Title: OnLine Monitoring for Instrument Calibration and Equipment Condition Assessment Regional Workshop on
1On-Line Monitoring for Instrument Calibration and
Equipment Condition AssessmentRegional
Workshop on Modernization Projects of NPP
Instrumentation and Control Systems Related to
Power Uprates and License Renewals
ProjectPortoroz, Slovenia14-18 April 2008
- Brandon Rasmussen
- brasmussen_at_kurztech.com
- Instrumentation Control Center
- operated by Kurz Technical Services, Inc.
- Harriman, TN 37748
2Applied On-Line MonitoringInstrument
Calibration Interval Extension
- The initial focus area for using empirical models
for On-Line Monitoring (OLM) in the power
industry was for calibration monitoring of
safety-related transmitters in the nuclear power
industry - Safety related instrumentation is recalibrated at
the end of every fuel cycle (18 months). - Safety parameters are measured by redundant
instrumentation, typically a group of 4
instruments. - Studies have shown that 90 of the
instrumentation is still within calibration after
18 months. - Using an on-line monitoring model to monitor
instrument calibration ensures that calibrations
will only be performed where necessary, resulting
in reduced radiation exposure, reduced outage
workloads, and cost savings.
3Instrument Calibration Interval Extension -
Illustration
200 Instruments
- It is required that at least 1 transmitter within
a redundant set be calibrated to prevent common
mode failure (Assume groups of 4)
50 Mandatory Calibrations
150 Instruments will be assessed
60 Condition Based Calibrations
- Conservatively estimate that 40 of the
instruments assessed will require calibration - Assessment is made through on-line analysis of
the sensor data
40
60
90 Instruments do not require calibration
- This conservative illustration indicates that
only 55 (110) of the safety related instruments
would require calibrations. - Significant benefits are available from the
reduced personnel radiation exposure and the
reduced outage workload (indirect costs). - Calibration costs are estimated between 1k and
3k per instrument (direct costs).
4Applied Instrument Calibration Monitoring
5Applied Instrument Calibration Monitoring
6Reasons Utilities Give for OLM of Instrument
Calibration
- Manual calibrations only validate the correct
operation of the instrumentation periodically
therefore, faulty sensors may remain undetected
for periods up to the calibration frequency. - Studies have shown that less than 5 of the 50 to
150 calibrated instrument channels were in a
degraded condition that required maintenance.
With the estimated cost of a typical manual
calibration ranging between 900 and 2000, the
cost of the potentially unnecessary work is
significant. - Performing maintenance on components that are
operating correctly provides an opportunity for a
fault to enter the system. - It also is expected to provide
- Monitoring under normal operating conditions.
- Enhanced detection of infant mortality.
- Enhanced maintenance planning capability.
- OLM may detect other operational abnormalities.
7Expected Outcomes of OLM for Instrument
Calibration
- The proposed methods will continuously monitor
the instrument channel condition and identify
those that have degraded to an extent that
warrants calibration. - The identified instrument channels will be
classified as either needing calibration at the
next outage or as entirely inoperable based on
the degree of degradation. - It is expected that the on-line methods will
reduce maintenance costs, reduce radiation
exposure, reduce the potential for
miscalibration, increase instrument reliability,
and may reduce equipment downtime.
8Expected Secondary Benefits
- OLM can also detect and identify
- Equipment degradation
- System changes
- Operational changes
- It is expected that these secondary benefits
could be the largest monetary driver for a
utility.
9Documentation of British Energys Implementation
- Plant Application of On-Line Monitoring for
Calibration Interval Extension of Safety Related
Instruments Volumes 1 2, EPRI, Palo Alto, CA,
and British Energy Group PLC, Suffolk, UK 2006.
1013486. - Updates were planned for 2007 and 2008.
10Implementation at British Energyhttp//mydocs.epr
i.com/docs/CorporateDocuments/EPRI_Journal/2007-Fa
ll/1016127_International.pdf
- On-Line Monitoring Improves Instrument
Calibration at British Energy - British Energys Sizewell B station became the
first to apply the EPRI guidelines to extend the
calibration interval of safety-related
instruments. - British Energy initially applied the OLM
techniques to about 200 instruments at Sizewell
B, focusing on the pressure, level, and flow
transmitters in the plants primary and secondary
protection systems. - Overall, 80 of the transmitters evaluated
during the first OLM cycle were found to be
within calibration tolerance throughout the fuel
cycle. - During the first outage, most of the
transmitters that were candidates for calibration
interval extension (70 of the total transmitters
evaluated) were, in fact, extended.
11Implementation at British Energyhttp//mydocs.epr
i.com/docs/CorporateDocuments/EPRI_Journal/2007-Fa
ll/1016127_International.pdf
- The additional 10 of transmitters that were
within tolerance were nonetheless scheduled for
calibration to maintain conservatism during the
initial implementation. - British Energy estimates that OLM, when fully
deployed, will routinely reduce outage duration
to 20 days from the 25 days normally required for
transmitter calibration, saving 1.5 million per
avoided outage day, or 7.5 million per operating
cycle. - Additional savings are expected from reductions
in labor costs, radiation exposure, and
calibration errors. - British Energys goal is to expand the OLM
application to nearly 2500 transmitters,
including many in the secondary system (steam
side) of the plant. - The project methodology and application for the
Sizewell project, together with a set of
supporting analyses and results, are published in
the EPRI report Plant Application of On-Line
Monitoring for Calibration Interval Extension of
Safety-Related Instruments Volumes 1 and 2,
(1013486), 2006.
12Instrument Calibration Monitoring Applications
- Electricite de France has practiced the use of
on-line monitoring to selectively calibrate
safety instrumentation for over a decade. - 2005 British Energy obtained regulatory
approval to implement OLM technology to extend
the calibration interval of their safety-related
transmitters at Sizewell B (Westinghouse PWR) - Software provided by Analysis and Measurements
Services Corporation, Knoxville, TN - EPRI / British Energy / AMS are collaborating to
document this implementation from 2006-2008 - The implementation followed a graded approach
over 3 fuel cycles
13New Research for NRC
- To prepare for expected plant specific license
amendments from utilities involved in the OLM
Implementation Project, in 2004 the NRC funded
the research to gain additional insights into the
potential challenges of OLM. - Objective provide guidance to the regulatory
review process of OLM. - Deliverables
- NUREG/CR-6895 Technical Review of On-Line
Monitoring Techniques for Performance Assessment - Volume 1 State of the Art (published 2006)
- Overview and literature survey.
- Volume 2 Theoretical Issues (final draft 2007)
- Technical methods and derivations
- Volume 3 Limiting Case Studies (final draft
2007) - Assumptions and demonstrations.
14U.S. NPP License Amendment Request
- VC Summer NPP (South Carolina, USA) was one of
the original test participants in the EPRI OLM
Implementation Users Group. - On March 9, 2005, VC Summer met with the NRC to
discuss the submission of a plant specific
license amendment for the application of OLM for
calibration interval extension. - On February 6, 2006, VC Summer submitted a 74
page License Amendment Request LAR 05-0667 - Subsequently, VC Summer withdrew the request.
15Anomaly Detection
- OLM systems for monitoring instrument
calibration, and on a larger scale equipment or
processes, attempt to identify anomalies in the
process data - Properly trained empirical process models can
detect early anomalies in the monitored
instrument channels. - An anomaly is defined as any change in an
individual instrument channel or a more
significant change in the process that is unusual
with respect to the expected operation of the
process. - The expected operation of the process is defined
by the training data set.
16Anomaly Detection
- Types of identifiable anomalies
- Gradual calibration drift in one or more channels
(Calibration Monitoring) - Abrupt changes in channels measurements, either a
single point or a series of consecutive points. - Abrupt changes in the channel variance.
- Changes between the channel inter-relationships
(may indicate a process drift or equipment
problem). - Common mode failure if redundant and
non-redundant instruments are included in the
same model.
17Extensions to Equipment Monitoring Anomaly
Interpretation
- The detection of anomalies in the data signifies
that a change has occurred in the learned normal
physical relationships of a system component. - The output of an On-Line Monitoring system is a
set of identified anomalies. In many cases, an
equipment degradation mechanism will manifest
itself through changes in the instrument channel
indications well before the eventual failure. - Linking a set of anomalies to a given equipment
degradation mechanism is referred to as anomaly
interpretation and is the basis for the extension
of empirical process modeling to equipment
condition assessment.
18Empirical Equipment Monitoring
- Monitoring equipment using empirical models has
the potential for the identification of early
warning indicators of component degradation or
failure. - This is the largest financial driver for
utilities to implement empirical OLM systems. - One of the largest cost benefits can be derived
from avoiding a lost-power event
19Summary of Lost Power Incidents at Selected U.S.
Nuclear Units 2000-2003
20Lost Power Incidents at Selected U.S. Nuclear
Units 2000-2003
1 Low Dollar Estimate 250,000/day Mean
Dollar Estimate 600,000/day High Dollar
Estimate 2,500,000/day
21Empirical Equipment Monitoring
22How are Assessments Made?
- Consider an Autoassociative models output
Assessments are made based on the behavior of the
residual values. In this case, the model
residuals indicate normal operation
23How are Assessments Made?
In this case, the model residuals indicate
abnormal operation. This information allows for
the most basic assessment that based on
historical operation, sensors A, B, C, D are
reporting data that is unexpectedly high.
24Anomaly Interpretation
- Assume that a failure mode or event (E) exists
whereby signals A, B, C, D increase prior to
the eventual failure. - Further, assume that these increases are
identifiable by an empirical model as anomalies,
and this mode is differentiable from competing
modes. - Logic Rule Interpretation
- If RA gt t AND RB gt t AND RC gt t AND RD gt t
- Then event E will occur.
- Probability Based Interpretation
- P(E RA gt t) 0.2, P(E RB gt t) 0.2,
- P(E RA gt t RB gt t) 0.6,
- P(E RA gt t RB gt t RC gt t ) 0.9,
- P(E RA gt t RB gt t RC gt t RD gt t ) 0.97
25Event Notification
26Simple Anomaly Interpretation Scheme
27Anomaly Pattern Example
28Making Assessments
- Assessments require Interpretation, without
interpretation the information is typically not
useable - In its most basic form it is the indication of an
anomaly - Assessment may requires some (or all) of the
following - System expertise to understand the information,
including knowledge of the empirical model and
knowledge of the monitored system - Software based tools to identify the cause of an
observed deviation - Information from related systems to support the
assessed condition - Physical investigation at the equipment site
additional diagnostic testing
29Typical Scheme of ECM
30Automated Assessments
- The ideal scenario is to devise an automated
anomaly interpretation scheme - Maps a set of observed anomalies to a specific
degradation or assessment - Requires extensive effort to link potential
available indicators to a specific fault
31Automated Anomaly Interpretation
32Requirements for Accurate Reliable Early Warning
- Successful OLM models will provide early warning
of impending failure or warning of sufficient
degradation. Several requirements must be met in
order for these capabilities to be available - Differentiability The various failure modes or
degradation mechanisms for a component or piece
of equipment must be differentiable from one
another so that the diagnosis is accurate and
unambiguous. - Anomalous Indications of the condition must be
identifiable by an OLM model. - Repeatability All occurrences of a given
degradation mechanism or failure must
consistently produce the same precursor
indications. - Timely Indications must occur and be properly
identified with enough lead time to provide some
advantage (e.g. event avoidance).
33Establishing a Failure Signature Database
- EVALUATE
- Failure Modes Degradation Modes
- consider probability of modes based on history or
manufacturers data - consider effect of failure with respect to
personnel safety, loss of production, maintenance
and related costs - DEFINE
- Potential indicators of these modes (with respect
to information contained in or produced by an OLM
system) - consider the requirements for accurate
identification (DART) - consider input from various modeling tools
deployed - construct interpretation logic to combine
resultant information from multiple monitoring
tools and other information streams to present a
concise diagnosis (in some cases, prognosis)
34Other Industries Utilizing Empirical Equipment
Process Models
- Aircraft
- Chemical processing
- Petroleum
- Heavy machinery
- Power Fossil, Nuclear
- Manufacturing
- Locomotive
- Wind Turbines
- Security Monitoring
- Pulp and Paper
- Pipelines
35On-Line Monitoring Benefits
- Information Filtering
- Performance Monitoring
- Calibration Monitoring Reduction
- Early Warning of equipment degradation
- Provide system engineers and maintenance staff
with necessary information to make informed,
cost-effective operations and maintenance
decisions based on the actual condition of the
system/equipment - Allow earlier mitigation or corrective actions
- Reduce the likelihood of unplanned plant trips or
power reductions - Reduce equipment damage
- Reduce likelihood of repetitive failures
- Knowledge Capture and retention through embedding
expert knowledge into automatic anomaly
interpretation - Cost-savings associated with calibration
reductions - Cost-savings associated with planned and
preventative maintenance rather than unplanned
and reactive