Advancement of Environmental Decision Support Systems Through HPCC
E. D. Brill, J. W. Baugh, and S. R. Ranjithan
US Environmental Protection Agency
8/96 to 8/00
Decision making for single- and cross-media environmental problems is a complex
and tedious process. Competing design objectives and constraints, coupled
with large quantities of data and complex simulation models, cause policymakers
to spend enormous amounts of time and effort just to find feasible management
strategies. The goal of this research is to overcome the current computational
resource limitations by developing tools for use within a high-performance computing
and communications (HPCC) environment, bringing DSSs closer to realizing the
decision-making power of a true joint-cognitive system.
Decision Support for Seismic Performance Evaluation
Abhinav Gupta, John Baugh Jr., and G. Mahinthakumar
National Science Foundation
9/00 to 2/02
The overall goal of this research is to develop formal computational approaches
that support comprehensive decision making in structural engineering.
Specifically, these approaches will be realized in a prototype decision support
system that draws on complementary strengths of the engineer and the computer
in a joint-cognitive system. Its design will be based on three major concepts:
(1) optimization for evaluating alternatives and supporting what-if analyses;
(2) sub-component approach to address structural model synthesis, scalability,
model updating, and uncertainty propagation; and (3) implementation in an object-oriented
framework of high performance distributed computing. A series of simple
and real-life test cases will be used to evaluate the proposed prototype.
Employing Artificial Neural Networks and Genetic Algorithms to Optimize Turbidity
and Natural Organic Matter Removal In Drinking Water Treatment
D. R. U. Knappe, S. Ranjithan, and J. Ducoste
ongoing
Many drinking water treatment plants conduct pilot studies to optimize treatment.
These pilot studies typically result in a wealth of data that is difficult to
interpret. This study showed that artificial neural network models could
describe pilot study results that were obtained over a wide range of treatment
conditions and raw water qualities. Pilot studies were conducted by the
City of Philadelphia, PA. Artificial neural network models are currently
used to determine optimal treatment strategies that consistently produce desired
water quality criteria at minimum cost.
Application of Genetic Algorithms to the Automotive Safety Industry and Research
in Improving the Liklihood of Identifying the Global Optimum
D. H. Loughlin and J. W. Baugh Jr.
Delphi-Delco Electronics Systems Corp.
9/98 to 5/01
This project involves the application of a GA-based optimization package to
the calibration of an automotive airbag release algorithm. In the first
phase of this project, a distributed optimization software package capable of
GA and local search was developed. This package was demonstrated to improve
airbag performance as well as to reduce overall calibration time. In the
second phase of this project, the GA component of the optimization software
was improved to utilize distributed computing resources more efficiently, reducing
calibration time.
Enhancements to the Strategy Development Tool to Support Regulatory Analysis
D. H. Loughlin, J. W. Baugh, E. D. Brill, and S. R. Ranjithan
NC Department of Environment and Natural Resources - Division of Air Quality
9/98 to 8/01
The Strategy Development Tool (SDT) is a prototype decision support system for
air quality management. Components of the SDT include tools for: 1) visualizing
emissions inventories, 2) designing and testing control strategies, 3) modeling
the costs and impacts of incentive-based control approaches such as emissions
trading programs, and 4) identifying low-cost management alternatives through
optimization. The goal of this project is to upgrade the features of the
SDT so that it can be used by the State of North Carolina in developing state
implementation plans towards meeting the Federal air quality standards.
In addition, we are using mathematical models and an approach called "modeling
to generate alternatives" to predict the results of emissions trading programs.
System-Wide Optimization of Wastewater Treatment Plants Using Genetic Algorithms
D. H. Loughlin, J. Ducoste, and F. de los Reyes
Hydromantis, Inc. and the National Science Foundation Research Experiences for
Undergraduates (REU) Program
10/00 to 12/01
In this project, we are using a genetic algorithm (GA) to perform system-wide
optimization of wastewater treatment unit processes. The GA simultaneously
selects efficient unit processes and optimizes their design parameters with
the goals of minimizing cost and ensuring that effluent standards will be met.
The GA-based optimization approach will be used for single and multi-objective
optimization, as well as for reliability-based design.
A Parallel Supercomputer Model for Lung Acoustics
G. Mahinthakumar (with R. Ward and C. Easterly, Oak Ridge National Laboratory)
Laboratory Directed Research and Development, Oak Ridge National Laboratory
(ORNL)
10/99 to 9/01
The overall goal of this project is to develop a computational model for the
airflow and sound generation through the human lung and to validate the results
using experimental data. This project is part of the larger multi-institutional
virtual human initiative centered at ORNL. The project team consists of
several senior researchers from ORNL, University of Tennessee, University of
Utah, and us at NCSU. Our primary responsibility in this project is to
develop a parallel computer model sound generation and propagation through the
human lung using the visible human data from the National Library of Medicine.
The graduate student working on this project will spend the 2000 and 2001 summer
semesters at ORNL (supported by the project) working with ORNL researchers and
using the state-of-the-art parallel computers at ORNL.
Development and Performance Testing of a Large-Scale Parallel Groundwater Flow
and Transport Code
G. Mahinthakumar
1/00 to 12/02
We have developed (over the past 5 years) a parallel groundwater flow and transport
simulator intended for high resolution numerical simulations involving multicomponent
groundwater transport and remediation. The main feature of the code is
that it can solve very large problems (in the order of tens of millions of cells)
efficiently on today’s parallel supercomputers. We will continue with
the development and performance testing of this code on the latest parallel
computers including those at ORNL, SDSC, NERSC, and NCSA.
Heuristic Approaches for Solving PDE-Governed Inverse Problems
G. Mahinthakumar
1/00, ongoing
We will build upon our past work on using genetic search algorithms (GA) for
solving groundwater biological activity zone identification problems (published
in December 1999 issue of ASCE J. of Environmental Engineering) to include other
types of inverse problems involving large-scale partial differential equations
(PDE) solutions. Examples include contaminant source identification (e.g.,
DNAPL pools), seismic detection of land mines, etc. Another focus of this
research is to investigate empirical (e.g., artificial neural networks) and
other approximate solution approaches (e.g., proper orthogonal decomposition)
to the forward PDE problem. Less costly approximate solutions can considerably
reduce the computational cost associated with the large number of forward PDE
solutions required in heuristic approaches such as GA by providing better starting
trial solutions.
Use of Chemical Waveforms for Subsurface Characterization Problems
G. Mahinthakumar and G. Moline
Faculty Research and Development Fund
1/01 to 6/01
We are investigating a new technology for chemical remote sensing that will
provide a quantitative measure of subsurface properties and mass transfer processes.
This technology is based on combining two important components: (i) knowledge
of the interaction between chemical tracers and subsurface constituents such
as nonaqueous phase liquids (NAPLs) (e.g., hydrocarbons and organic contaminants),
microbial populations, organic carbon (biomass), or the geological materials
themselves; and (ii) a highly structured chemical wave analogous to the energy
waves used in geophysical remote sensing methods. Our ongoing investigation
is primarily based on hypothetical three-dimensional numerical simulations but
lab experiments are planned in the future.
Soft Computing for Geotechnical Problems
M. S. Rahman
1998, ongoing
Uncertainty, imprecision, complexity, and nonlinearity are inherent in many
geotechnical problems. The conventional modeling of the underlying systems
becomes quite difficult. Recently, a new approach to modeling has emerged
under the rubric of "soft computing." This consists of many complementary
tools: fuzzy logic, neural network, probabilistic reasoning, genetic algorithm
and others. Among these, neural networks can handle complexity and nonlinearity,
while fuzzy logic provides a rational framework to incorporate imprecision.
In this research program, applications of “soft computing,” are being developed
for some important geotechnical problems. Fuzzy neural network models
are being developed for the prediction of liquefaction, liquefaction-induced
ground displacements, uplift capacity of suction caissons, and axial capacity
of offshore piles.
NSF CAREER Award: Development of a Computer-Based Methodology to Assist
in Environmental Systems Analysis and Decision Making and Its Applications in
Watershed Management
S. R. Ranjithan
National Science Foundation
5/98 to 5/02
This project will: 1) investigate ways to enhance the capabilities of genetic
algorithms (GAs) for complex environmental systems analysis, 2) develop and
integrate into existing courses an interactive training module to assist in
teaching the fundamentals of GAs and their uses in environmental systems analysis,
3) explore applications of the methodology in watershed management, and 4) integrate
the applications and their findings in courses related to environmental systems
analysis. The academic plans include the development of teaching modules
and interactive techniques for both students and practitioners. To demonstrate
the practical applicability, an array of realistic watershed management applications
will be investigated, including a case study of the Neuse River Basin in North
Carolina.
A Sign Inventory Study to Assess and Control Liability and Cost
W Rasdorf and J Hummer
NC Department of Transportation (NCDOT)
1/01 to 6/02
The State of North Carolina currently maintains on the order of 3 million signs
and hundreds of thousands of miles of pavement markings. In the year 2002
the Federal Highway Administration is expected to release a final rule which
will specify minimum levels for sign and pavement marking retroreflectivity.
Compliance with the standard will be costly and noncompliance will have serious
liability implications. The purpose of this project is to determine, quantify,
and present to NCDOT alternative approaches for meeting the standard.
The results of this study will enable the NCDOT to minimize this new financial
burden and reduce liability while improving safety.
Models and Algorithms for Quality and Accuracy of Spatial Data in Infrastructure
Engineering
W. Rasdorf (NC State University) and H. Karimi (University of Pittsburgh)
National Science Foundation
10/99 to 9/02
Two key data items in any infrastructure information management system (IIMS)
include (1) location of a component and (2) the resulting spatial relationships
among components that result from their locations. To gain additional
IIMS functionality and to support more useful analysis, we are investigating
linkages between spatial and attribute information. Such tools or capabilities
would improve infrastructure-related problem analysis as well.
Operating Policies for Improved Transit Productivity
J. R. Stone and J. W. Baugh
SYSTAN, Inc.
5/00 to 7/01
The purpose of this research is to develop approaches for paratransit operators
who wish to improve system productivity by implementing innovative operating
policies. Using a multiobjective optimization procedure based on simulated
annealing that was developed in the WSTA Mobility Management project, we will
evaluate policy issues in two cities: Santa Clara and Winston-Salem. By
simulating operating policies, we can estimate impacts on vehicle productivity
and on operating costs. The results of the research will be used to establish
general policy guidelines for paratransit routing and scheduling.
Return to the Department of Civil Engineering Home Page