2021 Abstracts

Session 1 - March 16, 2021

Progressive damage in carbon fiber epoxy laminate composites via 3D in-situ tomography
Alejandra Ortiz
Structures and Materials, Graduate Student

Fiber-reinforced polymer composites have allowed engineers to reduce structural weight while maintaining desired mechanical properties. This had led to an increased interest for a damage tolerant approach for composite structures. However, they exhibit various damage mechanisms that can have complex interactions, making hard to predict how these materials will behave under loading. Using In-situ synchrotron X-ray tomography, damage initiation and progression can be observed at the local microstructure scale of a T650/5320 laminate composite with two lay-up configurations loaded in monotonic tension. Specifically, this methodology allowed for the observation of the micromechanical damage mechanisms and their interactions at the notch tip. Coupled with image processing algorithms, the damage was tracked as loading increased and the crack grew and exited the tomography window for the [-45/45/-45]S configuration, and just before fracture for the [+45]6 configuration. The analysis showed that intralaminar matrix cracking, a discontinuity along the fiber direction that exists through the thickness of a ply, was the dominant mechanism during crack initiation. Delamination, a discontinuity between two plies, became prevalent during later stages of crack progression. And fiber breakage, a discontinuity in a fiber, was largely led by intralaminar cracking.

Agent Based Modelling (ABM), Direct Operating Cost Model (DOC) Model Development for Urban Air Mobility (UAM) Vehicles
Ryan Howard and Ethan Wright
Aerospace Systems, Undergraduate Students

As Urban Air Mobility (UAM) vehicles are certified and introduced to the market, prospective UAM operators, like Uber, must assess which vehicles are best suited for a specific metropolitan network. In our literature review, we found no selection studies for UAM vehicles. In this research, we pose a methodology for addressing this need, develop a Direct Operating Cost (DOC) model, and apply agent-based modelling (ABM) tools. The DOC sets the service price point which determines who will travel via UAM. Our DOC model includes costs for vehicle and battery acquisition, energy, infrastructure, pilot and crew, and vehicle maintenance. For two of our candidate UAM vehicles, the DOCs per flight hour for the Wisk Cora and Archer Maker vehicles are $661 and $443, respectively. These results are comparable to industry estimates [1][2]. The next steps in our research are integrating the vehicle performance characteristics and DOC model into an ABM and determining the number of UAM-preferred trips for each vehicle. UAM-preferred trips are trips with a UAM segment that have a lower effective cost than an automobile trip to the same destination. The candidate UAM vehicle that generates the highest number of UAM-preferred trips will be the best-suited vehicle.

Session 2 - March 23, 2021

Solution Methods for Liquid Propellant Trapping in General Tank Geometries
Logan Walters
Aerodynamics, Graduate Student

As satellites become smaller, the need for satellite propellant tanks to use optimally shaped pressure vessels is declining, allowing use of different tank geometries that may more efficiently allocate space. However, for certain tank geometries some propellant will become trapped, preventing its use. This work aims to bring awareness to several ways in which liquid trapping can occur and describe solution methods that can be applied to analyze when and how liquid trapping happens in general tank geometries. A fluid statics code called Surface Evolver is used to calculate possible fluid configurations for a range of propellant volumes and contact angles. The special case of a cylindrical tank with ellipsoidal end caps is considered, as it is easy to analyze and demonstrates common types of liquid trapping. For this case, the possible configurations are a spherical liquid-gas interface, an asymmetric liquid-gas interface, or a liquid ring. Surface Evolver results are compared against analytical solutions for the spherical liquid-gas interface and liquid ring. These show excellent agreement, demonstrating Surface Evolver’s capabilities for handling liquid trapping problems. As non-standard tank geometries become commonplace, liquid trapping will need to be considered in the design of future satellite tanks.

The Effects of Wind Tunnel Noise on Hypersonic Boundary Layer Transition
Geoffrey Andrews Aerodynamics, PhD Student

The design of practical hypersonic vehicles – from futuristic transports to defense systems, to planetary entry vehicles – depends on accurate prediction of boundary layer transition. The difference between a laminar and turbulent boundary layer can be the difference between safe flight and catastrophic failure, as the onset of transition correlates to an order of magnitude increase in surface heat flux and skin friction. However, though understanding the behavior of hypersonic boundary layers is key to successful flight, reliably predicting transition from wind tunnel data is a technical nightmare, as conventional hypersonic test facilities produce acoustic noise which not only contaminates measurements but also affects the fundamental physics involved. Understanding the influence of facility noise on hypersonic transition is a major difficulty which has plagued designers of hypersonic vehicles for decades. The current work addresses this problem by using large-scale numerical simulation to resolve the effects of acoustic noise on past experiments performed in a conventional hypersonic wind tunnel. These experiments produced results that contradicted theoretical predictions of the time; however, computational modeling of the tunnel’s disturbance environment shows accurate prediction of the experimental behavior, paving the way for better correlation between conventional hypersonic ground test facilities and flight data.

Session 3 - March 30, 2021

A New Hybrid LES_RANS Method with Adaptive Anisotropic Downstream RANS Model
Wanjia Zhang
Aerodynamics, PhD Student

Using liquid fuel to cool down the compressor exit Turbulent flow pervades many engineering applications, and its prediction remains a challenge. One promising predictive method combines large-eddy simulation (LES) with simulation based on Reynolds-Averaged Navier-Stokes equations (RANS). This study developed a method to overcome stability and accuracy issues associated with hybrid LES-RANS methods. The method developed involves extracting Reynolds stresses from the upstream LES solution and then using that information to reconstruct and convert the downstream linear RANS model with a scalar eddy viscosity to a nonlinear RANS model with an effective anisotropic eddy viscosity. The method developed was evaluated by computing film cooling of a flat plate with the coolant injected through a row of circular holes. Results obtained show instabilities at the LES-to-RANS interface to be completely eliminated. Results obtained also show the LES-RANS method developed can yield solutions almost as accurate as those from LES, even though a significant portion of the flow is computed by the adapted anisotropic RANS model instead of LES, which greatly increases computational efficiency. Since the modification of the downstream RANS model is based on information extracted from the upstream LES solution, the method developed is adaptive to the problem being studied.

Efficient Inverse Optimal Control Enabling Objective Learning from Minimal Data, and Pontryagin Programming
Wanxin Jin
Autonomy and Controls, PhD Student

The notable success of artificial intelligence inspires a rethink of the significance of classic control theory: how can the rich structures and analysis in classic control theory benefit machine learning in its efficiency, explainability? The control-induced learning focuses on integrating both fields and developing innovative algorithms to answer the above question. In this presentation, I will talk about my two contributions in this direction. One is the efficient inverse optimal control to enable objective learning from minimal data and learning under distributed settings. The other is the Pontryagin Differentiable Programming, which bridges the concepts of control theory, deep learning, and reinforcement learning, and provides a unified end-to-end learning framework to efficiently solve a broad range of learning and control problems. In the second part of this presentation, I will focus on how to incorporate human factors into robot learning processes, to further promote the learning efficiency while maintain high levels of autonomy. Here, I will introduce two contributions: one is learning from human’s sparse demonstrations, which allows a robot to learn a task objective only from human-specified waypoints; and the other is learning from human’s directional corrections, which enables a robot to learn its objective from human’s directional feedback.

Session 4 - April 6, 2021

Long Term Effects of Test Anxiety on STEM Success in and out of Aerospace Engineering
Justin Major
Structures and Materials, Graduate Students

Test anxiety (TA) is the uneasy feeling students may experience anticipating an exam. In a prior path analysis (n=561 students, n=1 institution) using data from a larger study of student success (SUCCESS project; n=2672 students; n=17 institutions), I investigated the impact of TA on students’ first-year performances in science, technology, engineering, mathematics (STEM), and STEM overall. I found that TA unduly impacted all student GPAs. Additionally, effects were heightened for female and non-binary students; no differences were found for underrepresented minority (URM; not White or Asian) or low-income students. More recent work applied similar methods to the entire SUCCESS sample to predict second-year GPA, including whether results differed for aerospace students. I found that TA similarly impacted students’ GPAs with heightened effects for female and non-binary students. I also found that URM GPAs, while TA did not directly affect them, were lower. Finally, I found that results were unchanged for aerospace students, though there is an indication that their GPAs tend to be higher. These results support ongoing conversations regarding the stakes of testing. This study suggests that decreasing stakes will lead to student performance benefits, with additional benefits for female and non-binary students who are highly underrepresented in engineering.

Model Based System Engineering Tools Impact on Performance of Hypersonic Vehicle sub-systems
Brandon Smith
Aerospace Systems, Graduate Student

Hypersonic vehicles present a challenging design problem due to strong interdependencies between subsystems in the presence of significant size, weight, and power (SWaP) constraints. These dependencies need to be examined during the early stages of hypersonic vehicle conceptual design so that alternative design options can be identified and the requirements for subsystems can be specified. In this work, we demonstrate the use of a set of model-based systems engineering tools (MBSE), i.e., System Modeling Language (SysML) Systems Operational Dependency Analysis (SODA), Systems Development Depends Analysis (SDDA), and Robust Portfolio Optimization (RPO) for exploring the impact different design alternatives have on the performance of individual hypersonic vehicle subsystems, specifically guidance navigation, and control (GNC) and composite structures. We will show how SysML representations facilitate identification of interdependencies between subsystems which can subsequently be analyzed using quantitative system analysis methods such as SODA, SDDA, and RPO with the objective of improving the accuracy and pace of the conceptual design of hypersonic vehicles.

Session 5 - April 8, 2021

Using Plasma Spectroscopy to Measure Thrust of Micro propulsion Systems
Mart Hartigan
Propulsion, Undergraduate Student

Current conventional thrust measurement techniques for micro-propulsion devices are infeasible for suborbital flight, requiring a novel approach to be flown efficiently. Such an approach has been developed involving plasma spectroscopy to test the Film Evaporation MEMS Tunable Array (FEMTA) during an upcoming suborbital flight demonstration. FEMTA is a water-based micro-propulsion device for attitude control of micro-satellites. The thrust profile of this device is in the micronewton range, requiring more precision than typical techniques on board a suborbital trajectory. This technique involves a FEMTA firing into the entrance of a cathode tube once exposed to the rarefied gas environment of the upper atmosphere, generating plasma whose ultraviolet emission can be analyzed using spectroscopy to calculate thrust. To meet the launch provider’s electromagnetic interference and safety requirements, the breakdown voltage of water – which is a function of pressure and cathode length – must be minimized. To minimize the number of physical tests needed, Direct Simulation Monte-Carlo (DSMC) methods are used to model the experiment and iteratively solve for optimal cathode length given the pressure profile of the FEMTA plume. Preliminary computational and experimental results show that plasma spectroscopy is a viable method for thrust measurement in a suborbital environment.

Characterization of Computer Overclocking Performance at Cryogenic Temperatures
Payton Case
Aerospace Systems, Undergraduate Student

Computer overclocking is a process in which a user modifies hardware, firmware, and software to increase the speed at which a processor completes computational tasks. With increasing avionics requirements in the aerospace industry, overclocking is an attractive solution to improve system performance without the development of new computer hardware. The objective of this work is to characterize the stability of overclocking performance improvements of Central Processing Units (CPUs) at cryogenic temperatures. A test-bench was constructed and fit with a water-cooling loop. The computer was then subjected to various standardized benchmarks to represent real-world performance of the CPU. These initial benchmarks exhibited a linear trend between cooling capacity and performance with an expected 2.4% improvement in benchmark completion time for every 10 W of increased cooling capacity. The testbench was then fit with a Liquid Nitrogen (LN2) cooling system. At CPU temperatures below 150 K, stable CPU frequency was increased from 4.0 GHz in the water cooling tests to as high as 6.3 GHz yielding a 32.3% improvement in mean benchmark completion time. A range of cryogenic temperatures is identified over which significant computational performance can be achieved while maintaining CPU stability.

Seminar 6 - April 13, 2021

Efficient Inverse Optimal Control Enabling Objective Learning from Minimal Data, and Pontryagin Programming
David “Matt” Boston
Structures and Materials, PhD Student

Adaptable structures can be used to provide additional functionality and performance benefits to systems. This is particularly true for aircraft that exhibit multiple roles or operate in multiple flight regimes with conflicting design goals. Providing spanwise adaptability has been shown to be an effective means of increasing performance by taking advantage of changes in wing aspect ratio. Structural adaptability can be implemented through mechanistic means or compliant structures. The latter has been increasingly investigated for structural adaptation due its ability to couple mechanistic functionality with load bearing capabilities. Exploiting elastic instabilities in compliant structures additionally allows designers to take advantage of large deformations and changes in stiffness for shape adaptation. An exploration of a multistable, compliant honeycomb implemented for use in a beam-like extensible spar is presented. The honeycomb is characterized for its stiffness response in a variety of unit cell arrangements. The honeycomb is then implemented in a beam-like structure and compared to an equivalent plain honeycomb. A concept, hybrid morphing spar composed of traditional beam sections and the beam-like honeycomb is additionally presented. This morphing spar is subjected to simulated aerodynamic loading and the resulting deformation is compared to a conventional, rigid structure.

 Wavelet-optical Flow Velocimetry using Chemiluminescence Images 
Oluwatobi Busari
Propulsion, PhD Student

The subject of this talk is a new application of computer vision to combustion imaging. Computer vision consists of techniques that teach computers how to see. Applications to artificial intelligence have historically included feedback control for self-driving cars and trained segmentation of images. The application to combustion flames is novel, but not unique. We proceed as follows: Because the application is novel, the new results are described in detail. For example, combustion flames are overlaid on a turbulent flow, so that interpreting the motion of the combustion flames requires the separate identification of flame dynamics and flow dynamics. Such inferences have been difficult to identify previously, and the simplicity here is merely exchanged for stricter limits of inferences. The results allow us to make such claims as the role of reactivity on the flame attachment modes of a turbulent transverse reacting jet. These measurements are useful in designing cutting edge gas turbine combustors.

Seminar 7 - April 15, 2021

Swarm Optimization Techniques Applied to Combative Swarm Contexts
Katharine Burn
Aerospace Systems, Graduate Student

Swarms of unmanned aerial vehicles (UAVs) are undergoing rapid technological advancement for potential applications related to many different fields including delivery, transport, and combat. Because of this, several methods have emerged to characterize swarms including predicting UAV trajectories, swarm structure, and swarm intent. This research aims to investigate swarm characterization techniques in the context of mitigating combative swarms of UAVs. A crucial part of autonomous combative swarm mitigation is the ability to examine threatening swarms and extract key knowledge. Multiple types of information might be gathered regarding swarms and swarm elements to isolate potential threats and determine possible responsive actions to take. In an effort to synthesize this information, potential methods of interest may draw from techniques originally developed for unrelated fields. For example, velocity vector fields, statistical physics, and non-equilibrium thermodynamics have been used in the development of various swarm characterization methods [1,2,3]. Through this research, various methods will be investigated to characterize combative swarms. Based on the desired information to be extracted, potential costs and benefits will be examined for the different characterization methods. Consequently, recommendations for swarm characterization methods used for future applications will be made.

Modeling Subsystem Interdependencies for High-Speed Flight Vehicles Using System Operational Dependency Analysis
Eli Sitchin
Aerospace Systems, Graduate Student

High-speed flight vehicles pose a significant design challenge to systems engineers due to the strong interdependencies between constituent subsystems at high Mach numbers, as certain subsystem failures can cascade and jeopardize both mission success and the safety of the vehicle. Vehicle designers therefore need to scrutinize these interdependencies early in the design process in order to identify vulnerabilities in the system architecture. In this work, we explore the use of Systems Operational Dependency Analysis (SODA) in modeling subsystem-level interdependencies in order to identify vulnerable constituent subsystems. We will also discuss how we can use other model-based systems engineering (MBSE) tools – such as SysML – to construct system architectures that designers can model in SODA. These MBSE tools can be combined with traditional systems engineering tools such as failure mode and effects analysis (FMEA) to determine dependency relationships and strengths prior to constructing a SODA model.