A fuzzy knowledge-based system for handling criticality analysis in power plants maintenance

Abstract

The inspection planning in electric power industry is used to assess the safety and reliability of system components and to increase the ability of failure situation identification before it actually occurs. It reflects the implications of the available information on the operational and maintenance history of the system. The output is a ranking list of components, with the most critical ones at the top, which indicates the selection of the components to be inspected. The objective of this paper is to demonstrate the use of a fuzzy relational database model for manipulating the data required for the criticality component ranking in inspection planning. The component criticality classification is formed by incorporating criteria like the system downtime in case of failure, results of previous inspections, cost of replacement, safety requirements, environmental aspects, qualitative past service history, expected change in operating conditions and alternative supply patterns. Often, numeric values are not available for the component criticality analysis, thus qualitative thresholds and linguistic terms must be used. The need for symbolic reasoning and the use of linguistic terms appoints the fuzzy logic approach as an appropriate tool for the elaboration of the criteria involved in the criticality analysis. Fuzzy linguistic terms for criteria definitions along with fuzzy inference mechanisms allow the operators expertise to be exploited. The proposed database model ensures the representation and handling of the above fuzzy information and additionally offers the user functionality for specifying the precision degree by which the conditions involved in a query are satisfied. In order to illustrate the behavior of the model a case study is given using real inspection data.