Case-based reasoning techniques applied to operation experience feedback in nuclear power plants

Abstract

This paper concerns the work being conducted to develop an information retrieval tool for operation feedback in French nuclear power stations, using CBR techniques, to enhance the interactivity and efficiency of information search operations. Experience feedback problem factors are discussed and the specific nature of the data gathered is presented, with particular emphasis being placed on the importance of free-format text. The principles of the REX plain-language consultation system, based on specialized semantic networks with various points of view, are also briefly described Reference is made to the conclusions of a study concerning prospects for the application of the main CBR techniques to nearest neighbour event searches, and to the development options adopted for the REX-CASE demonstrator. This prototype extends the text indexing on specific and structured suject terminology, normally used in REX, to include indexing on the n-uplets of terms significant in the similarity between texts. The various possible interrogation strategies are also presented.