Technical Thrust Area

Large Scale Structural Systems and Optimal Design

Description

The optimal design of large scale structural systems will enhance the survivability of man-made engineering systems and their interaction with natural phenomena. As such, a focus of this TTA is the reduction of the computation time of the forward problem solution using subscale models, reduced-order models, and/or the development of computational algorithms based on acceleration cards, e.g., GPUs. The computational power of the upcoming supercomputers will rely on these acceleration cards and thus porting current computational algorithms to support these cards is also of interest. Another focus of this TTA is the accurate modeling of failure of large scale structures across vast spatiotemporal scales: computational modeling of brittle or ductile failure initiation at the smallest scale and the accurate transition to the largest structural scale. Novel formulations and/or computational methods that effectively capture the complexity of fracture and damage phenomena are thus necessary. In addition, developments of novel optimization algorithms and design optimization formulations that efficiently handle a large number of design variables and various complex performance/fabrication constraints are vital. Finally, many large scale structural systems are subjected to a variety of environmental conditions, including mechanical, thermal, chemical, electromagnetic, and/or gravitational fields. Consequently, optimal design of such structures requires the development of new and advanced multiscale and multiphysics coupled computational algorithms.

Topics of interest in the context of materials and large scale structures include:

  • Topology optimization
  • Failure modeling
  • Multiscale and multiphysics modeling
  • High performance computing and algorithms for modern architectures
  • Subscale and reduced-order modeling

Committee 

Chair: Pablo Seleson, Oak Ridge National Laboratory
Vice-Chair: X. Shelly Zhang, University of Illinois at Urbana-Champaign
Members-at-Large: Patrick Diehl, Louisiana State University

George Gazonas, CCDC Army Research Laboratory

Members
Not a member? Click here to learn how to become one.

Early-Career Colloquium

The TTA is hosting a monthly early-career colloquium. The colloquium will be held the first Wednesday of the month at 1PM CST. Registration for the colloquium is free but required.

December 7, 2022: Prof. Xingsheng Sun, University of Kentucky

Title: Rigorous Multiscale Quantification of Material Uncertainties
Abstract
The macroscopic properties of materials and structures that we observe and exploit in engineering applications result from complex interactions between physics at multiple length and time scales: electronic, atomistic, defects, domains, etc. Multiscale modeling seeks to understand the interactions between these physics across scales. However, assessing such interactions can be challenging due to the complex nature of material properties and the prohibitive computational cost of integral calculations. This talk will focus on how to quantify the propagation of material uncertainties across multiple scales. To this end, we exploit the multiscale and hierarchical nature of material response and develop a framework to quantify the overall uncertainty of material response induced by the uncertainties at finer scales without the need for integral calculations. Specifically, we rigorously bound the uncertainty at each scale and then combine the partial uncertainties in a way that provides a rigorous bound on the overall or integral uncertainty. The bound provides a conservative estimate on the uncertainty. Importantly, this approach does not require integral calculations that are prohibitively expensive and is able to quantify the propagation of uncertainties through different paths across scales. Finally, the developed framework has been employed to investigate how material uncertainties in the single-crystal properties of magnesium affect the ballistic performance of armor structures.
Bio: Dr. Xingsheng Sun is a tenure-track Assistant Professor in the Department of Mechanical and Aerospace Engineering at the University of Kentucky (UK). Prior to joining UK, he was a Postdoctoral Scholar in Aerospace at California Institute of Technology from 2018 to 2021. His main research interest lies in long-term atomistic modeling and simulations, multiscale quantification of material uncertainties, mechanics of materials in extreme conditions, and application-driven materials by design. Dr. Sun holds a Ph.D. degree in Aerospace Engineering from Virginia Tech, a M.S. degree in Mechanical Engineering from Hunan University (China), and a B.S. degree in Mechanical Engineering from Dalian University of Technology (China).

REGISTER

November 2, 2022: Prof. Hayoung Chung, Ulsan National Institute of Science and Technology

Title: Understanding Multifunctional Behaviors using Sequential Multiscale Approaches
Abstract: Multiscale methods broadly refer to computational frameworks used to investigate the influence of small-scale parameters on macroscopic behaviors of multifunctional materials, namely low-dimensional materials and smart materials, to just name a few. The way that small-scale information, typically extracted from in-silico experiments (e.g., DFT, MD), is transferred to the macroscopic analysis model (e.g., finite element) distinguishes one multiscale method from the others. Among these, the sequential multiscale method is an alias of a one-way bridging model where the small-scale data forms a basis of a high-fidelity multiphysics surrogate that is utilized as a constitutive model of the macroscopic analysis. The method is especially useful in analyzing structural behaviors of multifunctional materials, where we can assume static equilibrium conditions on both lower and higher scales. However, despite its simplicity and wide usage, the multiscale method is far from being generalized due to its need for massive data and knowledge of the specific domain, which my team and I are trying to eliminate. In this talk, we first aim to explain the broad application of the sequential multiscale method, as it enables a more profound understanding of non-traditional multifunctional materials such as nanomaterials and liquid-crystal solids. After that, we present our recent endeavor toward a scalable and transferable multiscale model using the state-of-the-art machine learning approach. We finalize the talk by introducing our other endeavors toward fast and efficient topology optimization.
Bio: Hayoung Chung is an assistant professor in the Department of Mechanical Engineering at Ulsan National Institute of Science and Technology (UNIST) in Korea since 2019. He obtained his bachelor’s and doctoral degree at Seoul National University (SNU), Korea, in 2010 and 2017. His Ph.D. research focused on analyzing and tailoring multiphysics phenomena in photo-responsive liquid crystal networks. After graduation, he held postdoctoral positions at the Structural Department at UC San Diego, USA, where he worked on several projects revolving around efficient topology optimization of thermoelastic large-deforming structures. His current research interests include topology optimization, data-driven multiscale modeling, and computational mechanics to address issues found in 4D printing technology.

October 5, 2022: Prof. Brendan Keith, Brown University

TitleAdaptive Sampling for Constrained Optimization under Uncertainty
Abstract: Stochastic optimization problems with deterministic constraints commonly appear in machine learning, finance, and engineering applications. This talk presents an improved adaptive solution strategy for this important class of problems. The aim is to decrease the computational cost while maintaining an optimal convergence rate. The guiding principle is to adjust the batch size (or sample size) on the fly so that the error in the gradient approximation remains proportional to the error in the underlying optimization problem. After providing motivation and context, I will present new adaptive sampling algorithms that simultaneously maintain optimal sample efficiency and iteration complexity for risk-neutral and risk-averse optimization under uncertainty with deterministic constraints. I will then demonstrate the efficacy of these algorithms in multiple applications, drawing mainly from use cases found in engineering design. This talk will provide an introduction to adaptive sampling that aims to be accessible to a broad audience as well as showcase ongoing work in collaboration with Lawrence Livermore National Laboratory and UT Austin.
Bio: Brendan Keith is an Assistant Professor in the Division of Applied Mathematics at Brown University in Providence, Rhode Island. His research interests are mainly related to the modeling and simulation of problems arising in natural sciences and engineering, focusing on numerical methods for partial differential equations, scientific machine learning, and PDE-constrained optimization. In 2018, Brendan received his Ph.D. in Computational Science, Engineering, and Mathematics from the Oden Institute for Computational Engineering and Sciences at the University of Texas at Austin. He has held postdoctoral positions at TU Munich, ICERM, and Lawrence Livermore National Laboratory.


Career Path Panel at EMI 2022

The TTA jointly organized with the EMI Computational Mechanics Committee a Career Path Panel at the Engineering Mechanics Institute Conference 2022 (EMI 2022). The event featured panelists from the academia, national labs, and the industry, followed by a networking session. For further details click here.

Graduate Mentoring Program

The TTA organized a summer mentoring program for USACM graduate students. Students were matched with a mentor from the academia, DOE national labs, or the industry for a virtual one-to-one meeting to discuss topics of interest to the students, including getting guidance on career paths and receiving advice on how to make the best of WCCM-APCOM 2022 for those students planning on attending the congress.


 Past Early-Career Colloquium Talks

April 6, 2022: Dr. Masoud Behzadinasab, Brown University

TitlePeridynamics for Simulation of Fracture Mechanics Problems
Abstract: Understanding the fracture phenomena, i.e., catastrophic material failure, remains a major challenge in a variety of fields within science and engineering. While limited analytical solutions exist for fracture mechanics problems and experiments are often costly in many practical cases, computational methods and simulations are commonly used in practice for analysis, design, and optimization of materials, structures, buildings, and vehicles. Therefore, it is imperative to develop accurate, efficient, and robust predictive models for simulating the response of solids and structures in real-world scenarios such as in extreme conditions involving large deformation, fracture, and fragmentation (e.g. blast events). Computational fracture mechanics has been an active area of research for several decades. Novel methods have emerged over the last two decades for solving fracture mechanics problems such as the peridynamic (PD) theory, the extended finite element method (XFEM), and phase-field fracture models. The continuum PD model, which is based on an integro-differential formulation, has increasingly attracted the attention of researchers in computational solid mechanics due to its natural capabilities in handling material discontinuity and autonomous crack growth. In this talk, I will summarize the key developments in the field of peridynamics and arrive at a conclusion that two schools of PD are emerging in recent years. One school takes a more traditional view of "PD as a model" of a nonlocal continuum, while another approaches "PD as a discretization" methodology for local continua where the nonlocality is reduced under mesh refinement. The focus of this talk will be on correspondence-based PD, which enables the use of classical constitutive models. I will present: 1) a ductile fracture PD model and its application to the third Sandia Fracture Challenge, 2) a PD Kirchhoff--Love shell formulation for the failure analysis of thin-walled structures undergoing inelastic large deformations, 3) a recently developed PD formulation corresponding to the now classical, yet continuously evolving Microplane constitutive model (the M7 version) for the failure analysis of concrete materials, and 4) an immersed coupling of isogeometric analysis and peridynamics for the simulation of blast-induced fluid-structure interaction. Several numerical examples are provided to showcase the applications of peridynamics to fracture mechanics problems and how it can enhance our predictive capabilities in extreme events.
Bio: Masoud Behzadinasab is currently a Postdoctoral Research Associate in the Computational Mechanics group at Brown University led by Prof. Yuri Bazilevs. Prior to joining Brown in January 2020, he earned his PhD in Engineering Mechanics, supervised by Prof. John Foster, from the University of Texas at Austin. His research is in computational solid mechanics and mainly focused on developing novel methods for fracture mechanics problems. Masoud is a recipient of the 2019 IMECE Best Student Poster Award on Computational Mechanics and the inaugural 2021 Thomas J. R. Hughes Fellowship from the USNC/TAM.

March 2, 2022: Dr. Som L. Dhulipala, Idaho National Laboratory

Title: Active Learning for Accelerating Expensive Computational Tasks in an Open-Source High-Performance Computing Platform
Abstract: 
Active learning is a process by which the machine learning (ML) model, initially trained on a sparse dataset, actively queries the next training point by analyzing the prediction uncertainties. For expensive computational modeling tasks such as uncertainty quantification with high-fidelity finite element models, active learning has the capacity to bring significant computational gains while providing robustness guarantees on the predictions. This presentation will introduce a recently developed active learning with multifidelity modeling algorithm for efficient rare events estimation using computationally expensive models. Then, the massively parallel open-source Multiphysics Object Oriented Simulation Environment (MOOSE) is introduced and the implementation of active learning algorithms in MOOSE is discussed.  These active learning capabilities are demonstrated by applying them to several examples such as Navier-Stokes equations and nuclear fuel failure. This presentation is concluded with ongoing work related to the consideration of multiple low-fidelity models in multifidelity active learning and the application of active learning to a heat-pipe microreactor failure analysis. 
Bio: Dr. Som Dhulipala is a Computational Scientist at Idaho National Laboratory. His interest lies in the area of uncertainty quantification as applied to finite element analysis and critical infrastructure resilience. Currently, he is researching the theory and applications of multifidelity modeling, Bayesian deep learning, Markov models, and efficient sampling algorithms for computational problems critical to INL and the DOE missions. He also develops the open-source software MOOSE stochastic tools module for conducting uncertainty quantification studies using multiphysics models.

February 2, 2022: Prof. Narasimha Boddeti, Washington State University

Title: Optimal Design and Manufacture of Architected Fiber Reinforced Composites
Abstract: 
Fiber reinforced composites (FRCs) are important advanced materials that are widely used in aerospace, naval and construction industries. FRCs possess a simple microstructure characterized by the fiber modulus, orientation, aspect ratio and volume fraction. Through deliberate variation of these microstructural parameters, a desired effective material response can be architected into the structure leading to an efficient use of the material and superior structural performance (i.e., lighter weight, stiffer, stronger, and tougher composites) akin to natural materials such as wood and bone. Advances in manufacturing technologies such as material extrusion additive manufacturing and automated fiber placement are now enabling the realization of such complex spatial variation of fiber-based microstructures. However, most design approaches either do not take complete advantage of this extended design freedom or fail to take fabrication constraints into account. We developed a novel design to manufacture framework that combines efficient design synthesis for architected composite structures via multiscale topology optimization and process planning for fabrication through novel computational geometry algorithms. Our approach enables design of complex, optimized composite shell structures and can be adapted to various manufacturing methods.
Bio: Narasimha (Na-ra-sim-ha) is an assistant professor in the School of Mechanical and Materials Engineering. He obtained his bachelor’s degree in mechanical engineering from IIT, Kharagpur, India (2002-2006) and doctoral degree from University of Colorado, Boulder, USA (2009-2014). He was a postdoctoral scholar at CU Boulder and Singapore University of Technology and Design before joining WSU in 2020. His PhD research focused on adhesion mechanics of graphene while his current research focuses on developing design automation methods for additive manufacturing. His research interests include topology optimization, additive manufacturing, computational mechanics, architected materials, and soft robotics.

December 1, 2021: Prof. Jifu Tan, Northern Illinois University

Title: Multiphysics Modeling of Biological Flows with Cell Suspensions
Abstract: Fluid flows containing solid particles are common in engineering and biology, e.g., complex flow with cellular suspensions. The dynamic behavior in such phenomena is complex and poorly understood due to interactions between individual particles as well as interactions between the particles and the surrounding fluid and bounding walls. Moreover, the presence of highly deformable particles, such as blood cells, vesicles and polymers, make it particularly challenging to accurately describe the dynamics in such systems. In this presentation, a massively parallel Multiphysics simulation platform developed on popular open-source code will be introduced to simulate such complex flows. The fluid flow was solved by the Lattice Boltzmann method (LBM), while the solid deformation and diffusion were simulated by particle-based solver. The coupling was achieved through the immersed boundary method (IBM) so that different physics can exchange information with each other. The developed simulation framework was validated with various analytical solutions and experiments and shown to scale almost linearly over thousands of processors. Applications are given in drug delivery, cancer cell detection in microfluidics, and blood flow in a retina vascular network, demonstrating an efficient way of simulating coupled multi-physics problems.
Bio: Dr. Jifu Tan currently is an assistant professor in Mechanical Engineering at Northern Illinois University. Before that, he was a postdoctoral associate in the Department of Chemical and Biomolecular Engineering at the University of Pennsylvania. He obtained his Ph.D. and M.S. in Mechanical Engineering from Lehigh University in 2015 and 2012, respectively. He received his B.S. degree in Civil Engineering from Beijing Jiaotong University in 2007. His primary research interest is fluid structure interaction and its application in engineering and medicine, such as nanoparticle delivery, blood flow modeling, thrombosis and bleeding simulation, microfluidic device design for cell separation. He is also interested in high performance computing, multiscale modeling, and data driven models. He is actively engaged in developing open-source code for scientific research.

November 3, 2021: Prof. Matthew J. Zahr, University of Notre Dame

Title: High-Order and Reduced-Order Methods for Improved Engineering Analysis and Design
Abstract: Optimization problems governed by partial differential equations are ubiquitous in modern science, engineering, and mathematics. They play a central role in optimal design and control of engineering systems, data assimilation, and inverse problems. However, as the complexity of the underlying PDE increases, efficient and robust methods to compute the objective function and its gradient become paramount. To this end, I will present a model reduction framework to reduce the time and resources required to solve optimization problems governed by PDEs. The framework is demonstrated in the context of aerodynamic shape optimization and structural topology optimization. In addition, I will demonstrate the role of optimization in computational physics extends beyond traditional design and control problems. I will introduce a novel optimization-based numerical method for high-order accurate approximation of PDE solutions with non-smooth features, e.g., flows with shock waves and fracture of solid media. I will demonstrate the method with a suite of two- and three-dimensional compressible flow problems and discuss the extension to fracture. In all cases, discontinuities in the flow are fit to high-order accuracy with curved mesh elements, which leads to accurate solutions on extremely coarse meshes.
BioMatthew is an assistant professor in the Department of Aerospace and Mechanical Engineering at the University of Notre Dame. He received his PhD in Computational and Mathematical Engineering from Stanford University in 2016 and from 2016-2018 was the Luis W. Alvarez Postdoctoral Fellow in the Department of Mathematics at Lawrence Berkeley National Laboratory. His research interests include high-order methods for computational physics, PDE-constrained optimization, model reduction, and numerical methods for resolving shocks and discontinuities. In 2020, he received the AFOSR Young Investigator Award.

October 6, 2021: Prof. Evgueni T. Filipov, University of Michigan

Title: Simulating Shape Morphing Origami: Kinematics, Mechanics, and Multi-Physics
Abstract: Origami principles for folding of thin sheets can create a variety of deployable, reconfigurable, and adaptable three-dimensional structures. Practical applications range from metamaterials and biomedical micro-robotics, to large-scale deployable architecture. This talk will present my group’s work in creating analytical models for simulating morphing origami-inspired structures at multiple scales. The models are computationally efficient because they use a simplified bar and hinge framework to capture the geometry of the origami, yet they are capable of simulating kinematics, mechanics, and multi-physical behaviors of the origami. The models can capture highly nonlinear behaviors including contact, multi-stability, active actuation, and electro-thermo-mechanical coupling. The talk will present scenarios of how we apply these tools to real world analysis and design problems. We simulate self-assembly of micro-robots, evaluate the stiffness-to-weight ratio of origami structures, and explore the complex stiffness anisotropy in origami with curved creases.
Bio: Evgueni Filipov is an Assistant Professor in the Department of Civil and Environmental Engineering at the University of Michigan, Ann Arbor. His research interests are focused on the underlying mechanics of origami-inspired deployable and reconfigurable structures. These mechanics are employed to improve stiffness, functionality, and manufacturing of the folded systems. He holds MS and PhD degrees in Civil Engineering from the University of Illinois at Urbana-Champaign, and a BS from Rensselaer Polytechnic Institute. He has received the NSF CAREER Award (2020), the DARPA Young Faculty Award (2018), the Cozzarelli Prize from the National Academy of Sciences (2015), and the NSF Graduate Research Fellowship. Learn more about his research at his lab’s website: http://drsl.engin.umich.edu/


Members Only (password required)

Discussion Forum