DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Post-esc-failure performance of a uam-scale hexacopter with dual three-phase motor
No full textMay 2024School of EngineeringDual three-phase motors are simulated on a 1200 lb hexacopter to examine its ability to tolerate any single electronic speed controller (ESC) failure. Redundancy is built into the powertrain, as each motor includes two sets of stators powered by electrically independent ESCs that drive a single motor shaft together. Control laws are tuned on the healthy aircraft to meet standard flying-qualities specifications. Power constraints are embedded in the controller to enforce both instantaneous and continuous power limits for each ESC. With the proposed powertrain, the aircraft is shown to be capable of trimming in all single ESC failure cases, including the case of aft failure in forward flight. Time-domain simulations demonstrate the aircraft's ability to re-establish trim and recover from any single ESC failure in hover and forward flight. Flying-qualities specifications are re-evaluated for the post-failure aircraft, showing minor degradation in several flying-qualities metrics. The post-failure aircraft's ability to follow commands without violating power constraints is verified through time-domain simulations. Thus, the benefits of implementing the dual three-phase motor on eVTOL aircraft are demonstrated by the hexacopter's ability to operate after any single ESC failure.M
Linearized conductivity reconstructions and ecg imaging
No full textMay 2024School of ScienceDuring the cardiac cycle, cells in the heart undergo a process of depolarization and repolarization which cause the chambers in the heart to contract and relax pushing blood through the heart and out to the lungs to be oxygenated and pushing oxygenated blood from the lungs to the rest of the body. The change in polarization is described by a current density that results in an electric potential that we can measure on the body's surface. The problem of determining the heart's electrical activity using measurements made on the body's surface is the problem of Electrocardiography (ECG). The problem has largely been studied using fixed approximations to the conductivity distribution in a person's chest. Conductivity changes in time are often overlooked such as the change in conductivity due to blood flow and air movement. To measure these effects, we can use data taken from an Electrical Impedance Tomography (EIT) system. EIT is an imaging method by which electrical currents are applied on the body's surface using a set of electrodes and the resulting voltages are measured. The goal of EIT is to be able to image the conductivity in a region of the body to identify structures or derive some information about the physiological activity that is happening in a specific area. Up to now, the ability to use EIT and ECG data to solve the problem of electrocardiography has been limited due to the speed at which EIT systems are able to collect data as well as the need for two separate systems to collect each type of data. However, the development of the ACT5 system allows us to simultaneously measure EIT and ECG data at a speed of approximately 27 images per second and a sampling frequency of 864 Hz. A typical cardiac cycle lasts around 0.8 seconds which means we can use the reconstructed conductivity distribution from the EIT data at approximately 21 instances in time to help solve the ECG problem. This thesis describes both the forward and inverse problems of EIT and ECG. Because the inverse EIT problem is ill-posed, we want to determine the best current patterns that allow one to produce the most informative images. We begin by looking at the stability and resolution of the linearized conductivity reconstruction algorithm using two common current patterns, trigonometric and dipole, applied in a two-dimensional circular setting using the Complete Electrode Model. This study extends a previous study done by David Dobson and Fadil Santosa that showed that the trigonometric patterns were more stable and offered more resolution when using the point electrode model. Next we look at the forward and inverse problems of Electrocardiography and how choices in boundary conditions, conductivity, and source modelling affect the accuracy of our reconstructions using a set of simulations. Finally, by showing that the conductivity has a non-negligible effect on the reconstruction of the heart's current sources, we motivate the EIT and ECG problem and show how we can develop an algorithm that can be applied to human subjects.
The results obtained from the subject data show how boundary effects and conductivity influence the path of the total cardiac vector, or the total current produced by the heart, over a single cardiac cycle. When analyzing the path we see clusters of small magnitude vectors that correspond to atrial depolarization, a ring of large magnitude vectors that correspond to ventricular depolarization and a well-defined loop of moderate magnitude vectors that correspond to ventricular repolarization. We also look at using multiple dipole sources to describe the electrical activity of the heart and how using multiple dipole sources can improve our ability to reproduce a partial body surface map. To help interpret the results, this data is presented alongside an approximated lead II electrocardiogram and partial body surface map interpolated from the voltages measured by the ACT5 system.Ph
Flight dynamics and handling qualites of electric quad-rotor aircraft for urban air mobility applications
No full textMay 2024School of EngineeringThe growth of Future Vertical Lift and Advanced Air Mobility programs has led to thedevelopment of a large variety of novel electric Vertical Take-Off and Landing (eVTOL)
aircraft concepts. In this thesis, the viability of the use of a quad-rotor configuration for
Urban Air Mobility (UAM) applications is examined based on flight dynamics and handling
qualities considerations.
An isolated, variable-RPM rotor is first considered, and an examination is conducted
of the effects of increasing rotor size on response time and power consumption, due to
the increasing rotor blade inertia. With variable-RPM, larger rotors may not be responsive
enough for effective operation without relatively over-sized power-trains. In order to examine
the effects of scaling up typical quad-rotor platforms on power-train requirements, variable-
RPM quadcopters of increasing size are simulated in hover. Motor weight fractions are
found to be notably higher for the larger quadcopters, highlighting the relative lack of control
authority of variable-RPM inputs for producing changes in thrust. The effects of disk loading
on the control authority are also considered. Increased disk loading results in increased trim
power, but is shown to significantly reduce estimated motor weight.
As an alternative to variable-RPM alone, a UAM-scale quadcopter with a hybrid control
scheme is proposed to improve the flying qualities of the quadcopter. Hover and forward
flight analysis is performed on a single-passenger, UAM-scale quadcopter with both variable
rotor speed and collective pitch control. With these redundant controls, power consumption
can be increased to improve authority of the pitch inputs for changes in rotor thrust. Hybrid
control mixing is implemented using a complementary filter, allowing the aircraft to use
pitch-control for short-term responses and RPM-control for trim. The benefits of this hybrid
control scheme are demonstrated through simulation of hot/high/heavy conditions, where
trimming with RPM-control allows the pitch actuators to maintain margin for maneuvers.
Hybrid control allows the aircraft to reap the benefits of pitch-control for maneuverability,
while maintaining stall margin by using RPM-control for trim.
In order to perform pilot evaluations of the handling qualities, a real-time capable
simulation model is developed for the hybrid control quadcopter. Four control modes are
flown by test pilots in a flight simulator. Three hybrid control configurations (Eco, Standard, and Sport modes) are evaluated, along with a baseline variable-RPM case. The RPM
control case is shown to be undesirable to pilots due to its increased delays associated with
rotor acceleration. Standard and Sport modes perform the best overall, though some pilot
comments suggest that Sport mode may be overly aggressive. Power-efficient Eco mode has
the lowest Handling Qualities Ratings due to its slow response and reduced magnitude of
commands. Overall, it is shown that the hybrid control quadcopter is able to significantly
outperform its variable-RPM counterpart in piloted simulation of Mission Task Elements.Ph
Deciphering Crypto Twitter
No full textCryptocurrency is a fast-moving space, with a continuous influx of new projects every year. However, an increasing number of incidents in the space, such as hacks and security breaches, threaten the growth of the community and the development of technology. This dynamic and often tumultuous landscape is vividly mirrored and shaped by discussions within “Crypto Twitter,” a key digital arena where investors, enthusiasts, and skeptics converge, revealing real-time sentiments and trends through social media interactions. We present our analysis on a Twitter dataset collected during a formative period of the cryptocurrency landscape. We collected 40 million tweets using keywords related to cryptocurrency and performed a nuanced analysis that involved grouping the tweets by semantic similarity and constructing a tweet and user network. We used sentence-level embeddings and autoencoders to create K-means clusters of tweets. We identified six groups of tweets and their topics to examine different cryptocurrency-related interests and the change in sentiment over time. For example, we identified different groups of tweets demonstrating coordinated behavior in the market or expressing distrust in centralized cryptocurrency exchanges. Moreover, we discovered sentiment indicators that point to real-life incidents in the crypto world, such as the FTX incident of November 2022. We also constructed and analyzed different networks of tweets and users in our dataset by considering the reply and quote relationships and analyzed the largest components of each network. Our networks reveal a structure of bot activity in Crypto Twitter and suggest that they can be detected and handled using a network-based approach. Our work sheds light on the potential of social media signals to detect and understand crypto events, benefiting investors, regulators, and curious observers alike, as well as the potential for bot detection in Crypto Twitter using a network-based approach
Gradient-enhanced robust design optimization for engineering systems under uncertainty
No full textMay 2025School of EngineeringEngineers are interested in numerical robust design optimization (RDO) during early phases of the design process for its ability to produce system configurations that are both optimally performant and robust to sources of uncertainty in fabrication, operation, and analysis. Prohibitive to this practice is the often intractable cost of accurately estimating statistical or probabilistic measures of the optimization objectives or constraints from repeated sampling of analysis codes, particularly for problems involving expensive high-fidelity or multi-disciplinary analyses. Given the recent proliferation of efficient gradient calculation methods within these analysis codes, we explore and contribute to gradient-based methods for accelerating robust design optimization, simultaneously leveraging gradient-enhanced uncertainty quantification (UQ) and gradient-based optimization techniques. To this end, we develop a novel gradient-based partition-of-unity surrogate model and adaptive sampling method tailored to robust design optimization. Furthermore, we propose a multi-fidelity robust optimization method that uses surrogate-based UQ and adaptive sampling-based gradient error estimates to mitigate the cost of sampling effort as the optimizer approaches convergence. For a suite of analytical test problems and an aerostructural design problem involving uncertainties related to shock-boundary layer interaction, the novel surrogate and adaptive sampling method demonstrate competitive to superior global accuracy per sample compared to standard surrogates, and the proposed multi-fidelity optimization method demonstrates greatly-reduced sampling effort to achieve design convergence overall when compared to single-fidelity benchmarks.Ph
Assessing membrane performance and modeling transport of solute through membranes: from nanofiltration to membrane chromatography
No full textAugust 2024School of EngineeringMembrane technology has played a key role across different fields, facilitating the treatment of essential consumables (e.g., drinking water, therapeutic medicines, dairy products and beverages). Membrane processes can be categorized into process that optimize solvent throughput, such as water (solvent-targeted), or maximize solute recovery, such as therapeutic medicines and vaccines (solute-targeted). Meeting the growing demand for these products requires a better understanding of membrane processes, including strategic membrane selection and modification.Membrane processes involve complex phenomena that are challenging to understand and control. As a result, membrane process development often involves empirical iterative or trial and error approaches, which consume significant time and resources. Mechanistic models provide a route to address these challenges, improve process understanding, and guide optimization. One way to optimize membrane processes for specific applications is to develop new membrane materials, to provide better permeability, better selectivity, or better resistance to changes in productivity during filtration (fouling). Therefore, it is critical to have a framework for evaluating membrane performance that is unbiased.
The work described in this thesis addresses these challenges in three phases. The first phase evaluates transient behavior during filtration, using nanofiltration (NF) as an application domain, and specifically as an example of a “solvent-targeted” application. While membrane processes are often designed for steady-state operation, they often exhibit transient or non-steady behavior caused by solute adsorption, leading to overestimates of selectivity during startup, and posing risks of water quality deterioration due to desorption when feed concentration changes. Dye removal by NF in both organic and aqueous phases is used as a model system to investigate transient selectivity (i.e., anomalously high rejection) during the early stages of filtration, due to molecule adsorption onto membrane surfaces, and steady-state rejection that arises after the membrane's sorption capacity has been saturated.
A comprehensive transport model was developed, including diffusion, convection, physical (steric) partitioning, adsorption, and varying boundary conditions. Static binding experiments were performed to evaluate dye adsorption onto each NF membrane, and the resulting equilibrium liquid and solid phase concentrations were correlated using the Langmuir isotherm. Flux decline because of solute adsorption was observed, which was described well by assuming that the permeability was reduced in proportion to the mass of solute adsorbed (i.e., the Langmuir fouling model). Integrating both the isotherm and fouling model into the transport model enabled the description of transient selectivity in the form of breakthrough curves that were asymptotic to steady state sieving. The transport model was evaluated using both local equilibrium Langmuir isotherm, and Langmuir adsorption kinetics, which account for second order adsorption and first order desorption. A comparison of these two approaches revealed that only the kinetic model could accurately capture the entire breakthrough curve and steady-state rejection values, suggesting a sorption rate limited process.
The second phase of this thesis focus on a “solute-targeted” membrane processes to capture mRNA, used for vaccines. We have developed a mechanistic mathematical model to describe the dynamic purification of mRNA using microporous membranes modified by graft polymerization of polythymidine (Oligo dT) to promote affinity interactions. The model accommodates different membrane configurations (flat sheet and hollow fiber) and was extended to describe other chromatography techniques including monoliths and traditional resin bead columns. To better understand the response of the membrane adsorber, we first measured the overall dispersion and other mixing phenomena in the system, including process tubing and the membrane column, using transport of conservative tracers. Dispersion and transport in the system were modeled using a combination of mixed (CSTR) and plug flow (PFR) domains (so-called ideal reactors).
The elution of captured mRNA using an elution buffer to weaken affinity interactions was described using a convection-dispersion-desorption model with Langmuir desorption kinetics, where the desorption rate constant was modified to account for the changing buffer composition (e.g., salt concentration) as a function of time during elution. The model was calibrated using pure mRNA solution and then validated by predicting capture and elution of mRNA from a complex mixture after in-vitro transcription (IVT) using the flat sheet membrane. The simulation results demonstrate the model’s predictive capability. We found that desorption is the rate-limiting mechanism during elution, providing guidance to improve elution process efficiency. The study demonstrates the usefulness of the developed model to help understand complex transport phenomena, and to help solve industrial challenges by expediting process development and facilitating scale-up studies.
The third and final phase of this thesis introduces a framework to assess membrane performance without bias, using productivity, fouling potential, and energy consumption as metrics. Membrane scientists and engineers invest significant effort in synthesizing new membranes, often involving the optimization of surface chemistries to enhance selectivity and minimize the potential for retained species that could decrease membrane permeability (i.e., induce fouling). Furthermore, they routinely conduct comparisons of membrane performance across various applications, selecting the most suitable membrane for scaling up in a particular application When different membranes under consideration have different permeability values (or resistances, Rm), traditional plots of flux versus time for constant pressure operation, or pressure versus volume in constant flux operation are biased, and often do not produce a valid assessment of performance. In this study, we elucidate how Rm affects fouling kinetics across various fouling mechanisms and experimental protocols. We demonstrate that traditional plots often obscure two important performance criteria: productivity, defined as the accumulated volume throughput, and energy consumption. The practical implications of our findings include the possibility of overestimating the performance of membranes with lower productivity or higher energy consumption. Additionally, screening studies may inadvertently select lower-performing membranes under the mistaken belief of superior performance.
We have introduced two novel methods to assess membrane performance in an unbiased manner. In the first approach, we have developed a graphical approach using normalized coordinates. This approach yields linear plots with slopes solely dependent on fouling parameters, and are unbiased, i.e., independent of membrane resistance Rm. Therefore, these normalized coordinates, customized for each fouling mechanism and operational mode, effectively isolate fouling potential. In a second approach, we developed new graphical approaches to visualize the potential trade-offs between better antifouling performance but lower membrane permeability by examining either productivity (volume throughput) for constant pressure operation, or specific energy for constant flux operation. Our study establishes a comprehensive framework for evaluating membrane performance under fouling conditions, incorporating fouling, energy consumption, and volume throughput as metrics, independent of membrane resistance. This framework offers valuable guidance for process design and the development of antifouling membrane materials.
Overall, the work reported in this thesis presents an investigation into various membrane processes and applications, including both "solvent-targeted" nanofiltration and "solute-targeted" membrane chromatography applications. These two applications were explored and interconnected through the modeling of solute transport behavior within the membrane. Additionally, we present a framework for evaluating membrane performance under fouling conditions, with a focus on optimizing solvent productivity.Ph
Effective Data Distillation for Tabular Datasets
No full textData distillation is a technique of reducing a large dataset into a smaller dataset. The smaller dataset can then be used to train a model which can perform comparably to a model trained on the full dataset. Past works have examined this approach for image datasets, focusing on neural networks as target models. However, tabular datasets pose new challenges not seen in images. A sample in tabular dataset is a one dimensional vector unlike the two (or three) dimensional pixel grid of images, and Non-NN models such as XGBoost can often outperform neural network (NN) based models. Our contribution in this work is two-fold: 1) We show in our work that data distillation methods from images do not translate directly to tabular data; 2) We propose a new distillation method that consistently outperforms the baseline for multiple different models, including non-NN models such as XGBoost
Efficient upwind and partitioned implicit/explicit finite difference schemes for the second-order wave equation on overset grids
No full textMay 2024School of ScienceIn this thesis, we discuss multiple finite-difference methods to solve the wave equation in second-order form on complex geometry discretized by overset grids. We begin by incorporating upwind dissipation into the modified equation (ME) scheme in an efficient and accurate manner in order to suppress instabilities that can grow from small perturbations in the computation, as the ME scheme does not have strong dissipative effects. For stiff problems, we derive an implicit reformulation of the ME scheme to give unconditional stability. A hybrid implicit-explicit ME scheme is also devised for such problems, especially for those with domains containing localized areas of finer grid resolution that can be caused by intricate geometry. From a numerical perspective, one significant challenge for discretizations of the wave equation is their lack of inherent dissipation mechanism. As a result, perturbations, such as those arising from overset grids, can lead to numerical instability. These instabilities can be eliminated through the addition of artificial dissipation devices, and for wave equations in second-order form, there are number of existing formulations. Current methods include ad-hoc dissipation (FDA), which adds an artificial dissipative operator into the ME scheme, and the more recently developed upwind schemes (UW). Both options have been shown to stabilize grid solutions, but existing FDA methods may loose an order of accuracy when strong instabilities are present and existing UW formulations are computationally expensive and complicated in implementation. Here we devise the upwind predictor-corrector (UPC) scheme at general even-order to introduce upwind dissipation into the ME scheme in a simple, accurate, and efficient manner. In this approach a higher-order upwind operator, derived from the modified equation analysis of the UW scheme, is incorporated into the ME scheme through a modular corrector-style stage. The higher-order operator maintains the underlying ME scheme’s accuracy as well as its time-step stability restriction, and takes on the dissipative properties of the UW scheme, while avoiding much of the UW scheme’s high cost and complex implementation. The scheme is extended to curvilinear coordinates, and numerical tests illustrate the UPC scheme’s accuracy and stability on single and overset grids in one, two, and three dimensions. Various CPU run time comparisons on rectangular and curvilinear grids demonstrate the UPC scheme’s computational speedups over the UW schemes. Explicit schemes, however, can be inefficient for stiff problems due to their time-step restrictions. Such stiffness may arise in problems with small grid cells associated with meshing for complex geometry or in algorithms requiring long time integrations. To address these situations, we derive the implicit modified equation (IME) schemes at second- and fourth-order accuracy. These schemes adopt the compact stencil of the ME scheme and introduce implicitness by applying spatial operators to a time-average approximation spanning three levels in time. Truncation error analysis reveals a parametric class of schemes, and stability analysis identifies those with unconditional stability. The upwind operator derived from the UW scheme is also incorporated into the IME schemes to ensure stability on overset grids. Three different methods to incorporate upwinding are investigated, and stability analysis reveals conditions for stability. Numerical results confirm the accuracy and stability of all the methods. Results on overlapping and curvilinear grids in one, two, and three dimensions demonstrate the stability of the dissipative schemes. For the case of stiffness induced by locally refined grids, we devise the partitioned implicit-explicit (PIE) scheme. In this approach we use local implicit time stepping in areas of the domain exhibiting stiffness, and explicit time integration elsewhere. The time step is then selected based on the explicit scheme, which in this case is the ME scheme. Numerical analyses and example calculations demonstrate the stability of the method. The PIE scheme is also made dissipative by replacing the IME and ME schemes with their upwind counterparts, the UPC scheme and the upwind IME schemes. Numerical tests on overset grids in one, two, and three multiple dimensions confirm the accuracy and stability of the schemes. Simulations on complex geometry show the PIE scheme’s superiority in computational efficiency over the UPC scheme in stiff problems.Ph
Implications of microglial and macrophage nadph oxidase 2 (nox2) enzyme activation in neuroinflammation
No full textDecember 2023Sustained neuroinflammation has been shown to be a major driving force in the progression of various neurodegenerative diseases. Microglia and infiltrating macrophages in the central nervous system play an important role in the progression of neuroinflammation through the release of a variety of inflammatory factors. Reactive oxygen species (ROS) released by these immune cells are one such type of inflammatory factor that result in oxidative stress, a hallmark of neuroinflammation. Thie release of ROS is facilitated by the enzyme NADPH Oxidase Isoform 2 (NOX2) present in both microglia and macrophages. Along with neuroinflammation, circadian disruption is another characteristic feature of neurodegenerative diseases. While the effect of neuroinflammation and circadian disruption on the progression of neurodegenerative diseases have been studied individually, knowledge of the interaction between the two is still lacking.In this thesis, the implications of NOX2 activation were explored with relevance to neuroinflammation by investigating its interaction with the microglial and macrophage circadian clock and other inflammatory mediators. First, the effect of NOX2 activation and inhibition on the microglial circadian clock was established. For this, the BV2 mouse microglial cell line was used as a representative cellular model. We were able to show that inhibition of NOX2 resulted in the retention of the circadian clock function in BV2 microglia even under pro-inflammatory activation. Second, the observations made in BV2 microglia were confirmed in primary mouse bone marrow-derived macrophages. We found that similar to BV2 microglia, NOX2 inhibition resulted in an active circadian clock in macrophages under pro-inflammatory activation. Third, having established the interaction between NOX2 activation and the circadian clock, we investigated how the pathological nature of the macrophages affected their response toward NOX2 activation and inhibition. For this, macrophages from mice subjected to circadian disruption and neurodegenerative disease-mediated neuroinflammation were subjected to NOX2 inhibition. We found that NOX2 inhibition was able to reduce ROS levels irrespective of the macrophage type. Finally, we investigated the interaction of NOX2 with another important inflammatory mediator in microglia, cyclophilin A. For this, the mouse BV2 and human HMC3 microglial cell lines were used as cellular models. We obtained preliminary information indicating the potential interaction between NOX2 and cyclophilin A in microglial cells. In combination, these studies provide mechanistic insights into the different ways in which NOX2 could be targeted to reduce microglia and macrophage-mediated inflammation with relevance to neuroinflammation.Ph
Deadlock Freedom in Actor Languages
No full textDecember 2023We introduce a framework using session types for denoting the relationships between multiple actors in a system. We prove that ascribing to this framework guarantees deadlock freedom, and demonstrate tests of how an implementation of such would work in the SALSA language