<p>Current industrial practice in the steel Industry involves the use of natural gas with high methane content as a primary energy source. Natural combustion produces greenhouse gases, and with the continued focus on managing and reducing harmful emissions from industrial processes, there is a need for research into alternative sources of energy. Among several alternatives that have been studied is hydrogen: a non-carbon-based fuel. This work uses a coupled computational fluid dynamics (CFD)-finite element analysis (FEA) combustion model to investigate hydrogen utilization as a fuel in a reheat furnace and how it impacts the quality of the steel produced by understanding the three dimensional (3D) flow behavior, furnace temperature profile, thermal stress distribution, heat flux, formation of iron oxides, emission gases and mode of heat transfer onto the steel slabs. The modeling process integrates the five different zones of a pusher type reheating furnace (top and bottom) and modeled using Ansys Fluent 2020R1 and Ansys Workbench 2022R1. Changes in these parameters are determined by comparison to a baseline case that uses methane as fuel and maintaining the same heat input in terms of chemical energy into the furnace. Global mechanism was used for hydrogen and two step mechanism was used for methane combustion. Results revealed a 2.6% increase in average temperature to 1478K across the furnace for hydrogen which resulted in 6.45% increase in maximum heat flux into the slabs. Similar flue gas flow patterns were seen for both cases and heat transfer mode from the combustion gases to the slabs was primarily by radiation (~97%) for both methane and hydrogen. 11.5% increase in iron oxide formation on the slab was recorded for the hydrogen case, however, the bulk of the iron oxide formed was more of wüstites which are the easiest form of iron oxide to descale. However, elevated nitrogen oxide (NOx) levels were recorded for hydrogen combustion which led to further study into NOx mitigation techniques. Application of the staged combustion method using hydrogen fuel showed potentials for NOx reduction. The use of regenerative burners further conserved exergy losses in hydrogen fuel application. Insignificant deviation from base case thermal stress distribution and zero carbon emission from the hydrogen case indicates the usability of hydrogen as an alternative fuel in reheating furnace operations. </p>

<p>The rapid growth of fields such as metamaterials, composites, architected materials, porous solids, and micro/nano materials, along with the continuing advancements in design and fabrication procedures have led to the synthesis of complex structures having intricate material distributions and non-trivial geometries. These materials find important applications including biomedical implants and devices, aerospace and naval structures, and micro/nano-electromechanical devices. Theoretical and experimental evidences have shown that these structures exhibit size-dependent (or, nonlocal) effects. This implies that the response of a point within the solid is affected by a collection of points; ultimately a manifestation of the multiscale deformation process. Broadly speaking, at a continuum level, the mathematical description of these multiscale phenomena leads to integral constitutive models, that account for the long-range interactions via nonlocal kernels. </p> <p><br></p> <p>Despite receiving considerable attention, the existing class of approaches to nonlocal elasticity are predominantly phenomenological in nature, following from their definition of the material parameters of the nonlocal kernel based on 'representative volume element' (RVE)-based statistical homogenization of the heterogeneous microstructure. The size of the RVE required for practical simulation, does not achieve a full-resolution of the intricate heterogeneous microstructure, and also implicitly enforces the use of symmetric nonlocal kernels to achieve thermodynamic consistency and mathematically well-posedness. The latter restriction directly limits the application of existing approaches only to the linear deformation analysis of either periodic or isotropic nonlocal structures. Additionally, the lack of a consistent characterization of the nonlocal effects, often results in inconsistent (also labeled as 'paradoxical') predictions depending on the nature of the external loading. In order to address these fundamental theoretical gaps, this dissertation develops a fractional-order kinematic approach to nonlocal elasticity by leveraging cutting-edge mathematical operators derived from the field of fractional calculus.</p> <p><br></p> <p>In contrast to the class of existing class of approaches that adopt an integral stress-strain constitutive relation derived from the equilibrium of the RVE, the fractional-order approach is predicated on a differ-integral (fractional-order) strain-displacement relation. The latter relation is derived from a fractional-order deformation-gradient mapping between deformed and undeformed configurations, and this approach naturally localizes and captures the effect of nonlocality at the root of the deformation phenomena. The most remarkable consequence of this reformulation consists in its ability to achieve thermodynamic and mathematical consistency, irrespective of the nature of the nonlocal kernel. The convex and positive-definite nature of the formulation enabled the use of variational principles to formulate well-posed governing equations, the incorporation of nonlinear effects, and enabled the development of accurate finite element simulation methods. The aforementioned features, when combined with a variable-order extension of the fractional-order continuum theory, enabled the physically consistent application of the nonlocal formulation to general continua exhibiting asymmetric interactions; ultimately a manifestation of material heterogeneity. Indeed, a rigorous theoretical analysis was conducted to demonstrate the natural ability of the variable-order in capturing the role of microstructure in the deformation of heterogeneous porous solids. These advantages allowed the application of the fractional-order kinematic approach to accurately and efficiently model the response of porous beams and plates, with random microstructural descriptions. Results derived from multiphysical loading conditions, as well as nonlinear deformation regimes, are used to demonstrate the causal relation between the kinematics-based fractional-order characterization of nonlocal effects and the natural role of microstructure in determining the macroscopic response of heterogeneous solids. The potential implications of the developed formalism on scientific discovery of material laws are examined in-depth, and different areas for further research are identified.</p>

<p>The design and discovery of novel materials are difficult not only due to expensive and time- consuming calculation and measurements of their properties, but also thanks to the infinite search spaces. With the increasingly abundant data from experiments and simulations, learning from data has the potential of bypassing complex physics-based simulations and experiments and providing fast approximations of the solution. Deep learning models are helpful in the design process that requires prohibitively expensive iterative computations. In addition, as efficient and accurate sur- rogate models, trained deep networks can incorporate techniques, such as sensitivity analysis and active learning, to provide guidance in searching promising candidates. Moreover, deep learning models need to account for the material structural information, such as molecule and atom align- ments, chemical bonds, and grain-level interactions, as it plays an important role in determining the macroscopic properties. In this thesis, we start with developing two standard deep learning model- based materials design frameworks for lithium-ion batteries and thermoelectric materials, and we then investigate the feasibility of standard deep learning models on data with graph-structured in- formation and identify the challenges. Finally, we propose a deep graph operator network that effectively capture the spatial dependency encoded in the graph structure to solve networked dy- namical systems.</p> <p><br></p> <p>In the first half of the thesis, we propose a hybrid convolutional neural network to infer lithium- ion battery microstructure properties, Bruggeman’s exponent and shape factor, given its voltage vs. capacity curves. The trained model accurately predicts the microstructural properties on both experimental and simulation data, and it can readily accelerate the processing-properties- performance and degradation characteristics of the existing and emerging chemistries of lithium- ion batteries. Also, we develop a AI-guided framework to discover and design thermoelectric materials, where we train classifiers based on the materials chemical and structural information embeddings and combine with variance-based sensitivity analysis to suggest candidates and con- duct fast screening.</p> <p><br></p> <p>In the second half of the thesis, we build a data-centric framework with a recurrent neural network-based classifier to achieve traffic incident detection on highway networks. We incorporate weak supervised learning and design labeling functions to create large amount of training data with probabilistic labels. The trained deep ensemble accurately detects incidents with predictive uncertainty. To capture the structural information in the network, we then propose a deep graph operator network that maps the input graph state function to the output graph state function. The proposed model enables resolution-independence and zero-shot transfer, where we do not require a set of fixed sensors to encode the graph trajectory and can use the trained model directly on larger graphs with high accuracy. We utilize the proposed model to solve power grid transient stability prediction and traffic forecasting problems.</p>

<p dir="ltr">Nature moves small things by chemical energy. Inspired by this, catalytic reactions driven microswimmers have been designed and believed to be promising to help transport drugs and other cargos at microscales. However, decorating the microswimmers with drugs and cargos would make them heavy and hard to move. An alternative solution to this would be designing self-pumping devices that can pump the fluid and things carried by the fluid all together without external resources. In this work, we have presented the first full numerical model of electrochemically-powered self-pumping in the Pt-Au coated polycarbonate membrane reported by Jun and Hess [1]. The simulations demonstrate that autonomous flow in self-pumping membranes is an electro-osmotic flow driven by a self-generated electric field. The injection and consumption of H⁺ on Pt and Au respectively lead to a charge asymmetry and an associated electric field that acts on the electric double layers (EDL) coating the pore walls driving fluid move, i.e. self-electro-osmosis. Key parameters controlling the performance of self-pumping are pore radius and background pH values, as they affect the EDL overlap and ionic strength. Other parameters such as porosity and pore length can both be tuned to find the local optimum for a membrane design. By tuning these parameters, the trade-off between increased ionic current and increased ionic strength could be balanced, contributing to an optimum self-pumping performance. When inclination or deformation occurs in cylindrical pores, the self-pumping flow does not significantly deviate from the trend. Membranes with complicated shape of contracting/expanding pores and cross-linked connecting pores should follow same pattern as cylindrical pores with similar pore size. In addition, if we replace the Pt/Au catalytic pairs by TiO₂/Au photocatalytic pairs, self-pumping membrane could be driven by light. The geometry of pore enhances the light absorption, enabling self-pumping membrane achieving high flow rate at large porosity with relatively large pores. At the end, we provide experimental evidence of self-pumping flow on TiO₂-Au plates as well as self-pumping membrane driven by light.</p>

<p dir="ltr">This thesis delves into the dynamics and driving factors of thermal transport via natural convection, the onset and severity of quiescent cavitation and its impact of auto-injector device performance, and the controlled release of rapidly swelling pharmaceutical tablets. In each of these instances showcases how variations in external conditions or the introduction of new variables can disrupt the equilibrium of fluid systems, leading to complex behaviors. Vertical thermal convection illustrates how temperature gradients induce fluid movement and patterns; cavitation inception focuses on the formation of vapor cavities due to pressure drops within a fluid; and rapid tablet swelling explores the interaction between solid materials and liquids, leading to significant changes in concentration and mass transfer. These studies collectively enhance our understanding of transport dynamics, highlighting pathways to achieve optimal transport and delivery conditions for various industrial and pharmaceutical processes.</p>

The present doctoral thesis summarises results of investigation focused on the characterisation of materials involved in Solid Oxide Cell technology. The main topic of investigation was the ceramic cell, also known as MEA. Particular attention was given to the role that bi-material interfaces, co-sintering effects and residual stresses play in the resulting mechanical response. The first main goal was to investigate the effects of the manufacturing process (i.e. layer by layer deposition) on the mechanical response; to enable this investigation, electrode layers were screen-printed one by one on the electrolyte support and experimental tests were performed after every layer deposition. The experimental activity started with the measurement of the elastic characteristics. Both elastic and shear moduli were measured via three different techniques at room and high temperature. Then, uniaxial and biaxial flexural strengths were determined via two loading configurations. The analysis of the elastic and fracture behaviours of the MEA revealed that the addition of layers to the electrolyte has a detrimental effect on the final mechanical response. Elastic characteristics and flexural strength of the electrolyte on the MEA level are sensibly reduced. The reasons behind the weakening effect can be ascribed to the presence and redistribution of residual stresses, changes in the crack initiation site, porosity of layers and pre-cracks formation in the electrode layers. Finally, the coefficients of thermal expansion were evaluated via dilatometry on bulk materials serving as inputs for finite elements analyses supporting experiments and results interpretation. The second most important goal was to assess the influence of operating conditions on the integrity of the MEA. Here interactions of ceramic–metal interfaces within the repetition unit operating at high temperatures and as well at both oxidative and reductive atmospheres were investigated. The elastic and fracture responses of MEA extracted from SOC stacks after several hours of service were analysed. Layer delamination and loss of mechanical strength were observed with increasing operational time. Moreover, SEM observations helped to detect significant microstructural changes of the electrodes (e.g. demixing, coarsening, elemental migration and depletion), which might be responsible for decreased electrochemical performances. All the materials presented in this work are part of SOC stacks produced and commercialised by Sunfire GmbH, which is one of the world leading companies in the field.

<div>Additive manufacturing (AM) or 3D printing has gained significant advancement in recent years. However the potential of 3D printed metals still has not been fully explored. A main reason is the lack of accurate knowledge of the load capacity of 3D printed metals, such as fatigue behavior under cyclic load conditions, which is still poorly understood as compared with the conventional wrought counterpart.</div><div><br></div><div>The goal of the thesis is to advance the knowledge of fatigue behavior of 15-5 PH stainless steel manufactured through laser powder bed fusion process. To achieve the goal, a combined numerical and experimental study is carried out. First, using a rotary fatigue testing experiment, the fatigue life of the 15-5 PH stainless steel is measured. The strain life curve shows that the numbers of the reversals to failure increase from 13,403 to 46,760 as the applied strain magnitudes decrease from 0.214\% from 0.132\%, respectively. The micro-structure analysis shows that predominantly brittle fracture is presented on the fractured surface. Second, a finite element model based on cyclic plasticity including the damage model is developed to predict the fatigue life. The model is calibrated with two cases: one is the fatigue life of 3D printed 17-4 stainless steel under constant amplitude strain load using the direct cyclic method, and the other one is the cyclic behavior of Alloy 617 under multi-amplitude strain loads using the static analysis method. Both validation models show a good correlation with the literature experimental data. Finally, after the validation, the finite element model is applied to the 15-5 PH stainless steel. Using the direct cyclic method, the model predicts the fatigue life of 15-5 PH stainless steel under constant amplitude strain. The extension of the prediction curve matches well with the previously measured experimental results, following the combined Coffin-Manson Basquin Law. Under multi-amplitude strain, the kinematic hardening evolution parameter is incorporated into the model. The model is capable to capture the stresses at varied strain amplitudes. Higher stresses are predicted when strain amplitudes are increased. The model presented in the work can be used to design reliable 3D printed metals under cyclic loading conditions.</div>

<p> As space exploration continues to advance, so does the drive to inhabit celestial bodies. In<br> order to expand our civilization to the Moon or even other planets requires an enormous amount of research and development. The Resilient Extra-Terrestrial Habitat Institute is a NASA funded project that aims to develop the technology needed to establish deep-space habitats. Deep-space inhabitation poses many challenges that are not present here on earth. The Moon, for example, has temperatures that range from -233−123°C. Aside from the extreme temperatures, a variety of thermal loads will need to be handled by the Environmental Control and Life Support System (ECLSS). Aside from the research and architecture of the International Space Station’s ECLSS, very little information is known about disturbances related to the thermal management of extra- terrestrial habitats.<br> </p> <p>RETHi is developing a Cyber-Physical Testbed (CPT) that represents a one-fifth scale<br> prototype of a deep space habitat. In order to answer difficult research questions regarding ECLSS and thermal management of a deep-space habitat, a heat pump was modeled and validated with the physical part of the CPT. Once validated, the heat pump model is able to accurately predict the steady state behavior given the indoor and outdoor conditions of the testbed. When coupled with the interior environment (IE) model, it gives insight into the system’s requirements and response. Experimental testing was conducted with the heat pump in order to validate the model. After the model was validated, a series of parametric studies were conducted in order to investigate the effects of varying thermal loads and dehumidification. Since the groundwork was laid through model development and experimentation, future work consists of designing a more versatile heat pump to test a variety of disturbance scenarios. Although the heat pump model is specifically designed for the CPT, it proves to be versatile for other closed and pressurized environments such as aircraft and clean rooms according to the analysis of dehumidification and dependence on pressure. </p>

<p>Electric load forecasting is a critical tool for power system planning and the creation of sustainable energy systems. Precise and reliable load forecasting enables power system operators to make informed decisions regarding power generation and transmission, optimize energy efficiency, and reduce operational costs and extra power generation costs, to further reduce environment-related issues. However, achieving desirable forecasting performance remains challenging due to the irregular, nonstationary, nonlinear, and noisy nature of the observed data under unprecedented events. In recent years, deep learning and other artificial intelligence techniques have emerged as promising approaches for load forecasting. These techniques have the ability to capture complex patterns and relationships in the data and adapt to changing conditions, thereby enhancing forecasting accuracy. As such, the use of deep learning and other artificial intelligence techniques in load forecasting has become an increasingly popular research topic in the field of power systems. </p> <p>Although deep learning techniques have advanced load forecasting, the field still requires more accurate and efficient models. One promising approach is federated learning, which allows for distributed data analysis without exchanging data among multiple devices or cen- ters. This method is particularly relevant for load forecasting, where each power station’s data is sensitive and must be protected. In this study, a proposed approach utilizing Federated Deeponet for seven different power stations is introduced, which proposes a Federated Deep Operator Networks and a Lagevin Dynamics-based Federated Deep Operator Networks using Stochastic Gradient Langevin Dynamics as the optimizer for training the data daily for one-day and predicting for one-day frequencies by frequencies. The data evaluation methods include mean absolute percentage error and the percentage coverage under confidence interval. The findings demonstrate the potential of federated learning for secure and precise load forecasting, while also highlighting the challenges and opportunities of implementing this approach in real-world scenarios. </p>

<p>Breast cancer affects 12.5\% of women over their life time and tissue expansion (TE) is the most common technique for breast reconstruction after mastectomy. However, the rate of complications with TE can be as high as 15\%. Even though the first documented case of TE happened in 1957, there has yet to be a standardized procedure established due to the variations among patients and the TE protocols are currently designed based on surgeon's experience. There are several studies of computational and theoretical framework modeling skin growth in TE but these tools are not used in the clinical setting. This dissertation focuses on bridging the gap between the already existing skin growth modeling efforts and it's potential application in the clinical setting.</p> <p><br></p> <p>We started with calibrating a skin growth model based on porcine skin expansions data. We built a predictive finite element model of tissue expansion. Two types of model were tested, isotropic and anisotropic models. Calibration was done in a probabilistic framework, allowing us to capture the inherent biological uncertainty of living tissue. We hypothesized that the skin growth rate was proportional to stretch. Indeed, the Bayesian calibration process confirmed that this conceptual model best explained the data. </p> <p><br></p> <p>Although the initial model described the macroscale response, it did not consider any activity on the cellular level. To account for the underlying cellular mechanisms at the microscopic scale, we have established a new system of differential equations that describe the dynamics of key mechanosensing pathways that we observed to be activated in the porcine model. We calibrated the parameters of the new model based on porcine skin data. The refined model is still able to reproduce the observed macroscale changes in tissue growth, but now based on mechanistic knowledge of the cell mechanobiology.  </p> <p><br></p> <p>Lastly, we demonstrated how our skin growth model can be used in a clinical setting. We created TE simulations matching the protocol used in human patients and compared the results with clinical data with good agreement. Then we established a personalized model built from 3D scans of a patient unique geometry. We verified our model by comparing the skin growth area with the area of the skin harvested in the procedure, again with good agreement.</p> <p><br></p> <p>Our work shows that skin growth modeling can be a powerful tool to aid surgeons design TE procedures before they are actually performed. The simulations can help with optimizing the protocol to guarantee the correct amount of skin is growth in the shortest time possible without subjecting the skin to deformations that can compromise the procedure.</p>