107 多変量解析手法によるパターン認識技術を用いた設備の異常兆候検出手法に関する検討(セッション2 センサ技術・信号処理II)

概要

論文の詳細を見る
The authors have endeavored to newly develop a monitoring technology used for detecting abnormalities in machinery at an early stage during which signs of abnormality appear, and for promptly determining the cause of abnormality. A technology verification test was practiced using the Mahalanobis-Taguchi System, which is defined as a multivariate data analysis technique. Based on measurement data obtained from a simulator, the individual characteristics of data gathered under conditions of normal, rated operation and data gathered under abnormal conditions were analyzed and evaluated. As a result, it was confirmed that all verified abnormal conditions can be detected during the stage in which signs of abnormality appear. It was also found that all abnormal conditions can be detected before the simulator gives off an alerts indicating an abnormal level ; and, in the case in which the simulator does not give off an alert, the proposed early detection technique can be used to detect problematic conditions deviating from those of a regular, normal state.
一般社団法人日本機械学会の論文
2006-11-29