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.

The microelectronics industry is actively focusing on advanced packaging technologies, notably on three-dimensional stacking of heterogeneous integrated (3D-HI) circuits for enhanced performance. Despite its computational performance benefits, this approach faces challenges in thermal management due to increased power density and heat generation. Conventional cooling methods struggle to address this issue effectively. This study investigates microfluidic intralayer cooling techniques using analytical correlation and computational fluid dynamics (CFD) principles to propose a method capable of managing thermal performance across varying load conditions. The proposed configuration achieved a dissipation of 40 W/cm2 with a volumetric flow rate of 200 mL/min, maintaining chip temperature at 315K. Additionally, extreme hotspot conditions generating 1kW/cm2, along with the presence of thermal resistance from redistribution layers (RDLs), are analyzed. This research aims to establish a model for understanding geometric property variations under different heat flux conditions in 3D heterogeneous integration of semiconductor packaging.

This research aims to identify optimal pin fin shapes that minimize flow pressuredrop and maximize heat transfer performance while investigating the influence of genetic algorithm (GA) parameters on these shapes. The primary goal is to discover innovative pin fin configurations through the use of a GA, moving away from traditional circular cylindrical designs. The study also examines GA parameters, including population size, generation size, selection methods, crossover rates, tournament size, and elite counts. A physical condition considered in this study is a rectangular channel with a square cross-section integrated with 10 pin fins, operating at a Reynolds number of 2316, and subjected to a heat flux of 5 W/cm2 at the bottom surface. Overall, the research seeks to enhance the energy efficiency of a liquid cooling system, with potential applications in the thermal management of computing devices. By enabling operating at significantly lower power, the optimized cooling system promises to reduce energy consumption and operational costs.

This project involved research into solar thermal and geothermal energy generation as possible solutions to the growing U.S. energy crisis. Background research into this topic revealed the effects of climate and environmental impacts as major variables in determining optimal states. Delving into thermodynamic engineering analyses, the main deliverables of this research were mathematical models to analyze plant efficiency improvements in order to optimize the cost of operating solar thermal and geothermal power plants. The project concludes with possible future research areas relating to this field.

This project involved research into solar thermal and geothermal energy generation as possible solutions to the growing U.S. energy crisis. Background research into this topic revealed the effects of climate and environmental impacts as major variables in determining optimal states. Delving into thermodynamic engineering analyses, the main deliverables of this research were mathematical models to analyze plant efficiency improvements in order to optimize the cost of operating solar thermal and geothermal power plants. The project concludes with possible future research areas relating to this field.

Ethylene is one of the most widely used organic compounds worldwide with ever increasing demand. Almost all the industries currently producing ethylene globally use the method of steam cracking, which, though highly selective and cost effective, is energy intensive along with having a high carbon footprint. This study aims to analyze micro-scale partial oxidation of propane as a novel approach towards ethylene generation which is simpler, less energy consuming, operates at lower temperatures and causes minimum CO2 emission. The experimental study endeavors to maximize the ethylene production by investigating the effect of variables such as temperature, flow rate, equivalence ratio and reactor diameter. The micro-scale partial oxidation of propane is studied inside quartz tube reactors of 1 mm and 3 mm diameter at a temperature range of 800 to 900 oC, at varying flow rates of 10 to 100 sccm and equivalence ratios of 1 to 6. The study reveals ethylene yield has a strong dependence on all the above factors. However, the factors are not completely independent of each other. Adjusting certain factors and levels results in greater ethylene yields as high as 10%, but propane to ethylene conversion efficiency is approximately constant for most conditions. Low CO2 concentrations are also recorded for most of the factor and level combinations, indicating the potential to achieve lower CO2 yields compared to conventional approaches. The investigation indicates promise for application in the field of ethylene generation.

Nanomaterials redefine the lens through which the world is viewed today. The miniaturization of devices and systems to the nanoscale explodes the realm of what is possible as the interactions with neighboring atoms and molecules increase. This interactivity creates ripple effects that lead to superior mechanical, thermal, electrical, and optical properties that are highly desired across several industries. Two-dimensional (2D) materials are a branch of this family, and the focus of this paper revolves around a recent addition to this category called MXenes. The versatile properties of these 2D nanomaterials have made them unique, as they have the desired performance that can be utilized in several industries, especially energy management, wastewater treatment, and microelectronic devices. Followed by the MAX phase synthesis, hydrofluoric (HF) acid has been the primary etchant utilized to derive these 2D nanoparticles. However, alternative etchants via reactions are desirable to achieve similar selective etching without involving highly toxic HF. Therefore, this study investigated MXene synthesis and applications in 3D printing, followed by the formation of the precursor MAX, an optimized in-situ etching method, and streamlined post-etching processes to maximize 2D MXene yield. The etched powders were then analyzed using scanning electron microscopy (SEM), x-ray diffraction (XRD), atomic force microscopy (AFM), and energy-dispersive x-ray spectroscopy (EDS) characterization methods to verify and validate the MXene dimensions, chemistry, and crystal structures. Simple applications, such as the dispersion feasibility for customizing micropatterns via 3D printing, were also demonstrated as examples. Finally, this research showed the simple processing of 2D MXenes and their potential in structural support, heat dissipation, microelectronics, optical meta-surfaces, and other areas.

Applications like integrated circuits, microelectromechanical devices, antennas, sensors, actuators, and metamaterials benefit from heterogeneous material systems made of metallic structures and polymer matrixes. Due to their distinctive shells made of metal and polymer, scaly-foot snails, which are found in the deep ocean, exhibit high strength and temperature resistance. Recent metal deposition fabrication techniques have been used to create a variety of multi-material structures. However, using these complex hybrid processes, it is difficult to build complex 3D structures of heterogeneous material with improved properties, high resolution, and time efficiency. The use of electrical field-assisted heterogeneous material printing (EFA-HMP) technology has shown potential in fabricating metal-composite materials with improved mechanical properties and controlled microstructures. The technology is an advanced form of 3D printing that allows for printing multiple materials with different properties in a single print. This allows for the creation of complex and functional structures that are not possible with traditional 3D printing methods. The development of a photocurable printing solution was carried out that can serve as an electrolyte for charge transfer and further research into the printing solution's curing properties was conducted. A fundamental understanding of the formation mechanism of metallic structures on the polymer matrix was investigated through physics-based multiscale modeling and simulations. The relationship between the metallic structure's morphology, the printing solution's properties, and the printing process parameters was discovered.The thesis aims to investigate the microstructures and electrical properties of metal-composite materials fabricated using EFA-HMP technology and to evaluate the correlation between them. Several samples of metal-composite materials with different microstructures will be fabricated using EFA-HMP technology to accomplish this. The results of this study will provide a better understanding of the relationship between the microstructures and properties of metal-composite materials fabricated using EFA-HMP technology and contribute to the development of new and improved materials in various fields of application. Furthermore, this research will also shed light on the advantages and limitations of EFA-HMP technology in fabricating metal-composite materials and study the correlation between the microstructures and mechanical properties.

Progressive miniaturization in electronics demands advanced materials with excellent energy conversion and transport properties. Opportunities exist in novel material morphologies such as hierarchical structures, multi-functional composites and nanoscale architectures which may offer mechanical, thermal and electronic properties tailored to a wide range of applications (e.g., aerospace, robotics, biomedical etc.). However, the manufacturing capabilities have always posed a grand challenge in realizing the advanced material morphologies. Furthermore, the multi-scale modeling of complex material architectures has been extremely challenging owing to the limitations in computation methodologies and lack of understanding in nano-/micro-meter scale physics. To address these challenges, this work considers the morphology effect on carbon nanotube (CNT)-based composites, CNT fibers and thermoelectric (TE) materials. First, this work reports additively manufacturable TE morphologies and analyzes the thermo-electric transport behavior. This research introduces innovative honeycomb TE architectures that showed ~26% efficiency increase and ~25% density reduction compared to conventional rectangular TE architectures. Moreover, this work presents 3D printable compositionally segmented TE architecture which provides record-high efficiencies (up to 8.7%) over wide temperature ranges if the composition and aspect ratio of multiple TE materials are optimized within a single TE device. Next, this research proposes computationally efficient two-dimensional (2D) finite element model (FEM) to study the electrical and thermal properties in CNT based composites by simultaneously considering the stochastic CNT distributions, CNT fractions (upto 80%) and interfacial resistances. The FEM allows to estimate the theoretical maximum possible conductivities with corresponding interfacial resistances if the CNT morphologies are carefully controlled, along with appreciable insight into the energy transport physics. Then, this work proposes a data-driven surrogate model based on convolutional neural networks to rapidly approximate the composite conductivities in a second with accuracy > 98%, compared to FEM taking >100 minutes per simulation. Finally, this research presents a pseudo 2D FEM to approximate the electrical and thermal properties in CNT fibers at various CNT aspect ratios (up to 10,000) by simultaneously considering CNT-CNT interfacial effects along with the stochastic distribution of inter-bundle voids.

The design of energy absorbing structures is driven by application specific requirements like the amount of energy to be absorbed, maximum transmitted stress that is permissible, stroke length, and available enclosing space. Cellular structures like foams are commonly leveraged in nature for energy absorption and have also found use in engineering applications. With the possibility of manufacturing complex cellular shapes using additive manufacturing technologies, there is an opportunity to explore new topologies that improve energy absorption performance. This thesis aims to systematically understand the relationships between four key elements: (i) unit cell topology, (ii) material composition, (iii) relative density, and (iv) fields; and energy absorption behavior, and then leverage this understanding to develop, implement and validate a methodology to design the ideal cellular structure energy absorber. After a review of the literature in the domain of additively manufactured cellular materials for energy absorption, results from quasi-static compression of six cellular structures (hexagonal honeycomb, auxetic and Voronoi lattice, and diamond, Gyroid, and Schwarz-P) manufactured out of AlSi10Mg and Nylon-12. These cellular structures were compared to each other in the context of four design-relevant metrics to understand the influence of cell design on the deformation and failure behavior. Three new and revised metrics for energy absorption were proposed to enable more meaningful comparisons and subsequent design selection. Triply Periodic Minimal Surface (TPMS) structures were found to have the most promising overall performance and formed the basis for the numerical investigation of the effect of fields on the energy absorption performance of TPMS structures. A continuum shell-based methodology was developed to analyze the large deformation behavior of field-driven variable thickness TPMS structures and validated against experimental data. A range of analytical and stochastic fields were then evaluated that modified the TPMS structure, some of which were found to be effective in enhancing energy absorption behavior in the structures while retaining the same relative density. Combining findings from studies on the role of cell geometry, composition, relative density, and fields, this thesis concludes with the development of a design framework that can enable the formulation of cellular material energy absorbers with idealized behavior.