Multiple Objective Optimization - For many engineering
design and production planning problems, there are multiple problem objectives, and often, these objectives
are conflicting and can not be measured or observed without uncertainty. This
is further complicated by the
consideration of risk-aversion, i.e., a decision-maker may prefer a solution
with a slightly less desirable estimated objective function if it is known
that it is a more "certain" estimate. We are developing models to
efficiently identify a subset of promising solutions, we call the "pruned"
Pareto set.
Reliability of Electricity Transmission/Distribution Networks -
Electricity networks are generally very reliable but also very complex and
getting older. We have developed models to estimate network reliability and
availability and to identify critical components in the network that most
significantly impact the overall system. Additionally, we are working on
electricity network optimization and economic decision-making models based
on planned retirements of aging equipment or components.
Replacement Analysis of Aging Assets - A complex system is often
composed of aging assets and requires an optimal plan or strategy to
determine replacement and restoration (or overhaul) times to minimize total
cost of ownership, or to determine an optimal plan to increase the capacity
of the system or efficiently change to a different technology. This is
further complicated by secondary objectives, such as the minimization of
energy consumption or waste. We have developing applied optimization models
to provide solutions to this difficult problem.
Applied Operations Research - I am interested in other types of
optimization problems including scheduling and production planning, and
maintenance planning.
FUNDED RESEARCH PROJECTS
Future Electric Power System in a Carbon Constrained
World (Principal Investigators: David Coit,
Frank Felder) - grant funded by the Rutgers Energy Institute (REI)
and Center for Energy, Economic & Environmental Policy (CEEEP)
($60,000, 9/07 to 8/09) - This grant involves investigating the future of
electricity transmission & distribution systems, including distributed
resources, that are needed in a carbon constrained world using
multi-objective optimization under uncertainty to model explicitly the
trade-offs necessary to produce an energy efficient future. This design and
modeling work will lead to a systematic investigation of public policies
that would be necessary to effectuate the necessary transformation of the
electric power system.
Stochastic Optimization of System Reliability with Risk-Averse Decision
Makers (Principal Investigator: David Coit) - NSF CAREER grant
($310,000, 9/99 to 8/04) - This grant involves the development of
optimization algorithms with the goal of maximizing system reliability when
critical components within the system have uncertain reliability.
Traditional approaches require explicit knowledge of reliability for all
available components, thereby ignoring the variability and the risk implicit
in using uncertain estimates. This research is explicitly considering
uncertainty and variability with component and system reliability estimation
and the propagation of uncertainty.
Probabilistic Decision Support for Evaluating Technology Insertion and
Assessing Aviation Safety System Risk (Principal Investigator: James
Luxhoj) - NASA grant ($1,164,000, 3/03 to 3/06) - I am collaborating
with Jim Luxhoj (PI) in this grant involving the modeling of aviation safety and
risk, and the assessment of risk mitigation technologies and insertions. An
Aviation System Risk Model (ASRM) is being developed and enhanced based on
Bayesian belief networks. ASRM is a risk-based decision support system
designed to evaluate the safety risk by studying observed accidents and
using analytical generalization. Bayesian probability theory is being used
for model quantification.
Impact of Parameter Uncertainty on Asset Criticality and System Reliability
(Principal Investigator: David Coit) - Industry grant ($30,000, 1/04
to 12/04) - This project is to model the uncertainty related to reliability
evaluation of electricity transmission systems. Reliability and uncertainty
metrics are being developed for outage rates, repair times and downtimes for
various electricity transmission configurations. The metrics relate
component reliability estimation uncertainty to the effect at system level.
They can be used to prioritize data collection and reliability improvement
activities.
Relating Field Data to Accelerated Life Testing (Principal
Investigators: David Coit, E. A. Elsayed) - funded by NSF ($100,000,
8/00 to 7/02) - This grant was to investigate the relationship between the
actual wear, degradation and failure of vehicle controllers experienced in
the field with that expected from accelerated testing conducted in the
laboratory. DaimlerChyrsler Electronics, Huntsville, AL, is the controller
designer and manufacturer. This is a TIE grant conducted with a corresponding
grant awarded to Auburn University's Center for Advanced Vehicle Electronics
(CAVE).
Reliability Modeling and Integrated Prognostics for Highly Reliable Systems
(Principal Investigator: David Coit) - funded by the NSF/Industry QRE
Center ($30,000, 10/00 to 12/01) - In this study, we reviewed and
compiled new and novel approaches to the improvement of reliability of
fielded systems. For many modern systems, traditional reliability
engineering analyses, metrics and design practices are not leading to the
very high reliability requirements often needed or expected. Although, many
approaches were evaluated, particular emphasis was focused on different
approaches to implement integrated prognostics to anticipate failures.
Reliability Prediction Based on Degradation with Multiple Changing Stresses
(Principal Investigator: David Coit) - funded by the NSF/Industry
QRE Center ($30,000, 10/99 to 12/00) - Degradation modeling represents
an opportunity to observe and analyze performance deterioration and to
predict reliability prior to the occurrences of failure. Parametric drift or
deterioration is modeled as a dynamic (time-variant) probability
distribution. Reliability predictions are made by considering shifts in the
distribution compared to a failure threshold. In this project, currently
available methods were extended to consider changing stress levels and
uncertain stress conditions.
Repairable Systems Reliability: Planning and Assessment Tools
(Principal Investigator: David Coit) - funded by the NSF/Industry QRE
Center ($30,000, 10/98 to 12/99) - This project involved the development
and automation of planning and assessment tools for repairable system
reliability with non-constant rate of occurrence of failure. The
distribution of failure times is often non-stationary for many products
during (1) development testing where there is reliability growth, and (2)
extended field usage where there is deterioration. The problem is further
compounded because a system is designed with many subsystems and components,
each with different failure patterns.
People Sensitive Processes (Principal Investigator: David Coit) -
funded by U. S. Army ARDEC ($22,700, 8/98 to 12/99) - People related
processes are an integral part of most manufacturing processes, whether they
are highly automated or almost entirely manual in nature. Effective
processes with significant manual processing must depend on consistent,
sustained levels of human performance if they are to consistently produce
units free from error. Those factors which either directly or indirectly
influence human performance or impact the occurrences of human errors were
compiled and mapped to activities at an Army contractor manufacturing
facility.
System Reliability Models with Uncertain Component Reliability Estimates
(Principal Investigator: David Coit) - funded by the NSF/Industry QRE
Center ($30,000, 10/97 to 10/98) - Sufficient data are often not
available for all components in a system for highly accurate system-level
reliability predictions. System reliability models and sensitivity measures
have been developed based on component reliability and the variability
associated with the component reliability estimates or predictions. The
models explicitly acknowledge the uncertainty with component reliability
estimates or predictions. Sensitivity measures quantify the relationship
between a lower-bound on system reliability and the component reliability
estimation variance.
SOFTWARE
RelSens
Reliability Sensitivity Tools- These tools characterize and quantify
the propagation of uncertainty on system reliability. This software tool
provides system reliability models and sensitivity measures based on
component reliability and the variance associated with the component
reliability estimates or predictions. The models explicitly acknowledge the
uncertainty with component reliability estimates or predictions. Sensitivity
measures quantify the relationship between a lower-bound on system
reliability and the component reliability estimation variance.
RRAT Repairable System Reliability Assessment Tools - These
tools model trends for repairable systems reliability. This is useful during
reliability growth studies or when studying system deterioration or
degradation. This software tool analyzes and models system failure data.
Three different trend tests are included to test the significance of any
trend. A nonhomogeneous Poisson process (NHPP) model is used to model the
rate of failure for repairable systems. Specifcally, the Crow/AMSAA model is
used.
If
you download software, please notify me via
email
