model consists of two stages to accomplish the task of victim detection in IR images.Item Accident Debris Detection with UAV Using Deep Learning and Estimate Debris Perimeter(2021-11) Alam, Homayra; Valles, Damian; Dong, Zhijie Sasha; Stapleton, BillRoad debris clean-up process can be improved by the utilization of drones, Deep Learning, and detection to optimize the operation and re-open roads for traffic. Common debris is unsecured items that fly out from vehicles after an accident. The cleaning procedure of the road debris after an accident is cumbersome and sensitive. It demands much workforce and a time-consuming process to haul debris properly. The project aims to detect debris on the road using a drone to minimize the time cleaning procedure by generating a perimeter in correlation with GPS information. This thesis provides a framework for development of a Deep Learning model with Computer Vision using feature extraction and object recognition to detect debris and calculate the perimeter area. Currently, clean-up crews perform their task by observation to identify debris objects in different weather conditions. Debris can be comprised of dangerous chemicals, vehicle fluids, pesticides that may cause specialized crews for proper removal and delay the re-opening of roads. In contrast, the drone can fly faster and detect desired objects using the DL model more precisely. The perimeter information from the drone’s data will also optimize the cleaning procedure, improve time, and clean efficiency.Item Additive Manufacturing of High Temperature Thermoset Composites(2022-05) Omer, Liam Michael; Tate, Jitendra; Talley, Austin; Hacopian, EmilyHigh temperature materials (HTM) are an invaluable part of multiple industrial processes such as plastic injection molding surfaces, airplane turbine blades, and a variety of other nonuniform molding surfaces. The composites used in these applications requires work intensive processes using reinforcement systems that inhibit freeform geometry and through thickness variation in material properties. Additive Manufacturing offers the potential to break through the barriers of state-of-the-art (SOTA) materials by offering freedom of design and function. Additive manufacturing reduces production time by providing a model to product workflow with limited requirements for tooling and machine setup in advance. However, current work in printing thermoset composite parts is limited. This research proposes the creation of an additive manufacturing process for high temperature thermoset composites based upon UHTR, Milled Carbon Fiber, and Graphitized Microballoons. The material system will introduce a novel curing process using high frequency inductive heating of fillers to provide an even, effective cure without the need for post processing.Item Advanced ZigBee Network With Greater Range and Longevity(2021-12) Shifa, Tasneem Khan; Stern, Harold; Compeau, Rich; Stapleton, WilliamWith the rapid development of Internet of Things networks (IoT), a new type of wireless standard, ZigBee, has emerged in order to satisfy the demand of low power dissipation, low cost, and easy deployment among wireless communication devices. In the currently developed ZigBee system, all transmitters use a common spreading code which can result in a large number of message collisions. A promising new system using multiple spreading codes has previously been proposed to increase system throughput, reduce collisions, and increase energy efficiency or range, but it has only been evaluated with constant message lengths and single hop topology. Systems with such restrictions represent only a small subset of IoT networks. For our research, we aim to evaluate the system with variable message length and multiple hopping topology. We will consider a large network with many sensors which are out of the limited range of the coordinator but which can transmit messages through a router, which involves two hops. Therefore, our new proposed multiple hop system has larger range and longevity compared to the single hop proposed system and reduced collisions compared to the current ZigBee system.
We have implemented the code in MATLAB and run multiple simulations in terms of varying the amount of message traffic, message length and number of CAP slots. By comparing each data set and its graphical representation, the results show that our new proposed system has higher success rates than the current system. Our findings determine suitability for a much larger set of IoT systems and applications and may suggest protocol changes that can produce further improvements to increase reliability and security, range, operating life, and throughput of ZigBee systems. This will be significant for enabling new applications and attracting more customers. So, with the design of high-performance ZigBee wireless communication networks, it will have a broad application space in real life.
Item All Inkjet-Printed High On/Off Ratio Two-Dimensional Materials Field Effect Transistor(2018-08) Jewel, Mohi Uddin; Chen, Maggie Yihong; Droopad, Ravi; Yu, QingkaiThis thesis introduces the development of a novel ink, design, fabrication, and characterization of an all inkjet printed high current on/off ratio field effect transistor (FET). The inks were obtained through the liquid phase exfoliation of nitrogen-doped graphene (NDG), and molybdenum disulfide (MoS2) nanosheets into appropriate solvents. A stable and efficient method of inkjet printing is developed for NDG nanosheets. The concentration of nanosheets and the presence of MoS2 were determined from UV-Vis spectra of the inks. The morphology of percolation clusters using NDG was studied using the thickness profile and scanning electron microscopy (SEM) images. The solvent-induced defects in NDG nanosheets were characterized by Raman spectroscopy. There were little or no solvent-induced defects in the nanosheets recovered by curing after printing. Barium titanate (BaTiO3) was prepared and used as a high k (~20.5) dielectric for the printed transistors. The NDG transistors were designed, fabricated, and characterized on the glass substrate. Due to the low on/off ratio of NDG transistors, NDG thin films were electrochemically doped with MoS2 by multiple printing passes. The incorporation of semiconducting MoS2 into NDG was confirmed by energy dispersive spectroscopy (EDS) for further analysis. A transistor with high current on/off ratio was obtained by NDG-MoS2 heterostructures channel. To our best knowledge, this is the highest on/off ratio for a fully inkjet printed transistor based on 2D materials.Item An augmented reality facet mapping technique for ray tracing applications(2018-12) Siddaraju, Varun Kumar; Koutitas, George; Aslan, Semih; Valles, DamianThis research presents a novel spatial mapping technique that is capable of extracting the vector map of an indoor environment based on the images captured from a smartphone camera and the spatial maps captured from the Microsoft HoloLens. The extracted vector map follows the facet model concept and can be used as input in ray tracing algorithm. The ray tracing algorithm is used for visualizing and predicting the indoor wireless channels. The proposed solution offers three different algorithms, the first algorithm (Low cost 2D image to facet model algorithm) uses the edge and corner detection algorithms to compute the coordinates of the walls and doors of the indoor environment. The second algorithm (Minimum- maximum algorithm) computes the spatial map corner vertices by using the data processing techniques. The third algorithm (Spatial understanding algorithm) uses the Microsoft HoloLens Toolkit’s “spatial understanding” feature to compute the spatial maps for detecting and measuring the individual wall dimensions. Finally, using the corner coordinates, spatial corner vertices and individual wall dimensions from all the three algorithms, a simple 3D vector map is designed. The output of all the algorithms is a facet model that can be used by ray tracing algorithms which are embedded in Augmented Reality (AR) applications. The overall process provides a better human-to-network interface and an improved user experience that is expected to provide a new way for indoor network planning of residential 5G systems.