136879 research outputs found
Sort by
Compositional Path Design for Graded Alloys Using Reinforcement Learning
No full textCompositionally Graded Alloys belongs to the category of Functionally Graded Materials (FGMs), distinguished by their varying spatial composition within the structure that results in material alloys of superior properties over traditional alloys. In the recent years, these compositionally graded alloys have gained significant recognition, primarily because of the advancements in the additive manufacturing techniques like Directed Energy Deposition (DED) which make the production of these alloys feasible. However, a linear gradient path of these alloys result in the inclusion of deleterious phases within the alloys micro-structure that can adversely affect the final alloy properties which may result in cracks & in the work done by Kirk et al [1] an innovative gradient path planning algorithm inspired by the state-of-the-art robotic route planning algorithms was adopted and a successful gradient path was designed in Fe-Ni-Cr material system which avoided the deleterious phases that was impacting alloy gradient when printed from 316L stainless steel to pure Cr. One significant drawback of this technique is that it limits the flexibility of material space exploration that could be done by the designer, any new composition exploration can only occur after re-configuring the path planner to compute the feasible gradient throughout the material domain and this limitation leaves designers with few alternatives. Q-Learning is a model-free Reinforcement Learning algorithm was used to solve this design space exploration problem in the ternary materials environment. An innovative method of encoding absolute position of the agent in barycentric coordinates along with the relative heading state to goal was formulated to model the system thus enabling design space exploration. The effectiveness of the proposed method was measured in a holdout dataset which produced a validation accuracy of 97%
Contrast-Independent Partially Explicit Time Discretization for Multiscale Problems and Its Application
No full textPartial differential equations are widely used in modeling and simulations. In many applications, there are high contrast changes in media properties. The implicit methods are typically used for temporal discretization and are unconditionally stable such that the time step size can be large during the computation. However, the implicit methods are more complicated to solve, especially for nonlinear problems. In contrast, the explicit methods are relatively easier to compute, while they require a much smaller time step size. This work focus on developing a contrast-independent partially explicit time discretization scheme for multiscale problems, specifically, nonlinear problems and also its application. The proposed method divides the spatial space into two components: contrast dependent (fast) and contrast independent (slow) spaces defined via multiscale space decomposition. Following this decomposition, temporal splitting is proposed that treats fast components implicitly and slow components explicitly. As a result, the scheme contains two equations, one implicit and the other explicit. The space decomposition and temporal splitting are chosen such that it guarantees a stability and formulate a condition for the time stepping. With the appropriate construction of spaces and stability analysis, we find that the required time step in our proposed scheme scales as the coarse mesh size, which creates a significant computational saving.
We first apply this approach on the parabolic diffusion reaction equations. We present numerical results and show that the proposed methods provide accuracy similar to implicit methods and the required time step size is independent of the contrast.
Nonlinear time fractional partial differential equations has wide application in physics and engineering. For the case of time fractional diffusion equations, the constraints on time steps are more severe and we extend our partially explicit methods to help alleviate this problem.
In our scheme, the implicit solution part can still be expensive, especially for nonlinear problems. Therefore, we introduce a modified partial machine learning algorithm to replace the implicit solution part of the original algorithm. Then we compute the explicit part of the solution using our splitting strategy. In addition, we use Proper Orthogonal Decomposition based model reduction to further improve our algorithms.
We extend the proposed scheme to solve multi-physics problems. We propose a partially explicit scheme with physics-based splitting. We take the convection diffusion equation as an example. In this scheme, we do a physics-based splitting, i.e. the convection equation is solved using the exact solution, and then we solve the diffusion equation via the partially explicit scheme
Comparing Single-Level Regression Based Methods for Analyzing Nested Data When the Extent and Location of Clustering Are Systematically Varied
No full textCollecting and analyzing clustered data is inevitable in educational research. Often, applied researchers will use traditional multilevel modeling to account for the non-independence of observations that result from clustering. However, multilevel models are more complex and have additional data requirements or assumptions relative to a single-level model that applies a correction to the standard errors. Using Monte Carlo Simulations, this work examines the efficacy of various single-level regression-based alternatives for analyzing clustered data when the data-generating model assumes predictors at the lowest level of clustering have both within and between-level variance as well as the effect of different centering choices ��� aspects missing from many contemporary works. The results of this study indicate that correcting standard errors using Taylor Series Linearization or the CR2 correction are the most flexible single-level regression-based methods for analyzing clustered data, as long as within-level predictors are group-mean centered, and the group mean for each cluster is re-introduced. Under these conditions, unbiased standard errors for within and between level estimates are recovered, even with few clusters. However, if within-level predictors are not centered, these corrections lead to increasingly negatively biased standard errors for the within-level estimate as the extent of clustering in the predictors increases, especially when the number of clusters is small. Next, fixed-effects modeling always recovered unbiased standard errors as long as slopes could be assumed to be non-randomly varying; even slight violations of this assumption lead to negatively biased standard errors for within-level estimates. Finally, examining nested data without correcting the standard errors (regular ordinary least squares regression) or using the DEFT correction is not recommended. These methods only lead to unbiased standard errors for within or between-level estimates under conditions that are not likely to be satisfied in real data analysis
Impact of Immune Reactions and Crosslinking Chemistry on Hydrogel Properties and Performance
No full textHydrogels are hydrophilic, three-dimensional structures that can be synthesized from synthetic and natural polymers for use in various tissue engineering applications. Their efficacy, however, is heavily dependent on the immune reaction to the material used. Another feature of hydrogels is their tunability in physicochemical properties, which can be modulated based on the crosslinking chemistries used to synthesize the structure. While often treated as being interchangeable, recent evidence suggests crosslinking chemistry���s potential to significantly impact material properties. Here, we sought to investigate the implications of polymer immunogenicity and crosslinking chemistry on hydrogel design and application.
Our first goal was to investigate poly(ethylene glycol) (PEG), a synthetic polymer considered to be bioinert, and how sensitization to the polymer impacts tissue engineering efficacy. To this end, PEG-based microporous annealed particle (MAP) hydrogels were assembled in situ of critical-sized calvarial defects through bioorthogonal tetrazine click reactions, and bone formation and morphology was found to be significantly influenced by PEG sensitization. Next, a head-to-head comparison of annealing chemistry used during MAP hydrogel assembly was further characterized between tetrazine-norbornene click reactions and radically-mediated thiol-norbornene click reactions. Differences in materials properties, such as storage moduli and susceptibility to enzymatic degradation, emerged as a result of annealing chemistry and were further modulated with respect to TNCP concentration. We deduced tetrazine-norbornene click products (TNCPs) induce secondary interactions that contribute to these changes, but these distinctions were negligible when applied in vivo.
Next, we evaluated whether the modulatory effects of TNCP-induced interactions could be applied in polymer-based biomaterials other than PEG. Specifically, we applied various concentrations of TNCPs to a hyaluronic acid (HA)-based bulk hydrogel and observed tetrazine-mediated changes to material properties. We also were able to leverage these interactions to assemble a supramolecular HA hydrogel that demonstrated shear-thinning and self-healing behavior that suggest potential as an injectable material, which was assessed in vitro using a genetically engineered strain of bacteria. Finally, TNCP-induced secondary interaction were leveraged in development of a hydrogel platform where a mock therapeutic was directly conjugated onto the HA backbone. Retention of the conjugated molecule and minimal degradation of the platform was observed after one week
Cryo-Em Structures of Single-Strand RNA Pepeviruses and the Interaction with Their Host Receptor, Type IV Pilus
No full textPositive-sense single-stranded RNA bacteriophages (ssRNA phages) have recently been reclassified because of the increase in the number of genomes discovered in metagenomes and metatranscriptomes. ssRNA phages infect gram-negative bacteria and are known to be structurally and genetically simple RNA viruses. Their infectivity is dependent on host retractile pili. Despite the discovery of ssRNA phages since 1960, only the structures of mature virions for F-pilus-specific E. coli phage (coliphage) MS2 and Q�� were solved by cryo-electron microscopy (cryo-EM). Mature virions of coliphages are composed of a single copy of maturation protein (Mat) and a capsid containing a single strand of genomic RNA (gRNA). Mat is an essential protein required for phage adsorption onto pili. However, ssRNA phages are diverse, as they infect several gram-negative bacteria and target different types of pili. Pseudomonas aeruginosa pepeviruses, PP7 and LeviOr01, infect their hosts through type IV pili (T4P). Here, we present the cryo-EM structures of pepevirus PP7 and LeviOr01, which are distinct from those of coliphages. These T4P targeting phages unexpectedly contain two Mat proteins in mature virions. One of the Mat proteins is internalized (MatIN) inside the virion, whereas the other is exposed (MatEX) for T4P binding. LeviOr01 exhibits plasticity in its assembly through Mat, in which particles with defined gRNA containing only a single MatIN and no Mat proteins are observed. Although mature PP7 virions always consist of two Mat proteins, a region of PP7-gRNA surprisingly displays high plasticity by adopting several conformations. This discovery underlines the importance of structural studies in mature virions and suggests that the structural simplicity of ssRNA phages might require reconsideration. We also solved the complex structure of PP7 bound to purified T4P from PAO1 to better understand host recognition and the entry mechanism. These findings provide insights into the exploitation of pepeviruses for phage therapy and biotechnological applications
A Deep Learning-Based Methodology to Re-Construct Optimized Re-Structured Mesh from Architectural Presentations
No full textDuring the mid���20th century, the emergence of novel technologies presented a paradigm shift for human intelligence. Computers, in particular, became instrumental tools for designers and architects, enabling the creation of intricate systems with freeform structures. These computers were capable of generating the final shape of designs by employing predetermined algorithms. Nevertheless, the prevalence of intricate freeform constructions is a notable characteristic of modern architecture. The design and calculation of these forms pose a complex challenge, as they deviate significantly from the original target industries in terms of aesthetics, statics, scale, and manufacturing technologies. While these structures may be intricate and, in some instances, praiseworthy, their architectural application necessitates a distinct approach. The formation of novel forms, shapes, and relationships within architectural design compositions can manifest creativity, leading to the exploration and discovery of innovative notions. Architects and designers are, therefore, inclined towards circumstances in which disparities hold significance. Architects may employ or adapt preexisting shapes as a foundation for their design endeavors. Architects draw inspiration from a particular image and integrate the associated notion into their design process, resulting in a more innovative architectural building. Current methodologies enable the meticulous hand alteration of forms through the utilization of documents or photographs. The proliferation of Machine Learning technology is augmenting architectural responsibilities, enabling swift and inventive results, expediting the design process, and developing a forward-looking approach to adjust and retrace actions instantaneously. The objective of this study is to assess the viability and accuracy of incorporating machine learning capabilities into defining an algorithmic-based methodology, with the goal of enhancing the design creativity process in architecture. This will be achieved through the utilization of image-to-mesh 3D reconstruction deep learning techniques, specifically applied to complex, irregular architectural structures. Additionally, a generative mesh optimization algorithm will be employed to generate a customizable mesh surface that can be further manipulated, paneled, and subjected to morphological operations
Influence of Rock Types on Porosity-Permeability Relations in Clastic and Carbonate Reservoirs with Application to CO2 Storage Site Characterization
No full textAccurate site characterization is essential for evaluating geological carbon dioxide storage potential. Geoscientists can model and monitor the behavior of injected carbon dioxide and rock interactions with knowledge of spatial variation of porosity and permeability. This thesis aims to estimate and understand permeability in carbonate and clastic reservoirs with geological analysis and acoustic well log data.
Jennings and Lucia (2003) model is used to calculate rock fabric numbers (����) in a carbonate reservoir in the Michigan Basin. By integrating information about the cored sections, three distinct classes were identified from rock fabric numbers. With Sun model (2004), a shear-frame flexibility factor (������ ) is calculated from acoustic properties and is used to relate permeability to rock pore structures. The shear-frame flexibility factor (������ ) is related to rock fabric numbers (����) through a linear transformation. The relations between shear-frame flexibility factor and rock fabric numbers will be very useful to estimate permeability from acoustic log and 3D seismic data, which will help predict CO2 pathways in potential CO2 storage sites.
This research also indicates that with sonic logs, we can calculate volume of shale in clastic reservoirs and relate to permeability. Permeability is controlled more by clay content in higher porosity zones. Higher volume of shale values indicates lower permeability values and that mechanical strength and pore structure play a greater role when constraining permeability values. Volume of shale can be related to elastic properties with Sun model (2004). The shear-frame flexibility factor can help constrain ranges of permeability in clastic reservoirs more accurately when porosity is at least 20%.
In addition, a fluid substitution model can be produced with Gassmann���s equations (1951). The impact of different fluid saturation changes is caused by CO2 injection on elastic properties and can be detected from synthetic seismic modeling and related to post-stack inversion results.
The theoretical results of this thesis are valuable for site characterization and locating potential CO2 storage, especially with the use of these rock physics models. With these results, geoscientists can use these methods to better comprehend the behavior of injected CO2 and rock interactions in the reservoir
Hardware Approaches for Enabling Multi-Channel Multinuclear MRI/MRS
No full textThis dissertation addresses the challenges faced in exploring X-nuclear possibilities in magnetic resonance imaging and spectroscopy. While proton-based studies are common, the interest in other X-nuclei has grown rapidly. However, limited support from existing MR systems and the challenges in designing multinuclear RF coils hinder these studies. This dissertation proposes a multinuclear RF coil setup and implements two system modification approaches to overcome these challenges.
In X-nuclear studies, the use of double-tuned or multi-tuned RF coils is crucial for examining multiple nuclei without adjusting the setup. This work introduces a three-frequency volume transmit and array receive RF coil setup. The volume transmit coils include a 1H birdcage coil and a double-tuned 2H and 23Na saddle coil, designed to operate individually. A broadband decoupled four-channel receive array is inserted, allowing simultaneous signal reception from all three nuclei. This setup is evaluated through parallel imaging at 1H, 2H, and 23Na frequencies, offering a potential solution for triple-tuned X-nuclear RF coils.
Expanding MR scanners to multinuclear array receiving capability is also explored. This dissertation presents a hardware phase correction solution for receive-only frequency translation, addressing phase incoherence caused by different frequencies between transmit and receive. The proposed hardware utilizes passive phase detectors, a direct digital synthesizer (DDS), and an Arduino microcontroller to automatically correct the received signal phase during the scan and eliminates the need for storing individual echoes or FIDs and performing retrospective phase correction.
Furthermore, the dissertation introduces a cost-effective multinuclear add-on system for traditional MR scanners, enabling simultaneous multinuclear experiments. This system includes multiple transmit and receive mixing channels and a four-channel flexible Local Oscillator (LO) source. By interfacing with the spectrometer, the scanner can transmit and receive at different frequencies simultaneously, reducing experiment time. The results demonstrate simultaneous multinuclear transmit and receive capabilities with 2H and 23Na gradient echo images and interleaved transmit and simultaneous receive ability for 1H, 2H, and 23Na FIDs, showing comparable signal-to-noise ratio performance as the single-frequency operation
The Effect of Mechanical Degradation on Sustained Fluoride-Releasing O-Rings: An In-Vitro Study
No full textTraditional methods for caries prevention focus on operator application or patient compliance, but the current study aims to establish a viable alternative to the current white spot lesion prevention techniques. O-rings incorporated with fluoride have been previously studied and assessed for a continuous release of fluoride, however, their ability to release fluoride after being mechanical degraded has not been assessed. The proposed research will provide more evidence on the efficacy of the O-rings and their ability to withstand mechanical degradation.
To assess the mechanical degradation these CaF��� O-rings they were brushed in a toothbrushing simulator for 2-weeks and 8-weeks in DIW at a force of 225g. Control and brushed calcium fluoride O-rings were placed on upper lateral incisor brackets, brushed, weighed, and soaked in distilled water. The amount of released fluoride was tested over a period of 7 weeks.
There were statistically significant differences between the control group and the 2-week and 8-week brushing groups throughout the entire duration of the study. The control and experimental groups dropped below the therapeutic range after the first week. When comparing the control group and experimental groups we see there were statistically significant difference between control and 2-week as well as control and 8-week groups at all time points in the study. All three groups experienced a tapering off of fluoride release. When looking at the cumulative release, the control group showed a statistically significant larger cumulative release of fluoride rate compared to both the 2- and 8-week groups (p<0.001). Weight analysis showed that there is a significant loss of material after mechanical abrasion when compared to before brushing. Additionally, there were several revisions to the protocol that were necessary to produce consistent results.
The 2-week and 8-week groups showed significantly less fluoride release over the duration of the study and had lower release rates when compared to the control, which can be affected by the coating protocol. In the first week of experimentation the 2-week group had a higher release rate, but this difference did not continue for the duration of the study
Superconductivity Near a Quantum-Critical Point: Analysis of the Discrete ��-Model at Finite Temperatures
No full textNear a Quantum-Critical Point (QCP) in a metal, strong Fermion-Fermion interaction mediated by a soft collective Boson gives rise to two competing tendencies, non-Fermi Liquid behaviour and Superconductivity which differs from the standard BCS theory. In this thesis, we consider a class of models, known as ����� Model, in which the effective interaction potential takes the form V (���) ��� 1/|���| �� . We introduce some numerical and semi-analytical techniques to analyze the behaviour near QCP. Based on the mapping between Eliashberg theory and the classical spin chain, we get a discrete Hamiltonian in terms of the angles made by individual spins. 2 different numerical approaches are introduced to solve the infinite number of coupled non-linear equations resulting from minimizing the Hamiltonian, which will yield the topologically distinct minima and saddle points. We find that the minimum energy state is always a superconducting state with winding number n = 0. Other superconducting states were also found, they represent topologically different pairings with different winding numbers. These states represent the saddle points of the free energy functional. We focus our analysis mainly on the case 1 < �� < 2 at finite temperatures