摘要 |
<p><P>PROBLEM TO BE SOLVED: To solve the problem wherein a precision of failure detection is low, and a failure of a type and a site different from those of a failure that should be detected, is detected in a failure analysis system in which cases are registered, and a rule and a threshold are learned among operation management systems in which a failure in a system is detected and a type or a site of a failure is presented to a maintenance operator. <P>SOLUTION: Inclusion of a value representing a degree of abnormality of a device to be monitored and situation information concerning the device to be monitored in a feature space used when learning a type and a site of a failure, provides a distance in the feature space between cases with labels of a type and a site of a failure or a label of normality when creating a hyperplane for classifying a failure, thereby enabling creation of a hyperplane that detects a failure with a high degree of precision and classifies them. <P>COPYRIGHT: (C)2009,JPO&INPIT</p> |