2009 NCSA Private Sector Program Annual Meeting

2009 NCSA Private Sector Program Annual Meeting

Speaker Abstracts

Technical Workshop

Welcome
Merle Giles, Director, NCSA Private Sector Program

Design of Complex Systems
Professor Andrew Alleyne, UIUC College of Engineering Associate Dean for Research

Simulation and modeling of nonpremixed flame extinction and reignition
Professor Carlos Pantano-Rubino, UIUC Department of Materials Science & Engineering

Flame extinction and reignition in turbulent reacting flows is present in a number of propulsion and energy conversion applications but, surprisingly, it remain poorly understood. In diffusion flames, where reactants are not premixed, extinction appears when the rate of strain is sufficiently large owing to excessive heat losses. The chaotic nature of turbulent flows produce flame geometries that are highly complex and flames can develop holes that evolve and interact in time; they grow, collapse, merge and split. We present the results of some theoretical developments and those of a three-dimensional direct numerical simulation (DNS) of a spatially evolving planar turbulent reacting jet with substantial extinction and reignition fronts. A total of eight chemical species are integrated in time along with the fluid mechanical fields on a large computational grid with 100 million grid points. A novel data reduction and identification algorithm was developed to postprocess the large DNS database and extract the shape of the evolving flame surface including its edges and their propagation velocity. Statistical and morphological information of the flame will be discussed.

Multiscale Modeling of Heterogeneous Solid-Propellant Combustion
Dr. Thomas L. Jackson, UIUC Senior Research Scientist, Center for Simulation of Advanced Rockets, Adjunct Professor, Department of Aerospace Engineering

Over the past decade a considerable amount of effort has been dedicated to computational simulations of heterogeneous solid propellant combustion. A variety of tools have emerged from these studies. The most important are the modeling of propellant morphology, homogenization, and an unsteady three-dimensional coupled combustion code with complete coupling of the gas-phase combustion processes and the solid-phase heat conduction across the unsteadily regressing non-planar propellant surface. The combustion code is both parallel and scalable. These tools have been applied to the study of both non-aluminized and aluminized heterogeneous propellants. After a brief introduction, various issues with regards to heterogeneous propellant flames will be discussed. These include the particle packing problem; the manner in which one-dimensional unsteady descriptions can be constructed from the multidimensional framework, for use as a subgrid component of rocket simulations; fluctuations arising from the propellant morphology and its effect on chamber flow acoustics; the phenomena of erosive burning; etc.

Towards a Continuum Multiphysics Simulation of Metal Solidification Processes
Dr. Seid Koric, NCSA Engineering Apps. Analyst & Technical Coordinator, Industrial Projects

A coupled computational thermo-mechanical model has been developed to simulate the continuous casting of complex shaped sections, such as used for steel beam blanks. An efficient numerical procedure to integrate the constitutive equations at the local level is combined with a global finite-element solution of temperature and stress. It includes realistic constitutive behavior of the liquid/mushy zone, delta-ferrite, and austenite phases of the solidifying steel shell using a fixed grid approach. Heat transfer is computed in the shell, the complex-shaped mold, and across the interfacial gap between them, and is fully-coupled with the stress model to include the effect of shell shrinkage and gap formation on lowering the heat flux. Current work, to incorporate results from turbulent thermal-fluid flow simulations of liquid pool into this thermal-stress model, is introduced.

Combustion & HPC: Challenges to be facilitated
Dr. Ahmed Taha, NCSA Research Programmer & CFD Analyst

The presentation sheds some light on the importance of HPC in conducting meaningful combustion simulations. Due to the complexity of the physics of the combustion phenomena and its essential linkage with other complicated flow and energy characteristics such as turbulence and chemistry schemes, the combustion simulation needs huge computational resources. These demands and challenges could be facilitated by the availability of the advancing HPC capabilities. Different applications in the combustion simulations serving both the academic and industrial arenas will be presented.

Multiphysics Simulations of Gas Turbine Combustors, and HPC Needs
Dr. M.S. Anand, Manager, Aerothermal and Combustion Methods, Rolls-Royce

A brief overview of aspects of multiphysics simulations for gas turbine combustors and the current state-of-the-art in modeling will be presented from an industry perspective. Some of the computational challenges and needs related to high performance computing will be highlighted.

Q&A Panel with each speaker
Moderator: Seid Koric, NCSA

back to the agenda

Poster: Tele-immersive spaces for future business communication
Presenter: Peter Bajcsy, NCSA
Authors: Peter Bajcsy, Kenton McHenry, Rob Kooper, Rahul Malik, Andrew Spencer, Suk Kyu Lee, and Hye Jung Na
Wednesday, May 13
5:00 p.m.

This poster presents a prototype system for real-time 3D scene reconstruction using visible and thermal infrared spectrum cameras that are connected with other similar systems located at multiple geographical locations. Our tele-immersive system, called TEEVE, enables interaction between users in the same virtual space by digitally cloning the users in real-time and immersing them into a common 3D virtual space. The goal of this research and development is to remove physical distances, enable collaboration, and to provide technologies for 3D rehabilitation. Our work addresses the problems related to design, robustness and deployment of such systems. We present components for optimal placement of multiple cameras, color calibration, robust foreground detection, 3D information integration, data compression, and rendering of cloned and synthetic objects.

back to the agenda

Poster: Mining large size complex PDF documents for industrial knowledge management and preservation
Presenter: Rob Kooper, NCSA
Authors: Rob Kooper, William McFadden, Jason Kastner, Michal Ondrejcek, Kenton McHenry and Peter Bajcsy
Wednesday, May 13
5:00 p.m.

This poster addresses the problems of comprehensive document comparisons and computational scalability of document mining using cluster computing and the Map and Reduce programming paradigm. Contemporary documents, such as the Adobe Portable Document Format (PDF), contain many types of digital objects that have to be extracted, represented and summarized for further appraisal analyses and data mining purposes. While the volume of contemporary documents and the number of embedded object types have been steadily growing, there is a lack of understanding (a) how to compare documents containing heterogeneous digital objects, and (b) what hardware and software configurations would be cost-efficient for handling document processing operations such as document appraisals.

We have designed a computer-assisted framework for content based appraisal of contemporary documents. The novelty of our work is in designing a methodology and a mathematical framework for comprehensive document comparisons including text, image and vector graphics components of documents, as well as in prototyping the appraisal framework with integrity verification. We present example results of grouping, ranking and integrity verification to illustrate accuracy improvements and automation of appraisal. We have also identified the document statistical summarization to be suitable for parallel execution on computer clusters using the Map and Reduce programming paradigm. This work reports dependencies between the computation speed and the number of cluster nodes, the number of map and reduce operations per node, the partition of documents, and the hardware specifications of clusters.

back to the agenda

Poster: Automated conversion and classification of 3D CAD models for industrial data management and retrieval purposes
Presenter: Kenton McHenry, NCSA
Authors: Kenton McHenry, Rob Kooper, Michal Ondrejcek, Jason Kastner and Peter Bajcsy
Wednesday, May 13
5:00 p.m.

This poster addresses the problems of understanding 3D CAD file conversions using third party software, conversion automation of a large volume of 3D files and comparative evaluations of 3D file content. The plethora of 3D file formats poses a difficult problem for long term retention and management of 3D files during design, construction, commissioning, in-service and retirement of products. Our work is motivated by understanding conversions to several file formats vying to become the universal 3D format and by estimating the information loss of 3D format conversions over format and conversion software. We have designed a framework called Polyglot for converting 3D file formats and for measuring the information loss as one converts from one 3D file format to another. Polyglot provides capabilities for finding the conversion path, executing conversions, viewing 3D content interactively and scripting executions of a large volume of conversions.

back to the agenda

Poster: Discovery of relationships between engineering drawings and 3D CAD models for business intelligence gathering
Presenter: Michal Ondrejcek, NCSA
Authors: Michal Ondrejcek, Jason Kastner, Rob Kooper, Kenton McHenry and Peter Bajcsy
Wednesday, May 13
5:00 p.m.

This poster presents a framework for relationship discovery from 2D engineering drawings and 3D CAD Models. The framework consists of modules for automated file system analysis, file content analysis, integration of the results from analyses, storage of metadata and data-driven decision support for discovering relationships among files.

The system level analysis is performed by using Aperture and leads to metadata about file name, MIME type, and location on disk. The file content analysis includes filtering for file type detection (e.g., file format identification using DROID and PRONOM) and type-specific content analysis (such as, information extraction from 2D engineering drawings using Optical Character Recognition (OCR), and keyword based extraction of information from 3D CAD models). The integration component consolidates metadata extracted from the file system and form the file content by establishing mappings between ontologies used in metadata RDF-based representations. All RDF-based representations of metadata are stored using Tupelo in an underlying content repository. The extracted metadata in a content repository are mined by computing statistics of common descriptors across file types, for instance, for the 2D engineering drawings and 3D CAD file types.

We report our preliminary design of the framework and the performance of prototype modules for a test collection of electronic records documenting the Torpedo Weapon Retriever (TWR 841). This test collection presents a problem of unknown relationships among files that currently include 784 2D image drawings and 22 CAD models.

back to the agenda

Poster: Sparse Linear Solvers on Massively Parallel Machines
Presenter: Seid Koric, NCSA
Authors: Anshul Gupta, Thomas George, and Seid Koric
Wednesday, May 13
5:00 p.m.

For the class of problems solved by implicit numerical methods, many new implementations in computational solid mechanics and fluid dynamics may produce ill-conditioned sparse matrices, which are not easily solvable by current iterative solvers. In addition, most of these codes don't scale beyond 1,000 cores today. A comprehensive set of scalability results is performed in this study with a set of realistic test matrices on a few different massively parallel computers. A well designed factorization algorithm in wsmp direct solver has showed a scaling on over 16,000 cores while achieving a performance of over 7 trillion floating operations per second, un unprecedented number for sparse solvers. These solver capabilities will enable much more accurate modeling of complex scientific and engineering systems on the future petascale computing platforms.

back to the agenda

Poster: A Rational Integrated Approach to Designer Materials, Aerodynamics, Structural Control and Vehicle Performance: Analytical and Computational Issues
Presenter: Hank Lee, UIUC
Authors: Harry H. Hilton, Daniel H. Lee and Craig G. Merrett
Wednesday, May 13
5:00 p.m.

Traditional protocols as well as past and current conventional wisdom dictate that structural members be analyzed and designed based on readily available "off the shelf" materials. Similarly, aerodynamic shapes are selected from catalogued airfoils and performance is placed on a "best can do" rational. Prescribed limitations, such as constraints on cost, weight, some dimensions, service performances, etc., then leave open only one or two dimensional adjustments for each structural member in order to attempt to satisfy the desired operational effects.

On the other hand, the Computational Structural/Solid Mechanics Group at NCSA/TRECC has developed a novel general analytical approach that is based on the calculus of variations, where material properties, morphed airfoil shapes and servo-controls are tailored/engineered to produce prescribed structural, aerodynamic, aeroelastic and aero-viscoelastic effects as well as flight performance. In essence, this replaces the present prevailing reliance on geometric sizing and supersedes it with optimal material and aerodynamic behavior for each structural element and aerodynamic shape. Effectively, one does not set out to build a better mouse trap mechanism, but rather analytically determine optimal properties, which if manufactured will produce the "best" mouse trap.

The fundamental paradigm for flight and submarine light-weight structures is the pervasive requirements for analyses and designs that yield high-strength yet lightweight material construction and high lift/low drag devices with integrated with fluid/solid interactions. To that end, proper materials, analytical and computational tools and novel cost effective approaches must be initiated and utilized. Composites have found wide acceptance in fulfilling a significant number of these requirements and are being used to replace many traditional metal (aluminum, magnesium, titanium, etc.) structural components. However, they generally do not behave elastically but rather viscoelastically thus bringing into focus the additional time dimension; relaxation and creep properties and degradation of moduli and failure stresses with time. Ultimately, viscoelasticity introduces new phenomena, such as damping, creep and time dependent failure concepts leading to structural lifetime or survival time criteria as well as deformation induced alterations in aerodynamic responses.

The ultimate benefits stemming from analyses based on designer materials, aerodynamics and controls are that more efficient overall optimal systems can be realized in terms of overall material properties, sizing, performance, response to loads and temperatures, etc. It is estimated that the application of these protocols to the over all optimum analysis of a large transport or military airplane would require the simultaneous solution of some 800,000,000 simultaneous transcendental equations. An enterprise of such magnitude would be ideally suited for for the new NCSA-IBM Blue Waters peta-scale mainframe.

Of course, the as yet unexamined additionally needed new procedures for manufacturing structural materials to prescribed specifications based on their a priory mechanical properties require to be addressed separately by material scientists. If nothing else, at least new data bases of needed material properties can be established through computer simulations of the analytical results which will serve as guides to materials manufacturers as to what properties are desirable and needed. A parallel statement can be made for new airfoil geometries.

back to the agenda

Poster: Blue Dolphin: An Advanced Information System for Real-Time Decision Making
Presenters: Barbara Minsker and David Hill, NCSA
Authors: Barbara Minsker and David Hill
Wednesday, May 13
5:00 p.m.

NCSA is creating sustainable corporate value through unprecedented data synthesis across product life cycles, including their real-time interactions with external systems. Understanding and optimizing systems-level interactions requires dynamic coupling of end-to-end data and modeling components, which today is a cumbersome manual process that must be repeated for every new analysis. Relevant data and models are created at many organizational, spatial, and temporal scales, and integrating the information is a major challenge. We are creating cyberinfrastructure called Blue Dolphin to address these challenges through unparalleled capabilities for exploiting the vast stream of available data, including data robots that automatically harvest, transform, and fuse data from multiple internal and external sources; automated provenance that creates traceability from data to decisions; model and tool integration into an infrastructure "spinal cord" supported by background high-performance computing; and executive dashboards for real-time decision making. We have built a world-class collaboration with industry and other partners to create production-ready technologies that will provide breakthrough transformations from data to decision support. We are seeking collaboration and testbed opportunities to expand our capabilities.

back to the agenda

Poster: Data Integration in Project of Virtual World Exploration Data Bank
Presenter: Y. Dora Cai, NCSA
Authors: Y. Dora Cai, Marshall Scott Poole and Tim Cockerill
Wednesday, May 13
5:00 p.m.

VWE is a multidisciplinary Social Sciences project dedicated to the study of behavior and communication using data from the massively multiplayer online game: EverQuest2. A twenty-person team of scholars from four universities is engaged in the study. The project has integrated heterogeneous data sets from various game logs into a massive data warehouse. The project is supported by a very large database expected to reach 60 TB.

It has taken considerable effort to construct and maintain such a large database. The main challenges that we have faced include: the need to upload massive amounts of data, the need to handle the game logs captured at one second resolution with thousands of potential variables, the need to manage heterogeneous data sets, the need to cleanse noise and maintain data quality, and the need to maintain data integrity and link the large variety of game logs. To meet these challenges, we have applied four advanced database technologies: (1) create a global maximum schema to cover all variables, (2) use parallel processing for data loading and data retrieval, (3) apply load-merge techniques to upload and partition database tables, (4) utilize stored database procedures and functions to speed up database performance.

The advanced database technologies we have implemented in this project can be applied to other data integration and data warehouse projects.

back to the agenda

Poster: Managing Complex Workflows with the NCSA/LEAD Workflow Broker
Presenter: Jay C. Alameda, NCSA
Authors: Albert L. Rossi, Jay C. Alameda and Robert B. Wilhelmson
Wednesday, May 13
5:00 p.m.

We will be presenting the results of many years of development, the NCSA/LEAD Workflow Broker, which is capable of managing hundreds to thousands of jobs in support of computational science engineering applications. The Workflow Broker represents a service stack, focused on running very wide workflows with automated parametric run generation and management as a core feature, and a user interface This has been tested in adaptive weather forecasting and is being used to support atmospheric science research, currently pushing 900 jobs in single submissions to be able to better understand the impact of key parameters on weather phenomena. The Workflow Broker is currently being adapted to support workflows in concert with IBM's LoadLeveler resource manager, in the context of the NSF-funded Blue Waters project.

back to the agenda

Poster: Semantic Data Exploration - Computational Knowledge Spaces
Presenter: Terry McLaren, NCSA
Author: Joe Futrelle, Luigi Marini, Rob Kooper, Kailash Kotwani, Terry McLaren, Jim Myers, and Alejandro Rodriguez
Wednesday, May 13
5:00 p.m.

Integration of data across disciplines and organizations is critical to end-to-end system solutions for science and business problems. NCSA is developing key capabilities to manage disparate data sources, transform and synthesize data for new uses, perform and document series of computational analyses on these data, share the resulting conclusions with distributed collaborators, and preserve the data and its processing history for the long-term. This poster describes the technologies and provides examples that achieve these capabilities.

back to the agenda

Opportunities and Challenges in Computational Modeling and Simulation
Thom Dunning, NCSA/IACAT/University of Illinois at Urbana-Champaign
Thursday, May 14
9:45 a.m.

The past several decades has seen an extraordinary increase in the predictive capabilities of computational modeling and simulation. Computational simulations are able to treat more and more complex systems with ever increasing accuracy. This increased fidelity is a result of advances along several lines: more accurate computational models, improved mathematical techniques, and increased computing power. However, changes in the direction of computing technologies are posing new challenges. Increases in computing power now depends on exploiting more and more concurrency in applications. We discuss both the opportunities and the challenges.

back to the agenda

Where future is leading (Petascale and beyond)
Chris Maher, IBM
Thursday, May 14
9:45 a.m.

Petascale is upon us, Exascale is next

During the summer of 2008 the Roadrunner hybrid supercomputer, built in partnership between IBM and Los Alamos National Laboratory, became the first system to achieve a petflop of sustained performance on Linpack. Within the next few years a number of systems will be developed with peak performance of 8-20 times that of Roadrunner, using powerful multi-core processors, connected to each other over very high performance fabrics. These systems, which are under intense development today, will be able to sustain a petaflop and beyond of performance on a wide range of important scientific problems. Hard work and innovation will be required to make applications scale and perform on these systems. However, the language extensions and tools being developed today will enable application developers to fully harness the enormous capabilities of these systems. The Blue Waters project that NCSA has embarked on with IBM will be a breakthrough system in terms of scale, performance and productivity. But this is just the beginning. Technologists and scientists at IBM and in universities and laboratories around the world are already mapping the course to exascale computing. In this presentation we will discuss the capabilities and tools being developed for petascale computing and the challenges ahead to achieve exascale computing.

back to the agenda

What could we do with 'infinite' compute power to change everyday life?
Thomas Lange, Proctor & Gamble
Thursday, May 14
10:30 a.m.

Petascale and beyond is now at our fingertips. So what? How could it improve our everyday life, and what are we doing to make that happen? At P&G, examples of how we are using HPC to develop and engineer products to better serve the worlds consumers will be presented. With products ranging from Tide™ to Pantene™, Pringles™ to Olay™, Bounty™ to Gillette™, P&G touches 3 billion lives a day—hopefully making each a little better. The question of how would P&G could use unlimited computing to improve what we already do will be explored? In the broader sense, for those who make things or offer services we all experience everyday; from cars to airplanes, from groceries to medical devices, from sports equipment to banking...we will consider ideas for how to 'cash in' Moore's law for better lives.

back to the agenda

The CREATE-AV Computationally Based Engineering Software Development Project
Robert L. Meakin, DoD High Performance Computing Modernization Program
Thursday, May 14
2:00 p.m.

Over the past century of flight, aircraft have become indispensable systems for reconnaissance, transport of manpower and materiel, and projection of both force and humanitarian aid. The evolving global threats of terrorism and frequent unpredictable natural catastrophes that characterize our time have created a gathering need for improved aeronautical systems that allow us to rapidly adapt to and fulfill our obligations. This imperative comes in a time of fiscal constraint, reduced staff, challenging deployment schedules, and a growing need for timely, high quality, and actionable decision data. The traditional aircraft design/test/fix paradigm is largely based on historical experience and extensive ground-based and flight testing. The process can be costly, lengthy, and risky, demanding high degrees of expertise of the executing workforce. Ultimately, this is an unsustainable paradigm for the rapid development of new aircraft systems, which require technology integration that may differ significantly from historically well-understood vehicle classes. Further, globalization of the world marketplace has created fierce competition in all areas of technology, effectively precluding even the possibility of a sustainable continuance.

Recently, a new program was initiated through the Office of the Secretary of Defense to develop and deploy Computationally Based Engineering (CBE) software products to improve defense acquisition processes for Air Vehicles, Ships, and Radio-Frequency antennas. The program is known as CREATE, which stands for Computational Research and Engineering for Acquisition Tools and Environments. The Air Vehicles (AV) element of the program is known as the CREATE-AV Project. The purpose of the initiative is to exploit the power of next-generation computer systems in order to increase the capacity of the acquisition engineering workforce, enable streamlining of associated workflows, and minimize the need for rework due to early detection of design faults and performance anomalies. Of course, ground-based and flight test methods of generating decision data will always be essential. However, CBE-based methods of generating decision data, has the potential to significantly augment conventional methods and enable very significant process improvements. The greatest potential for CBE impact is realizable through the introduction of an ability to generate consistent, appropriate-to-need fidelity, physics-based decision data in each phase of acquisition. As a result, the most promising concepts can be rapidly and correctly identified, anomalous performance issues can be diagnosed early and remedial actions taken prior to fabrication. Subsequent to fabrication, use of CBE to conduct computational rehearsals of ground-based scale-model, full-scale prototype, as well as fielded vehicle flight tests, can lead to focused and effective testing, certification, and operational use.

Development and deployment of the kind of large, multi-scale, multi-effect CBE software products necessary to realize the potential outlined above represents a very formidable challenge. Such an enterprise requires long-term support that extends through deployment and well into maintenance phases of the product life. The present talk reviews the strategy being adopted by the CREATE-AV Project to accomplish these objectives.

back to the agenda