Acoustics, Dynamics, and Controls Commons^™

Analysis Of Closed-Loop Dynamical Systems Using Relu Nns, Pouya Samanipour

Theses and Dissertations--Mechanical and Aerospace Engineering

This thesis develops a framework for analyzing the stability and invariance properties of closed-loop dynamical systems whose dynamics are derived from neural networks with rectified linear units (ReLUs). The ReLU neural network defines continuous piecewise affine (PWA) functions, which makes them ideal for analysis using polyhedral geometry and optimization techniques. To begin, we introduce two exact algorithms for enumerating polytopic regions of affine behavior induced by ReLU networks: a half-space-based method that checks the feasibility of activation patterns and a geometric vertex-based method. Using these methods, we are able to perform precise region-wise analysis, and these methods serve as the …

Simulation-To-Hardware Validation Of Mpc For Binary Thruster 2d Cubesat Control, Aevar Amundinusarson Oefjoerd

Theses and Dissertations--Mechanical and Aerospace Engineering

Small satellites must execute precise, fuel-limited maneuvers under strict actuation and operational constraints. Model Predictive Control (MPC) is a suitable approach because it optimizes future actions while explicitly enforcing these limits. This work evaluates (1) whether MPC can operate in real time using binary (on/off), asymmetric thrusters, and (2) whether a high-fidelity digital twin can be used to tune and validate the controller prior to hardware testing.

An MPC framework was implemented on a planar satellite prototype that floats on air bearings, operates at 16.7 Hz (60 ms period), and uses eight binary thrusters exhibiting up to 13.2% variation in …

Knowledge Integrated Deep Learning For Enhanced Fault Diagnosis And Prognosis For Rotating Machinery And Its Applications On Marine Systems, Yunsheng Su

Dissertations, Master's Theses and Master's Reports

This dissertation develops a knowledge-informed deep learning framework for robust fault diagnosis and prognosis across various engineered systems, including bearings, lithium-ion batteries, and mooring systems in wave energy converters (WECs). A deep transfer learning (DTL) method is first developed to improve bearing fault diagnosis by fusing data from multiple sources and extracting key features using convolutional neural networks (CNNs). Building upon this, a knowledge-informed deep network (KIDN) framework is proposed, integrating physics-based features with deep learning to enhance diagnostic accuracy and generalizability. A constrained Gaussian process (CGP) model is further employed to guide adaptive network design and reduce model tuning …

Toward Precise Long-Range Underwater Acoustic Geo-Positioning: Utilizing Vehicle Data And Deepening The Gnss Analogy Through Uncertainty Modeling, Isaac B. Salazar

Open Access Master's Theses

Electromagnetic signals, such as those used by Global Navigation Satellite Systems (GNSS), attenuate dramatically underwater, but acoustic signals can travel hundreds of kilometers, and can be used for positioning in much the same way. Concepts from GNSS can be applied to the subsurface context, as at a basic level, the principles of geo-positioning are identical. Key challenges in translating satellite positioning models to the underwater acoustic domain include differences in signal type as well as instrumentation and propagation environment. Acoustic signals travel at much slower speeds and are subject to significant environmental variability due to complex ocean dynamics. Here, the …

Impacts Of A Multi-Layered Acousto-Elastic Bottom On Sound Propagation In A Seamount Environment, Haley H. Green

Open Access Master's Theses

The New England Seamounts Acoustics (NEMSA) experiment is a comprehensive oceanographic field study aimed at investigating underwater acoustic propagation and scattering effects within a complex marine environment. The primary objective of this research is to assess the influence of seafloor and sub-bottom geoacoustic properties on acoustic propagation, signal reflections, and transmission loss at Atlantis II Seamount. Two acoustic models, BOUNCE and BELLHOP, were employed to incorporate elastic wave effects within the sub-bottom layers and facilitate the analysis of range-dependent bathymetric features. Two model scenarios were analyzed and compared, one with a layered elastic bottom and another with a fluid bottom …

Head Mounted Surgical Robot, Mason Currens

Williams Honors College, Honors Research Projects

The purpose of this design project is to utilize a six degree of freedom, robotic system that will be head mounted device that can either assist in, or perform, facial and optical surgery. The six degree of freedom idea came from current systems used in flight simulations. By optimizing the system to fit the figure of a patient’ s head, many of the present errors in surgical robotics can be reduced. One of the major errors when using a facial tracking robot is sudden impulses, such as a cough or a twitch, that the robot can't precisely track. Mounting the …

Process To Design And Analyze Dynamic Environment Test Fixtures (Pdadye), Cora J. Taylor

Dissertations, Master's Theses and Master's Reports

Vibration qualification testing is an important requirement in many industries. In-lab testing is intended to evaluate a component’s life expectancy by exposing it to a dynamic environment similar to what it may experience in its field condition. To successfully perform these tests, it is imperative that the field boundary conditions (BCs) of the component of interest are replicated as closely as possible in the lab. One approach to achieving accurate dynamic environment testing is to design a fixture that sufficiently represents the impedance of the next-structure (field structure). The process to design and analyze dynamic environment test fixtures (PDADyE) is …

Stepper Motor Load Estimation Using A Neural Network, Soham S. Kulkarni

Dissertations, Master's Theses and Master's Reports

Bipolar stepper motors are common in scenarios that require low-cost precision including 3d printers. The ability to estimate stepper motor load provides an opportunity for health monitoring. For example, excessive load estimates may indicate a future problem in the driven system such as bearing wear. Stepper motor drivers often have built-in stall detection strategies that exploit the driver's precise knowledge of the high-frequency voltage pulses being sent to the motor. The goal of this work was to develop a retrofit approach to stepper motor load estimation using externally measured voltage, current, and speed. A stepper motor dynamometer was created to …

Dynamic Force Feedback Controller Improves Metrics Of Stepping In A Biomimetic Robot Leg, Isabella Mae Kudyba

Graduate Theses, Dissertations, and Problem Reports (ETD)

Force feedback is a type of robotic control where forces are measured, fed back into the control system, and actuation is adjusted to compensate for environmental forces, with the goal of quick and precise limb positioning in walking. Robots of today use single force sensors (e.g. pressure sensitive pads, load cells) per leg which are often bulky, expensive, and susceptible to crosstalk. Walking robots often employ model predictive controllers or policy controllers trained through reinforcement learning (RL) to actuate their limbs, but these methods are time and energy consuming. Biomimetic robots offer an alternative approach, investigating biological theories with the …

Reliability Of An Extended Version Of The 3m™ Eargage Tool To Assess Earcanal Size And Assist Earplug Selection, Bastien Poissenot-Arrigoni, Laurence Martin, Alessia Negrini, Djamal Berbiche, Olivier Doutres, Franck Sgard

Études primaires

Objective: Evaluate the ability of an extended version of the 3 MTM Eargage to estimate the earcanal size and assess the likelihood that a particular earplug can fit an individual’s earcanal, ultimately serving as a tool for selecting earplugs in the field. Design: Earcanal morphology, assessed through earcanal earmolds scans, is compared to earcanal size assessed with the extended eargage (EE) via box plots and Pearson linear correlations coefficients. Relations between attenuation measured on participants (for 6 different earplugs) and their earcanal size assessed with the EE are established via comparison tests. Study sample: 121 participants exposed to occupational noise …

Modeling The Noise Attenuation Effects Of Bubble Curtains At Coastal Virginia Offshore Wind Installation, Gavin Dies

Open Access Master's Theses

The Coastal Virginia Offshore Wind (CVOW) pilot project consists of two turbines roughly 40 km off the coast of Virginia Beach, Virginia. Water- and seabed-borne acoustic and seismic signals from impact pile driving at CVOW were recorded during the installation of these turbines in May 2020. In-water pressure signals were measured using various methods at multiple ranges. The water depth at the wind turbines and the measurement devices was approximately 26 m. During installation, one of the monopiles utilized a double bubble curtain, while the other monopile did not. This arrangement has allowed for an analysis of the bubble curtain’s …

Reverse Geometric Dispersion In The Beaufort Sea: Dissecting Sound Propagation At Long Ranges In Ducted Profiles, Jessica Desrochers

Open Access Dissertations

Acoustic propagation in the Arctic is evolving as surface temperatures increase and ice melts, resulting in a changing sound propagation environment. This study investigates the acoustic arrival structure in the Canada Basin and its connection to the underlying sound speed profiles. Two Seagliders were deployed in the summer of 2016 and 2017, diving in a sawtooth pattern collecting conductivity, temperature, and depth (CTD) data as part of the Canada Basin Acoustic Glider Experiment (CABAGE). These data are used to calculate sound speed profiles used to predict and understand acoustic arrival structures.

The warming of the Pacific Summer Water and surface …

Reactive Model-Free Control For Underactuated Vehicles Operating In Resource-Constrained Environments, Austin Lane Sponaugle

Graduate Theses, Dissertations, and Problem Reports (ETD)

Buoyancy-driven, underactuated vehicles offer a compelling solution for long-duration autonomous operations in constrained environments. By leveraging buoyancy to generate lift, these vehicles can remain aloft for extended periods with minimal power requirements, making them well-suited for applications in atmospheric monitoring, planetary exploration, and communications. However, their limited actuation authority, strong coupling with environmental forces, and operating domains introduce significant control challenges, including nonlinearity, sensitivity to external disturbances, and limited observability.

This work investigates the design and implementation of lightweight, reactive model-free control strategies that enable robust autonomy in such constrained settings. A digital Extremum Seeking Controller (ESC) is first implemented …

Smart Materials For Noise And Vibration Damping In High-Speed Rail Systems: A Comparative Analysis, Hyginus Chidiebere Onyekachi Unegbu, Danjuma Saleh Yawas, Bashar Dan-Asabe, Abdulmumin Akoredeley Alabi

Makara Journal of Technology

Effective noise and vibration control remains a critical challenge in high-speed rail systems, directly influencing passenger comfort and the longevity of infrastructure. This study evaluated four advanced materials—piezoelectric materials, shape memory alloys (SMAs), magnetorheological (MR) fluids, and damping composites—focusing on their potential for mitigating noise and vibration in high-speed rail applications. A combination of experimental and simulation-based analyses was employed to assess these materials based on their noise reduction coefficient, vibration transmissibility ratio, thermal stability, and durability under varying environmental conditions. The findings revealed that damping composites and SMAs demonstrated superior performance, offering enhanced noise attenuation and vibration control compared …

End-To-End Autonomous Quadcopter Using Reinforcement Learning, Mohamed Marwan Chawa

Theses and Dissertations

This thesis investigates the potential of Reinforcement Learning (RL) for achieving robust and adaptable quadcopter control, focusing on trajectory and attitude stabilization. We compare state-of-the-art RL algorithms, specifically Proximal Policy Optimization (PPO), against traditional Proportional-Integral-Derivative (PID) controllers across three tasks: hovering, slow trajectory following, and fast trajectory following. To enhance realism, we employ a modified PyFlyt simulation environment with a high-fidelity Crazyflie 2.x model, accounting for motor dynamics, noise, wind disturbances, and aerodynamic drag.

The challenge of operating a quadcopter can be divided into two distinct parts: planning a flight path and actually following that path. Our focus is on …

Hitch Cart “Landing Gear”, Rebekah White, Jose Raygoza, Randy Hernandez, Brandon Leon

Mechanical Engineering

This report aims to allow our sponsor, to review our design process of the Hitch Cart Landing Gear Prototype. In the design overview section of this report, we discuss the primary design modifications we made to the wheel mechanism of the existing hitch cart prototype, including the addition of the ACME screws and the folding brackets. This allows our sponsor to see the intended improvements made to the past prototype and understand the primary goal of our project. Then, in the implementation section, we cover the entire manufacturing process to allow our sponsor to understand what manufacturing steps must be …

Characterizing West Florida Shelf Reef Fish Communities Using A Scientific Echosounder, Edmund A. Hughes

USF Tampa Graduate Theses and Dissertations

The advancement and application of non-invasive fisheries survey technologies, such as the scientific echosounder, have significantly enhanced our understanding of fish populations and their habitats, more recently in regions where traditional sampling methods are limited, such as complex reef habitats. This research, conducted as part of the Continental Shelf Characterization, Assessment, and Mapping Project (C-SCAMP) on the West Florida Shelf, leverages acoustic and visual technologies to investigate the distribution of fish densities, fish target strength and length relationships, and the relationship between seafloor topography and fish biomass distribution.

The first study compared reef fish densities estimated from near-concurrent acoustic (Simrad …

Robust Data-Driven Run-To-Run Control Via One-Step Constrained Optimization For Automated Serial Sectioning, Rhianna M. Oakley

Mechanical Engineering ETDs

This Thesis develops a one-step predictive run-to-run controller (R2R-MPC) for automating a mechanical serial sectioning (MSS) system. MSS is a destructive material analysis process that iteratively removes slices of material and captures 2D images, reconstructing them into 3D representations. Commonly used in material science for characterizing materials and failure analysis, MSS typically operates in an open-loop fashion, which experiences high variability in material removal due to system disturbances. To address this, a robust closed-loop R2R-MPC is presented, modeling MSS process uncertainty using a linear differential inclusion identified from operational historical data. The R2R-MPC is posed as an optimization problem that …

The Impact Of Hearing Aid User Speech On Environment Classification And Own Voice Listener Preferences, Robert Andrew Budinsky Iii

USF Tampa Graduate Theses and Dissertations

Hearing aids (HA) have improved over the past few decades with advanced signal processing, such as environment classification and background noise reduction, but are still limited by challenges related to when the user is speaking. These challenges include the potential interference a HA user’s voice has on device environment classification and the potential influence a HA has on own voice (OV) sound quality. However, the effects of OV on HA environment classification and signal processing are not well established, nor are the desired HA program settings regarding OV listening preferences. In this study, a combination of electroacoustic and behavioral methods …

Computer Vision In A Robotic Arm, Jack Maxwell

College of Engineering Summer Undergraduate Research Program

We used a machine learning-based object detection algorithm to give a robotic arm the ability to "see" with its camera.