Skip Navigation
Oklahoma State University

2nd Annual OSU MAE Graduate Research Symposium: Program

Talks will be held in room 270 of the OSU Student Union. Posters will be displayed in room 280 of the OSU Student Union.


07:30 - 08:00

Attendee Check-in & Breakfast (opening remarks by Drs. Fisher, Fore, and Santhanakrishnan)

08:05 - 09:10

Talks: Session 1; Room: 270 Student Union

Session Chair: Mitchell Ford, Ph.D. student & Secretary of MAE Graduate Student Council

(08:05-08:18):     Mitchell Ford - Effects of varying membrane area on bristled wing clap and fling
(08:18-08:31):     Saad Saleem - Development of a heat exchanger test facility to validate a heat pump simulation model
(08:31-08:44):     Kylar Moody - Development of a turbo-electric power system for multi-rotor unmanned aircraft
(08:44-08:57):     Timothy Emerson - Towards engineering metamaterials-inspired smart composites
(08:57-09:10):     Ryan Self - Online inverse reinforcement learning for linear and nonlinear systems
09:10 - 10:25

Posters: Session 1 & Coffee Break; Room: 280 Student Union

S M Abdullah Al Mamun - Data-driven feature-based sparse reconstruction of nonlinear fluid flows
Yuan Zhang - Determine the roles of material heterogeneity and thickness variability on the stability of thin membranes
Eric Brinkman - Lessons learned from developing low-cost sensor systems for ground-source heat pumps
Michael Harlan - Switched optimal control and dwell time constraints: a preliminary study
Sandesh Thapa - Cooperative aerial manipulation with decentralized adaptive force control
Shady Ahmed - A survey of data-assisted reduced order modeling techniques for predicting extreme events
10:30 - 11:22

Talks: Session 2; Room: 270 Student Union

Session Chair: Abraham Lee, Ph.D. student & Vice-President of MAE Graduate Student Council

(10:30-10:43):     Abraham Lee - Refrigerant charge and oil retention measurements of heat exchanger coils (ASHRAE RP1785)
(10:43-10:56):     Katelynn Harmon - On mitigating diabetic ulcer formation with smart elastomers
(10:56-11:09):     Saadbin Khan - Assessment of atmospheric plume source inversion using sparse reconstruction and gaussian dispersion
(11:09-11:22):     Phanidhar Chiluka - Studies on trough instability in nonwovens
11:30 - 12:20

Lunch; Room: 280 Student Union

12:30 - 13:22

Talks: Session 3; Room: 270 Student Union

Session Chair: Romit Maulik, Ph.D. student 

(12:30-12:43):     Romit Maulik - Physics-based artificial neural network formulations for LES of Kraichnan turbulence
(12:43-12:56):     Sam Allison - Wind velocity estimation using quadcopter trajectory deviations
(12:56-13:09):     Real KC - Study of conformal vortex generators via wake survey
(13:09-13:22):     Tyler Estrada - Local reorientation of liquid crystals in liquid crystal elastomers
13:22 - 14:40

Posters: Session 2 & Coffee Break; Room: 280 Student Union

Md Yeam Hossain - Effect of inlet duct configuration on the fan performance of air handling units (ASHRAE RP-1743)
Habib Ollah Boloorchi Tabrizi - Estimation on visual inertial odometry for simultaneously localization and mapping
Harsha Vardhan Reddy Vaddireddy - A survey of symbolic regression approaches to distill equations/models from big data
Kevin Coleman - Exploring invariant observers
Sk. Mashfiqur Rahman - Hybrid analytics paradigm in fluid dynamics
Mansoor Ahmed - Air mixer design guidelines using CFD analysis (ASHRAE RP-1733)
14:45 - 15:37

Talks: Session 4; Room: 270 Student Union 

Session Chair: Yasaman Farsiani, Ph.D. student & President of MAE Graduate Student Council

(14:45-14:58):     Eranda Ekanayake - Entropy analysis of gait parameters as a potential tool to predict osteoarthritis onset
(14:58-15:11):     Alvin Ngo - Evaluation of cylindrical asymmetric surface dielectric barrier discharge actuators for surface decontamination and mixing
(15:11-15:24):     Karthik Madhamshetty - Extraordinary wave manipulation characteristics of nonlinear inertant acoustic metamaterials
(15:24-15:37):     Trevor Wilson - Computational investigation of a low profile vortex generator
15:45 - 16:45 Undergraduate Research Poster CompetitionRoom: 280 Student Union

16:45 - 16:50

Closing remarks by Yasaman Farsiani, Ph.D. student & President of MAE Graduate Student Council 

16:50 - 17:00
Take down all posters

The symposium is sponsored jointly by: OSU School of Mechanical and Aerospace Engineering, OSU Graduate College, and the OSU Office of the Vice President for Research.

 

Abstracts:


 
Advisor: Omer San
 
Research Area: Thermal and Fluid Sciences
 
Title: AIR MIXER DESIGN GUIDELINES USING CFD ANALYSIS (ASHRAE, RP-1733)
 
Abstract: Globally, the increasing energy demand is one of the biggest challenge with cooling being a major part of it. Multinational research teams are collaborating to develop strategies for energy-efficient techniques. Computation Fluid Dynamics (CFD) is very effective and efficient tool, used to design and analyze fluid problems. It is used to investigate the flow characteristics e.g. pressure, velocity and mixing in flow systems. In 1960s, National Bureau of Standards initiated research on mixing devices to improve the techniques for measuring the capacity of heating, ventilation, and air-conditioning (HVAC) equipment. In the literature, limited guidelines are available for air mixers which provides further opportunities for research in this area. Existing and new models of air mixers will be investigated for their mixing effectiveness and pressure drop. The results of these investigation will provide more accurate guidelines for HVAC equipment capacity. Static air mixers add turbulence to the airstream to increase the mixing action. The pressure loss across mixers depends upon many factors including flow rate, size of the mixer and duct etc. Initially, air mixer consisting of baffles and louvers will be investigated. The results of these investigations will later use as baseline.
 

 
Advisor: Omer San
 
Research Area: Thermal and Fluid Sciences
 
Title: A Survey of Data-Assisted Reduced Order Modeling Techniques for Predicting Extreme Events
 
Abstract: Extreme events are typically high-intensity short-lived low-frequency intermittent instabilities occurring in dynamical systems. Rogue waves, earthquakes, and tornadoes are examples of extreme events occurring in nature. The prediction of these extremes beforehand is crucial since they pose severe and catastrophic effects both on humans and structures. This prediction requires rigorous analysis of relevant systems and accompanying complex dynamics. However, complete description of such systems is cumbersome because of its intrinsic high dimensionality, manifesting the indispensable need for robust reduced order models. The challenge is that the reliability of classical model reduction methods is limited as the truncated modes are usually essential for describing the underlying dynamics required for extreme events prediction and detection of their triggers. This imperfection in physics-based approaches may be effectively mitigated exploiting the available and fast-growing data streams. In our poster presentation, model order reduction techniques for predicting rare extreme events, particularly for geophysical flows, are surveyed. Also, possible enhancements of these techniques, incorporating machine learning tools and making use of available data, are presented.
 

 
Advisor: He Bai
 
Research Area: Dynamics and Controls
 
Title: Wind Velocity Estimation Using Quadcopter Trajectory Deviations
 
Abstract: Measuring wind velocity at various locations and heights in an area is currently a difficult problem. It requires special equipment, such as anemometers, mounted on UAS, weather balloons, or ground towers to get accurate wind velocity measurements. In practice, however, variations in wind velocity will result in deviations from a desired trajectory when flying a multicopter. Our research focuses on leveraging these trajectory deviation measurements in order to estimate wind velocities without the use of extra instrumentation on the multicopter. Specifically, we train a long short-term memory (LSTM) neural network (NN) to estimate the current wind velocity given a simulated quadcopter’s position and orientation at the current time step and at n previous time steps. We then use the trained NN to predict the wind velocity for data not in the training set and we quantify the prediction errors. In order to verify that the approach is generalizable, we train NNs for two different quadcopter controllers and examine the prediction accuracy for constant velocity winds and for winds generated using the Dryden model.
 

 
Advisor: He Bai
 
Research Area: Dynamics and Controls
 
Title: Estimation On Visual Inertial Odometry for Simultaneously Localization and Mapping
 
Abstract: Coordinates of a robot from a first-person viewpoint have different solutions. For example, you can use Stepper-motor or Optocounter module on wheels. However, we are not sure if wheels will slip and imagine what if we have a bipedal or a flying robot? How can we have both a coordinate system and map of the environment? Our Research is about using a camera and Inertial Measurement Units (IMUs) to give a robot the ability for simultaneous localization and mapping (SLAM). We cannot be sure whether each of them has a better answer so we use an estimation algorithm to see where we are as a robot. We apply an Iterated Kalman Filter with our devices to have a result for comparison. We are also trying to give IMU and machine vision answers (coordinates) as an input to Convolutional Neural Networks (CNN) and use ground truth coordinates as target data to train it for a better estimation. In other words, we give the robots an awareness about their coordinates by removing uncertainty which stems from the noisy data of IMU or our machine vision algorithm. In the future, we plan to have environments mapped using multi-agents by a variety of robots and also add attention to the perception system.
 

 
Advisor: He Bai
 
Research Area: Thermal and Fluid Sciences
 
Title: Lessons Learned from Developing Low-Cost Sensor Systems for Ground-Source Heat Pumps
 
Abstract: There is still relatively little data available to properly analyze the performance of Ground Source Heat Pump (GSHP) systems in residential applications. While commercial instrumentation is available, it is often expensive to deploy in residential systems. As a result, there continues to be a need for low-cost, properly installed sensor systems in residential GSHP systems. To collect meaningful performance data from any heating and cooling system, accurate measurements of heat transfer and electrical power must be made. Specifically, high-accuracy temperature sensors, flow sensors, and power transducers must be properly calibrated and installed, with appropriate data acquisition software, in order to calculate the COP of any system. To support wider collection of performance data for residential applications, we are developing a low-cost energy meter & data acquisition system and verifying its accuracy. This paper describes the lessons learned during the first stages of the research. One such lesson involves recognizing and accounting for cross-coupling effects during multiple channel ADC calibration at high frequencies, and how this coupling can considerably affect the resulting conversion parameters associated with the sensor.
 

 
Advisor: Shuodao Wang
 
Research Area: Solid Mechanics
 
Title: STUDIES ON TROUGH INSTABILITY IN NONWOVENS
 
Abstract: Instability of webs within process machinery can cause various problems including decreased web quality or productivity. For nonwovens, combined thickness and modulus variation can exist in many different regions in a span of a web. Therefore, nonwovens are much more prone to instability problems than uniform polyester webs or woven webs. Understanding in the effect of non-uniformity is needed to solve trough/wrinkling problems in nonwovens. In order to understand the localized instability problems associated with nonwovens, two major challenges remain: (1) how to characterize the statistically complex non-uniformity of nonwovens and how to relate the non-uniformity to the thickness and material property variations. (2) how to model trough formation in nonwovens if the information in the previous step is obtained?. The research objective is to identify the threshold of density variation above which trough instability occurs in nonwovens. Here, the density is chosen as an indicator of variations in effective properties (e.g. thickness and moduli) in nonwovens. This is achieved by performing experiments to characterize the density distribution in nonwovens, and adopt finite element analysis to find out the variation in density is related to trough instability.
 

 
Advisor: He Bai
 
Research Area: Dynamics and Controls
 
Title: Exploring Invariant Observers
 
Abstract: The Extended Kalman Filter (EKF) has become a popular and widely used version of the standard Kalman Filter, because of its ability to handle systems with non-linear dynamics. Recent studies have suggested incorporating the notion of invariance for non-linear observers. The idea is to design an invariant observer that exploits symmetries in the dynamics of a system in order to improve estimation accuracy. This led to the introduction of the Invariant Extended Kalman Filter (IEKF). Our research has focused on comparing the EKF and IEKF for a unicycle robot to confirm that the IEKF has greater estimation accuracy than the EKF. The goal is to understand the limitations of these observers and realize the conditions that must be satisfied to optimize local observability of certain systems.
 

 
Advisor: Jerome Hausselle
 
Research Area: Solid Mechanics
 
Title: ENTROPY ANALYSIS OF GAIT PARAMETERS AS A POTENTIAL TOOL TO PREDICT OSTEOARTHRITIS ONSET
 
Abstract: Osteoarthritis [OA] is a degenerative joint disease that affects about 27 million americans. OA onset correlates with age, gender, and biomechanical imbalances. Non-linear gait analyses provide a deeper understanding of movement variability. Approximate Entropy (ApEn) and Fuzzy Entropy (FuzzyEn) are used to characterize the amount of variability in data series. This study aimed at comparing the efficiency of ApEn and FuzzyEn as tools to assess gait variability. Comparisons were made between 20 healthy younger males and 20 females (18-25 years old). Subjects walked on a instrumented treadmill at a self-selected speed while kinetic and kinematic parameters were recorded. Our sensitivity analysis showed that FuzzyEn outputs are less sensitive to input parameters than ApEn outputs. Peak vertical ground reaction forces and hip, knee, and ankle ranges of motion exhibited a complex dynamic behavior, regardless of gender. This study provides baselines of expected variabilities amongst young healthy subjects. Future work will focus on quantifying gait variability amongst older adults and correlating it to OA onset, which is the critical step towards the development of simple, efficient screening protocols to detect patients at risk of developing OA.
 

 
Advisor: James Manimala
 
Research Area: Manufacturing and Materials
 
Title: TOWARDS ENGINEERING METAMATERIALS-INSPIRED SMART COMPOSITES
 
Abstract: Imagine an engineered structure that not only provides load bearing conformity and integrity, but also suppresses undesirable vibrations, scavenges useable energy and monitors its own status! Imagine such a structure being nearly entirely passive adaptive. Metamaterials-Inspired Smart Composites (MISC) could help realize such novel structural materials. Prototype MISC test articles were fabricated having face sheets bonded to additively manufactured polymer cores equipped with harvesting coils, sandwiching a chemically-etched plate. This plate consists of periodic arrays of re-entrant cantilever beam resonators with center-loaded neodymium magnets acting as the multifunctional kernel. Experiments demonstrate isolation of a payload from mechanical disturbances within tunable frequency bands as well as ability to harvest usable electrical power. Harvesting predictions using a coupled electromechanical model correlated with experiments. The circuitry could double as an active control system for resonators and also as a structural health monitoring system. Potential applications include structural materials for equipment or vehicles used in adverse or remote environments, where maximizing energy recovery and structural awareness in addition to payload isolation is desirable.
 

 
Advisor: Aurelie Azoug
 
Research Area: Solid Mechanics
 
Title: Local reorientation of liquid crystals in liquid crystal elastomers
 
Abstract: Liquid crystal elastomers (LCEs) are smart elastomers that contract like artificial muscles when they actuate. To prepare actuating LCEs, the material needs to be oriented in the direction of actuation by stretching. Uniaxial stretching is easily achieved, but greatly limits the programmable contraction of the material. In this study, we aim at measuring the local orientation of LCEs originating from a heterogeneous strain distribution. To achieve this goal, we performed Digital Imaging Correlation (DIC) experiments on a holed LCE sheet using an inexpensive set-up. The resolution and sensitivity of the setup to speckle patterning parameters and software parameters was assessed beforehand by comparing DIC results of a model material to finite element simulations. We then measured the local ordering in a LCE specimen using the previously optimized settings. The strain field and the liquid crystal orientation were both extracted from the recorded images, as the material becomes transparent when oriented. The ability to program heterogeneous orientations in LCEs will lead to complex actuation patterns and to new applications for these promising materials.
 

 
Advisor: Arvind Santhanakrishnan
 
Research Area: Thermal and Fluid Sciences
 
Title: Effects of varying membrane area on bristled wing clap and fling
 
Abstract: The smallest flying insects such as thrips show preference for bristled wings and the use of clap and fling kinematics for flight at Reynolds numbers (Re) on the order of 10. At these low Re, overcoming viscous drag is the greatest challenge to Clap and Fling. Because of this, bristled wing design is of particular interest. This study examines the aerodynamic effects of varying the ratio of solid membrane area (MA) to total area (TA) of the wing. Morphological analysis of published images showed that MA/TA ranges from about 14-27% in thrips. Bristled wing models with MA/TA varying from 15-100% were tested in a robotic clap and fling model at Re ranging from 10 to 120. Flow along the wing chord was visualized using 2D PIV, and strain gauges were used to measure lift and drag forces. At low Re, reducing MA/TA resulted in lower circulation around the leading and trailing edges during both clap and fling. In contrast, at Re=120 (relevant to larger insects such as the fruit fly), vortex shedding decreased circulation about the leading edge during clap and the trailing edge during fling in wings with high MA/TA. Lowering MA/TA resulted in reduced lift and drag forces, but increased peak lift to peak drag ratios across the tested Re.
 

 
Advisor: Rushikesh Kamalapurkar
 
Research Area: Dynamics and Controls
 
Title: Switched Optimal Control and Dwell Time Constraints: A Preliminary Study
 
Abstract: Control systems for continuous dynamical systems have been well explored, but most modern control systems are switched, meaning they have continuous as well as discrete decision variables, and these systems can have constraints on the switching rate, such as a compressor. Optimal control problems for switched systems can be solved by embedding the switched problem into a larger formulation, implementing the control variables as continuous which makes the embedded problem solvable using conventional techniques. However, the optimal solution to the embedded problem may have control values that cannot be matched by the switched system, resulting in a suboptimal solution. This paper describes a method to ensure that the solutions to the embedded problem are a bang-bang type, which can be directly converted to a switched control signal. This is done by placing a cost on the control signal that is high when it cannot be represented by the switched control, and low when it can, and the weighting on this cost function is directly related to the dwell time constraint for the system. Simulations are run for a switched optimal control problem with and without the developed solution to showcase this method, and work on the theoretical framework to support these results is being done.
 

 
Advisor: Aurelie Azoug
 
Research Area: Solid Mechanics
 
Title: On mitigating diabetic ulcer formation with smart elastomers
 
Abstract: Diabetes has drastic consequences on the quality of life of over 29 million Americans, notably due to ulcers. Diabetic ulcers form because of abnormally high pressures under the foot. To heal ulcers, patients wear constraining pressure-offloading devices. We aim at harvesting smart elastomer properties to create user-friendly pressure-offloading footwear that provides treatment and prevention of ulcer recurrence. Success of this project requires two distinct steps. First, the effect of ulceration on plantar temperature and pressure distributions must be quantified, as they constitute the stimulus for the activation of the smart elastomer. Healthy subjects walked on an instrumented treadmill at a self-selected speed while wearing an artificial ulcer containing a temperature sensor. Second, the response of smart elastomers to pressure and temperature depends on their orientation. We compared the pressure offloading ability of smart elastomers in various states of orientation. Understanding the correlation between plantar temperature, pressure, and smart elastomer activation is the key to reaching a solution to pressure ulcers.
 

 
Advisor: Christian Bach
 
Research Area: Thermal and Fluid Sciences
 
Title: Effect of Inlet Duct Configuration on The Fan Performance of Air Handling Units (ASHRAE RP-1743)
 
Abstract: ASHRAE and AHRI provides standard for designing outlet plenum for air handling and fan coil units’ performance tests but only limited guidance for inlet plenum design. ASHRAE merely suggests installing an inlet plenum if the space within the test room permits. Based on the inlet plenum configuration the performance of the fan varies with the velocity profile of the incoming air, which is affected by the selected inlet plenum configuration. Based on our experience, different testing facilities use different designs for inlet ductwork and sampling. This leads to differences in the measured performance and can lead to false testing failures. This poster presents a part of the work for the ASHARE RP-1743 research project which aims to reduce the length of the inlet plenum while maintaining fan performance within a close band of the base case. For part of the project we also worked on the velocity profile for different inlet plenum configurations and tried to improve the uniformity of the profile to get better fan performance of air handling units. This poster presents the velocity profile for various inlet plenum configurations and test conditions and includes fan performance (power and flowrate) for different scenarios.
 

 
Advisor: Brian Elbing
 
Research Area: Thermal and Fluid Sciences
 
Title: Study of Conformal Vortex Generators via Wake Survey
 
Abstract: The use of vortex generators is nothing new to the aircraft industries as they are used as a mean of flow control. A new form of vortex generators: Conformal Vortex Generators (CVGs) from Edge Aerodynamic present a novel, yet simple mean of flow control in a commercial aircraft. This presentation presents the study of CVGs via wake surveys in a wind tunnel. For this study, multiple CVG configurations were applied separately to an LA203A wing model. The configurations were varied by changing the length, width, and the thickness of the CVGs. The speeds at which the tests were conducted were also varied, thus changing the chord length based Reynolds number of the flow. The CVGs were able to generate strong coherent structures that persisted into the wake region five chord lengths downstream of the wing. Increasing the Reynolds number increased the strength of those coherent structures. Varying the CVG configurations changed the intensity of the coherent structures but the overall drag produced by the CVGs remained somewhat similar. This presentation will compare the coherent structures through velocity profiles and overall drag production of the different CVG configurations, clean wing with no CVGs, and backwards facing step profiles placed at different chord lengths of the wing.
 

 
Advisor: Balaji Jayaraman
 
Research Area: Thermal and Fluid Sciences
 
Title: Assessment of Atmospheric Plume Source Inversion using Sparse Reconstruction and Gaussian Dispersion
 
Abstract: Atmospheric plume source inversion represents a classical inverse problem where the goal is to recover leak rate and source location under highly turbulent atmospheric conditions when the concentrations measured at a few sparse locations are known. In this way, this problem is not dissimilar to learning physical parameters from data. A key to solving such inverse problems is to reformulate the system into a sparse basis space where the underlying mathematical problem is not ill-posed, not akin to sparse reconstruction methodologies. Using this analogy, one can employ an idealized Gaussian Dispersion (GD) model to relate the observed concentration measurements at sparse locations to the leak rate and source locations. However, it is well known that the idealizations underlying such models make them very restrictive. On the other hand, we can represent a dispersion model using the unsteady advection-diffusion transport equation that governs the plume evolution. In this work we combine our earlier work on sparse feature-based reconstruction with this numerical transport equation model to recover the source leak rate and location. The performance of this approach will be compared against the tradition GD formulation.
 

 
Advisor: Christian Bach
 
Research Area: Thermal and Fluid Sciences
 
Title: Refrigerant Charge and Oil Retention Measurements of Heat Exchanger Coils (ASHRAE RP1785)
 
Abstract: Tuning of the charge level of a split system air conditioner is typically performed after installation for near optimal performance through measurement of condenser subcooling. However, the same method is insufficient for heat pumps since the function of the indoor and outdoor coil are reversed. A fixed charge used in both heat-pump and air-conditioning mode often leads to sub-optimal performance unless dedicated charge balancing devices are included into the systems – which is generally not possible due to cost constraints. This research presents a new method for accurately measuring refrigerant charge and oil retention. The resulting high-quality reference testing data will allow tuning of simulation models to improve charge prediction accuracy. The tuned models will then allow targeted matching of indoor and outdoor coils to eliminate performance penalties if modes are switched. Moreover, such tuned models will allow better seasonal performance predictions than achievable with current rating standards and accurate oil retention predictions for reliable operation.
 

 
Advisor: James Manimala
 
Research Area: Dynamics and Controls
 
Title: EXTRAORDINARY WAVE MANIPULATION CHARACTERISTICS OF NONLINEAR INERTANT ACOUSTIC METAMATERIALS
 
Abstract: Since being postulated over a decade ago, inerters have greatly helped enhance dynamic performance of mechanical systems in many applications. Their ability to lend a high dynamic mass presence with only a relatively low static device mass make them unique among mechanical elements. Based on notional device schemes, frequency and acceleration-dependent models for nonlinear inertant acoustic metamaterials (NLIAM) are advanced and examined. An inverse square law inertance model ensures a bandgap over almost the entire frequency bandwidth of interest even encompassing the extremely low frequency regime, although such an NLIAM could be unrealistic to attain in practice. A practical power law approximation is shown to deliver >100% widening of bandgap towards lower frequencies. Further, drawing inspiration from Duffing-type stiffness, first order correction to the dispersion relation for NLIAM with acceleration-dependent, cubically nonlinear inertance is obtained using a perturbation approach. Excitation amplitude-activated bandgap shifts enable this NLIAM to act as a passive adaptive filter for mechanical waves. Practical manifestations of NLIAM could help realize extraordinary wave manipulation capabilities especially suitable for low frequency structural dynamic applications.
 

 
Advisor: Balaji Jayaraman
 
Research Area: Thermal and Fluid Sciences
 
Title: Data-driven Feature-based Sparse Reconstruction of Nonlinear Fluid Flows
 
Abstract: In many flow applications, recovery of the flow field from sparse sensor measurement is an area of need for enhanced analysis, modeling and data inversion to discover new knowledge. Sparse reconstruction, like many inverse problems, is ill-posed due to the need to recover a high-dimensional flow state from a low-dimensional and often, undersampled measurements. However, it turns out that most flow systems tend to have a low-dimensional structure in a feature space spanned by an appropriate basis. In many compressive sensing applications, these bases are Fourier functions, which allow reducing dimensionality. In nonlinear fluid flows such Fourier functions may not work well. Hence, we employ data-driven feature maps such as those obtained using singular value decomposition (SVD) and autoencoder networks based on Extreme Learning Machines (ELM). The ELM autoencoder represents a single hidden layer feedforward network (SLFN) and is an alternative for optimal data-driven basis such as SVD. It is usually known that SVD basis while energy optimal, may not capture all the relevant dynamics while non-optimal ELM basis may be able to capture more of the dynamics. In this study, we explore how these two classes of methods perform as sparse reconstruction tools for flow systems.
 

 
Advisor: Omer San
 
Research Area: Thermal and Fluid Sciences
 
Title: Physics-based artificial neural network formulations for LES of Kraichnan turbulence
 
Abstract: We devise a physics-informed combination of artificial neural networks which is utilized for the convolution and deconvolution of coarse-grained fields to account for sub-grid scale turbulence effects. It is observed that an effective eddy-viscosity is characterized by our purely data-driven large eddy simulation framework without the explicit utilization of phenomenological arguments. In addition, our data-driven framework does not require the knowledge of true sub-grid stress information during the training phase due to its focus on estimating an effective filter and its inverse so that grid-resolved variables may be related to direct numerical simulation data statistically. Both a-priori and a-posteriori results are shown for the Kraichnan turbulence case in addition to a detailed description of validation and testing. Our findings indicate that the proposed framework approximates a robust and stable sub-grid closure which compares favorably to the Smagorinsky and Leith hypotheses for capturing theoretical kinetic-energy scaling trends in the wavenumber domain.
 

 
Advisor: Kurt Rouser
 
Research Area: Aerospace
 
Title: Development of a Turbo-Electric Power System for Multi-Rotor Unmanned Aircraft
 
Abstract: The desire for longer range and endurance with multi-rotor unmanned aircraft has spurred interest in turbo-electric power generation as an alternative to batteries. The proposed solution capitalizes on the inherent higher energy density of hydrocarbon fuels, compared to batteries. An alternative hybrid gas-electric power system based on a piston engine is challenged by vibration and cylinder wall cooling requirements that are averted with a turbine engine. Though turbine engines typically have a higher brake specific fuel consumption than piston engines, they offer an advantage in higher power-to-weight, increasing potential fuel capacity. This current research presents the preliminary design and evaluation of a turboelectric power system, identifying technical challenges and expected electrical system efficiency. Results are shown from static bench tests at various throttle settings, revealing a correlation with electrical system efficiency.
 

 
Advisor: Jamey Jacob
 
Research Area: Thermal and Fluid Sciences
 
Title: Evaluation of Cylindrical Asymmetric Surface Dielectric Barrier Discharge Actuators for Surface Decontamination and Mixing
 
Abstract: In recent years, dielectric barrier discharge (DBD) has been used as an effective means of microbial inactivation. Surface DBD in particular have exhibited promising results for surface decontamination. A cylindrical SDBD configuration was evaluated to determine if the inherent induced body force could be leveraged to impel plasma species, such as reactive oxygen species (ROS) and reactive nitrogen species (RNS), as an apparatus to sterilize surfaces. The cylindrical structure is evaluated in this study to observe whether an increase in mixing is possible to efficiently distribute the plasma species thereby improving bacterial inactivation efficiency. Exposed electrode numbers of 1, 2, 3, and 6 were tested and associated to their respective degree of mixing. The same actuators were tested on a five-strain mixture of Salmonella enterica subspecies to observe their degree of population reduction. The increase of induced airflow of SDBD actuators with increased numbers of electrodes correlates with increased bacterial inactivation. These results suggest that by improving the particle velocity, airflow mixing tendencies, and plasma volume at the same power inputs results in increased surface decontamination efficiency.
 

 
Advisor: Christian Bach
 
Research Area: Thermal and Fluid Sciences
 
Title: A Review of Designs and Test Strategies for Air Mixers (ASHRAE RP-1733)
 
Abstract: Moist air is the heat transfer medium in heating, ventilation, air-conditioning, and refrigeration (HVAC&R) applications. Performance measurement of HVAC&R equipment requires to measure air flowrate and unit change in enthalpy of the air between inlet and outlet. Accurate psychrometric performance testing of HVAC&R equipment requires mixing processes to ensure an accurate average of temperature and moisture content of the air at the measurement stations. In this paper, mixing effectiveness, design, and pressure drop of mixers are described by a comprehensive review of the literature. This paper provides methods to evaluate mixing effectiveness with the use of several metrics of the range of temperature values and a statistical analysis. Next, the effects of test condition on mixing effectiveness is presented with various parameters, including temperature pattern, flowrate, and flow velocity. This is followed by a review of various mixer’s designs and their performance that effectively create uniform temperature and humidity of the airstream at the measurement stations. It is expected that this comprehensive review will be helpful for the development of practical solutions in ASHRAE RP1733 that can be readily applied to suit specific needs in the development of HVAC&R equipment.
 

 
Advisor: Omer San
 
Research Area: Thermal and Fluid Sciences
 
Title: Hybrid Analytics Paradigm in Fluid Dynamics
 
Abstract: In most of the cases so far in fluid dynamics community, physics-based modeling approach has been adopted by the researchers to develop a mathematical approximation of the underlying physical phenomena. On the other hand, the data-driven modeling approach has recently become popular due to the advancement in computing power as well as the rapid growth in the availability of large datasets through computations or experiments. The idea of hybrid analytics is emerged by combining these two distinct modeling strategies in ways that only the benefits of both approaches can be leveraged simultaneously. In this study, we present the potentials of hybrid analytics approaches in fluids research by showing their implantations on some active research topics, such as model order reduction of geophysical flows, development of fast incompressible flow solver, adaptive mesh refinement, etc. This study also surveys some recent developments and prospects in hybrid modeling techniques by introducing complex artificial neural network architectures like convolution neural network (CNN) or generative adversarial networks (GANs). The central perspective of this presentation is to give a brief idea on how hybrid modeling can be done and what advantages we can extract out of this route of research.
 

 
Advisor: Craig Bradshaw
 
Research Area: Thermal and Fluid Sciences
 
Title: Development of a heat exchanger test facility to validate a heat pump simulation model
 
Abstract: The vast majority of modern day air-conditioning, heating, ventilation, and refrigeration applications utilize vapor compression systems. Modeling of the fin-and-tube heat exchanger coils in these systems is a critical component of developing a predictive simulation platform that enables model-based design of new equipment. In order to validate the results of these models, high-fidelity experimental data is needed. Once validated, these models can be used to size equipment that is energy efficient, which not only reduces consumer cost, but also is more environment-friendly. This paper presents the design and construction of a custom-designed pumped refrigerant loop. This pumped refrigerant loop will be combined with a small-scale wind tunnel and an existing psychrometric chamber facility to enable the acquisition of high-fidelity data to validate a numerical model that is being developed for evaluating the performance of fin-and-tube heat exchangers. The pumped refrigerant loop allows precise maintenance of the desired test conditions and flow rate of refrigerant and has been sized to test heat exchanger coils up to a capacity of five tons.
 

 
Advisor: Rushikesh Kamalapurkar
 
Research Area: Dynamics and Controls
 
Title: Online Inverse Reinforcement Learning for Linear and Nonlinear Systems
 
Abstract: This research focuses on the development of an online inverse reinforcement learning (IRL) technique for a class of linear and nonlinear systems. Based on the premise that the most succinct representation of the behavior of an entity is its reward structure, IRL aims to recover the reward (or cost) function by observing an agent perform a task and monitoring state and control trajectories of the observed agent. Most research that has been done on IRL has been offline, which only allows for repetitive tasks and unchanging environments. The development of online IRL techniques, by allowing the machine to update its reward function in real-time, would help machines adapt to changes in the environment by correcting previously inaccurate information, and allow for a more dynamic response to unforeseen alterations in task objectives. The IRL method developed in this research is model-based IRL, as opposed to model-free because it is more data efficient, however, model-based IRL is dependent on accurate model knowledge. Therefore, a state and parameter estimator is developed to facilitate real-time reward function estimation. Inclusion of this parameter and state estimation enables accurate reward function estimation in the presence of both unknown dynamics and unknown state variables.
 

 
Advisor: He Bai
 
Research Area: Dynamics and Controls
 
Title: Cooperative Aerial Manipulation with Decentralized Adaptive Force Control
 
Abstract: We consider a group of aerial manipulators (AM) collaboratively transporting a flexible payload. Each AM is a quadcopter with a single rigid link attached to it. We develop an adaptive decentralized control law for transporting a payload with an unknown mass. The algorithm provides desired thrust and attitude angles required for each quadcopter to cooperatively transport a flexible payload. It also guarantees that all the agents converge to a desired velocity and regulates the contact force. The sum of the estimates of the unknown mass from all the agents converge to the true mass such that each agent gets equal share of the payload's mass. We demonstrate the effectiveness of the algorithm in simulations.
 

 
Advisor: Omer San
 
Research Area: Thermal and Fluid Sciences
 
Title: A survey of symbolic regression approaches to distill equations/models from big data
 
Abstract: Extracting governing equations from data can be viewed as reverse engineering of Nature- using data to identify the physical models. This approach is crucial where data is abundant (geophysical flows, finance, nueroscience etc) but the accurate physical law/model is not available. In recent years, the use of machine learning (ML) methods complemented the need for formulating mathematical models through the application of data analysis algorithms that allow accurate estimation of observed dynamics by learning automatically from the given observations and building models. The two popular methods in engineering and science community that use ML framework are Artificial neural networks (ANNs) and Symbolic regression based approaches. One problem of ANN-type approaches is the difficult-to-interpret black-box nature of the learned models. On the other hand symbolic regression algorithms are capable of learning/finding an analytically tractable function in symbolic form from the data which is considered of high valuable property. In the current paper we focus on surveying the symbolic regression based approaches such as genetic programming (GP), fast function extraction (FFX) applied to real world nonlinear dynamics system such as turbulence modeling of geophysical flows.
 

 
Advisor: Brian Elbing
 
Research Area: Thermal and Fluid Sciences
 
Title: Computational Investigation of a Low Profile Vortex Generator
 
Abstract: Vortex generators are known to provide several aerodynamic benefits including delayed separation, increased lift, and stabilized shocks by energizing the boundary layer. But, these benefits come at the cost of increased parasitic drag due to the geometry of the traditional vortex generator. The current work studies a low profile vortex generator, termed a conformal vortex generator (CVG), which can successfully energize the boundary layer without the increase in parasitic drag. This study utilizes a commercial computational fluid dynamics (CFD) package (Star-CCM+) to analyze the flow over the CVGs. Due to the low profile of the CVG, quasi-Direct Numerical Simulation (qDNS) was used to capture both large and small scale flow features. Using a flat plate developing boundary layer as inlet boundary condition, the sensitivity of the downstream flow field to the CVG geometry is assessed by varying CVG geometry and flow conditions. These results, as well as comparison with water tunnel experimental results will be presented.
 

 
Advisor: Shuodao Wang
 
Research Area: Solid Mechanics
 
Title: Determine the Roles of Material Heterogeneity and Thickness Variability on the Stability of Thin Membranes
 
Abstract: A lot of important synthesized thin structures, ranging from very advanced thin forms of stretchable electronics, porous media, metal alloy systems, to non-woven fabrics widely used in diapers and wipes, are becoming more and more complex in material composition and thickness variability. Characterization of those soft and thin materials is extremely difficult due to the relatively small dimensions of sample. The objective of this research is to develop a non-contact, in vivo measurement techniques for thin, soft, and heterogeneous materials. Fluorescent microscopy will be used to acquire the micrographs of sample surface from two different angels. The micrographs will be used to reconstruct, and track the 3D surface topography evolution under dynamic loading conditions. The full-field deformation of the surface will be determined, and then the reverse problem will be solved in finite element codes to extract the dynamic properties. The technique can be applied to study the mechanical properties and dynamic responses of most of the heterogeneous thin materials.