Simulation-based expert system for nuclear reactor control and diagnostics. Progress report

Technical Report · Thu Jul 31 04:00:00 EDT 1986

OSTI ID:5554956

Lee, J C; Martin, W R

This research concerns the development of artificial intelligence (AI) techniques suitable for application to the diagnostics and control of nuclear power plant systems. The overall objective of the current effort is to build a prototype simulation-based expert system for diagnosing accidents in nuclear reactors. The system is being designed to analyze plant data heuristically using fuzzy logic to form a set of hypotheses about a particular transient. Hypothesis testing, fault magnitude estimation and transient analysis is performed using simulation programs to model plant behavior. An adaptive learning technique has been developed for achieving accurate simulations of plant dynamics using low-order physical models of plant components. The results of the diagnostics and simulation analysis of the plant transient are to be analyzed by an expert system for final diagnoses and control guidance. To date, significant progress has been made toward achieving the primary goals of this project. Based on a critical safety functions approach, an overall design for the nuclear plant expert system has been developed. The methodology for performing diagnostic reasoning on plant signals has been developed and the algorithms implemented and tested. A methodology for utilizing the information contained in the physical models of plant components has also been developed. This work included the derivation of a unique Kalman filtering algorithm for using power plant data to systematically improve on-line simulations through the judicious adjustment of key model parameters. A few simulation models of key plant components have been developed and implemented to demonstrate the method on a realistic accident scenario. The chosen transient is a loss of feed flow exasperated by a stuck open relief valve, similar to the initiating event of the Three Mile Island Unit 2 accident in 1979.

• Research Organization: Michigan Univ., Ann Arbor (USA). Dept. of Nuclear Engineering
• DOE Contract Number: AC02-85NE37945
• OSTI ID: 5554956
• Report Number(s): DOE/NE/37945-1; ON: DE86014668
• Country of Publication: United States
• Language: English

Related Subjects

22 GENERAL STUDIES OF NUCLEAR REACTORS
220400 -- Nuclear Reactor Technology-- Control Systems
220900* -- Nuclear Reactor Technology-- Reactor Safety
ACCIDENTS
ALGORITHMS
ARTIFICIAL INTELLIGENCE
COMPUTERIZED SIMULATION
CONTROL SYSTEMS
DIAGNOSTIC TECHNIQUES
EXPERT SYSTEMS
LOSS OF FLOW
MATHEMATICAL LOGIC
MATHEMATICAL MODELS
NUCLEAR FACILITIES
NUCLEAR POWER PLANTS
POWER PLANTS
REACTOR ACCIDENTS
REACTOR COMPONENTS
REACTOR CONTROL SYSTEMS
RESEARCH PROGRAMS
SIMULATION
THERMAL POWER PLANTS