Principal component analysis based fault classification

• US 7,096,153 B2
• Filed: 04/16/2004
• Issued: 08/22/2006
• Est. Priority Date: 12/31/2003
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Guralnik, Valerie, Foslien, Wendy K
• Primary Examiner(s): Bui, Bryan
• Application Number:
  US10/826,614
• Publication Number:
  US 20050141782A1
• Time in Patent Office: 858 Days
• Field of Search: 702/33, 702/34, 702/116, 702/179, 702/180, 702/182, 702/185, 702/186, 702/188, 702/189, 700 28- 30, 700/47, 700/49, 700/108
• US Class Current: 702/179
• 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 11/3013 where the computing system is an embedded system, i.e. a combination of hardware and software dedicated to perform a certain function in mobile devices, printers, automotive or aircraft systems testing or monitoring of control systems or parts thereof G05B23/02
• CPC Class Codes: G06F 11/3072 where the reporting involves data filtering, e.g. pattern matching, time or event triggered, adaptive or policy-based reporting
• CPC Class Codes: G06F 18/232 Non-hierarchical techniques
• CPC Class Codes: G06F 18/2433 Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection

Abstract

Principal Component Analysis (PCA) is used to model a process, and clustering techniques are used to group excursions representative of events based on sensor residuals of the PCA model. The PCA model is trained on normal data, and then run on historical data that includes both normal data, and data that contains events. Bad actor data for the events is identified by excursions in Q (residual error) and T2 (unusual variance) statistics from the normal model, resulting in a temporal sequence of bad actor vectors. Clusters of bad actor patterns that resemble one another are formed and then associated with events.

Claims

1. A system for identifying events in a process, the system comprising:

a controller coupled to sensors monitoring a process;

a principal component analysis model receiving data from the sensors monitoring the process and reducing a number of variables associated with the data from the sensors, the model further comprising;

a training module that is run on historical data to create a pool of vectors with values for the variables, wherein the training module further creates clusters of bad actors from the values based on statistics and associates the clusters with known events; and

a run time module that receives incoming data from the sensors, calculates statistics, determines if events are occurring, and identifies clusters to identify events.

2. The system of claim 1 wherein the run time module calculates statistics related to the principal component analysis model, determines and finds a nearest cluster of bad actors related to the event to identify the event.

3. The system of claim 1 wherein the run time module identifies a sequence of cluster matches and correlates the sequence of cluster matches to known events.

4. The system of claim 1 wherein the model further determines if a cluster needs to be split when new bad actors are added and splits the cluster into two clusters using a goodness of fit criteria.

5. The system of claim 4 wherein the model determines if a new event category is encountered and broadens limits for the sequence of clusters.

6. The system of claim 1 wherein a cluster is limited to a predetermined number of bad actors.

7. The system of claim 1 wherein the statistics comprise Q and T2.

8. The system of claim 1 wherein the model uses a feature scoring scheme to identify top contributors of bad actors.

9. The system of claim 8 wherein the feature scoring scheme is based on rank, value, and percent of contribution to a Q-residual sensor to identify a relative importance.

10. The system of claim 9 wherein the top-contributors are determined based on a majority percentage of the Q-residual.

11. The system of claim 9, where the top-contributors are determined based on only the contributors with absolute values that are drastically different from values of other contributors.

12. The system of claim 9 wherein the scoring scheme is based on predetermined limits.

13. The system of claim 12 wherein, the limits are computed statistically through change point detection methods.

14. The system of claim 8, wherein a predetermined minimum/maximum number of contributors are selected from rank, value, and percent of contribution to a Q-residual sensor to identify a relative importance.

15. A system for identifying events in a process, the system comprising:

means for monitoring a process via sensors;

means for receiving data from the sensors monitoring the process and reducing a number of variables associated with the data from the sensors to produce a principal component analysis model, the model further comprising;

means run on historical data for creating a pool of vectors with values for the variables, and creating clusters of bad actors from the values based on statistics and associates the clusters with known events; and

means for receiving incoming data from the sensors, calculating statistics, determining if events are occurring, and identifying clusters to identify events.