44
NuclearPlantJournal.com Nuclear Plant Journal, January-February 2014
Controlling O&M...
on synthetic and experimental data have
also shown the ability to account for
a reasonable amount of variability in
NDE measurements due to, for instance,
microstructural variationsormeasurement
noise. Ongoing experimental studies are
assessing the sensitivity of advanced
NDE techniques, and analysis techniques
that can relate the measurement back to
the level of accumulated creep strain in
the specimen.
Advantage of Integrating Diagnostics
and Prognostics with ERMs
Risk monitors extend probabilistic
risk assessment (PRA) frameworks
by incorporating the dynamic plant
configuration(e.g.,equipmentavailability,
operating regimes, and environmental
conditions) into the risk assessment.
PRA is itself a systematic safety analysis
methodology that follows four steps: (1)
identify undesirable consequences (e.g.,
reactor unavailability and core damage)
and initiating events that can lead to
these consequences; (2) systematically
identify accident sequences (defined by
event trees and fault trees) through which
the facility can move from the initiating
event to the undesired consequence;
(3) calculate the probability of
occurrence for each accident sequence;
and (4) rank the accident sequences
according to probability of occurrence
(or, alternatively, contribution to the
undesirable event) to manage the major
contributors to risk. Time-independence
of component failures is assumed in
traditional PRA modeling, and PRA
component failure rates are typically
assumed to be static over the life of the
component. Changes (i.e., degradation)
in the failure rate of a component that
might be expected to normally occur
over the component life are not explicitly
represented.
Enhanced risk monitors (ERMs)
integrate
equipment
condition
assessment (ECA) and prognostics
information to calculate time-dependent
failure probabilities. Essentially, an
ERM incorporates the time-dependent
failure probabilities from PHM systems
to dynamically update the risk metric of
interest. In this, the ERM differs from
other approaches that incorporate aging
models for key components. Rather
than include generic aging models (for
example linear aging models where the
failure probability increases linearly
over time), the ERM approach uses
condition of the component to calculate
the failure probability. This condition-
based approach can potentially enable
improved estimates of risk that are based
on actual component condition.
Predictive risk estimates in terms
of core damage frequency (CDF) for
a generic multi-module AdvSMR case
study were obtained using the ERM
methodology being developed at PNNL.
TheCDF is a commonlyused riskmetric in
nuclear plant PRA analysis that measures
the frequency of accidents that can cause
core damage. Results to date include
predictive risk estimates due to different
failure probabilities for components, and
account for periodic maintenance actions
that are assumed to return components to
“as-new” condition. ECAs are assumed
to be performed periodically and used to
predict the Probability of Failure (POF)
over a predefined time horizon (in this
case, over the assumed licensed period of
40 years). When compared to the CDF
when static failure rates are assumed for
each of the components, the ability to
dynamically update risk estimates due to
changing component failure probabilities
provides an advantage with respect to
better characterization of risk as well as
targeting maintenance actions as needed
instead of on a preventive time-based
schedule.
Importance analysis is generally
performed on the results of a PRA and
provides a quantitative perspective on risk
and sensitivity of risk to changes in input
values, such as specific component failure
rates. However, classical importance
measures are difficult to extend to ERMs
and so new measures of importance
are being developed. A proposed new
importance measure that considers the
relative importance of the event to the
total CDF as well as the value of total
CDF itself appears to improve the ability
to distinguish between risk-importance of
components under time-varying failure
probabilities. Another aspect of ERM
that is being studied in ongoing research
is the inclusion of uncertainty within the
ERM framework, including uncertainty
regarding the specific condition of the
component, uncertainty in the POF, and
uncertainty in the time to failure. Finally,
risk metrics other than CDF – for instance,
economic metrics - are being examined.
Summary
Technologies such as PHM systems
are important for controlling O&M costs
by providing enhanced awareness of
component or equipment condition and
predictive estimates of component failure.
Such estimates may be customized for
each AdvSMR unit and account for
its specific operational history. Such
information, when integrated with plant
control systems and risk monitors, helps
control O&M costs by enabling lifetime
management of significant passive
components, relieving the cost and labor
burden of currently required periodic in-
service inspection, and informing O&M
decisions to target maintenance activities.
ERMs that integrate equipment
condition and prognostics information
to calculate time-dependent failure
probabilities have the potential to enable
real-time decisions about stress relief for
susceptible equipment while supporting
effective maintenance planning. As a
result, ERMs are expected to improve the
safety, availability, and affordability of
AdvSMRs.
Ongoing research includes:
explicitly addressing uncertainty
quantification in PHM and ERM
systems;
incorporating stressor information
into the prognostics methodology,
to address variable loading scenarios
and reduce uncertainty in the
predictive estimates of RUL;
developing methods to directly
incorporate ECA for key components
in the ERM;
identification of economic and other
risk measures, in addition to the
traditional safety measures; and
developing PRA models that account
for changing load and demand
conditions (possibly resulting in
changing success criteria).
Contact: Pradeep Ramuhalli, PhD.,
Pacific Northwest National Laboratory,
902 Battelle Blvd., MSIN K5-26, Richland,
Washington 99352; telephone: (509)
375-2763, fax: (509) 375-6497, email:
.
