Edited book reporting recent results in AI research in power plant surveillance and diagnostics. High quality and applicability of the contributions through a thorough peer-reviewing process. Condition Monitoring and Early Fault Detection provide for better efficiency of energy systems, at lower costs. Inhalt Featured Topics: Analysis of important issues relating to specification, development and use of systems for computer-assisted plant surveillance and diagnosis.- Empirical and analytical methods for on-line calibration monitoring and data reconciliation.- Noise analysis methods for early fault detection, condition monitoring, leak detection and loose part monitoring.- Predictive maintenance and condition monitoring techniques.- Empirical and analytical methods for fault detection and recognition.
This publication reports on the work and findings of an IAEA coordinated research project. The technologies discussed in this project are intended to establish the state of the art in surveillance, diagnostic and prognostic (SDP) technologies for equipment and process health monitoring in nuclear facilities. The participants also identified technology gaps and research needs of the nuclear industry in the SDP area. The publication describes conventional SDP technologies as well as the latest tools, algorithms and techniques that have emerged over the past few years. These new tools have made it possible to identify problems earlier and with better resolution. The target audience of this publication is utility engineers, end users, researchers, managers and executives making decisions on implementation of the subject technologies in nuclear facilities, or determining the future direction of research and development in this area.
This report extends the application of on-line monitoring to equipment and process condition monitoring, encompassing an array of technologies including vibration monitoring, acoustic monitoring, loose parts monitoring, motor current signature analysis and noise diagnostics, as well as vibration analysis of the reactor core and the primary circuit. Furthermore, this report includes the application of modeling technologies for equipment and process condition monitoring. A majority of these technologies depend on existing data from existing sensors and first principles models to estimate equipment and process behavior using empirical and physical modeling techniques. In doing so, pattern recognition tools such as neural networks, fuzzy classification of data, multivariate state estimation and other means are used. These means are described in the report, and examples of their application and implementation are provided. The benefits of OLM for performance verification of process instruments were described in the first report and included such advantages as the ability to extend the calibration interval of pressure, level and flow transmitters, detection of blockages, voids and leaks in pressure sensing lines, detection of degradation of the dynamic response of process instruments, and the like. Examples of benefits of OLM for condition monitoring include: (1) the ability to determine the onset of failure of pumps, valves, motors and reactor vessel components; (2) residual life assessment of equipment; (3) equipment life extension and aging management; (4) the ability to establish objective schedules for preventive maintenance, equipment refurbishment or replacement; and (5) maintenance cost reduction, efficiency improvements, reduction of plant outages, and reduction of radiation exposure to plant personnel.--Publisher's description.
This book provides a hybrid approach to fault detection and diagnostics. It presents a detailed analysis related to practical applications of the fault detection and diagnostics framework, and highlights recent findings on power plant nonlinear model identification and fault diagnostics. The effectiveness of the methods presented is tested using data acquired from actual cogeneration and cooling plants (CCPs). The models presented were developed by applying Neuro-Fuzzy (NF) methods. The book offers a valuable resource for researchers and practicing engineers alike.
The papers presented on this occasion examined the most significant aspects of diagnostic strategies, emphasizing the importance of predictive maintenance in reducing production shortages and the costs of plant management. The contributions of these authors allow a critical comparison of the varied experiences in developing and applying the different diagnostic methodologies employed in several parts of the world. The following problems are discussed: characteristics of condition monitoring systems - data acquisition techniques and data processing methodologies; choice of transducers and of measurement point locations; data compression techniques; alarm levels evaluation (acceptance regions); strategies for detecting malfunction conditions; diagnostic methodologies for the on-line and off-line identification of the cause of fault; expert systems; definition of the guidelines for the presentation in control rooms of monitoring data and diagnostic results; rotordynamic models used, off-line, to confirm faults diagnosed on-line.
