https://s.uconn.edu/meseminar3/5

Abstract:

Energy demand and the need to reduce emissions have pushed combustion research towards the development of more efficient, environmentally-friendly engines. In lean-burn systems both high efficiency and low emissions can be achieved in principle by controlling the flame temperature; however the short resident times in practical systems and the unsteady coupling between turbulent mixing, thermochemistry and acoustics makes the realisation of such systems very challenging. Large eddy simulations are attractive to predict this unsteadiness but require modelling of the turbulence-flame interaction at the small scales, and this modelling becomes particularly challenging at high-pressure, realistic conditions due to the limited amount of available validation data and the need to keep the computational cost relatively low. In this seminar recent development of LES modelling in the context of presumed PDF approaches will be discussed and their predictive abilities and limitations analysed for conditions relevant for gas turbines. An application to a bi-stable combustor will be then presented at the end of the talk. 

Biographical Sketch:

Dr Ivan Langella is an Assistant Professor in Sustainable Aircraft Propulsion at the faculty of Aerospace Engineering at TU Delft. He obtained his Master degree in Naples, Italy, in 2011, and his PhD in Mechanical Engineering from University of Cambridge, UK, in 2016. Since then he has worked as a postdoctoral associate at the University of Cambridge until June 2018 on advanced combustion systems of aeronautical interests in collaboration with Rolls-Royce and DLR Germany. From June 2018 to April 2020 he has worked as Lecturer in Thermofluids science and Engineering at Loughborough University, UK, before joining the research group in Delft.

https://s.uconn.edu/meseminar2/26/21

Abstract:

This talk will present a machine learning framework that builds interpretable macroscopic surrogate elasto-plasticity models inferred from sub-scale direction numerical simulations (DNS) or experiments with limited data. To circumvent the lack of interpretability of the classical black-box neural network, we introduce a higher-order supervised machine learning technique that generates components of elasto-plastic models such as elasticity functional, yield function, hardening mechanisms, and plastic flow. The geometrical interpretation in the principal stress space allows us to use convexity and smoothness to ensure thermodynamic consistency. The speed function from the Hamilton-Jacobi equation is deduced from the DNS data to formulate hardening and non-associative plastic flow rules governed by the evolution of the low-dimensional descriptors. By incorporating a non-cooperative game that determines the necessary data to calibrate material models, the machine learning generated model is continuously tested, calibrated, and improved as new data guided by the adversarial agents are generated. A graph convolutional neural network is used to deduce low-dimensional descriptors that encodes the evolutional of particle topology under path-dependent deformation and are used to replace internal variables. The resultant constitutive laws can be used in a finite element solver or incorporated as a loss function for the physical-informed neural network run physical simulations.

Biographical Sketch:

Dr. Sun is an associate professor at Columbia University since 2014. He obtained his PhD from Northwestern in 2011 and worked as a senior member of technical staff at Sandia National Laboratories from 2011-2013. His research focuses on theoretical, computational, and data-driven mechanics for multiphase materials with complex microstructures. He is the recipient of several awards including the IACM John Argyris Award, ICE Zienkiewicz medal Prize, the ASCE EMI Da Vinci Award, the NSF CAREER award, Young Investigator Awards from Army Research Office and Air Force Office of Scientific Research.

 

https://s.uconn.edu/meseminar2/19

Abstract: In this talk, we will attempt to address the following two questions concerning Covid-19 type infectious respiratory disease spread: 

1. Can we identify the relative importance of the different dominant transmission routes of the SARS-CoV-2 virus? Initially Covid-19 was assumed to spread by large droplets whereas of late the airborne route has been recognized to be important. Knowledge of the most dominant transmission route is necessary to design corresponding disease mitigation strategies.

2. Can we construct a disease spread model from the flow physics of transmission? Present, widely used epidemiological models do not account for the flow physics that underpin disease transmission.

To these ends, the probability of infection caused by inhaling virus-laden droplets and their desiccated nuclei are individually calculated. At typical, air-conditioned yet quiescent indoor space, for average viral loading, cough droplets of initial diameter between 10-50 micron are found to have the highest infection probability while it is the airborne droplet nuclei, due to their persistence (especially in poorly ventilated spaces), contribute most to the disease spread. Combined with molecular collision theory adapted to calculate the frequency of contact between the susceptible population and the aerosol cloud, infection rate constants are derived ab initio, leading to a Susceptible-Exposed-Infectious-Recovered (SEIRD) model applicable for any respiratory event – vector combination. Viral load, minimum infectious dose, and dilution of the respiratory jet/puff by the entraining air are shown to mechanistically determine specific physical modes of transmission and variation in the basic reproduction number, from first-principle calculations.

Biographical Sketch: Prof. Swetaprovo Chaudhuri works in turbulent reacting flows and propulsion and is known for his contributions on turbulent flame stabilization, propagation, and structure using experiments, theory, and computations. After his BE from Jadavpur University (2006), he earned his PhD from the Department of Mechanical Engineering, University of Connecticut in 2010, working under Prof. B. M. Cetegen. He worked at Princeton University as a research staff with Prof. C K. Law and then at the Indian Institute of Science, as an Assistant/Associate Professor. Subsequently, he joined the University of Toronto Institute for Aerospace Studies as a tenured Associate Professor. Prof. Chaudhuri has authored/co-authored over hundred articles in top journals, conferences, and books, and has been honored by ASME, UConn, INSA, IAS, UTIAS. He is an elected Associate Fellow of AIAA (class of 2021) and a member of its Propellants and Combustion technical committee.

 

https://s.uconn.edu/meseminar

Abstract: Modern machine learning techniques, such as deep neural networks, are transforming many disciplines ranging from transportation to healthcare, by uncovering patterns in big data and making accurate predictions. They have also shown promising results for discovering design ideas, which is crucial for creating new products and enabling innovation. These automated computational design methods can support human experts, who typically create designs by a time-consuming process of iteratively exploring ideas using experience and heuristics. However, there are still challenges remaining in synthesizing new designs while navigating the exploration-exploitation trade-off. In this talk, we will discuss a few main challenges faced by data-driven generative models, which are unique to design problems, and show how novel architectures of performance-augmented Generative Adversarial Networks (named PaDGAN, MO-PaDGAN, and PcDGAN) address these challenges. By applying these algorithms to the airfoil synthesis problem, we will show how these methods outperform state-of-art design parametrization methods with large improvements in design performance, diversity, and novelty for single and multiple objectives. The talk will conclude by highlighting some broader applications and open challenges in data-driven design.

Biographical Sketch: Dr. Faez Ahmed is an Assistant Professor in Mechanical Engineering at the Massachusetts Institute of Technology (MIT), where he leads the Design Computation and Digital Engineering (DeCoDE) lab. His research focuses on developing new machine learning and optimization methods to study complex engineering design problems. His recent work includes proposing automated design synthesis methods to generate novel designs, creating the first provably-optimal algorithm for the diverse matching problem, and building computationally efficient ways for combining physics with human expert knowledge to design new products. Before joining MIT, Faez was a Postdoctoral Fellow at Northwestern University and completed his Ph.D. in Mechanical Engineering at the University of Maryland. He was employed in the railway and mining industry in Australia, where he pioneered data-driven predictive maintenance and renewal planning efforts.

Webex Link:  https://s.uconn.edu/meseminar

Abstract: Proton exchange membrane fuel cells (PEMFCs) provide clean and efficient conversion of chemical energy into electrical energy, and fuel cell electric vehicles offer attractive range, weight, and refueling times when compared to similar technologies. However, challenges in infrastructure, performance, durability, and cost hinder wide-spread adoption. PEMFC cost, performance, and durability are closely tied to the Pt catalyst in PEMFC electrodes. Efforts to lower the Pt loading or replace the catalyst altogether show poor performance or low durability. This difficulty highlights the challenges related to optimizing processes in the PEMFC catalyst layer (CL), a heterogeneous region made up of carbon, catalyst, gas, and ion-conducting polymer electrolyte phases where key limiting phenomena occur in PEMFCs.
In this presentation we cover the challenges and progress in developing structure-property-function relationships for the Nafion polymer phase in the PEMFC cathode catalyst layer. Work is presented in terms of developing a modeling tool for predicting and optimizing PEMFC performance based on CL design parameters (e.g., microstructure, nanostructure, and catalyst loading). To understand the influence of CL Nafion on PEMFC performance, Nafion properties as a function of film thickness, thermochemical conditions, and polymer structure at solid support interphases must be better understood. We use neutron reflectometry, combined with complementary thin-film techniques, to develop quantitative structure-property relationships for thin-film Nafion, for deploying in our model framework. Insights include an evaluation of modeling approaches for the PEMFC CL microstructure, quantifying Nafion water uptake and conductivity for varying thicknesses and solid substrates, and non-intuitive design principles for PEMFC CL performance. We conclude by highlighting remaining knowledge gaps and next steps for improved understanding and control in PEMFC catalyst layers.

Biographical Sketch: Steven C. DeCaluwe is an Associate Professor of Mechanical Engineering at the Colorado School of Mines in Golden, CO. He received his BS in mathematics and elementary education from Vanderbilt University (2000). After teaching elementary school for three years, he earned a PhD in mechanical engineering from the University of Maryland (2009) before serving as a postdoctoral fellow at the NIST Center for Neutron Research (2009–2012). He has been at the Colorado School of Mines Department of Mechanical Engineering, where he leads the CORES Research Group (cores-research.mines.edu). His research employs operando diagnostics and numerical simulation to bridge atomistic and continuum-scale understanding of electrochemical energy devices, with a focus on processes occurring at material interfaces and in reacting flows. Applications include lithium-ion batteries, beyond lithium-ion batteries (Li-O2 and Li-S), and proton exchange membrane fuel cells.

Webex Link:http://s.uconn.edu/meseminarf90

Abstract: The extracellular matrix (ECM) of an artery endows the tissue its load bearing and damage resistance capacities. This talk will focus on the complex interplay between the multiscale ECM structural inhomogeneity and mechanics of large elastic arteries. Our recent studies integrating multiphoton imaging and quantification, biomechanical characterization, and computational modelling showed that ECM structural inhomogeneity exists at multiple structural levels of the arterial wall. At the intralemellar level, varying fiber orientation distribution and undulation contributes to local ECM mechanical properties. At the interlamellar level, transmural variation in in-plane fiber orientation distribution determines the anisotropic mechanical behavior of the elastin network. Furthermore, the waviness gradient among the elastic lamellar layers plays an important role in maintaining tissue homeostasis. Finally, structural inhomogeneity in transmural interlamellar fibers and the development and propagation of aortic dissection will be discussed.

Biographical Sketch: Dr. Katherine Zhang is a Professor at Boston University’s Departments of Mechanical and Biomedical Engineering, and Division of Materials Science and Engineering, and the Associate Chair for the Graduate Program in the Department of Mechanical Engineering. She received her B.S. degree in Engineering Mechanics from Tsinghua University; and her M.S. and Ph.D. degrees in Mechanical Engineering from University of Colorado at Boulder, where she was also a postdoc for two years. In 2006, Dr. Zhang became an Assistant Professor at Boston University and established the Multi-Scale Tissue Biomechanics Laboratory. Her research focuses on vascular biomechanics and the multi-scale mechanics and mechanobiology of the extracellular matrix. Dr. Zhang was awarded the Clare Boothe Luce Assistant Professorship in 2006, the Young Faculty Award from DARPA in 2007, and the Faculty Early Career Development (CAREER) Award from the NSF in 2010. Dr. Zhang was elected a Fellow of the American Society of Mechanical Engineers (ASME) in 2018.

Webex Link: http://s.uconn.edu/meseminarf90

Abstract: Internet of Things (IoT) represents the convergence of
three major and irreversible technology trends, namely (i) embedded sensing/smart devices, (ii) pervasive connectivity, and (iii) real-time analytics and contextual intelligence. The ability to collect and share relevant data across a wide range of devices, coupled with the ability to make real- time decisions, results in an unprecedented opportunities for system modeling, monitoring, and prognosis. In this talk, several new data analytics techniques tailored for IoT-enabled smart and connected systems will be introduced, including modeling and prognosis of condition monitoring signals using B-spline based mixed effects model, degradation model considering environmental factors, and stochastic decision making. The advantageous features of the proposed methods are demonstrated through numerical studies and real world case studies.

Biographical Sketch: Shiyu Zhou is the Vilas Distinguished Achievement Professor in the Department of Industrial and Systems Engineering and the Director of IoT Systems Research Center at the University of Wisconsin-Madison. His research focuses on data-driven modeling, monitoring, diagnosis, and prognosis for engineering systems with particular emphasis on manufacturing and after-sales service systems. He has established methods for modeling, analysis, and control of Internet-of-Things (IoT) enabled smart and connected systems, variation modeling, analysis, and reduction for complex manufacturing processes, and process control methodologies for emerging nano-manufacturing processes. He has won a large number of highly competitive federal research grants. His research also attracted significant interests from industry and received significant direct funding support from various companies. He is a recipient of a CAREER Award from the National Science Foundation and the Best Application Paper Award from IIE Transactions. He is now the director of IoT Systems Research Center at UW-Madison and a fellow of IISE, ASME, and SME.