In the age of 5th and upcoming 6th generation fighter aircraft one key proponent of these impressive machines is the inclusion of stealth. This inclusion is demonstrated by thoughtful design pertaining to the shape of the aircraft and rigorous material selection. Both criteria aim to minimize the radar cross section of these aircraft over a wide bandwidth of frequencies corresponding to an ever-evolving field of radar technology. Stealth is both an offensive and defensive capability meaning that service men and women depend on this feature to carry out their missions, and to return home safely. The goal of this paper is to introduce a novel method to designing disordered two-phase composites with desired electromagnetic properties. This task is accomplished by employing the spatial point correlation function, specifically at the two-point level. Effective at describing the dispersion of phases within a two-phase system, the two-point correlation function serves as a statistical function that becomes a realizable target for heterogeneous composites. Simulated annealing is exercised to reconstruct two-phase composite microstructures that initially do not match their target function, followed by two separate experiments aimed at studying the impact of the provided inputs on its outcome. Once conditions for reconstructing highly accurate microstructures are identified, modifications are made to the target function to extract and compare dielectric constants associated with each microstructure. Both the real and imaginary components, which respectively affect wave propagation and attenuation, of the dielectric constants are plotted to illustrate their behavior with increasing wavenumber. Conclusions suggest that favorable values of the complex dielectric constant can be reverse-engineered via careful consideration of the two-point correlation function. Subsequently, corresponding microstructures of the composite can be simulated and then produced through 3-D printing for testing and practical applications.

This dissertation contributes to uncertainty-aware neural networks using multi-modality data, with a focus on industrial and aviation applications. Drawing from seminal works in recent years that have significantly advanced the field, this dissertation develops techniques for incorporating uncertainty estimation and leveraging multi-modality information into neural networks for tasks such as fault detection and environmental perception. The escalating complexity of data in engineering contexts demands models that predict accurately and quantify uncertainty in these predictions. The methods proposed in this document utilize various techniques, including Bayesian Deep Learning, multi-task regularization and feature fusion, and efficient use of unlabeled data. Popular methods of uncertainty quantification are analyzed empirically to derive important insights on their use in real world engineering problems. The primary objective is to develop and refine Bayesian neural network models for enhanced predictive accuracy and decision support in engineering. This involves exploring novel architectures, regularization methods, and data fusion techniques. Significant attention is given to data handling challenges in deep learning, particularly in the context of quality inspection systems. The research integrates deep learning with vision systems for engineering risk assessment and decision support tasks, and introduces two novel benchmark datasets designed for semantic segmentation and classification tasks. Additionally, the dissertation delves into RGB-Depth data fusion for pipeline defect detection and the use of semi-supervised learning algorithms for manufacturing inspection tasks with imaging data. The dissertation contributes to bridging the gap between advanced statistical methods and practical engineering applications.

Janus Transition Metal Dichalcogenides (TMDs) are emerging 2D quantum materials with an asymmetric chalcogen configuration that induces an out-of-plane dipole moment. Their synthesis has been a limiting factor in exploring these systems' many-body physics and interactions. This dissertation examines the challenges associated with synthesis and charts the excitonic landscape of Janus crystals by proposing the development of the Selective Epitaxy and Atomic Replacement (SEAR) technique. SEAR utilizes ionized radical precursors to modify TMD monolayers into their Janus counterparts selectively. The synthesis is coupled with optical spectroscopy and monitored in real-time, enabling precise control of reaction kinetics and the structural evolution of Janus TMDs. The results demonstrate the synthesis of Janus TMDs at ambient temperatures, reducing defects and preserving the structural integrity with the hitherto best-reported exciton linewidth emission value, indicating ultra-high optical quality. Cryogenic optical spectroscopy (4K) coupled with a magnetic field on Janus monolayers has allowed the isolation of excitonic transitions and the identification of charged exciton complexes. Further study into macroscopic and microscopic defects reveals that structural asymmetry results in the spontaneous formation of 2D Janus Nanoscrolls from an in-plane strain. The chalcogen arrangement in these structures dictates two types of scrolling dynamics that form Archimedean or inverted C-scrolls. High-resolution scanning transmission electron microscopy of these superlattices shows a preferential orientation of scrolling and formation of Moiré patterns. These materials' thermodynamically favorable defect states are identified and shown to be optically active. The encapsulation of Janus TMDs with hexagonal Boron Nitride (h-BN) has allowed isolation defect transitions. DFT coupled with power-dependent PL spectroscopy at 4K shows the broad defect band to be a convolution of individual defect states with extremely narrow linewidth (2 meV) indicative of a two-state quantum system. The research presents a comprehensive synthesis approach with insights into the structural and morphological stability of 2D Janus layers, establishing a complete structure-property correlation of optical transitions and defect states, broadening the scope for practical applications in quantum information technologies.

Phase-field (PF) models are one of the most powerful tools to simulate microstructural evolution in metallic materials, polymers, and ceramics. However, existing PF approaches rely on rigorous mathematical model development, sophisticated numerical schemes, and high-performance computing for accuracy. Although recently developed surrogate microstructure models employ deep-learning techniques and reconstruction of microstructures from lower-dimensional data, their accuracy is fairly limited as spatio-temporal information is lost in the pursuit of dimensional reduction. Given these limitations, a novel data-driven emulator (DDE) for extrapolation prediction of microstructural evolution is presented, which combines an image-based convolutional and recurrent neural network (CRNN) with tensor decomposition, while leveraging previously obtained PF datasets for training. To assess the robustness of DDE, the emulation sequence and the scaling behavior with phase-field simulations for several noisy initial states are compared. In conclusion, the effectiveness of the microstructure emulation technique is explored in the context of accelerating runtime, along with an emphasis on its trade-off with accuracy.Meanwhile, an interpolation DDE has also been tested, which is based on obtaining a low-dimensional representation of the microstructures via tensor decomposition and subsequently predicting the microstructure evolution in the low-dimensional space using Gaussian process regression (GPR). Once the microstructure predictions are obtained in the low-dimensional space, a hybrid input-output phase retrieval algorithm will be employed to reconstruct the microstructures. As proof of concept, the results on microstructure prediction for spinodal decomposition are presented, although the method itself is agnostic of the material parameters. Results show that GPR-based DDE model are able to predict microstructure evolution sequences that closely resemble the true microstructures (average normalized mean square of 6.78 × 10−7) at time scales half of that employed in obtaining training data. This data-driven microstructure emulator opens new avenues to predict the microstructural evolution by leveraging phase-field simulations and physical experimentation where the time resolution is often quite large due to limited resources and physical constraints, such as the phase coarsening experiments previously performed in microgravity. Future work will also be discussed and demonstrate the intended utilization of these two approaches for 3D microstructure prediction through their combined application.

Heusler alloys were discovered in 1903, and materials with half-metallic characteristics have drawn more attention from researchers since the advances in semiconductor industry [1]. Heusler alloys have found application as spin-filters, tunnel junctions or giant magnetoresistance (GMR) devices in technological applications [1]. In this work, the electronic structures, phonon dispersion, thermal properties, and electrical conductivities of PdMnSn and six novel alloys (AuCrSn, AuMnGe, Au2MnSn, Cu2NiGe, Pd2NiGe and Pt2CoSn) along with their magnetic moments are studied using ab initio calculations to understand the roots of half-metallicity in these alloys of Heusler family. From the phonon dispersion, the thermodynamic stability of the alloys in their respective phases is assessed. Phonon modes were also used to further understand the electrical transport in the crystals of these seven alloys. This study evaluates the relationship between materials' electrical conductivity and minority-spin bandgap in the band structure, and it provides suggestions for selecting constituent elements when designing new half-metallic Heusler alloys of C1b and L21 structures.

For the past two centuries, coal has played a vital role as the primary carbon source, fueling industries and enabling the production of essential carbon-rich materials, including carbon nanotubes, graphite, and diamond. However, the global transition towards sustainable energy production has resulted in a decline in coal usage for energy purposes, with the United States alone witnessing a substantial 50% reduction over the past decade. This shift aligns with the UN’s 2030 sustainability goals, which emphasize the reduction of greenhouse gas emissions and the promotion of cleaner energy sources. Despite the decreased use in energy production, the abundance of coal has sparked interest in exploring its potential for other sustainable and valuable applications.In this context, Direct Ink Writing (DIW) has emerged as a promising additive manufacturing technique that employs liquid or gel-like resins to construct three-dimensional structures. DIW offers a unique advantage by allowing the incorporation of particulate reinforcements, which enhance the properties and functionalities of the materials. This study focuses on evaluating the viability of coal as a sustainable and cost-effective substitute for other carbon-based reinforcements, such as graphite or carbon nanotubes. The research utilizes a thermosetting resin based on phenol-formaldehyde (commercially known as Bakelite) as the matrix, while pulverized coal (250 µm) and carbon black (CB) function as the reinforcements. The DIW ink is meticulously formulated to exhibit shear-thinning behavior, facilitating uniform and continuous printing of structures. Mechanical property testing of the printed structures was conducted following ASTM standards. Interestingly, the study reveals that incorporating a 2 wt% concentration of coal in the resin yields the most significant improvements in tensile modulus and flexural strength, with enhancements of 35% and 12.5% respectively. These findings underscore the promising potential of coal as a sustainable and environmentally friendly reinforcement material in additive manufacturing applications. By harnessing the unique properties of coal, this research opens new avenues for its utilization in the pursuit of greener and more efficient manufacturing processes.

The study aims to develop and evaluate failure prediction models that accurately predict crack initiation sites, fatigue life in additively manufactured Ti-6Al-4V, and burst pressure in relevant applications.The first part proposes a classification model to identify crack initiation sites in AM-built Ti-6Al-4V alloy. The model utilizes surface and pore-related parameters and achieves high accuracy (0.97) and robustness (F1 score of 0.98). Leveraging CT images for characterization and data extraction from the CT-images built STL files, the model effectively detects crack initiation sites while minimizing false positives and negatives. Data augmentation techniques, including SMOTE+Tomek Links, are employed to address imbalanced data distributions and improve model performance. This study proposes the Probabilistic Physics-guided Neural Network 2.0 (PPgNN) for probabilistic fatigue life estimation. The presented approach overcomes the limitations of classical regression machine models commonly used to analyze fatigue data. One key advantage of the proposed method is incorporating known physics constraints, resulting in accurate and physically consistent predictions. The efficacy of the model is demonstrated by training the model with multiple fatigue S-N curve data sets from open literature with relevant morphological data and tested using the data extracted from CT-built STL files. The results illustrate that PPgNN 2.0 is a flexible and robust model for predicting fatigue life and quantifying uncertainties by estimating the mean and standard deviation of the fatigue life. The loss function that trains the proposed model can capture the underlying distribution and reduce the prediction error. A comparison study between the performance of neural network models highlights the benefits of physics-guided learning for fatigue data analysis. The proposed model demonstrates satisfactory learning capacity and generalization, providing accurate fatigue life predictions to unseen examples. An elastic-plastic Finite Element Model (FEM) is developed in the second part to assess pipeline integrity, focusing on burst pressure estimation in high-pressure gas pipelines with interactive corrosion defects. The FEM accurately predicts burst pressure and evaluates the remaining useful life by considering the interaction between corrosion defects and neighboring pits. The FEM outperforms the well-known ASME-B31G method in handling interactive corrosion threats.

Computing the fluid phase interfaces in multiphase flow is a challenging area of research in fluids. The Volume of Fluid andLevel Set methods are a few algorithms that have been developed for reconstructing the multiphase fluid flow interfaces. The thesis work focuses on exploring the ability of neural networks to reconstruct the multiphase fluid flow interfaces using a data-driven approach. The neural network model has liquid volume fraction stencils as an input, and it predicts the radius of the circle as an output of the network which represents a phase interface separating two immiscible fluids inside a fluid domain. The liquid volume fraction stencils are generated for randomly varying circle radii within a 1x1 domain using an open-source VOFI library. These datasets are used to train the neural network. Once the model is trained, the predicted circular phase interface from the neural network output is used to generate back the predicted liquid volume fraction stencils. Error norms values are calculated to assess the error in the neural network model’s predicted liquid volume fraction stencils with the actual liquid volume fraction stencils from the VOFI library. The neural network parameters are optimized by testing them for different hyper-parameters to reduce the error norms. So as to minimize the difference between the predicted and the actual liquid volume fraction stencils and errors in reconstructing the fluid phase interface geometry.

With increasing advance complexity in the structure to be 3D printed, the use of post processing removal of support structures has become more complicated thing due to the need of newer tool case to remove supports in such scenarios. Attempts have been made to study, research and experiment the dissolvable and recyclable photo-initiated polymeric resin that can be used to build support structure. Vat photo-polymerization method of manufacturing was selected due to wide range of materials that can be selected and researched which can have the potential to be selected further for large scale manufacturing. Deep understanding of the recyclable polymer was done by performing chemical and mechanical property test. Varying light intensities are used to study the curing properties and respective dissolving properties. In this thesis document, recyclable and dissolvable polymeric resin have been selected to print the support structures which can be later dissolved and recycled.The resin was exposed to varying light projections using grayscales of 255, 200 and 150 showing different dissolving time of each structure. Dissolving time of the printed parts were studied by varying the surface to volume ratios of the part. Higher the surface to volume ratios of the printed part resulted in lower time it takes to dissolve the part in the dissolving solution. The mechanical strengths of the recycled part were found to be pretty solid as compared to the freshly prepared resin, good sign of using it for multiple times without degrading its strength. Cactus shaped model was printed using commercial red resin and supports with the recyclable solution to deeply understand the working and dissolving properties of recyclable resin. Without any external efforts, the supports were easily dissolved in the solution, leaving the cactus intact. Further work is carried on printing Meta shaped gyroid lattice structure in effort to lower the dissolving time of the supports while maintaining enough mechanical stress. Future efforts will be made to conduct the rheology test and further lower the dissolving time as much it can to be ready for the commercial large scale applications.

National Airspace Systems (NAS) are complex cyber-physical systems that require swift air traffic management (ATM) to ensure flight safety and efficiency. With the surging demand for air travel and the increasing intricacy of aviation systems, the need for advanced technologies to support air traffic management and air traffic control (ATC) service has become more crucial than ever. Data-driven models or artificial intelligence (AI) have been conceptually investigated by various parties and shown immense potential, especially when provided with a vast volume of real-world data. These data include traffic information, weather contours, operational reports, terrain information, flight procedures, and aviation regulations. Data-driven models learn from historical experiences and observations and provide expeditious recommendations and decision support for various operation tasks, directly contributing to the digital transformation in aviation. This dissertation reports several research studies covering different aspects of air traffic management and ATC service utilizing data-driven modeling, which are validated using real-world big data (flight tracks, flight events, convective weather, workload probes). These studies encompass a range of topics, including trajectory recommendations, weather studies, landing operations, and aviation human factors. Specifically, the topics explored are (i) trajectory recommendations under weather conditions, which examine the impact of convective weather on last on-file flight plans and provide calibrated trajectories based on convective weather; (ii) multi-aircraft trajectory predictions, which study the intention of multiple mid-air aircraft in the near-terminal airspace and provide trajectory predictions; (iii) flight scheduling operations, which involve probabilistic machine learning-enhanced optimization algorithms for robust and efficient aircraft landing sequencing; (iv) aviation human factors, which predict air traffic controller workload level from flight traffic data with conformalized graph neural network. The uncertainties associated with these studies are given special attention and addressed through Bayesian/probabilistic machine learning. Finally, discussions on high-level AI-enabled ATM research directions are provided, hoping to extend the proposed studies in the future. This dissertation demonstrates that data-driven modeling has great potential for aviation digital twins, revolutionizing the aviation decision-making process and enhancing the safety and efficiency of ATM. Moreover, these research directions are not merely add-ons to existing aviation practices but also contribute to the future of transportation, particularly in the development of autonomous systems.