Brian K. Hajek

A hybrid approach for detecting and isolating faults in nuclear power plant interacting systems

A fault detection and isolation (FDI) system is presented that can detect and isolate nuclear power plant (NPP) faults occurring in interacting systems. The proposed methodology combines two tools, observer-based residual generation and neural network pattern matching, into a powerful, hybrid diagnostic system. A computer-based model of a commercial boiling water reactor (BWR) is used as the reference plant. Two FDI methods are implemented on each of two BWR systems, and their performance characteristics are compared. One method uses conventional neural network techniques that use parameter values for input, and a second, hybrid methodology uses system models to create residuals for input to a neural network. Both FDI systems show good generalization abilities, but only the hybrid system decouples system interactions. Although implementation is impractical for all NPP systems, this hybrid technique is most useful in specific applications where operators have difficulty diagnosing faults in strongly interacting systems.

An Integrated Operator Advisor System For Plant Monitoring, Procedure Management, and Diagnosis

This paper reports on an integrated operator advisor system (OAS) built using generic task methodology. The operators activities of plant monitoring, data interpretation, procedure execution, and diagnosis have been implemented as the four generic tasks in the system. The OAS is capable of identifying the abnormal functioning of the plant in terms of threats to safety, preenumerated abnormal events, and deviations from normality. After the identification of abnormal functioning, the system will identify the procedures to be executed to mitigate the consequences of abnormal functioning and will help the operator by displaying the procedure steps and monitoring the success of actions taken. The system also is capable of diagnosing the cause of abnormal functioning. The diagnosis is done in parallel to the task to procedure execution.

DPRL: a language for representation of operation and safety maintenance procedures of nuclear power plants

1.0 Introduction The task of operation and safety maintenance performed by the reactor operator in a nuclear power plant is a complex knowledge based task and has to be done within given time constraints. The two essential tasks in the overall task are: (1) monitoring plant data to detect abnormal functioning of the plant, and (2) identifying and executing procedures to restore normal operation while maintaining plant safety. The knowledge required for both these tasks is available from the plant operating manuals, and the plant operators’ training and experience. Our objective in this paper is to describe the knowledge representation scheme developed for the second task. The objective of the first task is to detect known malfunction states by monitoring the plant conditions. The second task is initiated after the first task, i.e. after the known malfunction states have been detected. The objective of second task is to execute control procedures to: (1) restore normal plant operation, and (2) maintain plant safety. The plant environment on which the control procedures operate is dynamic and unpredictable and given the time limitations, it is not always possible to restore the normal operation and maintain safety simultaneously. It is for this reason that the plant safety of a nuclear power plant is defined independent of normal operation. The plant safety can be maintained if certain critical parameters remain within prescribed limits, whether or not the plant is producing power at its normal ratings. Because of the independence of safety and operation the actions required to

A generic task approach to a real time nuclear power plant fault diagnosis and advisory system

A generic task toolkit has been used in the development of an aid for operators of nuclear power plants. The toolkit consists of high-level programming tools that enable knowledge to be used in accordance with its need. That is, if diagnosis is the need, a framework for performing diagnosis is provided. The operator aid provides for monitoring the conditions in the plant, detecting abnormal events, and providing the operator with guidance and advice through procedures on what path should be following to mitigate the consequences. Using the two generic task tools, CSRL and DSPL, design work on a diagnosis and sensor validation system and a dynamic procedure management system is nearing completion.<<ETX>>

Experience with the hierarchical method for diagnosis of faults in nuclear power plant systems

Abstract Expert systems to aid nuclear power plant operators detect and diagnose abnormal conditions have been under development at The Ohio State University over the past eight years. These systems, genetically called operator advisors (OA), are designed to continuously monitor plant parameters and when an abnormality is detected, trigger a diagnostic module. Methods of diagnosis used by these OAs have employed variations of hierarchical classification. Hierarchical classification as a tool for diagnosis of complex systems is described and three systems using different methods for decomposing the hierarchy are presented. Decomposition of the hierarchy by both system and function is introduced and it is shown that a hybrid of the two methods is the optimum if there is good accessibility to operators and system designers, and if the goal is to emulate the operators approach to diagnosis. In the case of systems operating in different modes it is shown that the most direct method for applying hierarchical diagnosis is by developing a generic function based hierarchy and by incorporating multiple knowledge groups into the nodes in the hierarchy.

An Expert System for Sensor Data Validation and Malfunction Detection

Nuclear power plant operation and monitoring in general is a complex task which requires a large number of sensors, alarms and displays. At any instant in time, the operator is required to make a judgment about the state of the plant and to react accordingly. During abnormal situations, operators are further burdened with time constraints. The possibility of an undetected faulty instrumentation line, adds to the complexity of operators’reasoning tasks. Failure of human operators to cope with the conceptual complexity of abnormal situations often leads to more serious malfunctions and further damages to plant (TMI-2 as an example). During these abnormalities, operators rely on the information provided by the plant sensors and associated alarms. Their usefulness however, is quickly diminished by their large number and the extremely difficult task of interpreting and comprehending the information provided by them. The need for an aid to assist the operator in interpreting the available data and diagnosis of problems is obvious.

Development of a Probabilistic Technique for On-line Parameter and State Estimation in Non-linear Dynamic Systems

The DSD (Dynamic System Doctor) is a system-independent, interactive software under development for on-line state/parameter estimation in dynamic systems (1), partially supported through a Nuclear Engineering Education (NEER) grant during 1998-2001. This paper summarizes the recent accomplishments in improving the user-friendliness and computational capability of DSD

Determination of manganese in Naiad mollusc shells by neutron activation analysis

Abstract Naiads from two locations 11 river miles apart on the lower Muskingum River in Ohio were studied to determine whether location, calendar year and species could be correlated with the manganese concentration in the shells of the freshwater bivalve molluscs. Neutron activation analysis was used to determine the manganese concentration. The effect of each of these variables on the manganese content of the naiad shells was observed and discussed.