Automated self-learning diagnostic system

• US 6,970,804 B2
• Filed: 12/17/2002
• Issued: 11/29/2005
• Est. Priority Date: 12/17/2002
• Status: Active Grant

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Zhao, Feng, Siegel, Robert P.
• Primary Examiner(s): SHAH, KAMINI S
• Application Number:
  US10/320,908
• Publication Number:
  US 20040117153A1
• Time in Patent Office: 1,078 Days
• Field of Search: 702 57- 59, 702/108, 702/113, 702/115, 702/117, 702/183, 358/1.15, 340/3.43, 340/3.1, 399/8, 709/223, 371/15.1
• US Class Current: 702/182
• CPC Class Codes: H04L 41/0681 Configuration of triggering conditions
• CPC Class Codes: H04L 41/147 for predicting network behaviour
• CPC Class Codes: H04L 41/16 using machine learning or artificial intelligence

Abstract

A self-learning diagnostic system provides diagnostics capabilities which may be applied to a population of networked machines or components and assemblies in a product. The self-learning diagnostic system uses both the components'"'"' own historical data and the data for an entire population of networked machines of a given product in the field as the training set to adjust critical threshold parameters for detection and diagnosis. The system includes a nominal diagnostic threshold setting module which sets initial thresholds and an adjustment module, which adjusts thresholds continuously based on machine performance data. A service records evaluation module checks service records periodically for correlations and an adjustment module adjust service strategies based on correlation data.

Claims

1. A self-learning diagnostic system for a population of networked machines, comprising:

a nominal diagnostic threshold setting module for initiating self-learning, wherein initiating self-learning includes setting initial machine component thresholds based on a composite model of the machines within the population of networked machines;

a threshold adjustment module for continuously analyzing real time machine performance data received from each of the networked machines to identify performance trends among the networked machines, wherein said performance trends are utilized to develop new machine component thresholds;

a service records evaluation module; and

an adjustment module.

2. The self-learning diagnostic system according to claim 1, wherein said threshold adjustment module continuously adjusts said initial machine component thresholds.

3. The self-learning diagnostic system according to claim 1, wherein said service records evaluation module checks service records pedodically for data correlations, wherein said data correlations comprise the co-occurrence of faults with actions recorded in said service records.

4. The self-learning diagnostic system according to claim 1, wherein said adjustment module adjusts service strategies based on correlation data.

5. The self-learning diagnostic system according to claim 1, wherein the self-learning diagnostic system provides machine service rules for service engineers or customers.

6. The self-learning diagnostic system according to claim 1, wherein said adjustment module continuously adjusts detection thresholds as components within the population of networked machines age.

7. The self-learning diagnostic system according to claim 1, wherein said adjustment module continuously adjusts detection thresholds as environmental conditions change.

18. The self-learning diagnostic system according to claim 1, wherein said adjustment module continuously adjusts detection thresholds as usage patterns vary.

8. A method for operating a self-learning diagnostic system for a population of networked machines, comprising:

receiving initial machine data;

determining a nominal set of diagnostic set points and thresholds, wherein said nominal set of thresholds is based on a composite model of the machines within the population of networked machines;

perfonning self-learning diagnostics comprising;

collecting machine performance data from the population of networked machines;

analyzing real time performance data from the networked machines to identity performance trends;

utilizing said performance trends to develop new machine component thresholds;

updating set points with said new machine component;

checking machine service records for undetected failures;

updating or adding machine service rules; and

adjusting diagnostic thresholds;

repeating performance of self-learning diagnostics continuously during operation of the population of networked machines; and

providing feedback to engineering divisions.

9. The method for operating a self-learning diagnostic system for a population of networked machines according to claim 8, wherein updating set points and thresholds is based on statistical distribution of performance data.

10. The method for operating a self-learning diagnostic system for a population of networked machines according to claim 8, wherein checking machine service records further comprises checking machine service records for correlation with performance data, wherein said data correlations comprise the co-occurrence of faults with actions recorded in said service records.

11. The method for operating a self-learning diagnostic system for a population of networked machines according to claim 8, wherein updating or adding machine service rules is based on the correlation of machine service records with performance data.

12. The method for operating a self-learning diagnostic system for a population of networked machines according to claim 8, wherein providing feedback further comprises providing feedback to product divisions.

13. The method for operating a self-learning diagnostic system for a population of networked machines according to claim 8, further comprising performing said method in a centralized location.

14. The method for operating a self-learning diagnostic system for a population of networked machines according to claim 8, further comprising performing said method with distributed machine clusters.

15. The method for operating a self-learning diagnostic system for a population of networked machines according to claim 8, further comprising embedding said method in each machine of the population of networked machines.

16. The method for operating a self-learning diagnostic system for a population of networked machines according to claim 8, further comprising performing said method off-line.

17. The method for operating a self-learning diagnostic system for a population of networked machines according to claim 8, further comprising performing said method on-line using real-time machine performance data over a network connection.

19. A method for operating a self-learning diagnostic system for a population of networked machines, comprising:

receiving initial machine data;

determining a nominal set of diagnostic set points and thresholds, wherein said nominal set of thresholds is based on a composite model of the machines within the population of networked machines;

collecting machine performance data from field machines;

analyzing real time performance data from the networked machines to identify performance trends;

utilizing said performance trends to develop new machine component thresholds;

updating said nominal set points and thresholds based on said new machine component thresholds; and

providing feedback to engineering divisions.

20. An article of manufacture comprising a computer usable medium having computer readable program code embodied in said medium which, when said program code is executed by said computer causes said computer to perform method steps for operating a self-learning diagnostic system for a population of networked machines, said method comprising:

receiving initial machine data;

determining a nominal set of diagnostic set points and thresholds, wherein said nominal set of thresholds is based on a composite model of the machines within the population of networked machines;

performing self-learning diagnostics comprising;

collecting machine performance data from the population of networked machines;

analyzing real time performance data from the networked machines to identity performance trends;

utilizing said performance trends to develop new machine component thresholds;

updating set points with said new machine component thresholds;

checking machine service records for undetected failures;

updating or adding machine service rules; and

adjusting diagnostic thresholds;

repeating performance of self-learning diagnostics continuously during operation of the population of networked machines; and

providing feedback to engineering divisions.