A Methodology for the Design and Analysis of a Sensor Failure Detection Network

Abstract

Two topics are discussed: a method to evaluate and construct a sensor failure detection network and use of the network for signal validation of a complex system such as a nuclear power plant. The network is arranged in a tree structure and consists of plantwide sensor measurements and component models. Sensors are categorized into four classes via a logic state analysis to determine the effectiveness of the tree layouts and to reveal deficiencies in the sensor arrangements. Network building is automatic via a rule-based algorithm. Besides analytical redundancy and parity relations, plantwide consistency checks are implemented in the validation scheme to detect possible common-mode failures and modeling or process faults. Data are structured with an entity relationship and processed with an object-oriented technique. The working sequence is arranged using topological sorting to facilitate on-line, real-time applications. For a demonstration, the package is implemented on a microcomputer and applied to a pressurized water reactor plant for safety parameter validation. Its performance in detecting hypothetical sensor failures during power maneuver transients is presented.