Doctor of Philosophy, The Ohio State University, 2024, Electrical and Computer Engineering

We present an exposition on Koopman operator-based reduced-order modeling of high-dimensional electromagnetic (EM) systems exhibiting both linear and nonlinear dynamics. Since the emergence of the digital age, numerical methods have been pivotal in understanding physical phenomena through computer simulations. Computational electromagnetics (CEM) and computational plasma physics (CPP) are related yet distinct branches, each addressing complex linear and nonlinear electromagnetic phenomena. CEM primarily focuses on solving Maxwell's equations for intricate structures such as antennas, cavities, high-frequency circuits, waveguides, and scattering problems. In contrast, CPP aims to capturing the complex behavior of charged particles under electromagnetic fields. This work specifically focuses on the numerical simulation of electromagnetic cavities and particle-in-cell (PIC) kinetic plasma simulations. Studying electromagnetic field coupling inside metallic cavities is crucial for various applications, including electromagnetic interference (EMI), electromagnetic compatibility (EMC), shielded enclosures, cavity filters, and antennas. However, time-domain simulations can be computationally intensive and time-consuming, especially as the scale and complexity of the problem increase. Similarly, PIC simulations, which are extensively used for simulating kinetic plasmas in the design of high-power microwave devices, vacuum electronic devices, and in astrophysical studies, can be computationally demanding, especially when simulating thousands to millions of charged particles. Moreover, the nonlinear nature of the complex wave-particle interactions complicates the modeling task. Data-driven reduced-order models (ROMs), which have recently gained prominence due to advances in machine learning techniques and hardware capabilities, offer a practical approach for constructing "light" models from high-fidelity data. The Koopman operator-based data-driven ROM is a powerful met (open full item for complete abstract)

Committee: Mrinal Kumar (Advisor); Fernando Teixeira (Advisor); Ben McCorkle (Committee Member); Balasubramaniam Shanker (Committee Member) Subjects: Electrical Engineering; Electromagnetics; Engineering; Physics; Plasma Physics

PhD, University of Cincinnati, 2024, Engineering and Applied Science: Aerospace Engineering

A rigorous literature survey on combustion instability, acoustic modeling, dynamic flame modeling, and reduced-order modeling (ROM) was presented to acknowledge the difficulty of accurately predicting combustion instability for a lean, fully premixed combustor operating at speed beyond 5% of the speed of sound while producing significantly long flame. The objective is to validate the feasibility of predicting thermoacoustic instability using ROM for changes in boundary condition (exit blockage ratio) and flow rate on a combustor that produces a non-compact flame. A revised ROM was derived with convective flow rather than assuming stationary flow to predict combustion instability in the combustor that operates at a speed beyond 5% of the speed of sound. The structure of the ROM was revised to account for spatially distributed heat release rather than assuming a singular compact flame. A proposed end boundary condition model was used to improve the solution of the ROM. A bluff body combustor was tested to establish a combustor that can operate at speeds greater than 5% of the speed of sound. The bluff body combustor produced stable and unstable non-compact flame under the same flow condition (inlet mach number and equivalence ratio) as the combustor configuration changes (combustor length and blockage ratio). The stable flame combustor configuration is used for flame transfer function measurement. In contrast, the unstable flame combustor configuration is used for ROM prediction validation. The exit pressure reflection coefficients were measured for three different blockage ratios (69%, 56%, 0%) without combustion as the temperature, flow rate, and frequency change to validate the end boundary model proposed in the literature. The model used to characterize the acoustic exit boundary condition for the revised ROM was successfully validated using the multiple microphone method downstream of the bluff body flame holder. An inlet Mach number of 0.06-0.13 at an equ (open full item for complete abstract)

Committee: Jongguen Lee Ph.D. (Committee Chair); Kwanwoo Kim Ph.D. (Committee Member); Paul Orkwis Ph.D. (Committee Member); Prashant Khare Ph.D. (Committee Member) Subjects: Aerospace Materials

Doctor of Philosophy, The Ohio State University, 2023, Physics

Nuclear physics is the study of phenomena involving the strong interaction, through its fundamental theory of quantum chromodynamics (QCD) or through effective theories with composite degrees of freedom. It seeks to understand the origin and evolution of visible matter; the organizational principles and emergent phenomena of nuclei; how the fundamental interactions can be understood using the nucleus as a laboratory; and how nuclear physics research can benefit society. To make progress, experimental facilities, such as the Facility of Rare Isotope Beams (FRIB) and ATLAS, are continuously collecting data to further our understanding of atomic nuclei and astrophysical processes. At the same time, theoretical approaches are tasked with providing accurate models whose theoretical errors are well-known. By working together, theoretical results can help guide experimentalists in understanding which regions and measurements provide the most return. The results presented in this thesis contribute to this effort by describing the construction of surrogate models (i.e., emulators) that produce fast, but accurate predictions for bound and scattering systems with various interactions. Emulators provide the means necessary for obtaining well-defined errors of a theoretical model by accurately reproduce their high-fidelity counterpart in a fraction of the time it takes to solve the underlying physics equations. These emulators can make feasible the use of Bayesian parameter estimation in low-energy nuclear physics by using the emulators to make predictions with many different parameter sets, allowing one to properly and efficiently propagate the theoretical uncertainties to NN observables. When used in conjunction with Bayesian experimental design, it allows one to establish the optimal experimental design needed to produce quality measurements in a much faster way compared to traditional calculations. The emulators described in this thesis fall into two categories: model- (open full item for complete abstract)

Committee: Dick Furnstahl (Advisor); Yuri Kovchegov (Committee Member); Richard Hughes (Committee Member); Douglass Schumacher (Committee Member) Subjects: Physics

Doctor of Philosophy, The Ohio State University, 2021, Mechanical Engineering

Turbomachinery, like jet engines and industrial gas turbines in power plants, are very advanced and complex machines. Due to the complexity and cost of modern turbomachinery, there is active research in accurately predicting the physical system dynamics using computational models. Two big mechanisms that affect the structural response are the prestress effects from high rotational speeds and mistuning effects from tolerance deviations, wear, or damage. Understanding the role these two mechanisms play in the computational modeling of these systems is an important step toward a complete digital twin of an entire jet engine. There previously existed modeling methods that enabled each to be analyzed independently, but not simultaneously in an efficient manner. This will be one of the focus points of this dissertation. The other focus being an experimental investigation into exciting system resonances of a rotating bladed disk using air jets. These experiments will be used to validate the computational modeling method developed. This dissertation has three primary objectives. The first objective is to present reduced order modeling methods that allow for the efficient modeling of coupled systems and rotating systems, both with small or large mistuning. By efficiently including these mechanisms, more realistic boundary conditions can be used to help validate the reduced order models (ROMs) with experimental data. Both modeling methods create models a fraction of the size of the full model while retaining key dynamic characteristics of the full model. The second objective of this work is to show the capability of air jets in exciting synchronous and non-synchronous vibrations in a rotating bladed disk. Much previous research in this field focused on experiments with stationary systems. These tests can help isolate specific mechanisms that may be present in bladed disks, but may limit the applicability of the results to actual rotating systems. This work presents a method (open full item for complete abstract)

Committee: Kiran D'Souza (Advisor); Randall Mathison (Committee Member); Manoj Srinivasan (Committee Member); Herman Shen (Committee Member) Subjects: Mechanical Engineering

Doctor of Philosophy, The Ohio State University, 2019, Materials Science and Engineering

The design framework for complex materials property and processing models within the Integrated Computational Material Engineering (ICME) is often hindered by the expensive computational cost. The ultimate goal of ICME is to develop data-driven, materials-based tools for the concurrent optimization of material systems while, improving the deployment of innovative materials in real-world products. Reduced-order, fast-acting tools are essential for both bottom-up property prediction and top-down model calibrations employed for modern material design applications. Additionally, reduced-order modeling requires formal uncertainty quantification (UQ) from the processing stages all the way down to the manufacturing and component design. The goal of this thesis is to introduce a machine learning-based, reduced-order crystal plasticity model for face-centered cubic (FCC) polycrystalline materials. This implementation was founded upon Open Citrination, an open-sourced materials informatics platform. Case studies for both the bottom-up property prediction and top-down optimization of model parameters are demonstrated within this work. The proposed reduced-order model is used to correctly approximate the plastic stress-strain curves and the texture evolution under a range of deformation conditions and strain rates specific to a material. The inverted pathway is applied to quickly calibrate the optimal crystal plasticity hardening parameters given the macroscale stress-strain responses and evolutionary texture under certain processing conditions. A visco-plastic self-consistent (VPSC) method is used to create the training and validation datasets. The description of the material texture is given through a dimension reduction technique which was implemented by principal component analysis (PCA). The microstructures of engineering materials typically involve an intricate hierarchical crystallography, morphology, and composition. Therefore, an accurate, virtual representation (open full item for complete abstract)

Committee: Stephen Niezgoda (Advisor); Yunzhi Wang (Committee Member); Michael Groeber (Committee Member); Maryam Ghazisaeidi (Committee Member) Subjects: Materials Science

Master of Science, Miami University, 2016, Mechanical and Manufacturing Engineering

Heat exchangers have seen widespread use in aerospace applications to properly manage aircraft and propulsion system heat loads. However, these heat exchangers are typically non-conformal, costly to implement, and large in terms of their volume and weight relative to the flowrate. There is motivation therefore for heat exchangers to be better integrated into the existing flow paths of the engine, both for packaging and performance improvements. Recent advances in additive manufacturing have presented the potential to create highly conformal and better optimized heat exchangers. In this work, a reduced-order computational tool was developed to predict the thermal-hydraulic performance of an annular two-pass, cross-counterflow heat exchanger. Area ruling was used to keep the velocities nearly constant in the heat exchanger core. The outputs of the model include heat transfer rate, outlet temperature, pressure loss, heat exchanger weight and volume, and heat exchanger effectiveness as a function of the user-supplied geometry and fluid inlet conditions. Various geometries were then simulated and analyzed using the computational tool, and based on these results, a prototype heat exchanger was created using additive manufacturing for future experimental testing. Computational fluid dynamics (CFD) simulations were also performed to assess the overall accuracy of the reduced-order model.

Committee: Andrew Sommers Dr. (Committee Chair); Amit Shukla Dr. (Committee Member); Edgar Caraballo Dr. (Committee Member); Todd Bailie Dr. (Advisor) Subjects: Engineering; Mechanical Engineering

Master of Science, Miami University, 2015, Computational Science and Engineering

Proper Orthogonal Decomposition was heralded as an objective means of extracting coherent structures from turbulent flows. Prior to its introduction, coherent structures were subjectively defined and discussed among experts. Since its introduction many reduced order models have been developed with inconsistent and often flow dependent validation procedures. This work sets up a framework for a data driven approach to validation of reduced order models derived from steady turbulent flows. Here it is proposed that the `goodness' of a model can be scored by how similar experimental and simulated data move through the model space. This is achieved by generating a Markov model for both data sets, using clustering techniques and maximum likelihood estimates. Results show increasing scores correlate with improved turbulent kinetic energy and modal amplitude for 3 data sets and 14 models. Additionally the generation of a surrogate Markov model can be used to identify missing dynamics in a simulation.

Committee: Caraballo Edgar Dr. (Advisor); Mehdi Safari Dr. (Committee Member); Andrew Sommers Dr. (Committee Member) Subjects: Fluid Dynamics; Mathematics; Statistics

Doctor of Philosophy, The Ohio State University, 2011, Mechanical Engineering

Localized arc filament plasma actuators have demonstrated significant potential in controlling high-speed and high Reynolds number jets in open-loop. The two primary goals of jet control are either noise reduction or bulk mixing enhancement. This research develops the tools for implementing feedback for this flow control system. The particular jet considered is a Mach 0.9 axisymmetric configuration with Reynolds number 670,000. The jet near-field pressure is well-suited for real-time non-intrusive observation of the flow state. Its response to forcing is similar to that of the far acoustic field. Forcing near the jet column mode results in amplification; forcing close to the shear layer mode yields attenuation. As a preliminary effort, two model-free feedback control algorithms are developed and implemented for online optimization of the forcing frequency to extremize the near-field pressure fluctuations. The steady-state behavior of the jet under closed-loop control matches the optimal open-loop results. However, the responsiveness of the controllers is poor since the dynamics of the jet are neglected. The first step in model-based feedback control is the development of a reduced-order model for the unforced jet. A cylindrical domain spanning the end of the jet potential core is chosen for the significance of its dynamics to the applications at hand. A combination of proper orthogonal decomposition and Galerkin projection is used to reduce the Navier-Stokes equations into a small set of ordinary differential equations employing empirical data. Extensive validation is performed on two existing numerical simulation databases of jets spanning low and high Reynolds numbers, and subsonic and supersonic speeds. Subsequently, a 35-dimensional model is derived from experimental data and shown to capture the most important dynamical aspects. The short-term prediction accuracy is found to be acceptable for the purpose of feedback control. The statistics from intermediate-ter (open full item for complete abstract)

Committee: Mo Samimy PhD (Advisor); Andrea Serrani PhD (Committee Co-Chair); Datta Gaitonde PhD (Committee Member); Jeffrey Bons PhD (Committee Member) Subjects: Acoustics; Aerospace Engineering; Applied Mathematics; Fluid Dynamics; Mechanical Engineering

Doctor of Philosophy, The Ohio State University, 2008, Mechanical Engineering

Controlling the flow over aerodynamic bodies has been a challenging problem for many years. Different open loop control techniques have been used in several flow configurations with some degree of success. However, in most cases the effectiveness of the controller is limited to the design conditions. In the present work, Proper Orthogonal Decomposition (POD) is used to derive low dimensional models of the subsonic flow over a cavity, in an effort to develop a feedback control system that can control the characteristic of the flow field. The Galerkin method is used as an additional tool to capture the time evolution of the flow field, reducing the problem into a system of ordinary differential. The stochastic estimation method is then used to link the variables that can be physically measured with those involved in the model. Particle Image Velocimetry (PIV) data and surface pressure measurement for the unforced flow (baseline) and for several open loop forcing conditions are used to derive the models. Three different approaches are investigated for control input separation. Different combinations of the flow condition are used in the model derivation to determine which forced flow should be used as a general case. A feedback controller is designed and tested experimentally for each model. The results showed that the variation in the experimental SPL spectra between the different models was negligible. However, a closer look at other factors hinted that the actuation mode separation method (M1) using the white noise forcing is the best choice. This method of separation does not require a clear identification of the control input region in the data. Also, it generates the best results in terms of reducing the tone and the OASPL while using a lower power input to achieve it. The white noise forcing helps to simplify the derivation process, as there is no need to pre-identify a specific forcing case. The multiple time estimation provides the best results in terms of the (open full item for complete abstract)

Committee: Mo Samimy Dr. (Advisor); Michael Dunn Dr. (Committee Member); Chiu-Yen Kao Dr. (Committee Member); Walter Lempert Dr. (Committee Member); Andrea Serrani Dr. (Committee Member) Subjects: Aerospace Materials; Engineering; Mechanical Engineering