Inductive monitoring system constructed from nominal system data and its use in real-time system monitoring

• US 7,383,238 B1
• Filed: 02/24/2004
• Issued: 06/03/2008
• Est. Priority Date: 02/24/2004
• Status: Active Grant

• Current Assignee: USA As Represented By The Administrator Of The NASA
• Original Assignee: United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration
• Inventors: Iverson, David L.
• Primary Examiner(s): STARKS, WILBERT L
• Application Number:
  US10/789,029
• Time in Patent Office: 1,561 Days
• Field of Search: 706/25, 706/20, 382/157, 382/224, 194/317, 700/83, 715/503, 715/804
• US Class Current: 706/25
• 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
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

The present invention relates to an Inductive Monitoring System (IMS), its software implementations, hardware embodiments and applications. Training data is received, typically nominal system data acquired from sensors in normally operating systems or from detailed system simulations. The training data is formed into vectors that are used to generate a knowledge database having clusters of nominal operating regions therein. IMS monitors a system'"'"'s performance or health by comparing cluster parameters in the knowledge database with incoming sensor data from a monitored-system formed into vectors. Nominal performance is concluded when a monitored-system vector is determined to lie within a nominal operating region cluster or lies sufficiently close to a such a cluster as determined by a threshold value and a distance metric. Some embodiments of IMS include cluster indexing and retrieval methods that increase the execution speed of IMS.

Claims

1. A method of inductive learning comprising providing a computer that is programmed:

to provide or receive training data, including at least one of archived data, simulated nominal data and off-nominal data;

to provide vectors having a set of parameters determined from the training data;

to generate a cluster database comprising clusters that are associated with respective ranges of values for at least a subset of the set of parameters;

to index the clusters of the cluster database based on an indexing distance of each of the clusters from a predetermined indexing reference point;

to organize the clusters into a data structure of clusters based on the cluster indexing; and

to display a relationship between at least one of the vectors and the data structure in a visually perceptible format.

2. The method of claim 1 wherein said process of generating comprises:

determining a separation distance between a selected test vector and one of said clusters, and

producing a new cluster including the test vector, when the separation distance exceeds a threshold value.

3. The method of claim 2, wherein said computer is further programmed to determine a deviation distance by dividing said separation distance between said test vector and said one of said clusters by a value representing a range of values of at least one variable in said one of said clusters, and to associate the deviation distance with a severity of a deviation of the at least one monitored-system vector from a nearest cluster.

4. The method of claim 1 wherein said process of generating comprises:

determining a separation distance between a selected test vector and at least one of said clusters, and

expanding the at least one cluster to include the test vector when the separation distance is less than or equal to a threshold value.

5. The method of claim 4, wherein said computer is further programmed to determine a deviation distance by dividing said separation distance between said test vector and said one or said clusters by a value representing a range of values of at least one variable in said at least one of said clusters, and to associate the deviation distance with a severity of a deviation of the at least one monitored-system vector from a nearest cluster.

6. A method of monitoring a system comprising providing a computer that is programmed:

to provide or receive a cluster database comprising clusters that are associated with respective ranges of values for at least a subset of a set of cluster parameters;

to receive at least one monitored-system vector having monitored-system parameters, with parameter values generated by sensors that provide data measured on a monitored system;

to determine whether the at least one monitored-system vector is contained in any of the clusters based on at least a subset of the monitored-system parameters and the subset of the cluster parameters; and

when at least one of the monitored-system vectors is not contained in any cluster, to determine a deviation distance of the at least one monitored-system vector from a nearest cluster, to associate the determined deviation distance with a severity of a deviation of the at least one monitored-system vector from the nearest cluster, and to display in a visually perceptible format at least one deviation distance for the parameter values for the at least one monitored-system vector from the corresponding parameter values for the nearest cluster.

7. The method of claim 6, wherein said computer is further programmed:

to provide an additional database of clusters, associated with respective ranges of values for at least a subset of said set of parameters, the additional cluster database being annotated with diagnostic information; and

when at least one of said monitored-system vectors is not included in any cluster, to compare at least one of said monitored-system vectors with at least one of the clusters of the additional cluster database.

8. An apparatus for inductive learning comprising a computer that is programmed:

to provide or receive training data, including at least one of archived data, simulated nominal data and off-nominal data;

to provide at least one vector having a set of parameters based on said training data; and

to generate a cluster database comprising clusters associated with selected ranges of values for at least a subset of the set of parameters;

to index the clusters of the cluster database based on an indexing distance of each of the clusters from a predetermined indexing reference point;

to organize the clusters into a data structure of clusters based on the cluster indexing; and

to display a relationship between at least one of the vectors and the data structure in a visually perceptible format.

9. The apparatus of claim 8, wherein said process of generating comprises:

determining a separation distance between a test vector and one of said clusters, and

producing a new cluster if the separation distance exceeds a threshold value.

10. The apparatus of claim 9, wherein said computer is further programmed to determine a deviation distance by dividing said separation distance between said test vector and said one or said clusters by a value representing a range of values of at least one variable in said one of said clusters, and to associate the deviation distance with a severity of a deviation of the at least one monitored-system vector from a nearest cluster.

11. The apparatus of claim 8 wherein said process of generating comprises:

determining a separation distance between a test vector and at least one of said clusters, and

expanding the at least one of said clusters to include the test vector when the separation distance is less than or equal to a threshold value.

12. The apparatus of claim 11, wherein said computer is further programmed to determine a deviation distance by dividing said separation distance between said test vector and said one or said clusters by a value representing a range of values of at least one variable in the at least one of said clusters, and to associate the deviation distance with a severity of a deviation of the at least one monitored-system vector from a nearest cluster.

13. An apparatus for monitoring a system, comprising a computer, having a memory storing a cluster database comprising clusters, associated with respective ranges of values for at least a subset of a set of cluster parameters, where the computer is programmed:

to provide or receive one or more monitored-system vectors having monitored-system parameter, with parameter values generated by sensors that provide data measured on a monitored system;

to determine whether the monitored-system vector is contained in any of the clusters based on at least a subset of the monitored-system parameters and the at least a subset of cluster parameters; and

when at least one of the monitored-system vectors is not contained in any cluster, to determine a deviation distance of the at least one monitored-system vector from a nearest cluster, to associate the determined deviation distance with a severity of a deviation of the at least one monitored-system vector from the nearest cluster, and to display in a visually perceptible format at least one deviation distance for a parameter value for the at least one monitored-system vector from a corresponding parameter value for the nearest cluster.

14. The apparatus of claim 13, wherein said computer is further programmed;

to provide an additional database of clusters that are associated with respective ranges of values for at least a subset of said parameters, the additional cluster database being annotated with diagnostic information; and

when at least one of said monitored-system vectors is not included in any of said clusters, to compare said at least one of said monitored-system vectors with the clusters of the additional cluster database.