Technical Review of On-Line Monitoring Techniques for Performance Assessment: Theoretical Issues (NUREG/CR-6895, ORNL/TM-2007/188, Volume 2)
On this page:
Download complete document
This page includes links to files in non-HTML format. See Plugins, Viewers, and Other Tools for more information.
Manuscript Completed: October 2007
Date Published: May 2008
J.W. Hines, D. Garvey, R. Seibert, and A. Usynin
Department of Nuclear Engineering
The University of Tennessee-Knoxville
Knoxville, TN 37966-2210
Oak Ridge National Laboratory
Managed by UT-Battelle LLC
Oak Ridge, TN 37831-6156
S.A. Arndt, NRC Project Manager
Division of Engineering Technology
Office of Nuclear Regulatory Research
U.S. Nuclear Regulatory Commission
Washington, DC 20555-0001
NRC Job Code N6080
In 1995 the U.S. Nuclear Regulatory Commission (NRC) published a summary of the state-of-the-art for the area of on-line monitoring prepared by the Analysis and Measurement Services Corporation as NUREG/CR-6343, On-Line Testing of Calibration of Process Instrumentation Channels in Nuclear Power Plants. The conclusion of this report was that it is possible to monitor calibration drift of field sensor and associated signal electronics and determine performance of the instrument channels in a nonintrusive way.
In 1998, the Electric Power Research Institute (EPRI) submitted Topical Report (TR) 104965, On-Line Monitoring of Instrument Channel Performance for NRC review and approval. This report demonstrated a nonintrusive method for monitoring the performance of instrument channels and extending calibration intervals required by technical specifications (TS). A safety evaluation report (SER) was issued in 2000 in which NRC staff concluded that the generic concept of on-line monitoring (OLM) for tracking instrument performance as discussed in the topical report is acceptable. However, they also listed 14 requirements that must be addressed by plant-specific license amendments if the TS-required calibration frequency of safety-related instrumentation is to be relaxed. The SER did not review or endorse either of the two methods addressed in the topical report.
This report, published in three volumes, provides an overview of current technologies being applied in the United States for sensor calibration monitoring. Volume 1—State-of-the-Art, provides a general overview of current sensor calibration monitoring technologies and their uncertainty analysis, a review of the supporting information necessary for assessing these techniques, and a cross reference between the literature and the requirements listed in the SER. This volume entitled Volume 2—Theoretical Issues provides an independent evaluation of the application of the most commonly employed OLM methods. In empirical, model-based OLM, current measurements are applied to an algorithm that uses historical plant data to predict the plant's current operating parameter values. The deviation between the algorithm's predicted parameter values and the measured plant parameters is used to detect any instrument faults, including instrument drift. Many algorithms can be used to accomplish OLM; however, only auto-associative neural networks (AANN), auto-associative kernel regression (AAKR), and auto-associative multivariate state estimation technique (AAMSET) are investigated in this study. These techniques were chosen because they were either considered by EPRI's OLM working group that started in the 1990s, applied in EPRI's OLM Implementation Project which began in 2001, or are currently available as commercial products. Volume 3—Limiting Case Studies explores the inherent assumptions in the model whose impact on OLM is not yet known and also investigates other special limiting cases where the model behavior is unknown. The case studies reported in Volume 3 apply the modeling and uncertainty analysis techniques to a wide variety of plant data sets to consider the effects of these modeling assumptions and limitations.
The inclusion of a particular method should not be construed as an NRC endorsement for that technique. The theory behind each of these modeling techniques is explained in detail. The uncertainty of the model's prediction is also explained, as well as several methods for quantifying it. This expanded discussion on uncertainty is presented because OLM model uncertainty is one of the most critical issues surrounding the general acceptance of OLM as a valid technique for sensor performance assessment. If the uncertainty of the model is unknown, there can be no confidence in the model's predictions, making OLM for calibration extension pointless. Therefore, both analytical and Monte Carlo based uncertainty equations are developed for the three modeling techniques. Finally, the AANN, AAKR, and AAMSET models, along with their corresponding uncertainty values, are compared using both simulated and actual nuclear data. Although this comparison is theoretically thorough, it may not completely address the practical effects of the modeling assumptions and limitations. Therefore, Volume 3 of this series presents the results of applying the models and uncertainty analyses to a wide variety of plant data including those that violate some of the modeling assumptions; it recommends how these limiting test cases can be identified, quantifies the effects of not meeting certain assumptions, and presents precautions to take to assure the assumptions are met.