Abstract
PROBLEM TO BE SOLVED: To reduce false detection of abnormality in abnormality detection using operating sound of a monitoring target device, as one embodiment that.
SOLUTION: An abnormality detection program causes a computer to perform processing for detecting abnormality of a monitoring target device from second sound data obtained by collecting operation sound of the monitoring target device on the basis of abnormality evaluation data generated using a data point group belonging to a local cluster specified based on a first abnormality degree and a second abnormality degree calculated for each data point among a plurality of data points plotted by sampling, at a plurality of timings, sound data obtained by collecting operation sound of the monitoring target device on a multi-dimensional space including a plurality of feature amounts based on components of different frequency regions. The first abnormality degree is calculated based on the number of dividing times required for dividing at random into the plurality of data points and then dividing into individual data points. The second abnormality degree is calculated based on the number of dividing times required for dividing into the plurality of data points so that the distribution of representative values obtained from each data point satisfies a predetermined condition in two groups after division, and then dividing into individual data points.
SOLUTION: An abnormality detection program causes a computer to perform processing for detecting abnormality of a monitoring target device from second sound data obtained by collecting operation sound of the monitoring target device on the basis of abnormality evaluation data generated using a data point group belonging to a local cluster specified based on a first abnormality degree and a second abnormality degree calculated for each data point among a plurality of data points plotted by sampling, at a plurality of timings, sound data obtained by collecting operation sound of the monitoring target device on a multi-dimensional space including a plurality of feature amounts based on components of different frequency regions. The first abnormality degree is calculated based on the number of dividing times required for dividing at random into the plurality of data points and then dividing into individual data points. The second abnormality degree is calculated based on the number of dividing times required for dividing into the plurality of data points so that the distribution of representative values obtained from each data point satisfies a predetermined condition in two groups after division, and then dividing into individual data points.
Translated title of the contribution | Abnormality detection program, abnormality detection method, and information processing apparatus |
---|---|
Original language | Japanese |
Patent number | JP 6828626 B2 |
IPC | G01H 3/04,G01M 99/00 |
Filing date | 11/07/17 |
Publication status | Published - 10 Feb 2021 |
Externally published | Yes |