Waveform segmentation device and waveform segmentation method

A waveform segmentation device comprising:

• a sensor installed at a factory or a plant;
• and
• processing circuitry configured to:
• receive waveform data from the sensor, the waveform data representing sensor detection data;
• estimate a state level of the waveform data;
• identify a plurality of segmentation points and segment the waveform data at a plurality of segmentation points based on the estimated state level;
• estimate state levels of a plurality of pieces of partial waveform data based on the plurality of pieces of partial waveform data obtained by segmenting the waveform data at the plurality of segmentation points;
• update the estimated state level based on the estimated state levels;
• adjust positions of the plurality of segmentation points for segmenting the input waveform data based on a plurality of representative segmentation points for segmenting representative waveform data selected from among a plurality of pieces of waveform data belonging to a waveform group;
• evaluate the identified plurality of segmentation points or the adjusted plurality of segmentation points;
• calculate a number of partial waveforms required to segment each waveform data included in the waveform group by calculating a frequency of the number of partial waveforms in the waveform group, and determining the number of partial waveforms having a highest frequency as the number of partial waveforms required to segment each waveform data included in the waveform group;
• extract a waveform feature value from between two adjacent segmentation points among the segmentation points;
• generate an abnormality detection model based on the waveform feature value;
• perform abnormality detection on the sensor detection data using the abnormality detection model;
• and
• transmit the waveform feature value to a plurality of networks.

A waveform segmentation method comprising:

• receiving, from a sensor installed at a factory or a plant, waveform data from the sensor, the waveform data representing sensor detection data;
• estimating a state level of the waveform data;
• segmenting the waveform data at a plurality of segmentation points based on the estimated state level;
• estimating state levels of a plurality of pieces of partial waveform data based on the plurality of pieces of partial waveform data obtained by segmenting the waveform data at the plurality of segmentation points;
• updating the estimated state level based on the estimated state levels;
• adjusting positions of the plurality of segmentation points for segmenting the input waveform data based on a plurality of representative segmentation points for segmenting representative waveform data selected from among a plurality of pieces of waveform data belonging to a waveform group;
• calculating a number of partial waveforms required to segment each waveform data included in the waveform group by calculating a frequency of the number of partial waveforms in the waveform group, and determining the number of partial waveforms having a highest frequency as the number of partial waveforms required to segment each waveform data included in the waveform group;
• extracting a waveform feature value from between two adjacent segmentation points among the segmentation points;
• generating an abnormality detection model based on the waveform feature value;
• performing abnormality detection on the sensor detection data using the abnormality detection model;
• and
• transmitting the waveform feature value to a plurality of networks.