In-core instrumentation and core assessment Proceedings of a Specialists' Meeting Mito-shi, Japan, 14-17 October, 1996 Foreword This Specialists' Meeting on In-core Instrumentation and Reactor Core Assessment is the fourth in a series initiated by the NEA Committee on Reactor Physics and continued by the NEA Nuclear Science Committee. The first meeting was held in Fredrikstad, Norway, in 1983, the second at Cadarache, France, in 1988 and the third in Pittsburgh, USA in 1991. The purpose of the meeting was to review the improvements in the methods used to gather and interpret information on the conditions of the reactor core. Thirty-nine papers were presented in six sessions covering radiation sensors, safety analysis, system and validation, other miscellaneous measurements, core monitoring and core performance. One hundred and nine participants from nineteen countries and International Organisations attended the meeting. Participation from non-OECD countries was made possible thanks to the collaboration of the Working Group on Nuclear Power Plant Control and Instrumentation (IWGNPPCI) of the International Atomic Energy Agency. The NEA wishes to thank the authors and session chairmen for their contribution to the success of the meeting. It also wishes to express its appreciation to the Japan Atomic Energy Research Institute (JAERI) for arranging and hosting the meeting. TABLE OF CONTENTS SESSION 1: CORE MONITORING � A Summary: Oldrich Erben, Chairman An On-line Adaptive Core Monitoring System J.A. Verspeek, J.C. Bruggink, J. Karuza (The Netherlands) A Benefit Assessment of Using In-core Neutron Detector Signals in Core Protection Calculator System (CPCS) S. Han, S-J. Park, P-H. Seong (Korea) Computer Based Core Monitoring System P. Swaminathan, P. Sreenivasan (India) Analytical Evaluation of the Uncertainty of On-line Axial Power Distribution Measurement with the Four-section Ex-core Detector J. Matsumoto, K. Seki, Y. Komano (Japan) SESSION 2: CORE MONITORING � B Summary: W.A. Boyd, Chairman TOPRE & HOTPOINT In-core Monitoring Systems for WWER-440 Nuclear Power Plants T. Polák, O. Erben (Czech Republic) The Extended On-line Core Monitoring Technology with the Latest VERONA-u Version F. Adorján, L. Bürger, I. Lux, J. Végh, Z. Kálya, I. Hamvas (Hungary) F. Adorján, L. Bürger, I. Lux, J. Végh, Z. Kálya, I. Hamvas (Hungary) Application of the Core Surveillance System SCORPIO at Sizewell B Ø. Berg, M. McEllin, M. Javadi (Norway and UK)) RINGHALS 2 Core Monitoring Experience T. Andersson, Ø. Berg, K. Romslo (Norway) Experience and Evaluation of Advanced On-line Core Monitoring System "BEACON" at IKATA Site N. Fujitsuka, H. Tanouchi, Y. Imamura, D. Mizobuchi, T. Kanagawa, M. Masuda (Japan) The BEACON On-line Core Monitoring System: Functional Upgrades and Applications W.A. Boyd, R.W. Miller (USA) SESSION 3: CORE PERFORMANCE ASSESSMENT Summary: Øivind Berg, Chairman Reactor Internals Vibration Monitoring in Korean Nuclear Power Plant T-R. Kim, S-H. Jung, J-H. Park, S. Choi (Korea) Utilisation of Self-powered Neutron Detectors for Reactivity Control V. Borissenko (Ukraine) On-line Estimation of Local and Total Core Flow Rate by Neutron Noise Analysis in BWR M. Mori, S. Kanemoto, M. Enomoto, S. Ebata (Japan) Space-dependent Dynamics of PWR T. Suzudo, E. Türkcan, J. Verhoef (Japan and The Netherlands) SESSION 4: INSTRUMENTATION Summary: Jean-Pierre Trapp, Chairman Direct Experimental Tests and Comparison between Sub-miniature Fission Chambers and SPND for Fixed In-core Instrumentation of LWR G. Bignan, J.C. Guyard, C. Blandin, H. Petitcolas (France) Characteristics of Self-powered Neutron Detectors Used in Power Reactors W.H. Todt (USA) High Temperature Fission Chambers: State-of-the-art J-P. Trapp, S. Haan, L. Martin, J-L. Perrin, M. Tixier (France) Application of the Gamma Thermometer as BWR Fixed In-core Calibration System R. Raghavan, C.L. Martin, A.L. Wirth, T. Itoh, Y. Goto, R. Arai (USA and Japan) Experience with Fixed In-core Detectors at Seabrook Station J.P. Gorski, R.J. Cacciapouti (USA) NAJA: A New Non-destructive Automatic On-line Device for Fuel Assembly Characterisation and Core Loading Conformity Control G. Bignan, D. Janvier (France) SESSION 5: ADVANCED INSTRUMENTATION TECHNOLOGIES Summary: Joachim Runkel, Chairman Some New Optical Techniques for Reactor Instrumentation M. Nakazawa (Japan) Development of a Distributed Monitoring System for Temperature and Coolant Leakage F. Jensen, E. Takada, M. Nakazawa, H. Takahashi, T. Iguchi, T. Kakuta, S. Yamamoto (Japan and Sweden) In-core Measurements of Reactor Internals Vibrations by Use of Accelerometers and Neutron Detectors J. Runkel, E. Laggiard, D. Stegemann, P. Heidemann, R. Blaser, F. Schmid, H. Reinmann (Germany and Switzerland) Overview and Future Development of the Neutron Sensor Signal Self-validation (SSV) Project J-C. Trama, A. Bourgerette, E. Barat, B. Lescop (France) Fuel Rod Performance Measurements and Re-instrumentation Capabilities at the Halden Project O. Aarrestad, H. Thoresen (Norway) SESSION 6: SAFETY ANALYSIS SESSION 6: SAFETY ANALYSIS Summary: Masaharu Kitamura, Chairman Application of Process-monitoring Techniques to Neutron Noise Signals from Simulated-coolant-boiling Experiements H. Schoonewelle, T.H.J.J. van der Hagen, J.E. Hoogenboom (The Netherlands) Stochastic Pattern Recognition Techniques and Artificial Intelligence for Nuclear Power Plant Surveillance and Anomaly Detection L.G. Kemeny Theoretical Modelling of Fuel Assembly Vibrations for VVER-type Reactors V. Kinelev, S. Perov, V. Sulimov (Russia) Impact of Core Inertial Properties on Dynamic Characteristics of WWER-1000 Reactor Barrels V. Gribkov (Russia) SESSION 7: SIGNAL PROCESSING Summary: Erdinc Türkcan, Chairman Monitoring the State of the Coolant in a Boiling Water Reactor G. Roston, R. Kozma, M. Kitamura (Japan and Argentina) Decay Ratio Studies in BWR and PWR using Wavelet Ö. Ciftcioglu, E. Türkcan (The Netherlands and Turkey) ALPES, a Demonstrator for On-line Core Temperature Visualising and Processing A. Lebrun, J-P. Trapp, S. Sala (France) New Neutron-temperature Noise Methods and their Experimental Check on the Reactor VVER-1000 V.I. Pavelko, D.F. Gutsev (Russia) Reactor Thermal/Hydraulic Processes Monitoring and Aid to Diagnosis, Using Acoustical Signal and On-line Calculations K.N. Proskouriakov (Russia) SESSION 8: MISCELLANEOUS Summary: Shigeru Kanemoto, Chairman Development of a Real-time Plant Simulation System for BWRs K. Tominaga, S. Arita, Y. Ishii, H. Sano (Japan) Investigation of the Pellet-cladding Interaction Related Issues Including Fuel Rod Failure by Methods for Identification System with Distributed Parameters S. Utenkov (Russia) A Technical System to Improve the Operational Monitoring of the Ukranian Nuclear Power Plant Zaporozh�ye (Unit 5) M. Beyer, H. Carl, P. Schumann, A. Seidel, F-P. Weiß, J. Zschau, K. Nowak (Germany) ANNEX 1. List of participants AN ON-LINE ADAPTIVE CORE MONITORING SYSTEM J.A. Verspeek, J.C. Bruggink, J. Karuza N.V.G.K.N. Nuclear Power Plant P.O. Box 40, 6669 ZG Dodewaard, The Netherlands Abstract An on-line core monitoring system has been in operation for three years in the Dodewaard Nuclear Power Plant. The core monitor uses the on-line measured reactor data as an input for a power distribution calculation. The measurements are frequently performed, the power distribution is recalculated every two hours and whenever a significant change in certain measured parameters occurs. The system is used for monitoring as well as for predicting purposes. The limiting thermal hydraulic parameters are monitored as well as the pellet-clad interaction limits. The data are added to a history file used for cycle burn-up calculations and trending of parameters. The reactor states are presented through a convenient graphical user interface. At the Dodewaard NPP the assessment of the power distribution is entirely based on the neutron Traversing In-core Probe (TIP) measurements and calculations. The core monitor calculation of the power distribution is calibrated with the TIP traces measured at least once a week. This is done by adapting the calculated TIP traces to the measured TIP traces in an iterative process. Corrections are added to the nodal k(cid:181) values in such a way that at the end of the process the calculated TIP traces match exactly the measurements. These corrections of k(cid:181) have an impact on the power distribution which is slightly changed in accordance with the TIP-measurements. The D k(cid:181) ’s are stored and used for calculation of the power distribution until the next calibration with the measurements is performed. When no D k(cid:181) ’s are used, the RMS-value of the difference between measured and calculated nodal TIP traces is typically 4%. Using the old D k(cid:181) ’s in forecasting the future TIP traces gives a RMS of 2%. Introduction The Dodewaard NPP has a BWR (183 MWth) which is cooled by natural circulation. This first Dutch nuclear power plant was started up in 1968. At the moment a plant upgrade project is in progress. As a part of this project a core monitoring system is being installed for continuous observation of operation limiting parameters. This system has been developed since 1992 and will be implemented in the control room in the future. Description of the core monitoring system A core monitoring system has been developed by the Dodewaard Physics Group. The core monitor calculation of the power distribution is based on the nodal code LWRSIM [1] (a code developed for this purpose) which uses a one group kernel method. The CASMO code [2] provides the k(cid:181) , M2 and detector response functions for LWRSIM. CASMO is a 2D assembly burn-up code making use of the transport theory. The core monitor can run in several modes depending on the purpose necessary. The system is continuously running in the MONITOR mode to watch the thermal hydraulic parameters and create a history file with all the interesting measured and calculated quantities. In this mode no human interaction is necessary. The measured reactor data are collected by a data logging computer and then retrieved by a separate workstation for use in the core monitoring system. The 3D power distribution inside the core is calculated on-line with the measured reactor data as input. From the power distribution thermal hydraulic parameters like Maximum Linear Heat Generation rate (MLHGR) or Minimum Critical Power Ratio (MCPR) are derived and monitored. The parameters related to pellet clad interaction limits are monitored as well. Because there are no local power range monitors in the Dodewaard reactor there is no possibility of a continuous measurement of the power distribution. The neutron TIP recordings provide the only way to monitor the power distribution and the thermal hydraulic quantities. Therefore the TIP measurements are done on a weekly basis. The TIP traces are recorded on paper and electronically on a hard disk. A calibration of the calculated 3D power distribution by the core monitor is started automatically as soon as the TIP measurements are finished. The physicist only has to check that the measurements and calibration have been done correctly and that all the parameters have remained within operating limits. All the other modes are started manually. They do not interfere with the on-line MONITOR mode, although the reactor states saved by the MONITOR mode, can be read in and used. The REPORTER mode is used to view and print the reactor core states. The CALIBRA mode is used to manually start a calibration of the core monitor with the measured TIP curves. Predictive capabilities of the system are used in the WIZARD mode to help the physicist or operator in finding the best operating strategy and keeping the reactor state within Technical Specification limits. The WIZARD mode includes search options for control rod pattern, thermal power and k . This mode is used to make eff predictions starting with the actual reactor state as calculated by the core monitoring system in the MONITOR mode. A back-end graphical user interface is developed for convenient data presentation of reactor states, trending of parameters and forecasting calculations. The user interface is running on all PCs which are connected to the core monitoring system through a local computer network. The user interface gives full access to all the parameters on a nodal level (36 axial nodes for 164 assemblies). Elementary condensing operations like averaging over all axial nodes, or taking the maximum nodal value of a specific assembly can be performed through a control panel on screen. Parameters can be plotted versus time after having read the saved reactor states from the past or predicted future reactor states. Description of the adaptive method An adaptive method based on an idea by Congdon et al. [3] is used to match the calculated TIP traces to the measured TIP traces. In an iterative process corrections are added to the nodal k(cid:181) values in such a way that at the end of the process the calculated TIP traces exactly match the measurements. Figure 1. Geometry and TIP positions of the Dodewaard core TIP measurements are performed at a limited number of positions in the core (see Figure 1). At places where no measurement is done a TIP trace is constructed by using the quadrant symmetry of the core and by inter- and extrapolating the measured TIP traces. The TIP readings are translated to a flux F at the position of the assemblies using m bilinear interpolation from the four nearest TIP traces. The k(cid:181) values are calculated in the first quadrant of the core where most of the TIP measurements are done. For this purpose PITsevruc 3A and 6D are copied to their imaginary mirror positions 4A and 6C in the first quadrant (not indicated). TIP curve 2E in the third quadrant is not used for the adaptive process but to check the symmetry of the core power distribution. The same D k(cid:181) values used in the first quadrant, are applied to the mirror positions in the other three quadrants. The diffusion equation applies to both the flux derived from the measured TIP curves F and the flux derived from the simulated TIP curves F , where in the equation for F an m s m extra correction term D k(cid:181) is added to the k(cid:181) value: )1( k¥ ( -1) (cid:209) 2 F s+B2sF s = 0, B2s = keMff2 k¥ +D k¥ ( -1) k (cid:209) 2F m+B2mF m = 0, B2m = Meff2 Here B2 stands for buckling and M2 stands for migration area. After subtracting these equations an expression is obtained for the difference in flux Y = D F = F - F : s m D )2( (cid:209) 2 Y +D B2Y = 0, D B2 = B2 -B2 = - k¥ s m k M2 eff Solving this expression for D k(cid:181) yields: )3( 1 D k¥ = -keffM2D B2 = keffM2Y (cid:209) 2Y The term (cid:209) 2 Y / Y in the last expression is estimated with a finite difference approximation for the second derivative. Finally the estimation of the correction term D k(cid:181) si used in an iteration loop. When the iterative process is converged the simulated TIP curves are equal to the measured TIP curves. The 3D power distribution and derived quantities are slightly changed by this process and are now in accordance with the measured TIP values. The D k(cid:181) ’s are stored and used for future on-line power distribution calculations until the next calibration is performed. The adaptive process is a mathematical way to substitute the differences in TIP values with differences in k(cid:181) values. It is done to make the calculation of the power distribution in complete agreement with the measurements. The price that has to be paid for this is a set of k(cid:181) correction values the most of which are small. Since the adaptive process works with almost any set of TIP curves, the Root Mean Square (RMS) value of the TIP differences before adaptation or the RMS value of the k(cid:181) differences after adaptation should be considered to assure that no obvious mistake in either the measurement or the calculation has been made. Results In an earlier study the correlation between the nodal D k(cid:181) and other nodal parameters was examined [4]. No correlation was found for any of the examined parameters (relative power, control rod fraction, void fraction, burn-up, flow) except axial and radial positions. These last two dependencies will be discussed below. Figure 2 shows the radial distribution of the average absolute D k(cid:181) values from the calibration with a typical set of TIP curves in cycle 27. The value of the D k of an (cid:181) assembly is indicated by a degree of darkness. As shown, in this case the largest corrections are needed in the assemblies around the TIP tube 6B. This indicates that the differences between predicted and measured TIP values are the largest for TIP curve 6B. A possible explanation is that one or more of the assemblies next to this TIP tube has a bowed channel. Channel bowing has a large impact on the detector response function and results in deviations between measurement and calculation of TIP curves. Because the amount and directing of channel bowing is not known beforehand it can not be modelled. Figure 2. Radial distribution of average absolute DD k(cid:181)(cid:181) from TIP measurements on July 22, 1996 It turns out that the radial distribution of D k(cid:181) changes gradually during a cycle and stepwise from one cycle to the other. This strokes with the fact that assembly properties like burn-up or channel bowing vary gradually during a cycle. In a next cycle the properties of an assembly at a specific location in the core can change stepwise when the original assembly was replaced by another one. In order to examine a possible correlation of D k(cid:181) with channel bowing, at the end of cycle 24 the four channels around the TIP tube where the largest corrections were needed were measured for bowing. The bowing of these channels was not as large as to have to be rejected, whereas by other channels this was the case. Because the measurement of the channel bowing is quite inaccurate no correlation was shown between channel bowing and the radial distribution of D k(cid:181) as well. Figure 3 shows the axial profile of D k from a calibration with a typical TIP (cid:181) measurement in this cycle. Figure 3. Axial distribution of DD k(cid:181)(cid:181) from TIP measurements on July 22, 1996 The core average D k(cid:181) is shown, as well as the axial distribution for the assemblies H313 and H129 which have the highest and the lowest average D k(cid:181) respectively. The local minimums at node 10, 19 and 29 can be explained by the fact that there are extra materials at these positions that have an influence on the detector response function. This influence is difficult to model and give rise to systematic errors in the simulated TIP traces. There are spacers at positions 9.8, 19.5 and 29.1 in the core. Furthermore at positions 9.7, and 21.0 there are placeholders between the inner and outer TIP tubes and at node 32 and higher there is a spring and other material located. The high value of D k(cid:181) at node 1 is caused by an incorrect axial bottom albedo coefficient that is used in the 3D nodal power distribution calculation code LWRSIM. This coefficient can be changed in such a way to minimise the difference in D k(cid:181) for node 1. The axial and radial deviations discussed here can also be seen in a direct comparison of the measured and calculated TIP values. The information obtained in previous calibrations and stored in the D k(cid:181) values, can be used to improve the prediction of TIP curves. A part of the errors is space-dependent but time-independent. This holds true for the errors caused by the influence of extra materials in the core on the detector response function. These errors can be filtered out by storing the D k(cid:181) values obtained during a calibration and reusing them in future calculations. To show this in Figure 4 the RMS values for the difference between the measured and calculated TIP values is plotted versus cycle burn-up in cycle 26. Figure 4. RMS of DD TIP values in cycle 26 for different prediction modes Three different prediction modes are shown for the calculated TIP values. The first is a straightforward calculation without doing any calibrations. This yields a RMS value of about 4 %. The second one uses the D k(cid:181) values from the previous calibration normally one week old. This gives a RMS of about 2%. The last one uses the D k(cid:181) values of a calibration in the beginning of the cycle which is representative for the rest of the cycle. The first few calibrations are not representative because they are done at startup conditions with low power or no control rods inserted. At the beginning the prediction is as good as in the case that the k-correction from the previous point is used, but gradually the information stored in the correction factors D k(cid:181) for the time dependent part of the errors is lost and the RMS value increases. A gradual increase in the RMS value from 2% to 3% towards the end of the cycle can be noted. Nevertheless, at the end of the cycle a prediction with the corrections from the beginning of the cycle is better than the prediction without using D k(cid:181) values. Objectives for the future At the moment the core monitor is in a verification phase. During this cycle it has been operating successfully. It has been used by the Physics Group as a working tool and for reporting to the Dutch licensing authorities. In the near future it will also be available to the operators in the control room. For this purpose some additional developments have to be accomplished: