×

APPARATUS ANOMALY MONITORING METHOD AND SYSTEM

  • US 20110276828A1
  • Filed: 01/14/2009
  • Published: 11/10/2011
  • Est. Priority Date: 01/14/2009
  • Status: Active Grant
First Claim
Patent Images

1. An apparatus anomaly monitoring method with using information processing of a computer, the method of performing processes of, based on a plurality of state data items of a target apparatus obtained by measuring a state of the apparatus with a plurality of sensors, monitoring and judging an anomaly of the state of the apparatus, whereinthe method includes:

  • a first step of performing a model creating process of creating a model for the monitoring and judgment based on the plurality of state data items of the apparatus at a normal time; and

    a second step of performing a monitoring execution process of inputting the plurality of state data items of the apparatus at a predetermined time unit, monitoring and judging the anomaly of the state of the apparatus with using the model, and, when the anomaly is detected, outputting detection information,the first step includes;

    a step of categorizing the plurality of state data items into an objective-variable data item and two or more explanatory-variable data items other than the objective-variable data item in regression analysis; and

    a step of creating an individual model of predicting the objective-variable data item from the one explanatory-variable data item for each of the explanatory-variable data items as two or more individual models configuring an ensemble of the models to configure the ensemble of the models, andthe second step includes, for inputs of the plurality of state data items;

    a step of computing an individual predicted value of the objective-variable data item for each of the individual models configuring the ensemble of the models with taking the explanatory-variable data item as an input;

    a step of computing an individual error span between the predicted value and a measurement value of the objective-variable data item for each of the individual predicted values;

    a step of computing an ensemble error span in combination with the plurality of error spans obtained for each of the explanatory-variable data items; and

    a step of detecting the anomaly of the apparatus by comparing the ensemble error span and a threshold.

View all claims
  • 1 Assignment
Timeline View
Assignment View
    ×
    ×