Monitoring diagnostic device and monitoring diagnostic method

摘要

An abnormality monitoring process milt (2) divides sensor data collected from a monitoring-target apparatus (8) into sensor data for each of a plurality of condition modes based on a condition-mode transition point detected by a condition-mode transition point detecting process unit (21), and sorts the divided sensor data into a plurality of groups. Next, for each condition mode and each group, each piece of sensor data is compared with past statistic data, thereby detecting an abnormality. A causal diagnosis process unit (3) diagnoses an abnormality cause using link models before and after an abnormality is detected built based on a correlation coefficient between two pieces of sensor data in each group.

主权项

1. A monitoring diagnostic device comprising:
a first process unit that collects, from a monitoring-target apparatus having a plurality of components each having a normal operation behavior, a time-series physical quantity indicating a time transition in a physical condition of each of the plurality of components; a second process unit that detects an abnormality in the monitoring-target apparatus using the time-series physical quantities of the plurality of components collected by the first process unit, the abnormality indicating at least one of the plurality of components is behaving differently from the normal operation behavior; and a third process unit that diagnoses a cause of the abnormality, the second process unit being configured to execute:
a data-item group sorting process of sorting data items identifying the time-series physical quantities of the plurality of components into a plurality of groups; andan abnormality determining process of determining a group abnormality for each sorted group based on the time-series physical quantity of the component identified by the data item belonging to that group and statistical data on the time-series physical quantity of the component, the group abnormality indicating the component for a respective group is behaving differently from the normal operation behavior; and the third process unit being configured to execute:a link-model building process of building, for each sorted group, a link model including an inter-data-item link model representing a dependency relation between the time-series physical quantities of the components identified by the data item belonging to that group, and an inter-group link model representing a dependency relation between representative values of the groups; andan abnormality cause estimating process of estimating a cause of the abnoi orality based on a difference link model between a link model built based on the time-series physical quantity of the component before the abnormality is detected through the abnormality determining process and a link model built based on the time-series physical quantity of the component when the abnormality is detected through the abnormality determining process.