MICDE 2017 posters
The version of the browser you are using is no longer supported. Please upgrade to a supported browser.Dismiss

View only
Still loading...
First Name Last Name Role DepartmentTitleAbstractCoauthors
RyanFarberGraduate StudentAstronomyGalactic Winds and Cosmic Ray Transport in a Multiphase Insterstellar MediumMaking up roughly one third the pressure budget of the ISM, cosmic rays are likely to play a fundamental role in galaxy evolution. Recent 3D magnetohydrodynamic simulations have shown that advected cosmic rays puff up galactic disks and suppress star formation. Additionally, cosmic rays diffusing away from the galactic midplane can drive gas out of the galaxy with mass loss rates comparable to the star formation rate, thus regulating star formation. Yet, the impact of cosmic rays decoupling from cold, neutral gas in a multiphase interstellar medium has hithertofore not been studied. Preliminary work suggests that cosmic ray decoupling produces significantly more explosive feedback, dramatically affecting the evolution of the ISM and the efficiency of cosmic ray driven outflows. Mateusz Ruszkowski, Karen Hsiang-Yi, Ellen Zweibel
HoutanJebelliGraduate Student
Civil Engineering, Electrical Engineering
Measurement of Construction Workers’ Emotions Using a Wearable EEG DeviceConstruction workers are exposed to high mental work load and stress because of their involvement with physiologically and psychologically demanding tasks performed in a hazardous work environment. Working under stress significantly influence workers performance, awareness, and motivation, which could lead to serious health and safety problems. Recently, a electroencephalogram (EEG) sensor has drawn significant attention to measure the stress level of human subjects in the clinical domain. However, using EEG at the field is challenging since the recorded EEG signals can be easily contaminated from diverse artefact sources in construction fields. To address this issue, we propose an effective and comprehensive signal processing framework for; 1) acquiring high quality EEG signals by using signal processing methods such as filtering methods (e.g., low pass filter, high pass filter, and notch filter) and an independent component analysis method to remove prevalent artefacts at a jobsite (e.g., eye blinking, vertical eye movements, and facial muscular movements) ; and 2) monitoring construction workers’ stress level at the field. We applied the proposed framework to real world construction projects analyzing the data from eight construction workers from the field. The results confirmed the potential of a wearable EEG device to capture high quality EEG signals and measure workers’ stress at the field, which opens a door to assess workers’ stress level at construction sites.Houtan Jebelli1, SangHyun Lee2
ZechengGanPost-DocMathematicsBoundary Integral and Image-Moment Hybrid Method for Simulations of Solvated ProteinsSolving the Poisson/Poisson-Boltzmann (PB) equation with varying coefficients has been a bottleneck for simulating proteins within the framework of implicit solvent model due to the dielectric inhomogeneity. We present our recent work on a numerical Boundary Integral Equation (BIE) method, and a semi-analytical Image-Moment hybrid method for efficient and accurate simulations of electrostatic fields in systems consisting multiple proteins in solvent. The BIE is coupled with treecode for fast kernel summation, and is applicable to arbitrary shaped dielectric interface; while the hybrid method combines analytical image charge solution of dielectric spheres with the Method of Moments, and is accelerated by the fast multipole method (FMM). Numerical results from both methods are presented, showing that the induced charge on the dielectric interfaces can significantly change the interaction energy of solvated proteins.Zecheng Gan, Weihua Geng, Robert Krasny
AryaFarahiGraduate StudentPhysicsOn the Search for Lead Pipes in Flint We detail our ongoing work in Flint, Michigan to detect pipes made of lead and other hazardous metals. After elevated levels of lead were detected in the drinking water of Flint residents, as well as increased levels of blood in area children, the state government directed over $25 million to replace “water service lines,” which are the pipes connecting each home to the water system. In the absence of accurate records, and with the high cost of determining buried pipe materials, we put forth a number of predictive tools to aid in the search and removal of lead infrastructure. We discuss how our approach generalizes beyond Flint, and how similar methods are becoming increasingly important as the federal government has begun to increase spending on water and other infrastructure development.Arya Farahi, Jared Webb, Eric Schwartz, Jacob Abernethy
Sambit DasGraduate StudentMechanicalLarge Scale Electronic Structure Studies on the Energetics of Dislocations in Al-Mg Materials System and Its Connection to Mesoscale ModelsWe study the dislocation core in Aluminum and Magnesium using a local real-space formulation of orbital-free density functional theory, implemented using finite-element discretization. The framework affords the use of Dirichlet bulk boundary conditions enabling direct calculation of isolated dislocation core energetics. So far our studies on edge and screw dislocations in Aluminum, suggest that the core size—region with significant contribution of electronic effects to dislocation energetics—is around seven to ten times the magnitude of the Burgers vector. This is in stark contrast to estimates based on atomistic calculations. Interestingly, our study further indicates that the core-energy of both edge and screw dislocations are strongly dependent on external macroscopic strains. Similar studies in Magnesium are ongoing. Next, we use the electronic structure core energetics information to develop a continuum energetics model for a discrete dislocation network, which accounts for the core energy dependence on macroscopic deformations, and from the variations of the core energy with respect to the nodal positions of the network, we obtain the nodal core force which can directly be incorporated into discrete dislocation dynamics frameworks. Using case studies involving simple dislocation structures, we demonstrate that the contribution to the core force from the core energy dependence on macroscopic deformations can be significant in comparison to the elastic Peach-Koehler force even up to distances of 10-15 nm between dislocation structures. Thus, these core effects, whose origins are in the electronic structure of the dislocation core, can play an important role in influencing dislocation-dislocation interactions to much larger distances than considered heretofore.Sambit Das, Vikram Gavini
BikashKanungoGraduate StudentMechanicalLarge-scale all-electron density functional theory calculations using an enriched finite-element basis
EricHarperGraduate Student
Materials Science and Engineering
Freud: a software suite for high-throughput simulation analysisComputer simulation is an indispensable tool for the study of a wide variety of systems. As simulations scale to fill petascale and exascale supercomputing clusters, so too does the size of the data produced, as well as the difficulty in analyzing these data. We present Freud, an analysis software suite for efficient analysis of simulation data. Freud makes no assumptions about the system being analyzed, allowing for general analysis methods to be applied to nearly any type of simulation. Freud includes standard analysis methods such as the radial distribution function, as well as new methods including the potential of mean force and torque and local crystal environment analysis. Freud combines a Python interface with fast, parallel C++ analysis routines to run efficiently on laptops, workstations, and supercomputing clusters. Data analysis on clusters reduces data transfer requirements, a prohibitive cost for petascale computing. Used in conjunction with simulation software, Freud allows for smart simulations that adapt to the current state of the system, enabling the study of phenomena such as nucleation and growth, intelligent investigation of phases and phase transitions, and determination of effective pair potentials.Eric S. Harper, Matthew P. Spellings, Joshua A. Anderson, Sharon C. Glotzer
ShaowuPanGraduate StudentAerospaceData driven subgrid scale model for scalar transport for planar propagating combustion flameWe propose a machine learning framework to predict subgrid scale scalar flux in premixed turbulent combustion. Based on data obtained from high fidelity DNS, we explore different machine learning configurations with artificial neural network and Ababoost regression. Our a-priori results showed that the data driven approach offers excellent predictions of the correct SGS scalar flux especially at large filter sizes where analytical models tends to fail.Shaowu Pan, Karthik Duraisamy
YuqingZhouGraduate StudentMechanicalMulti-Component Topology Optimization for Additive ManufacturingTopology optimization for additive manufacturing has been limited to the component-level designs with the component size smaller than the printer’s build volume. To enable the design of structures larger than the printer’s build volume, this poster presents a gradient multi-component topology optimization framework for structures assembled from components built by additive manufacturing. For each component, manufacturing constraints on the build volume, support materials, and enclosed voids are imposed during the simultaneous optimization of overall base topology and component partitioning. The preliminary results on a minimum compliance structure show promising advantages over classic monolithic topology optimization approaches.Yuqing Zhou, Kazuhiro Saitou
MinaJafariGraduate StudentChemistryReliable and Efficient Reaction Path and Transition State Finding for Surface Reactions with the Growing String MethodOne of the challenges in computational chemistry is fast and reliable calculation of transition states and reaction paths. The ultimate goal is to design algorithms that systematically search the whole chemical space without guidance from the user and predict a network of plausible reactions. Towards this goal, one of the most robust and efficient methods for theoretical investigation of chemical systems is the growing string method (GSM). GSM was recently improved significantly by employing internal coordinates (Zimmerman, J. Chem. Phys. 2013, 138, 184102) and integrating an exact TS search scheme (Zimmerman, J. Chem. Theory Comput. 2013, 9, 3043). While this tool has been shown to be highly effective for molecular reactions, new developments are needed to handle surface reactions. Therefore this talk will describe the development of GSM for computing reaction paths and exact transition states of surface reactions, and its benchmarking against the climbing image nudged elastic band (CI-NEB) method. The double-ended and single-ended GSM’s are tested by computing reaction paths and transition states for a test set of 45 elementary reactions. Based on the results, the GSM is at least 2 times faster than CI-NEB, while CI-NEB fails to converge RP within 2000 gradient calculations in many of these cases.Mina Jafari, Paul M. Zimmerman
JosephKleinhenzGraduate StudentPhysicsInchworm Quantum Monte CarloWe generalize the recently developed inchworm quantum Monte Carlo method to the full Keldysh contour with forward, backward, and equilibrium branches to describe the dynamics of strongly correlated impurity problems with time-dependent parameters. We introduce a method to compute Green’s functions, spectral functions, and currents for inchworm Monte Carlo and show how systematic error assessments in real time can be obtained. We then illustrate the capabilities of the algorithm with a study of the behavior of quantum impurities after an instantaneous voltage quench from a thermal equilibrium state.Joseph Kleinhenz, Qiaoyuan Dong, Andrey Antipov, Guy Cohen, Emanuel Gull
GregoryTeicertGraduate StudentMechanical EngineeringData-interactive computational materials physics: Studies of precipitate morphology by a combination of experiment, nonlinear elasticity and machine learningWe present the initial results and workflow in an ongoing study of the morphology of Mg-Nd precipitates, driven by large data sets of energy computations for a range of geometry features and material parameters. The energy computations account for the eigenstrain in the precipitate and the interfacial energy. The values used for the interfacial energies and precipitate elasticity constants are calculated rather than observed, therefore the uncertainty associated with the calculation error is incorporated. The study is enhanced by training a deep neural network to the data to represent the energy surface. Predictions for the equilibrium precipitate shape for a given set of material parameters correspond to energy minima with respect to the geometry features. By comparing predictions to experimental data, we find that the precipitate shape is influenced significantly by seeking equilbrium with respect to the strain and interfacial energies.Gregory Teichert, Emmanuelle Marquis, and Krishna Garikipati
ZhenlinWangGraduate StudentMechanical EngineeringIntercalation driven porosity effects in coupled continuum models for the electrical, chemical, thermal and mechanical response of battery electrode materialsWe present a coupled continuum formulation for the electrostatic, chemical, thermal and mechanical processes in battery materials. Our treatment applies on the macroscopic scale, at which electrodes can be modelled as porous materials made up of active particles held together by binders and perfused by the electrolyte. Starting with the description common to the field, in terms of reaction-transport partial differential equations for ions, variants of the classical Poisson equation for electrostatics, and the heat equation, we add mechanics to the problem. Our main contribution is to model the evolution of porosity as a consequence of strains induced by intercalation, thermal expansion and mechanical stresses. Recognizing the potential for large local deformations, we have settled on the finite strain framework. In this first communication we have carried out a detailed computational study on the influence of the dynamically evolving porosity, via the electrostatic and reaction-transport coefficients, upon ion distribution, electrostatic potential fields, and charge-discharge cycles. Zhenlin Wang, Jason Siegel, Krishna Garikipati
Paul M.DoddGraduate StudentChemical EngineeringScalable Provenance and Metadata Management with signacResearchers in computational materials science are regularly posed with the challenge of managing large and heterogeneous data spaces. The amount of data increases in lockstep with computational efficiency multiplied by the amount of available computational resources, which shifts the bottleneck within the scientific process from data acquisition to data post-processing and analysis. We present a framework designed to aid in the integration of various specialized formats, tools and workflows. The signac framework provides all basic components required to create a well-defined and thus collectively accessible data space, simplifying data access and modification through a homogeneous data interface, largely agnostic of the data source, i.e., computation or experiment. The framework's data model is designed not to require absolute commitment to the presented implementation, simplifying adaption into existing data sets and workflows. This approach not only increases the efficiency for the production of scientific results, but also significantly lowers barriers for collaborations requiring shared data access.Carl S. Adorf, Paul M. Dodd, Sharon C. Glotzer
VictorWuGraduate Student
Industrial and Operational Engineering
Multicriteria Optimization for Brachytherapy Treatment PlanningBrachytherapy is a common method for treating cancer patients with radiation: radioactive seeds are implanted and dwell at various locations determined by a planner in the tumor. Treatment planning (TP) optimization determines the duration, i.e., dwell time, of a seed at each location, resulting in a dose distribution delivered to the treatment site. Delays in the brachytherapy treatment planning (TP) process are associated with increased anesthesia use, patient discomfort, and risk of treatment delivery inaccuracy, which may worsen toxicity or disease control.  Current commercial brachytherapy TP systems use an iterative ‘guess-and-check’ approach that is inefficient for exploring trade-offs in dose between targets and healthy structures.  This work introduces an efficient Pareto-style planning approach and intuitive graphical user interface that enables a planner or physician to directly explore dose-volume histogram (DVH) metric trade-offs without iteratively re-optimizing. Plans were generated using a “truncated” conditional value-at-risk (CVaR, a mean tail dose) to approximate intrinsic DVH metrics in a multi-criteria optimization (MCO) framework.  The MCO problem was solved repeatedly by varying the bounds on criteria represented as constraints and optimizing one criterion to generate a library of high-quality candidate plans.   By interpolating the dwell times from the library plans, we efficiently increased library density for a trade-off surface (represented via GUI) without additional re-optimization and maintained deliverability.  Process feasibility was verified by planning retrospectively cervix, prostate, and breast treatment sites. Plans obtained using a truncated CVaR approximation had generally superior DVH results than plans from a DVH-based commercial TP system.   Choice of CVaR tail and truncation sizes influenced the quality of the approximation.  Truncation improved CVaR approximations of upper tail DVH metrics (e.g., PTV D1cc).  Average candidate plan generation time was <30 seconds. Linearly-interpolated dwell times of candidate plans result in typically <0.5% loss in DVH metrics quality.  The brachytherapy treatment planning approach efficiently generates a trade-off surface consisting of high-quality plans that span a wide range of DVH values for each structure of interest.  The method is generalizable to any number of criteria, and library generation is trivially parallelizable.  Represented as an intuitive GUI, this tool could improve both TP time and quality for brachytherapy.Victor W. Wu, Marina A. Epelman, Mustafa Y. Sir, Kalyan S. Pasupathy, Michael G. Herman, Christopher L. Deufel
YuchenJiangGraduate StudentIndustrial and Operational EngineeringProduction Planning Problems with Joint Service-Level Guarantee: A Computational StudyWe consider a class of single-stage multi-period production planning problems under demand uncertainty. The main feature of our paper is to incorporate a joint service-level constraint to restrict the joint probability of having backorders in any period. This is motivated by manufacturing and retailing applications, in which firms need to decide the production quantities ex ante, and also have stringent service-level agreements. The inflexibility of dynamically altering the pre-determined production schedule may be due to contractual agreement with external suppliers or other economic factors such as enormously large xed costs and long lead time. We focus on two stochastic variants of this problem, with or without pricing decisions, both subject to a joint service-level guarantee. The demand distribution could be non-stationary and correlated across different periods. Using the sample average approximation (SAA) approach for solving
chance-constrained programs, we re-formulate the two variants as mixed-integer linear programs (MILPs). Via computations of diverse instances, we demonstrate the effectiveness of the SAA approach, analyze the solution feasibility and objective bounds, and conduct sensitivity analysis for the two MILPs. The approaches can be generalized to a wide variety of production planning problems, and the resulting MILPs can be efficiently computed by commercial solvers.
Yuchen Jiang, Juan Xu, Siqian Shen, Cong Shi
EduardoCoronaPost-DocMathematicsFast numerical algorithms for robust, high-fidelity simulation of terramechanicsHigh-fidelity simulations of soil mechanics and wheel-soil interactions is crucial to the success of integrated simulation environments for unmanned ground vehicles (UGVs) in off-road settings. However, physics-based simulations are extremely challenging owing to the multi-scale and multi-physics nature of these problems.

We employ the discrete element method (DEM) to model soil as granular media, and focus on complementarity-based formulations of frictional contact. Efficient quadratic cone programming techniques are available to solve the underlying non-linear complementarity problem (NCP). However, their robust and efficient implementation remains elusive for large-scale systems.

In this work, we use the Tensor Train decomposition, an approximate hierarchical matrix factorization technique, to produce fast and easy to update direct solvers for the Newton step solve of primal-dual interior point methods (PDIP) as applied in this context. For a number of relevant examples, we present results suggesting significant speed ups for the NCP solution for large-scale, densely-packed rigid body dynamics.
Eduardo Corona, Paramsothy Jayakumar, Shravan Veerapaneni
SarahCherngGraduate StudentSPH-EpidemiologyExtending Statistical Associations with Mechanistic Modeling to Convey Theoretical PathwaysPrevious simulation models of social networks have attempted to disentangle the effects between peer selection and peer influence on substance use, social media sharing, and delinquency, and find that many studies can potentially over- or under- estimate the level of peer influence effects due to difficulties disentangling them from peer selection effects (i.e., potential reverse causality). In this study, we hypothesize that peer influence diffusion through social adolescent networks (i.e., assuming peer selection) is a key mechanism that generates differences in peer influence of smoking behavior by network density. To explore this hypothesis, we apply the pertinent pathways posited through statistical analyses in a mechanistic computational model to isolate the independent effects of peer influence by network structure. This aim presents an explanatory mechanism integrated into a mechanistic model, that when incorporated with empirical data, can bring insights into the meaning of the statistical associations observed between urbanicity, network density, peer influence, and smoking behavior. Furthermore, these results may provide insights into a range of potential social policy levers that can help further reduce US smoking prevalence in adolescents.
Sarah T Cherng, Yu-Han Kao, Rafael Meza
XiChenGraduate StudentPhysicsSimulation of the NMR Response in the Pseudogap Regime of the CupratesThe pseudogap in the cuprate high-temperature superconductors was discovered as a suppression of the Knight shift and spin relaxation time measured in nuclear magnetic resonance (NMR) experiments. However, theoretical understading of this suppression in terms of the magnetic susceptiblility of correlated itinerant fermion systems was so far lacking. Here we study the temperature and doping evolution of these quantities on the two-dimensional Hubbard model using cluster dynamical mean field theory.
We recover the suppression of the Knight shift and the linear-in-T spin-echo decay that increases with doping.
The relaxation rate shows a marked increase as T is lowered but no indication of a pseudogap on the Cu site, and a clear downturn on the O site, consistent with experimental results on single layer materials but different from double layer materials.
The consistency of these results with experiment suggests that the pseudogap is well described by strong short-range correlation effects.
Xi Chen, J.P.F.LeBlanc, Emanuel Gull
HannaTerletskaPost-DocPhysicsNumerical studies of competing phases of matter in 2D extended Hubbard model.Physical systems with strong electron-electron interactions exhibit a wide variety of competing quantum phases of matter, including superconductivity, Mott insulating behavior, good metallic and charge ordering. The charge ordered phase is now considered to be the hallmark of many low-dimensional transition metal oxides. In this work, we perform numerical studies of the extended Hubbard model which the minimal model for charge order phenomena due to the electron-electron interactions. We use Quantum Monte Carlo based algorithm to study this model at half-filling on a two-dimensional square lattice. We show that the model exhibits metallic, Mott insulating, and charge ordered phases. Using the broken symmetry solution, we determine the precise location of the charge ordering phase transition line and the properties of the charge ordered and charge disordered phases as a function of temperature, local interaction, and nearest neighbor interaction. We also assess the regime of applicability of simpler approximation schemes for systems with non-local interactions.Hanna Terletska, Tianran Chen and Emanuel Gull
JiaLiGraduate StudentPhysicsDecomposition of Coulomb Interaction Tensors In Quantum Many-Body SystemsElectron correlations play a crucial role in the physical properties of quantum materials, and many numerical methods are being developed to study them. However, due to the typically large system size and the long range nature of the Coulomb interaction, the interaction tensor involved in many methods causes difficulties in computing resources. In this poster we present systematic numerical decomposition of the Coulomb interaction tensor, which would efficiently reduce the consumption in memory and CPU time, and would consequentially enable the numerical study of more complicated physical problems.Jia Li, Emanuel Gull
MerylSpencerGraduate StudentPhysicsActomyosin fibers and cell packing geometry influence cell elongation under stressWe have discovered actomyosin fibers in epithelial cells in the notum of Drosophila pupa. Previously such structures were only observed in the cortex separating cells. Using the vertex model framework we theorize that the fibers may form as a response to high external stress on the cells and function to stop the cells from becoming too elongated based on the cell orientation. We can test our hypothesis by using image analysis techniques to identify and skeletonize our data. We then input the cell packing into a computation model where cells are strained with constant edge lengths in order to determine an overall cell orientation and compare cell orientation and number of fibers.Meryl Spencer, Jesus Lopez-Gay, Yohanns Bellaïche, and David K. Lubensky
JiaxingWuGraduate StudentApplied PhysicsOscillations contribute to memory consolidation by changing criticality and stability in the brainOscillations are a near universal feature of every level of brain dynamics and have been shown to contribute to many functions, such as neural binding, information encoding and processing, and subsystem integration. Despite almost a century of active research and many proposed hypotheses, the role of oscillations in shaping network dynamics is still not fully understood. To investigate the fundamental mechanism underpinning oscillatory activity, the properties of heterogeneous networks are compared in situations with and without oscillations, both computationally and experimentally. Our results show that both network criticality and stability are changed in the presence of oscillations. Criticality describes the network state of neuronal avalanche, a cascade of bursts of action potential firing in neural network. The branching parameter σ is defined as the average number of subsequent active neurons at the next time point triggered by one neuron. σ<1, σ=1, σ>1 correspond to subcrtitical, critical and supercritical state respectively. Preliminary results indicate that an increase of σ is associated with better learning performance. Stability measures how stable the spike timing relationship between neuron pairs is over time. Using a detailed spiking model, we found that the branching parameter σ changes relative to oscillation and structural network properties, corresponding to transmission among differnt critical states. Also, analysis of functional network structures shows that the oscillation helps to stabilize neuronal representation of memory. Further, quantitatively similar results are observed in biological data recorded in vivo. By inhibiting parvalbumin-expressing (PV+) interneurons, delta (0.5-4Hz), theta (4-12Hz) oscillations are blocked, leading to poor learning behavior. Previously, supracritical state is thought to be associtated with a high Excitatory/Inhibitory ratio. However, the change in branching parameter reflects that the system does not have to transit to supracriticality, even though the network becomes more excitatory caused by inhibition of PV+. In summary, we have observed that, by regulating the neuronal firing pattern, oscillations affect both criticality and stability properties of the network, and thus contribute to memory formation.Jiaxing Wu, Nicolette Ognjanovski, Sara J. Aton, Michal Zochowski
Sijia LiuPost-DocEECS
Data-enabled graphical model to build chemical reaction mechanisms
Chemical reaction networks are chemical systems that involve multiple species and reactions. The dynamics of chemical reaction networks are modeled by continuous-time Markov chains. The stochastic kinetics of chemical reaction networks, under some conditions, can be reduced to a set of ordinary differential equations known as the mass-action kinetics. Existing methods do not handle the case when the species molar concentration values are corrupted by measurement uncertainty (e.g., noise and uncertainty in initial conditions) or the case when there exist missing species/reactions. This poster discusses the estimation of true initial conditions under noisy concentration
values and handling of missing species/reactions.
Omar Khalil, Sijia Liu, Yaya Zhai, Alfred Hero, Paolo Elvati and Angela Violi
ChengyuDaiGraduate StudentPhysics
Efficient Phase Diagram Sampling Strategy by Bayesian Active Learning
Why sample the state space by naive strategy of grid search? We address the problem of efficient phase diagram sampling by adopting and improving state of the art active learning techniques, resulting in an order of magnitude reduction in sampling cost. The main technical issue is that traditional grid search sampling often has low efficiency, wasting time on statepoints that are not informative about the phase boundaries. We propose to solve this problem by a new framework to adaptively choose the next most informative statepoints to sample. To enable active learning, we first interpolate the sampled statepoints' phases to the whole parameter space by Gaussian Process at each round. Then we define an acquisition function that quantifies the informativeness of next candidate trial statepoint and choose the next trial point to maximize the acquisition function sequentially. We also generalize this approach to allow batch sampling to better take advantage of parallel test setting.Chengyu Dai, Isaac R. Bruss, Sharon C. Glotzer
QicangShenGraduate StudentNuclear Engineering and Radiological SciencesStability Analysis of a Modified SN-CMFD SchemeUse fourier analysis to investigating a certain methodQicang Shen, Xuyun Lin, Thomas Downar
QuintonSkillingGraduate StudentBiophysicsNetwork criticality and oscillations increase functional stability during consolidation of new representations in brain networksForming and storing new neural representations to memory is an integral part of behavior It has been shown that information processing and memory storage capacity peak when neural dynamics in the brain reside near a critical point in a phase transition between order and disorder. However, it is less well known how such dynamics may benefit the storage of new representations as memories as they compete with existing representations in neural circuits. In this study, we examine two model systems poised near criticality and the effects on spatiotemporal dynamics driven by external input. The first model, given by a modified Hopfield formalism, allows easy access to changing proximity to a critical point between order and disorder while simultaneously examining the ability to store new memories. The second model is a simplified integrate-and-fire model based on the Bak-Tang-Wiesenfeld formalism. The benefit of this model is that it can be easily tuned to yield critical point spiking behavior and the analysis of functional connectivity stability at such a point. We conclude by comparing our model results to in vivo spiking data recorded from mouse hippocampus in a contextual fear conditioning environment. We show that residing near a critical point is important to stabilize spatiotemporal dynamics.Quinton Skilling, Jiaxing Wu, Nicolette Ognjanovski, Sara Aton, Michal Zochowski
JaredFergusonGraduate StudentApplied PhysicsAdaptive Mesh Refinement in 2D forced shallow-water and idealized 3D simulationAdaptive Mesh Refinement (AMR) techniques have the potential to address the challenges of modeling in uniform general circulation models (GCMs) tropical cyclones and other extreme weather and climate phenomena. By dynamically placing refined grids over salient transient features, models with AMR can provide sufficient local resolution while limiting the computational burden. This work explores and seeks to validate the use of an AMR approach in a high-order finite-volume dynamical core by using a series of forced 2D shallow-water and idealized 3D dycore test cases. Shallow-water based tests cases that use forcing mechanisms to mimic tropical cyclone-like vortex strengthening and orographically triggered features are implemented to quantify improvements gained from AMR grids, assess how well transient features are preserved across grid boundaries, and determine criteria that maximize the AMR effectiveness. Several of the test cases are extended to the non-hydorstatic 3D dycore to characterize accuracy and stability of its implementation.Jared Ferguson, C Jablonowski, H Johansen, P. McCorquodale, P. Colella, W. Langhans, P Ullrich
YuxiChenGraduate StudentClimate & Space SciencesMHD with Embedded Particle-in-Cell (MHD-EPIC): Capturing Kinetic Effects in Global SimulationsA new modeling capability to embed the implicit Particle-in-Cell (PIC) model iPIC3D into the extended magnetohydrodynamic model BATS-R-US has been developed recently in University of Michigan. The PIC domain can cover the regions where kinetic effects are most important, such as the magnetopause reconnection sites. The BATS-R-US code,which is a MHD model, with its block-adaptive grid can efficiently handle the rest of the computational domain where the MHD or Hall MHD description is sufficient. The current implementation of the MHD-EPIC model allows two-way coupled simulations in two and three dimensions with multiple embedded PIC regions. The MHD and PIC grids can have different grid resolutions and grid structures. The MHD variables and the moments of the PIC distribution functions are interpolated and message passed in an efficient manner through the Space Weather Modeling Framework (SWMF). Both BATS-R-US and iPIC3D are massively parallel codes fully integrated into, run by and coupled through the SWMF. This model has been successfully applied to the global magnetosphere simulations of Ganymede, Mercury and Earth.Yuxi Chen, Gabor Toth
JoshuaKammeraadGraduate StudentChemistryLearning to represent chemical dataChemists think about chemistry in terms of heuristics at various levels of abstraction, some of which are general to broad areas of chemistry and others which are more precise but domain specific. Accurate chemical data can be generated using quantum simulation but this is computationally expensive. Machine learning methods may provide a manner of identifying and using patterns in chemical data at lower computational cost. However, machine learning often requires substantial domain specific, ad hoc tuning. We seek to explore how human computer interaction can be utilized to train algorithms to focus on the predictive patterns in chemical data and how to utilize the results of algorithmic learning to lead humans to previously undiscovered chemical heuristics. We look at employing an autoencoder to generate valuable chemical representations and discuss further potential strategies for expanding the applicability of machine learning to chemical reactivity prediction.Joshua Kammeraad, Paul Zimmerman
EllenMulvihillGraduate StudentChemistryPost-Marcus electronic transition dynamics via the generalized quantum master equationElectronic transition (ET) processes play an important role in many technologically relevant processes, such as redox reactions, batteries, and solar cells. The exponential scaling of the computational effort with system size makes a fully quantum-mechanical description of a ET process impossible when it takes place in a complex condensed phase molecular system. As a result, the most commonly used approach for modeling the rates of ET processes is based on Marcus theory or Fermi's golden rule (FGR). However, Marcus theory has many restrictions for the systems it can adequately describe. The development of Generalized Quantum Master Equation (GQME) methods fills this gap within charge transfer dynamics. GQME methods work both in between and at the weak and strong coupling limits, offering a necessary flexibility when the level of coupling in the system is either unknown or between these extremes. In this poster, we develop and compare different versions of GQME methods and their various approximations.Ellen Mulvihill, Alexander Schubert, Xiang Sun, Eitan Geva
BobbieWuGraduate StudentMathematicsPairwise hydrodynamic interactions of soft particles in applied electric fieldsThis poster will present a simulation method for the soft-particle electro-hydrodynamics (EHD), which involves: (1) boundary value problem of a coupled system of Laplace equation and Stokes equation, (2) accurate numerical scheme based on the boundary integral equation method (BIEM), and (3) investigation of EHD phenomena including particle deformation and interactions.Bowei Wu
EricWalkerPost-DocChemistryReaction Path Discovery under Potential BiasElectrocatalysis may form the basis for a fully renewable means to convert the greenhouse gas carbon dioxide into fuels or value-added chemicals. Since the conversion efficiency depends intimately on the mechanism of reaction—which is not fully understood—ab initio quantum chemical calculations may provide molecular details that allow for the rational design of electrocatalysts. This work extends the method of Goodpaster, et al., which treats surfaces under applied potential and implicit solvent, to the growing string method for reaction path discovery. The growing string method is capable of finding the activation energy and reaction energy from a starting reactant and a set of driving coordinates for bond breaking or forming. Several vital surface reactions are conducted to demonstrate the efficacy of this method for CO2 reduction under potential bias on a Cu(100) surface. Furthermore, the same reaction is investigated on a silver surface to explain the experimentally observed behavior of Ag(100) to switch between carbon monoxide or hydrogen production as a function of potential bias. The reactions path discovery examples on Cu and Ag will demonstrate the power of the growing string method for investigating surface reactions under electrochemical bias, and provide foundational insights into the rational design of electrocatalysts.Eric Walker, Paul Zimmerman
SergeiIskakovPost-DocPhysicsDiagrammatic Monte-Carlo for Dual Fermion approachThe dual fermion series is a diagrammatic approach for correlated lattice models that includes non-perturbative local and perturbative non-local dynamic correlations. In this talk we show results from a simulation of the 2D Hubbard solved with dual fermions, where we stochastically sample the dual fermion perturbation series using a diagrammatic Monte Carlo method. We present a description of the method, compare to other methods, and show several applications to correlated systems.Sergei Iskakov, Emanuel Gull
TimothyBrooksGraduate StudentAerospaceDesign Optimization of Highly Flexible Wing Aircraft Using Unconventional Manufacturing TechniquesThe demand for more efficient transport aircraft by airlines has pushed aircraft manufacturers to design a higher aspect ratio wing in order to reduce induced drag. However, increasing the aspect ratio of the wing incurs a weight penalty, and also leads to aeroelastic issues due to the higher flexibility. To address this issue, many have started to investigate how to utilize the next generation of materials to enable higher aspect ratio wings that do not incur a prohibitive structural weight penalty, while not being susceptible to undesirable aeroelastic phenomena.
One material technology of current interest are tow-steered composites. Tow steering is a new composite manufacturing technique whereby a machine is used to manufacture composites with fiber angles that vary continuously throughout the structure. Since the stiffness of a composite laminate is highly dependent on the orientation of these fiber angles, this gives the designer much more control over the stiffness properties of the structures. There are currently machines that manufacture tow-steered composite structures, but manufacturers have not yet taken full advantage of this capability due to a lack of numerical tools to design these new composites.
My research looks at designing a tow-steered composite wing through numerical optimization. This process will optimize both the composite layup as well as the local structural thicknesses. In this work a high-fidelity aerostructural analysis is used, accounting for the wing’s flexibility through the coupling of the aerodynamic and structural disciplines. In addition to the structural sizing and composite fiber tow angles the wing shape is also optimized simultaneously. Using this technique, optimal solutions yield wings with passive load alleviation. Passive load alleviation utilizes the wing flexibility to tailor the load on the wing to be more structurally favorable at the flight conditions that size the structure, while keeping an aerodynamically favorable pressure distribution at the cruise flight conditions.
Timothy Brooks, Joaquim Martins
StevenKiyabuUndergraduate StudentMechanical EngineeringComputational Screening of Hydration Reactions for Thermal Energy StorageReversible chemical reactions offer the prospect of materials that can store greater amounts of heat than current materials that store latent or sensible heat. Hydration reactions are one particular area of interest due to their high energy densities and reversibility at moderate temperatures. However, to our knowledge, there is no comprehensive characterization of the theoretical energy densities of hydration reactions. The present study uses first principles calculations, namely density functional theory as implemented in the Vienna Ab-initio Simulation Package, to characterize the energy storage densities and turning temperatures of 175 hydration reactions involving metal halide hydrates and metal hydroxides. From this, we were able to identify several theoretically promising hydration reactions for various temperature ranges.Steven Kiyabu, Jeffrey S. Lowe, and Donald J. Siegel
JosephCiccheseGraduate StudentChemical EngineeringHow to optimize tuberculosis antibiotic treatments using a computational granuloma modelTuberculosis (TB), one of the most common infectious diseases, requires a treatment of multiple antibiotics taken over at least six months. The treatment length often results in poor patient-adherence, and the emergence of multi-drug resistant TB indicates that new antibiotic treatments are needed. New antibiotics are being developed or repurposed to treat TB, but because there are numerous potential antibiotics, the design space for new treatments is too large to search exhaustively. Here we propose a method of combining an agent-based and multi-scale model of TB granuloma formation and treatment with surrogate-assisted optimization (SAO) to identify optimal TB treatments. We tested the ability of SAO to locate optimal treatments in single-antibiotic test problems. We found that using SAO provides an efficient and practical strategy for TB treatment optimization, and show how SAO could be used to locate optimal treatments when considering combinations of antibiotics administered together. Joseph M. Cicchese, Elsje Pienaar, Denise E. Kirschner, Jennifer J. Linderman
AyoubGouasmiGraduate StudentAerospaceReconstructing Memory Effects in Coarse-grained Multiscale Problems Using the Mori-Zwanzig FormalismReduced Order Models (ROMs) of complex, nonlinear dynamical systems often require closure, which is representing the contribution of the unresolved physics on the resolved physics. The Mori-Zwanzig (M-Z) procedure allows one to write down formally closed evolution equations for the resolved physics. In these equations, the unclosed terms are recast as a memory integral involving the past history of the resolved variables only. While the M-Z procedure does not directly reduce the complexity of the original system, these equations can serve as a mathematically consistent starting point to develop closures based on approximations of the memory. However, a priori knowledge of the memory kernel, which is not explicitly known for nonlinear systems, is of paramount importance to assess the validity of a memory approximation. Unraveling the memory kernel requires the determination of the orthogonal dynamics which is a projected high-dimensional partial differential equation not tractable in general. A method to estimate the memory kernel a priori, using full order solution snapshots, is proposed. The main idea is to solve a pseudo orthogonal dynamics equation, that has a more tractable form, instead of the original. The method is exact in the linear case where the kernel is known explicitly. Results for the Viscous Burgers equation are presented and discussed.Ayoub Gouasmi, Eric Parish, Karthik Duraisamy
ChaseDwelleGraduate StudentCivil & Environmental EngineeringUncertainty quantification for urban flooding with reduced order modelingThis work serves as a demonstration of performing uncertainty quantification in urban hydrology. Due to the large temporal and spatial scales involved, hydrology is a difficult problem for traditional uncertainty quantification frameworks. We investigate uncertainties in a flooding problem by i) representing uncertainty for both model parameters and stochastic fields, ii) efficiently propagating these uncertainties with reduced order modeling, using Bayesian compressive sensing to tame the dimensionality of the problem, and iii) analyze sensitivities of urban flooding to hydrologic processes.M. Chase Dwelle, Jongho Kim, Valeriy Ivanov
VyasRamasubramaniGraduate StudentChemical EngineeringThe Effects of Depletion on the Stability and Assembly of Colloidal CrystalsThe depletion force offers a powerful route to influence the entropic self-assembly of colloidal particles. Various recent works have demonstrated that depletion can promote the formation of alternative crystal structures to those that arise in identical systems in the absence of depletion. In this study, we aim to develop a quantitative understanding of this effect. We apply Monte Carlo simulations to a well-studied family of truncated tetrahedra to assess the relative stability of various structures. We show that the presence of depletion alters the free energy landscape for particles of a particular shape, and we show that depletion preferentially favors different structures than those known to form from truncated tetrahedra in the absence of depletion. Free volume calculations show that that the free volume varies significantly for different crystal structures. Several order parameters are also analyzed and used to identify the phases present. We perform melting simulations to determine how much depletion is required to stabilize various structures, and we calculate free energies of various crystals to determine their relative stabilities. Vyas Ramasubramani, Andrew Karas, Jens Glaser, Sharon Glotzer
ElizabethAgeeGraduate StudentCivil & Environmental EngineeringResolving three-dimensional root water uptake and lateral interactions for forest systemsA growing body of research has highlighted the importance of root architecture and hydraulic properties to the maintenance of plant water uptake and transpiration streams under water limitation and drought. Detailed studies of single plant systems have shown the ability of root systems to adjust zones of uptake due to the redistribution of local water potential gradients, thereby delaying the onset of stress under drying conditions. While computational complexity has previously hindered the implementation of microscopic root system structure and function in larger scale hydrological models, newer hybrid approaches allow for the resolution of these properties at the plot scale. Using a modified version of PFLOTRAN, a massively parallel model of multiphase flow and reactive transport, we model three-dimensional root water uptake in variably saturated soils for a one-hectare temperate forest plot. Using the model environment as a virtual laboratory, we use an ensemble simulations to provide insights on the impact of root morphology, peer interaction, and root hydraulic properties on individual and community response to water limitation. Results demonstrate the ability of interacting systems to shift areas of active uptake based on local gradients, allowing individuals to meet water demands despite competition from their peers. Further, the degree to which root systems shift, and the potential hydraulic stress of these shifts, is directly correlated with the degree of interaction with other members of the community. These results illustrate how inter- and intra-species variations in root properties may influence not only individual response to water stress, but also help quantify the margins of resilience for forest ecosystems under changing climate.Elizabeth Agee, Lingli He, Gautam Bisht, Valentin Couvreur, Simone Fatichi, Ashley Matheny, Gil Bohrer, and Valeriy Ivanov
AllisonKellyGraduate StudentChemistryMechanochemical Reaction Path Finding with the Force-Growing String MethodThe application of mechanical force has the ability to distort the potential energy surface of a reaction making computational treatments of mechanochemical reactions difficult. Current methods are either inefficient or unable to trace along the force-modified potential energy surface which means accurate reaction paths and transition state saddle points cannot be identified for reactions under an applied force. The Force-Growing String Method (F-GSM) is presented here as a reliable and efficient method for identifying reaction paths and transition states on the force-modified potential energy surface. The use of internal coordinates in F-GSM is proven effective in choosing suitable coordinate systems for complicated mechanochemical reactions. Additionally, F-GSM is able to operate in a single-ended fashion, which makes it a valuable explorative tool for probing reactions whose reaction paths or products are unknown.Allison Kelly and Paul Zimmerman
KokiSagiyamaPost-DocMechanical EngineeringA Numerical Study of Branching and Stability of Solutions to Three-dimensional Martensitic Phase Transformations using Gradient-regularized, Non-convex, Finite Strain Elasticitykoki sagiyama, shiva rudraraju, krishna garikipati
LouisJoslynGraduate StudentBioinformaticsDoes Exhaustion Explain Low T-cell Functionality in the Mycobacterium tuberculosis Granuloma?Each year, approximately 2 million people die from Tuberculosis, an infectious disease caused by Mycobacterium tuberculosis (Mtb). The hallmark of Mtb infection are granulomas. These are a collection of host cells whose purpose is to contain or clear infection, creating a complex hub of immune and bacterial cell activity. Yet, given cellular activity and potential for frequent interactions between host and bacterial cells, a surprisingly low quantity of Mtb-responsive T cells (~ 8% of granuloma T cells) was observed in a recent study of Mtb infection within non-human primate (NHP) granulomas (Gideon et al. 2015). Various mechanisms could limit T cell function, one hypothesis is T cell exhaustion. While T cell exhaustion lacks a formal definition in the literature, continual antigen stimulation causes a subpopulation of T cells to enter a state characterized by low cytokine production, low proliferation and a series of inhibitory receptors – the most popular of which are PD1, TIM3, LAG3 and CTLA4 (Kahan, Wherry and Zajac 2015; Wherry 2011). In this work, we utilize experimental data to calibrate and inform an agent-based model (Ray et al. 2008; Segovia-Juarez et al. 2004) that captures environmental, cellular, and bacterial dynamics within granuloma formation in lungs during Mtb infection. Specifically, we updated a previous version of this model to evaluate the direct impact of an exhausted T cell phenotype. We compared results with a range of 5-95% exhaustion to predict the impact that T cell exhaustion would have on granuloma-scale outcomes. Together, the conclusions of our model coupled with the results of experimental work suggest that T cell exhaustion alone cannot be responsible for the low quantity of Mtb-responsive T cells found in granulomas. Louis Joslyn, Eileen Wong, JoAnne Flynn, Denise Kirschner
Anand PratapSinghGraduate StudentAerospaceData Driven Augmentation of Physical ModelsReynolds Averaged Navier-Stokes (RANS) based models are built on a simplified understanding of physics and are calibrated on canonical flows. Such models underperform and are inaccurate in the presence of complex flow features. This work focuses on augmenting such models using data-driven techniques which include full field inversion and machine learning. We outline the development of the augmentation using experimental data and present results for flow over an airfoil.Anand Pratap Singh, Karthik Duraisamy
EricParishGraduate StudentAerospaceA Dynamic Subgrid Scale Model for LES Based on the Mori-Zwanzig FormalismIn this work we present a closure model for Large Eddy Simulation based on the Mori-Zwanzig formalism. The model is constructed by approximating the memory integral in the Mori-Zwanzig formalism with a Markovian term and by exploiting similarities between two levels of coarse-graining via the Germano identity. The model is parameter-free and has a structural form imposed by the mathematics of the coarse-graining process. We outline the development of the model and present results for decaying turbulence and turbulent channel flow.Eric Parish and Karthik Duraisamy
AlauddinAhmedPost-DocMechanical EngineeringDiscovery of High-Performing MOFs via Machine Learning We describe a data-driven approach to developing metal-organic frameworks (MOFs) with high hydrogen storage capacities. MOFs are crystalline nanoporous materials containing a metal cluster bonded to organic linkers. High-throughput (HT) grand canonical Monte Carlo simulations and a Voronoi networks-based technique are used to compute the capacities and properties of nearly 500,000 compounds. The large properties-performance database resulting from these calculations presents a unique opportunity for machine learning (ML). Using ML, we predict the necessary structural properties of a MOF needed to achieve specific capacity targets. We refer to this approach as ‘reverse crystal engineering.’ This method illustrates the possibility of developing purpose-built materials with specified functionalities. Alauddin Ahmed and Donald J. Siegel
EhsanMirzakhaliliGraduate StudentMechanical Engineering
A computational model of the calcium dynamics in Caenorhabditis elegans ASH sensory neuron
C. elegans is widely used as a model system for monitoring stimulus-evoked Ca2+ transients in neurons. The ASH sensory neuron is the subject of several such studies, primarily due to its key importance as a polymodal nociceptor. However, despite the pivotal role of ASH in C. elegans, the overall biology and the characteristics of its Ca2+ transients (e.g., the "off" response), no mathematical model has been developed to describe the full mechanism of ASH Ca2+ dynamics.
We propose a computational model which captures the Ca2+ transients in the C. elegans ASH neuron upon its activation. The model is built on biophysical cascades that unfold as part of the neuron's Ca2+ signaling events and homeostatic mechanism (e.g., TRPV channels and voltage-gated channels activation, Ca2+ release from intracellular stores, IP3 dynamics, PMCA and SERCA pumps function). The state of the ion channels is described based on Hodgkin-Huxley formulation and the remaining molecular states are based on kinetic equations with phenomenological adjustments.
We use experimental data of osmotic stimulus-evoked ASH Ca2+ transients, detected with a FRET sensor (TN-XL) to build our model. The experiments include data from young and aged worms, both untreated and exposed to oxidative stress. The model must account for all these cases with minimum changes in the parameter sets. In the first step, we use a multi-objective genetic algorithm to find a set of parameters for all the cases. Next, we use the results from the multi-objective genetic algorithm as an initial guess to find the parameters for young untreated worms using a hybrid method that consists of a genetic algorithm and nonlinear least-squares. Next, the optimum parameter sets for young untreated worms are used to develop a large data sets that account for models that show aging and exposure to oxidative stress. Finally, we use relevant biological criteria to propose multiple scenarios that might lead to aging and exposure to oxidative stress in the worms.
Our model includes for the first time the changes in ASH cytoplasmic Ca2+ flux observed both upon delivery and withdrawal of the stimulus (i.e., the "on" and "off" responses). Our model can also be used to predict the ASH Ca2+ response to stimulation pulses that are challenging to achieve experimentally (stimuli sequences of varying durations/lengths, or ramp stimuli). This effort is the first to propose a quantitative dynamic model of the Ca2+ transients generating mechanism in a C. elegans neuron, based on essential biochemical pathways of the Ca2+ homeostasis machinery.
Ehsan Mirzakhalili, Bogdan Epureanu, Eleni Gourgou
Mohamed Amine
A universal gradient-enhanced surrogate model for airfoil design optimization
Computational fluid dynamics (CFD) based aerodynamic optimization mainly consists of four stages: the shape parametrization, the mesh creation (deformation), the flow solution and the optimization.
The efficiency of such process is measured by the time needed to complete the optimization and by the quality of the final solution.
With the growth of complexity in numerical simulation models for a more accurate representation of the physics, computational tools have become ever more complex and computationally expensive.
In addition, a large number of inputs is often considered.
As substitutes of time-consuming numerical simulation models, surrogate models which are a mathematical tool that imitates the true behavior of the physics, are widely used.
However, classical surrogate models do not scale well with the large number of inputs or are not computationally efficient.
To address these issues, we first make an airfoil database that covers a large airfoil space design especially for the transonic regime.
Next, we have developed a new kriging-based surrogate model approach adapted to high-dimensional problems.
This surrogate uses the gradient information provided by the CFD-adjoint results on the training points, and the partial-least squares reduction method (PLS) for constructing the kriging covariance function.
The gradient-enhanced kriging combined with PLS model (GE-KPLS) was validated on a most of academic and engineering functions tested with up to 100 dimensions, in terms of accuracy and efficiency.
Mohamed Amine Bouhlel, Jichao Li, Joaquim R.R.A. Martins
ShuyingLiGraduate StudentMechanical Engineering
Supramolecular Mode-Selective Control of Dynamic Disorder in Small Molecular Organic Semiconductors: Tetracene, Rubrene, and Tetrathiatetracene
The transport of charge carriers in small molecular organic semiconductors such as tetracene, rubrene, and tetrathiatetracene are highly sensitive to disorder in the crystal structure. Disorder can be of two types: dynamic or static. Plenty of work has been done to suppress the influence of the latter on charge transport in these narrow-band semiconductors. Consequently, it is possible to fabricate ultrapure, defect-free organic semiconductors with carrier mobility ~100 〖cm〗^2/Vs at low temperatures. On the other hand, the effects of dynamic disorder on charge transport cannot be easily eliminated since the time-scale for these interactions is on the order of the quasi-particle lifetime. This is further complicated by the relatively weak strength of the interaction ~k_B T, which is usually indistinguishable from the noise background of experimental probes such as x-ray diffraction and transmission electron microscopy. Thus, to obtain fundamental insights on the role play by dynamic disorder on charge transport in small molecular organic semiconductors, we must approach this problem computationally. By performing molecular dynamics simulations coupled with first principle calculations of important charge transport parameters such as the transfer integral and the frequency of inter/intra-molecular vibrations, we demonstrate how dynamic disorder in the family of high mobility organic semiconductors with tetracene core can be controlled.Shuying Li, Shantonio W. Birch and Kevin P. Pipe
Graduate StudentBiomedical Engineering
Accurate 3D Reconstruction of Tissue Bulk Attenuation Coefficient in Limited-Angle Ultrasound Tomography
Attenuation imaging by ultrasound transmission tomography has shown the ability to discriminate different breast tissue types and should be good for detection because attenuation artifacts are major indicators of cancer. Severe computational limitations in this necessarily 3D process as well as the definition of attenuation coefficient still delay its use in the clinic. Also limited is derivation of other tissue properties that could be mined for detection, diagnosis, prognosis, and treatment tracking and assessment. Since attenuation images are dominated by losses at tissue boundaries due to diffraction for which a priori information exists, 3D corrections for an image of tissue bulk attenuation coefficient that is corrected for boundary losses are presented. In this study, the attenuation reconstructions with boundary loss correction were investigated in 3D simulations. The simulations consisted of two linear ultrasound transducers arranged in limited-angle tomography. Each transducer had 40 finite line-source elements with 2 MHz center frequency, and both transducers were placed below and above an image volume of (axial × lateral) = (5 cm × 4 cm) with a volume thickness of 3 cm. Three imaging scenarios are presented in this study: imaging of one spherical lesion, two adjacent spherical lesions, and two adjacent spherical lesions with enclosed curved skin layers. All lesions were placed off the center of transducer image plane. The attenuation due to boundary losses was calculated by the two additional wave propagation simulations in which the speed of sound distribution of the present objects was known a priori: the simulation with non-absorbing object insertion and the simulation without object insertion. The 2D attenuation image reconstruction assumed wave propagation direction as a bent-ray in the transducer image plane and utilized probabilistic object location and shape for limited-angle imaging. The results show that applying boundary loss correction using the 3D simulations of wave propagation in non-absorbing media yields better attenuation estimates and less artifacts around object boundaries. Higher accuracy should be achieved if the attenuation image is reconstructed in 3D. This is the first attempt to utilize a priori pulse echo and speed of sound distributions to correct stacks of limited angle, 2D attenuation data for boundary losses both in-plane and out-of-plane. This study also shows promise in allowing significant performance improvement in breast tissue characterization in the same view as mammograms.Rungroj Jintamethasawat; Oliver D. Kripfgans, Ph.D.; Brian J. Fowlkes, Ph.D.; Paul L. Carson, Ph.D.
NegarFarzanehGraduate StudentBioinformatics
A computational Approach to Automated Abdominopelvic Traumatic Injury Decision Support and Severity Assessment
Trauma is the leading cause of death among American under 46. First destination of trauma patients to receive treatment is ICU, where making the right or wrong decision within a minute can change the outcome. Considering that enormous amount of data is produced for each patient, its complexity makes it even harder to use traditional methods. Therefore, most of this invaluable information ends up being under used.
Our goal is to extract information from different types of data such as CT images and clinical and demographic data from the patients suffering from abdominal injuries to develop an outcome prediction model. We develop image processing techniques to extract information from CT images and integrate them with vital signs, injury severity scores, demographic data, and other patient information to train a predictive mode. This model could be used for the purpose of estimating ICU days. The main stage in the image processing is identifying the border of each organ and measuring the amount of injury. Our algorithm yielded the Dice similarity value of %93.5 and Jaccard of %87.9 in liver segmentation. ICU day prediction yielded the AUC index of 0.80 for classifying to 1) less than 4 days and 2) higher 4 days.
Negar Farzaneh, S.M.Reza Soroushmehr, Samuel Habbo-Gavin, Kayvan Najarian