설비오류진단에서 가중 증거 척도를 이용한 다중센서 데이터 융합

Abstract

This study proposes a multisensor fusion method to accurately detect the evidence of fault target when machine fault evidences are highly conflicting and when the machine fault evidences between sensors have no similarities. This fusion method uses a combined weight which is composed of evidence entropy and belief distance. In this study, improved evidence entropy (IEE) system independently and broadly measures uncertain information of rarely occurring fault evidence. IEE measure includes four major factors including the mass function, the numbers of all fault targets, focal elements and intersection between focal elements. Among these four factors, the redundant intersection part of all fault targets was removed to reflect more uncertain information volume and quality. Based on the degree of conflict, Euclidean distance and cross entropy, belief distance measure optimizes consistency and similarity between sensor evidences. Combined weight is adjusted with the fault target evidences and fused by orthogonal combination rule. Numerical cases in machine fault diagnosis are demonstrated to describe the practicality and accuracy of the proposed multisensor data fusion rule.