Automatic fault classification for model-based process monitoring

• US 7,533,070 B2
• Filed: 05/30/2006
• Issued: 05/12/2009
• Est. Priority Date: 05/30/2006
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Guralnik, Valerie, Foslien, Wendy K.
• Primary Examiner(s): Holmes; Michael B
• Application Number:
  US11/442,857
• Publication Number:
  US 20070282777A1
• Time in Patent Office: 1,078 Days
• Field of Search: None
• US Class Current: 706/16
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
• CPC Class Codes: G05B 23/0281 Quantitative, e.g. mathematical distance; Clustering; Neural networks; Statistical analysis
• CPC Class Codes: G06F 18/232 Non-hierarchical techniques

Abstract

A computer implemented method, system and program product for automatic fault classification. A set of abnormal data can be automatically grouped based on sensor contribution to a prediction error. A principal component analysis (PCA) model of normal behavior can then be applied to a set of newly generated data, in response to automatically grouping the set of abnormal data based on the sensor contribution to the prediction error. Data points can then be identified, which are indicative of abnormal behavior. Such an identification step can occur in response to applying the principal component analysis mode of normal behavior to the set of newly generated data in order to cluster and classify the data points in order to automatically classify one or more faults thereof. The data points are automatically clustered, in order to identify a set of similar events, in response to identifying the data points indicative of abnormal behavior.

Claims

1. A computer implemented method, comprising:

applying a statistical model to a set of data in order to identify data points within set of data that are indicative of abnormal behavior; and

thereafter automatically clustering said data points utilizing a three-phase clustering algorithm that produces an event classifier for classifying at least one fault among said data points, thereby providing for the automatic construction of a library of faults for use in an early detection system.

2. The method of claim 1 wherein said statistical model comprises Principal Component Analysis (PCA).

3. The method of claim 1 wherein during a first phase of said three-phase clustering algorithm all data points among said set of data belonging to a same event are grouped together in order to produce a set of event clusters.

4. The method of claim 3 wherein said data points are clustered based on a time proximity.

5. The method of claim 4 further comprising determining said time proximity utilizing a probability density function by calculating a distribution value.

6. The method of claim 3 wherein during a second phase of said three-phase clustering algorithm said event cluster is segmented into a plurality of sub-clusters representing a single event in order to determine an event state in each event thereof.

7. The method of claim 6 wherein said event cluster is segmented into said plurality of sub-clusters utilizing a time-segmentation calculation and an optimization of clustering criteria.

8. The method of claim 7 further comprising segmenting said event cluster by:

repeatedly selecting a cluster with a most non-optimized clustering criterion among a current set of clusters; and

determining an optimal location for dividing a sub-cluster for segmentation into said plurality of sub-clusters.

9. The method of claim 8 further comprising:

terminating said segmentation of said event cluster when a smallest sub-cluster spans a time period shorter than a specified time period.

10. The method of claim 8 further comprising:

terminating said segmentation of said event cluster when a gain in a resulting clustering criterion is negligible and below a particular threshold in order to improve a resulting segmentation thereof.

11. The method of claim 3 wherein during a third phase of said three-phase algorithm at least one genuine cluster is determined by repeatedly combining together said plurality of sub-clusters across all events in order for the states of all events to be clustered in order to determine a set of event definitions.

12. The method of claim 11 wherein each sub-cluster is represented by a centroid and treated as a single data point.

13. The method of claim 12 wherein an agglomerative hierarchical algorithm is utilized to determine said at least one genuine cluster.

14. A system, comprising:

a data-processing apparatus;

a module executed by said data-processing apparatus, said module and said data-processing apparatus being operable in combination with one another to;

apply a statistical model to a set of data in order to identify data points within set of data that are indicative of abnormal behavior; and

thereafter automatically cluster said data points utilizing a three-phase clustering algorithm that produces an event classifier for classifying at least one fault among said data points, thereby providing for the automatic construction of a library of faults for use in an early detection system.

15. The system of claim 14 wherein said statistical model comprises Principal Component Analysis (PCA).

16. A program product residing in a computer, comprising:

instruction media residing in a computer memory for applying a statistical model to a set of data in order to identify data points within set of data that are indicative of abnormal behavior; and

instruction media residing in a computer memory for thereafter automatically clustering said data points utilizing a three-phase clustering algorithm that produces an event classifier for classifying at least one fault among said data points, thereby providing for the automatic construction of a library of faults for use in an early detection system.

17. The program product of claim 16 wherein said statistical model comprises Principal Component Analysis (PCA).

18. The program product of claim 16 wherein during a first phase of said three-phase clustering algorithm all data points among said set of data belonging to a same event are grouped together in order to produce an event cluster.

19. The program product of claim 16 wherein during a second phase of said three-phase clustering algorithm said event cluster is segmented into a plurality of sub-clusters representing a single event in order to determine an event state in each event thereof.

20. The program product of claim 16 wherein during a third phase of said three-phase algorithm at least one genuine cluster is determined by repeatedly combining together said plurality of sub-clusters across all events in order for the states of all events to be clustered in order to determine a set of event definitions.