METHOD AND DEVICE FOR MONITORING THE STATE OF A FACILITY

• US 20120290879A1
• Filed: 07/28/2010
• Published: 11/15/2012
• Est. Priority Date: 08/28/2009
• Status: Active Grant

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Shibuya, Hisae, Maeda, Shunji
• Granted Patent: US 8,682,824 B2
• Time in Patent Office: Days
• US Class Current: 714/26
• CPC Class Codes: G05B 23/021 adopting a different treatment of each operating region or a different mode of the monitored system, e.g. transient modes; different operating configurations of monitored system
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

This invention provides method for detecting advance signs of anomalies, event signals outputted from the facility are used to create a separate mode for each operating state, a normal model is created for each mode, the sufficiency of learning data for each mode is checked, a threshold is set according to the results of said check, and anomaly identification is performed using said threshold. Also, for diagnosis, a frequency matrix is created in advance, with result events on the horizontal axis and cause events on the vertical axis, and the frequency matrix is used to predict malfunctions. Malfunction events are inputted as result events, and quantized sensor signals having anomaly measures over the threshold are inputted as cause events.

Claims

1. A method for monitoring the state of facility that detects an anomaly based on a time-series sensor signal output from the facility or an apparatus, comprising:

a learning process of extracting a feature vector based on the sensor signal, selecting a feature to be used based on data check of the feature vector, selecting learning data to be used based on data check of the feature vector, creating a normal model based on the selected learning data, checking sufficiency of the learning data used for creating the normal model, and setting a threshold in accordance with the sufficiency of the learning data; and

an anomaly detecting process of extracting the feature vector based on the sensor signal, computing an anomaly measurement through the comparison between the normal model and the feature vector, and identifying the anomaly through the comparison between the anomaly measurement and the threshold.

2. The method for monitoring the state of facility according to claim 1, wherein the data check in selecting the feature and selecting the learning data is performed based on the mean and distribution for one operating cycle for each feature of the feature vector.

3. The method for monitoring the state of facility according to claim 1, wherein the data check in selecting the feature and selecting the learning data is performed based on the number of times of deviation from a waveform model for one operating cycle for each feature of the feature vector.

4. The method for monitoring the state of facility according to claim 2, wherein the feature is selected by inputting information on a warning or failure time of the facility or apparatus, adding a normal or anomaly label to the feature vector based on the information on the warning or failure time, and retrieving a combination of features based on the case in which a ratio of an anomaly measurement of the feature vector added with the anomaly label to an anomaly measurement of the feature vector added with the normal label is the maximum.

5. The method for monitoring the state of facility according to claim 3, wherein the feature is selected by inputting information on a warning or failure time of the facility or apparatus, adding a normal or anomaly label to the feature vector based on the information on the warning or failure time, and retrieving a combination of features based on the case in which a ratio of an anomaly measurement of the feature vector added with the anomaly label to an anomaly measurement of the feature vector added with the normal label is the maximum.

6. The method for monitoring the state of facility according to claim 2, wherein the learning data is selected by creating a plurality of normal models by random sampling from data other than data of a target cycle for each one operating cycle with respect to all learning data, computing a plurality of anomaly measurements of the data of the target cycle by using all the plurality of normal models, computing the predetermined number of means from the plurality of anomaly measurements of each data, acquiring a maximum value of the mean during the target cycle, and excluding data of a cycle in which the maximum value of the mean is larger than a predetermined threshold, from the learning data.

7. The method for monitoring the state of facility according to claim 3, wherein the learning data is selected by creating a plurality of normal models by random sampling from data other than data of a target cycle for each one operating cycle with respect to all learning data, computing a plurality of anomaly measurements of the data of the target cycle by using all the plurality of normal models, computing the predetermined number of means from the plurality of anomaly measurements of each data, acquiring a maximum value of the mean during the target cycle, and excluding data of a cycle in which the maximum value of the mean is larger than a predetermined threshold, from the learning data.

8. A method for monitoring the state of facility that detects an anomaly based on a time-series sensor signal and an event signal output from the facility or an apparatus, comprising:

a learning process of dividing a mode for each operating state based on the event signal, extracting a feature vector based on the sensor signal, selecting a feature to be used based on data check of the feature vector, selecting learning data to be used based on data check of the feature vector, creating a normal model for each mode based on the selected learning data, checking sufficiency of the learning data used for creating the normal model for each mode, and setting a threshold in accordance with the sufficiency of the learning data; and

an anomaly detecting process of dividing the mode for each operating state based on the event signal, extracting the feature vector based on the sensor signal, computing an anomaly measurement by comparing the feature vector with the normal model, and identifying the anomaly by comparing the threshold with the anomaly measurement.

9. The method for monitoring the state of facility according to claim 8, wherein the learning data is selected by acquiring information on a warning or failure, information on start and stop sequences, and information on occurrence of an event or event array having a low frequency, based on the event signal and determining exclusion by the predetermined basic unit based on the acquired information.

10. A method for monitoring the state of facility that detects an anomaly based on a time-series sensor signal and an event signal output from the facility or an apparatus, comprising:

a learning process of dividing a mode for each operating state based on the event signal, extracting a feature vector based on the sensor signal, creating a normal model for each mode based on the feature vector, checking sufficiency of the learning data used for creating the normal model for each mode, and setting a threshold in accordance with the sufficiency of the learning data; and

an anomaly detecting process of dividing the mode for each operating state based on the event signal, extracting the feature vector based on the sensor signal, computing an anomaly measurement by comparing the feature vector with the normal model, and identifying the anomaly by comparing the threshold with the anomaly measurement.

11. The method for monitoring the state of facility according to claim 10, wherein in the learning process, the feature vector at the time when the anomaly is judged by the anomaly identification is quantized to be set as a cause event and a failure event which occurs from the time of the anomaly judgment to the passage of a predetermined time is set as a result event, and a frequency matrix of a combination of the cause event and the result event is created, and

in the anomaly detecting process, the occurrence of the result event is predicted based on the frequency matrix created in the learning process and the feature vector at the time when the anomaly is judged by the anomaly identification.

12. The method for monitoring the state of facility according to claim 11, wherein at the time of predicting the occurrence of the result event, a normal sensor signal computed based on the normal model and the feature vector and an actual sensor signal are overlapped and displayed with respect to a predetermined time before or after the anomaly judgment.

13. The method for monitoring the state of facility according to claim 11, wherein at the time of predicting the occurrence of the result event, a normal sensor signal computed based on the normal model and the feature vector and an actual sensor signal are overlapped and displayed with respect to learning data in which the predicted result event and an input cause event coincide with each other.

14. A method for monitoring the state of facility, comprising:

mode-dividing a time-series event signal output from the facility or an apparatus in accordance with an operating state of the facility or apparatus;

acquiring a feature vector from a time-series sensor signal output from the facility or apparatus;

creating a normal model for each divided mode by using the mode dividing information and information on the feature vector acquired from the sensor signal;

computing an anomaly measurement of the feature vector for each divided mode by using the created normal model;

judging an anomaly by comparing the computed anomaly measurement with a predetermined threshold; and

diagnosing whether the facility or apparatus is anomalistic by using the judged anomaly information and the sensor signal.

15. The method for monitoring the state of facility according to claim 14, wherein in the normal model creating process,

learning data is selected from the acquired feature vector for each divided mode, and

the normal model is acquired by using the learning data selected for each mode.

16. The method for monitoring the state of facility according to claim 15, further comprising:

a process of checking sufficiency of the learning data for each selected divided mode,

wherein a threshold of an anomaly measurement for judging the anomaly is set for each divided mode in accordance with a result of checking the sufficiency of the learning data for each divided mode in the process.

17. The method for monitoring the state of facility according to claim 15, wherein the computed anomaly measurement and the set threshold of the anomaly measurement are displayed on a screen.

18. A device for monitoring the state of facility, comprising:

a mode dividing means inputting a time-series event signal output from the facility or an apparatus to mode-divide the event signal in accordance with an operating state of the facility or apparatus;

a feature-vector computation means inputting the time-series sensor signal output from the facility or apparatus to acquire a feature vector from the input sensor signal;

a normal-model creation means creating a normal model for each divided mode by using the mode dividing information from the mode dividing means and information on the feature vector of the sensor signal acquired by the feature-vector computation means;

an anomaly-measurement computation means computing an anomaly measurement of the feature vector acquired by the feature-vector computation means for each divided mode by using the normal model created by the normal-model creation means;

an anomaly judgment means judging an anomaly by comparing the anomaly measurement computed by the anomaly measurement computation means with a predetermined threshold; and

an anomaly diagnosis means diagnosing whether the facility or apparatus is anomalistic by using the information on the anomaly judged by the anomaly judgment means and the time-series sensor signal output from the facility or apparatus.

19. The device for monitoring the state of facility according to claim 18, wherein the normal-model creation means includes:

a learning-data selection unit selecting learning data for each mode divided by the mode dividing means from the feature vector acquired by the feature-vector computation means; and

a normal-model creation unit acquiring the normal model by using the learning data selected for each mode by using the learning-data selection unit.

20. The device for monitoring the state of facility according to claim 19, further comprising:

a learning-data check means checking sufficiency of the learning data for each divided mode, which is selected by using the learning-data selection unit of the normal-model creation means,

wherein a threshold of an anomaly measurement for judging the anomaly is set for each divided mode in accordance with a result of checking the sufficiency of the learning data for each divided mode in the learning-data check means.

21. The device for monitoring the state of facility according to claim 20, further comprising:

a display means displaying the anomaly measurement computed by using the means for setting the threshold and the set threshold of the anomaly measurement on the screen.