1、Page 1 of 13Implementation of On-Line Monitoring forTransmitter Calibration Period Extensionat Sizewell B David Lillis/Stephen OrmeBritish EnergyAbstractSizewell B is the United Kingdoms only Pressurised Water Reactor and although based on a generic Westinghouse 4 loop plant is significantly more in
2、strumented and has considerably less maintenance employees than its counterparts. This has been a key factor in the drive to the consideration of techniques, which either automate or remove unnecessary labour intensive control and instrumentation calibrations. These automated activities are typicall
3、y termed, condition based or On-Line Monitoring (OLM) calibration techniques.The development of OLM based maintenance regimes invariably requires the acquisition, screening and processing of large amounts of data over time necessitating the use of software based analytical packages. The simple act o
4、f introducing computer-based applications introduces unique challenges to the user in establishing that such applications perform as expected. Such challenges can be significantly exasperated when used to support maintenance regimes involving safety critical equipment and subject to close scrutiny b
5、y the Nuclear Regulator.This paper provides a general overview of the role of On-Line Monitoring (OLM) in reducing and/or eliminating several time consuming and often labour intensive calibration activities of safety category sensors associated with the reactor protection systems and post fault moni
6、tors at Sizewell B.In particular this resume looks at the most recent implementation of OLM at Sizewell B in support of a change in calibration frequency of once per fuel cycle to once every 4th fuel cycle, for pressure sensors associated with the reactor protection system.The increase in calibratio
7、n period was approved by the regulator in March 2005 and a phased implementation commenced during the recent refuelling outage (April 2005).1. INTRODUCTIONSizewell B is the United Kingdoms only Pressurised Water Reactor, it is owned and operated by British Energy and is based on the generic SNUPPS (
8、Standardised Nuclear Power Plant System) design used for Wolf Creek (Kansas) and Calloway (Missouri). Although based on a generic Westinghouse 4 loop plant, Sizewell B is significantly more instrumented and has considerably less maintenance employees than its US counterparts. This has been a key fac
9、tor in the drive to the consideration of techniques, which either automate or remove unnecessary labour intensive control and instrumentation activities. These activities are collectively termed condition based calibration/maintenance techniques.The development of condition based maintenance regimes
10、 invariably requires the acquisition, screening and processing of large amounts of data over time necessitating the use of software based analytical packages. The simple act of introducing computer-based applications introduces unique challenges to the user in establishing that such applications per
11、form as expected. This challenge can be significantly exasperated when used to support maintenance regimes involving safety critical equipment.The testing and validation of software based systems is attracting increasing scrutiny from the UK regulator and represents a significant burden for the impl
12、ementation of any new systems and may be so significant as to make the introduction prohibitively expensive, even though the benefits are obvious.Page 2 of 13In March of this year (2005) British Energy received regulator permission for the use of on-line monitoring (OLM) to facilitate a phased chang
13、e in calibration frequency from 2 years to 8 years for pressure and differential pressure transmitters associated with the Reactor Protection System.This paper provides a brief overview of how OLM has been used at Sizewell B in the past and in particular for the support of transmitter calibration pe
14、riod extension, the implementation thereof and the regulatory challenges presented and subsequently resolved.For the purpose of this paper the techniques described are directed toward safety category instrumentation associated with the Reactor Protection System (RPS) and Post Fault Monitoring (PFM),
15、 although the principles in general are equally applicable to non-safety equipment.2. REVIEW OF OLM TECHNIQUES IN USE AT SIZEWELL BThe industrys desire to minimise maintenance costs combined with the advances in computer based analysis applications have been the main driving forces for a move from t
16、ime based intrusive maintenance to that of informed conditioned based maintenance. The underlying principle being adopted that if equipment can be shown to be operating within its required acceptance limits by a monitoring regime, then do not subject it to unnecessary intrusive maintenance. This all
17、ows the maintenance workforce to be more efficiently employed in targeting that equipment known to require attention as opposed to current practice which calls for calibration verification of almost all sensors every fuel cycle.The following section details those condition-based processes already in
18、 use at Sizewell B.2.1 Pressure Sensor Response Time Testing using Noise Analysis Although not approved in the UK at the time of Sizewells start-up, a methodology had been developed for measuring sensor response time by examining its dynamic performance whilst in service, using noise analysis techni
19、ques. Because of the potential manpower savings all sensors were subjected to this new methodology during initial hot functional testing and when the reactor was taken to 100% power for the first time. These results were compared with the traditional hydraulic ramp technique results already performe
20、d as part of the standard start-up tests and shown to exhibit the same or a more conservative response time thereby validating the technique and establishing its pedigree for subsequent use.The noise analysis technique is based on monitoring the natural fluctuations (noise) that exist at the output
21、of pressure sensors while the plant is operating. These fluctuations are due to turbulence induced by the flow of water in the system, spatial variation heat transfer in the core, and other naturally occurring phenomena.The noise is extracted from the sensor output and provides a “fingerprint“ or “s
22、ignature“ for each sensor. The frequency spectrum of this fingerprint is analysed to produce a bode plot from which the time constant of the sensor can be calculated from the -3dB break point (1). = 1/(2f) (where f = -3dB frequency) (1)The test is performed where the sensor field wires reach their s
23、ignal conversion and signal conditioning equipment, i.e. at the RPS cubicles. The test is performed “at power“ and does not require any access to the sensors or any wires to be disconnected and can be performed on several sensors at once. As the time constant obtained from the noise analysis techniq
24、ue includes impulse tubing it will always yield a conservative result, which respect to the hydraulic ramp method. Page 3 of 13Noise analysis has revealed several degradations in sensor response time, including partially blocked lines, stuck or broken valves, diaphragm fluid loss and even a sensor w
25、hos response was too fast. It should be noted that it was only the last fault (which turned out to be a missing electronic filtering component) that could have been detected via the original hydraulic ramp method.The main disadvantage of the response time measurement via noise analysis is that it re
26、quires the presence of process noise to make it viable, and hence cannot currently be used on processes such as tank levels and reactor building pressures. However Sizewell B is currently looking at a proposal to use “pink” noise injection (see note below) to facilitate the extension of this techniq
27、ue to these services. (Note, in this context “pink” noise is used to describe noise which is restricted to a limited spectrum)This use of noise analysis to monitor the time response of sensors represents the first use of on-line maintenance/calibration techniques for RPS and PFM instrumentation at S
28、izewell. Interestingly several US plants have discontinued time response testing as they had seen no evidence of degradation, although of course these plants were only using the ramp method and would not have seen the failure mechanisms highlighted at Sizewell2.2 Temperature Sensors Response Testing
29、 using Noise AnalysisFollowing on from the success of the pressure sensor time response testing via noise analysis described in section 2.1, Sizewell B has also validated the technique for primary circuit coolant Resistance Temperature Devices (RTDs).The noise analysis process is almost identical to
30、 that employed for pressure sensors and again it was validated against a well established standard, in this instance that of Loop Current Step Response (LSCR).Although the LSCR testing must be performed whilst the plant is at power and could be construed as on-line monitoring, it is necessary to rem
31、ove the sensor from service for the performance of the test and hence is labour intensive and renders the equipment inoperable for the duration. The noise analysis technique is faster, non-intrusive and can be performed on several sensors simultaneously.2.3 Dynamic Temperature Sensor Cross Calibrati
32、onBecause of the critical path threat to the refuelling outage by the traditional iso-thermal method of calibrating temperature sensors, Sizewell B pursued and were successful in receiving approval from the regulator for an alternative automated dynamic technique of cross calibration for Refuelling
33、Outage 6 (RFO6) in October 2003.This involved gathering of temperature sensor data (RTD Early detection and resolution of problems Reduction in actual critical path and removal of risk to the critical path should a calibration problem be found.Although the improvements are apparent, as stated earlie
34、r such software based analysis packages are closely scrutinised by the regulator and significant resource was expended in meeting their concerns for example: Demonstrating that the software performed as expected, especially at the boundaries of its pass/fail determination. Demonstrating that the sof
35、tware correctly handled bad data inputs. Demonstrating that the software application was not compromised by the operating system. Demonstrating the reliability of the software.During the first use of the technique in the RFO6 cooldown, 3 problem RTDs were identified, 2 related to channel input condi
36、tioning printed circuit board and 1 recalibration requirement. All issues were resolved prior to heat-up resulting in the removal of the planned 8-hour isothermal plateau from the critical path and saving a 36-hour outage extension for re-calibration.During the most recent outage, RFO7, in March 200
37、5 the technique revealed operator induced loop current inaccuracies during equipment maintenance activities.“Built in” advanced correction features in the software which compensate for hot to cold leg and/or loop to loop temperature differences allow the software technique to be used at power operat
38、ion.3. TRANSMITTER CALIBRATION PERIOD EXTENSIONThe following section provides a review of the most recent use of OLM at Sizewell B in support of the extension to the calibration period of pressure and differential pressure transmitters associated with the Reactor Protection System and Post Fault Mon
39、itoring.3.1 Calibration RequirementAll RPS i.e. the sensor is not subjected to its normal running pressure and temperature. Many sensors are found to require no adjustment. Sensors are exposed to the risk of calibration errors and/or incorrect return to service.3.2.3 Increased Risk of Operator Error
40、 The mere act of removing a sensor from service and performing intrusive calibration checks exposes the sensor to the risk of operator error.If all sensors on the same parameter are calibrated at the same time such errors can become common mode (i.e. rogue operator or test equipment) which would be
41、difficult to detect on return to service.There is also evidence that subjecting maintenance staff to the pressures of adherence to critical path time duration or radiological exposure results in calibration errors, perhaps not outside acceptable limits, but sufficient to be observable.3.3 Motivation
42、 for Extending the Calibration Period for RPS SensorsAt Sizewell B there are approximately 330 pressure (including differential pressure) sensors which are subject to Tech Spec calibrations of which about 50% reside within the Reactor Building and are only accessible during a refuelling outage. Many
43、 of the remainder are within a radiological controlled area. A significant amount of time is spent each fuel cycle (approximately 5000 man-hours) checking the calibration of these sensors and it has been found that in most cases no significant drift has occurred and no adjustment is required. Hence
44、it can been argued that the requirement for calibrating the sensors at refuelling frequency is weak and it incurs increased risk to the sensors (in terms of intrusive maintenance) and involves operators taking up unnecessary dose.Several instances have been observed where sensors that were performin
45、g correctly immediately prior to an outage have been returned to service incorrectly e.g. bad valve alignment, or that the new calibration merely introduced a different bias caused by the operator and/or the calibration equipment used, i.e. the sensor is no better or worse than before it was re-cali
46、brated.Whilst the above highlights the safety advantages of extending the calibration period there is also a significant commercial issue for Sizewell B with respect to the management goal of reducing refuelling outage time to around 20 days. This goal is directly challenged by the current transmitt
47、er calibrations, which take approximately 25 days to complete.This last point is the main reason for Sizewell Bs OLM efforts being directed toward the safety category (tech spec related) transmitters as there is an easier cost benefit justification compared to its counterparts. Who have far fewer Pa
48、ge 6 of 13transmitters and are unlikely to be able to demonstrate significant savings to warrant implementation. The “down side” in pursuing the safety category route however is the effort required in obtaining regulator acceptance.3.4 History of Calibration Period Extension using OLM at SizewellThe
49、 interest in the use of On Line Monitoring (OLM) techniques for sensor calibration testing in nuclear power plants peaked in the late 1980s in the USA following nearly a decade of research and development efforts in the area of validation of the techniques for analysing the signal data.Since 1987, interest continued with several utilities testing a number of on-line monitoring systems for sensor calibration verification in nuclear power plants.This OLM work cumulated in September 2000 when EPRI issued a topical report 1, which pro
