136879 research outputs found
Sort by
Assessment of Corrosion Prevention and Mitigation Techniques for Concrete Bridge Decks in Texas
No full textCorrosion in reinforced concrete is a self-accelerating process. Once rust forms on steel due to cracks and environment factors, it expands in size, causing larger cracks. The corrosion in reinforced concrete bridge decks has resulted in significant economic losses. To repair and maintain the durability of bridges, the Texas Department of Transportation spends millions of dollars every year. This study provides a comprehensive study of corrosion mitigation methods, such as high-performance concrete (HPC), calcium nitrite corrosion inhibitor (CNI), linseed oil, silane, and epoxy-coated rebar, for concrete bridge decks throughout Texas. To assess the performance of the mitigation methods, 61 concrete bridges across Texas were selected for both field evaluation and laboratory experiments using concrete cores extracted from the bridges. The field evaluation included visual inspection and non-destructive evaluation, such as infrared thermography, ground penetrating radar, ultrasonic tomography, half-cell potential, and corrosion rate mapping. In the laboratory, experiments included surface resistivity, bulk resistivity, ultrasonic pulse velocity, water absorption, chloride content measurement, carbonation depth measurement, coating adhesion testing, and electrochemical impedance spectroscopy. Using the results from both field and laboratory evaluations, a decision-making tool for selecting corrosion mitigation methods in reinforced concrete, based on different environment conditions of Texas, was developed. The tool seeks to contribute to the formulation of a comprehensive strategy to control corrosion and reduce corrosion-related financial losses as much as possible. This research aims to provide various approaches for evaluating the performance of different mitigation methods applied to concrete bridge decks, as well as to determine the durability and effectiveness of the mitigation methods that have been in use on real bridges for an extended period
Systematic Uncertainty Quantification of MCNP Predicted Nuclide Concentrations in Fuel Burnup Simulations
No full textMonte Carlo N-Particle transport code (MCNP) is often used to simulate nuclear fuel burnup and depletion because it is efficient in solving the radiation transport equation for complex geometries. MCNP simulates fuel burnup and estimates the concentrations of actinides and fission products generated in the fuel, which are useful in nuclear forensics as well as safeguards monitoring. During fuel burnup simulations, the uncertainties in the predicted nuclide concentrations due to the uncertainty in the nuclear data used by MCNP are not propagated and predicted. The nuclide concentration is calculated through CINDER 90 isotope generation and depletion module in MCNP. The CINDER90 module uses the neutron reaction rates and flux values computed by MCNP for each burnup time step. The reaction rates can be broken down into three terms: neutron flux, number density of the target isotope that is transmuting, and microscopic neutron interaction cross section. The number density and neutron flux are provided by MCNP; however, the microscopic cross sections are not directly provided by MCNP in the output and will contain systematic uncertainty in varying degrees depending on the microscopic cross section of the target isotope of interest. Systematic uncertainty is not propagated through each MCNP burnup time step. Propagating the effects of systematic uncertainty using a Backward Euler numerical scheme allows for the reporting of the systematic relative error in the predicted nuclide concentrations, which the study undertaken in this thesis. This Backward Euler methodology was executed through python scripting and a program was developed to output the systematic relative error for user desired isotopes of interest utilizing on the results of MCNP fuel burn up simulation. It was concluded that the Backward Euler methodology and the Bateman equations successfully replicated the MCNP estimated concentrations given the appropriate one group cross sections. Additionally, it was determined that for select isotopes of interest the systematic uncertainty for the associated concentration can be estimated
Data Modeling, Computing, and Generation: New Techniques by and for AI
No full textThis dissertation investigates approaches in data handling within the domain of Artificial Intelligence (AI), covering data modeling, computing, and generation. It explores four primary tasks, each addressing distinct challenges and presenting novel solutions in their respective fields.
In the realm of data modeling, the Side Information Boosted Symbolic Regression (SIBSR) and Symbolic Modeling techniques are introduced. SIBSR incorporates side information into the symbolic regression process, enhancing the search for accurate mathematical relationships in complex datasets. Symbolic Modeling extends this approach to multi-dataset scenarios, particularly in financial asset pricing, providing adaptable and interpretable models that capture the dynamics of financial markets.
For data computing, the focus shifts to neuromorphic systems with the analysis of new Analog Error-Correcting Codes (ECCs) and the design of neural network-based decoders. These advancements address the challenges of reliability and accuracy in analog data processing, marking a progression of error-correcting from digital to analog and benefits in neuromorphic computing environments.
In data generation, Reinforcement Prompting, a novel methodology that leverages Large Language Models (LLMs) for the generation of synthetic data, is proposed. This approach mitigates issues of data privacy and scarcity of labeled datasets, especially in the finance domain. This method demonstrates that models trained on the generated synthetic data maintain performance integrity comparable to those trained on real financial data.
The dissertation presents a comprehensive exploration of these methods, substantiated by experimental evaluations and theoretical analysis. The research contributes to the advancement of AI in data handling, offering new perspectives and tools in data modeling, computing, and generation. The findings underscore the transformative potential of AI in understanding, processing, and generating data more effectively and ethically across various domains
The Impact of Preovulatory Estradiol on the Oviductal and Uterine Environments, and Profit per Pregnancy Associated with the Detection of Estrus
No full textThe expression of estrus and preovulatory estradiol concentrations influence pregnancy success in beef cattle; however, it is not clear how estrus and/or estradiol impact oviductal and uterine environments. The goals of this dissertation were to evaluate how preovulatory estradiol impacts oviductal gene and protein expression, how physiological estradiol exposure with/without estrus impacts pregnancy-associated factors, and how utilizing the detection of estrus with timed-artificial insemination (TAI) impacts expense and profit per pregnancy. In Chapter II, oviducts (n=6) were collected from synchronized beef cows with High (10.12��0.62 pg/ml, n=3) or Low (5.97��0.62 pg/ml, n=3) estradiol concentrations at fixed-time artificial insemination (FTAI), and differentially expressed genes were identified. Oviductal gene expression was hormonally regulated, with differential gene expression between High and Low cows. In Chapter III, oviducts and ovaries were collected from synchronized beef cows with High (���6.0 pg/mL; n=4) or Low (���4.5 pg/mL; n=5) estradiol concentrations. Oviducts were fixed and embedded for localization of prostaglandin E2 synthase (PTGES) and prostaglandin E2 receptor 2 (PTGER2) localization by immunofluorescence. There was no difference in immunoreactivity of PTGER2 protein; however, immunoreactivity of PTGES protein tended to be more intense in oviducts from High compared with Low cows. In Chapter IV, beef cows (n=603) were synchronized and grouped based on estrus by FTAI (day 0), or the administration of gonadotropin-releasing hormone (GnRH) and/or exogenous estradiol, and estrus by day 7. While estrus or physiological estradiol did not impact pregnancy or interferon-stimulated gene expression by day 19, pregnancy rate and pregnancy-associated glycoprotein abundance by day 24 differed between cows that did or did not express estrus by day 7. Differences in day 55 and 90 pregnancy were observed due to estrus by day 7, but not due to estradiol exposure without estrus. In Chapter V, beef cows and heifers were synchronized using a TAI protocol with (6d; n=437) or without (7d; n=429) estrus detection. Pregnancy rates did not differ; however, decreased expense and increased profit per pregnancy were observed with the 6d compared with the 7d protocol. These data indicate estrus supports pregnancy through maternal environment changes, and utilizing detection of estrus with TAI is more economical
OSHA Citations and Beyond: Strategies for Ensuring Combustible Dust Compliance
No full textManaging dust in the ever-changing world of industrial operations cannot be emphasized enough. Dust, frequently observed due to numerous industrial processes, poses several complex challenges that require careful consideration and proactive control. The appropriate handling of dust emerges as a crucial component of industrial processes, from the standpoints of safety and health to the optimization of operational efficiency. This project outlines the OSHA (Occupational Safety & Health Administration) citation data that was gathered against the citations obtained in response to the "dust" query from 2019 to 2023. The data from the citations was studied to understand which type of industries had received the most citations from OSHA. Furthermore, the equipment of concern was studied and analyzed, and the citations received from 2019 to 2023 were then compared with 2018, explaining the relevance of the shifting industries, citations, and industrywide concerns about handling dust within the above-mentioned timeframe. In total, there were 45 citations related to dust in 2018, which decreased over the years. In 2023, 42 citations were found under the general duty clause in the term "dust." There were also some citations where NFPA (National Fire Protection Association) standards were outlined, referred to, and explained. In the years considered in the project, over 95 percent of citations were classified as ' serious, ' with scarce instances of 'failure to abate' and ���willful��� violations. All penalties were compared case-wise, industry-wise, and year-wise to understand the cost relationships with incidents. New citations after the year 2018 were also assessed and evaluated. Further measures and standard requirements to ensure safety compliance against dust hazards were discussed in detail. This project emphasizes the critical significance of managing dust in industrial operations, highlighting its multifaceted challenges and pivotal role in safety, health, and operational efficiency. Through a detailed analysis of OSHA citation data from 2019 to 2023, industry trends, equipment concerns, and penalty assessments were investigated, providing a comprehensive understanding of the evolving landscape and imperative measures for ensuring safety compliance in handling dust hazards
Fetal and Infant Mortality Review (FIMR) Programs and Infant Mortality Outcomes
No full textInfant mortality is defined as the death of a baby before his or her first birthday. The United States had the highest infant mortality rate (IMR) of any high-income country at 5.4 deaths per 1,000 live births in 2020. Having a substantial number of poor birth outcomes in a community is a multi-faceted problem and often requires an interdisciplinary approach to understand the challenges impacting fetal and/or infant death. One such coordinated approach to address this problem is the Fetal & Infant Mortality Review (FIMR) program.
Despite the proposed function of FIMR programs to combat high infant mortality (IM), there is no published evidence linking the existence of FIMR programs to improved IM outcomes. Existing published research has been useful in clarifying what FIMR is in general terms.
In Chapter 2, a scoping literature review provides a focused approach to determining the relationship between FIMRs and intended outcomes of reduced IM by reviewing the literature for existing studies on FIMR and IM outcomes. Out of 97 screened articles, 12 were empirical articles on FIMR programs, and 7 articles included an evaluation of a FIMR program or process and were included in the review.
In Chapter 3, a quantitative study assesses differences in IMR in communities with an existing FIMR program examining IMR pre- and post FIMR implementation. Results demonstrated a decrease in IMR in communities post-FIMR implementation thus showing that there appears to be an association between FIMR programs and IM outcomes.
In Chapter 4, a qualitative study evaluates the FIMR process, how it works, and what makes it effective or ineffective. Eleven people participated in the virtual one-on-one interviews. A phenomenological approach was used, and findings revealed reported benefits of the FIMR program as well as areas for needed improvement.
This research concluded that more evaluative studies are needed for assessing FIMR program outcomes specific to IM, not solely its processes; and a nationally standardized approach for the operation of FIMRs is needed with room to tailor to specific communities. Findings need to be disseminated widely and used as an opportunity for greater accountability and improvements to FIMR programs
John Bickham field notebook: AK4001-AK4500.pdf
No full textBound book, each page corresponds to a karyotype slide data.Data pages for AK4501-AK5000 corresponding to unique identifiers of specimens/samples examined for biological research. Specimens are primarily housed at Texas A&M University; Biodiverstiy Research and Teaching Collection
Advancing Stress Detection: Towards Real-Time, Naturalistic, and Personalized Artificial Intelligence Approaches for Stress and Mental Disorder Detection
No full textThis dissertation presents a comprehensive exploration of Machine Learning (ML) and Deep Learning (DL) methodologies for the detection, prediction, and analysis of stress and stress-related mental disorders (MDs). The research begins with a systematic review of existing ML algorithms, preprocessing techniques, and data types used in stress detection, highlighting the superior performance of Support Vector Machine (SVM), Neural Network (NN), and Random Forest (RF) models. The review also underscores the prevalent use of physiological parameters, such as heart rate measurements, as stress predictors, and identifies significant research gaps, including model interpretability, personalization, the incorporation of naturalistic settings, and real-time processing capabilities.
Addressing these gaps, the dissertation presents a series of studies. The first study explores a machine learning-based method for continuous monitoring and identification of stress in a naturalistic setting among college students, using a mobile health (mHealth) application and wearable wrist-worn sensors. The XGBoost model was found to be the most reliable in this context, with an accuracy of 84.5%.
The second study proposes a stress detection methodology that utilizes an array of deep learning models, including feedforward neural networks (FFDNN), Conv1D, long short-term memory (LSTM), hybrid Conv1D-LSTM, and Conv1D-LSTM with attention mechanism. The hybrid Conv1D-LSTM model augmented with an attention mechanism outperformed the other models, achieving a cross-validation accuracy of 0.89 and an area under the curve (AUC) score of 0.94. The final study explores the potential of individualized ML and DL algorithms in accurately identifying stress responses in individuals afflicted with Post-Traumatic Stress Disorder (PTSD). The findings demonstrate that individualized models, particularly advanced deep learning models such as LSTM and hybrid CNN-RNN models, consistently surpass standard models in accurately identifying stress in PTSD patients.
In conclusion, this dissertation underscores the importance of employing advanced ML and DL techniques, such as hybrid models and attention mechanisms, and individualized approaches to improve stress detection performance. The findings contribute significantly to the field of digital health, suggesting innovative strategies for monitoring and managing stress in various populations
Evaluation of Multiple Approaches for Solidification Modeling of Advanced Nuclear Reactor Coolants
No full textThis dissertation focuses on developing and validating numerical tools for modeling solidification phenomena in advanced nuclear reactor coolants. The primary objective is to offer cost-effective alternatives to experimental studies and enable rapid simulation of multiphysics reactor transients. To achieve this, a hierarchical structured framework is established. High-fidelity models or experiments inform intermediate-fidelity data, which, in turn, provide training data for multidimensional coarse-mesh models capable of running on a single core and delivering fast results. The first phase of this research focuses on developing a solidification model using the Lattice Boltzmann Method (LBM) for 3D internal transient solidification modeling of Generation-IV nuclear reactor coolants. This model, based on the Total Enthalpy Model coupled with the Partially Saturated Method, addresses both high and low Prandtl numbers under laminar forced conditions. Key findings include the utility of the LBM total enthalpy Partially Saturated Method for solidification modeling, faster computational times compared to conventional Finite Volume Computational Fluid Dynamics (FV-CFD) methods for low Prandtl numbers, and the identification of stability limitations for high Peclet numbers.
In the second phase, we develop an intermediate-fidelity model based on a FV-CFD RANS enthalpy-porosity method, addressing high and low Prandtl numbers. We validate it against experimental data for high Prandtl numbers and against a high-fidelity LBM-LES model for low Prandtl numbers. Incorporating interface turbulence viscosity damping addresses RANS models��� tendency to overestimate heat transfer, enhancing their practical utility despite slightly reduced predictive accuracy compared to LES models.
In the third phase, we introduce a fine-to-coarse mesh upscaling technique enhanced by physics-based closure terms. Employing a data-driven strategy, we fine-tune the model���s closure coefficients utilizing the FV-CFD RANS intermediate-fidelity model on fine meshes. The calibrated coarse-mesh model reliably predicts essential performance metrics, including pressure drop, velocity profile, outlet temperature, and solid thickness distribution. This multidimensional approach marks a notable progression from conventional 1D methods, offering substantial time savings compared to fine-mesh models.
The highlights of this dissertation include (i) an innovative solidification model coupled with turbulence modeling based on LBM (ii) the validation and enhancement of an intermediate-fidelity RANS solidification model by incorporating a turbulence viscosity-damping source, preventing turbulence overproduction at the interface (iii) the development of a fine-to-coarse mesh upscaling solidification model, incorporating physics-based closure terms calibrated through a data-driven approach
Spatial Transformer Network You Only Look Once (STN-YOLO) for Improved Object Detection
No full textObject detection plays a crucial role in the field of computer vision by enabling machines to identify and locate objects within images or videos. There are several state-of-the-art object detection approaches and the most common model is You Only Look Once (YOLO). YOLO is a single-shot algorithm that directly classifies an object in a single pass by having only one neural network predict the bounding boxes and class probabilities. The baseline YOLO model may encounter challenges in detecting objects within cluttered or partially occluded scenes. Furthermore, the model may face difficulty in detecting small objects or those with low contrast. To improve the performance of the baseline model, a spatial transformer network (STN) is incorporated into YOLO baseline model.
The results showcase the efficacy of the end-to-end pipeline, highlighting the innovative application of STN integrated with YOLO that demonstrates improved object detection performance through quantitative metrics (e.g., precision, accuracy, recall and intersection over union). This study also investigates of the impact of different localization network in the STN like RESNET18 and VGG16 on the object detection. The objective of incorporating the STN module into YOLO is to enhance the model���s capability not only to attend to the most relevant regions in an image but also to perform spatial transformations such that the image is aligned before the detection is performed. The detected image can be used to perform analysis by extracting relevant information in terms of features or statistics from images