Method and system for automatically diagnosing faults in rendering devices

• US 8,249,830 B2
• Filed: 06/19/2009
• Issued: 08/21/2012
• Est. Priority Date: 06/19/2009
• Status: Active Grant

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Minhas, Rajinderjeet Singh, Monga, Vishal, Wu, Wencheng, Vats, Divyanshu
• Primary Examiner(s): Khuu, Cindy H
• Assistant Examiner(s): HWANG, TIMOTHY
• Application Number:
  US12/487,803
• Publication Number:
  US 20100325487A1
• Time in Patent Office: 1,159 Days
• Field of Search: 702/185
• US Class Current: 702/185
• CPC Class Codes: G05B 23/0283 Predictive maintenance, e.g. involving the monitoring of a system and, based on the monitoring results, taking decisions on the maintenance schedule of the monitored system; Estimating remaining useful life [RUL] preventive maintenance, i.e. planning maintenance according to the available resources without monitoring the system G06Q10/06
• CPC Class Codes: G06F 11/2257 using expert systems
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]

Abstract

A method and system for automatically determining an optimal re-training interval for a fault diagnoser based on online monitoring of the performance of a classifier are disclosed. The classifier generates a soft measure of membership in association with a class based on a training data. The output of the classifier can be utilized to assign a label to new data and then the members associated with each class can be clustered into one or more core members and potential outliers. A statistical measure can be utilized to determine if the distribution of the outliers is sufficiently different than the core members after enough outliers have been accumulated. If the outliers are different with respect to the core members, then the diagnoser can be re-trained; otherwise, the output of the classifier can be fed to the fault diagnoser.

Claims

1. A computer-implemented method for automatically diagnosing new faults in a device, said method comprising:

said computer generating a measure of membership with respect to each feature vector belonging to at least one known class among a plurality of known classes utilizing a classifier;

said computer clustering a plurality of members associated with said at least one known class into two clusters including core class members and at least one outlier cluster in order to detect a presence of a new defect and a drift in the probabilistic distribution of the plurality of members between the core class and the at least one outlier cluster in a known defect in the device overtime, wherein the device is a rendering device and the known defect is an image quality defect;

said computer determining if a probabilistic distribution of said plurality members of said outlier cluster associated with said class quantifies a change in distribution when a set of spurious features becomes sufficiently different from said corresponding distribution of said core class members to therefore indicate said new faults and said drift in said known defect and thereafter diagnose said new faults and said drift in said known defect in said device, wherein said new faults are unknown image quality defects; and

said computer retraining said classifier with updated characteristics of said new defect.

2. The method of claim 1 further comprising said computer initially mapping a plurality of known faults to user-defined classes to derive said plurality of known classes.

3. The method of claim 1 further comprising said computer configuring said classifier based on training data if said distribution of outliers is different than said core members, wherein said classifier assigns image quality based feature vectors to said at least one known class among said plurality of known classes.

4. A computer-implemented method for automatically diagnosing new faults in a device, said method comprising:

said computer mapping known faults to user-defined classes to derive a plurality of known classes;

said computer configuring a classifier based on training data, wherein said classifier assigns image quality based feature vectors to at least one known class among said plurality of known classes;

said computer generating a measure of membership with respect to each feature vector belonging to said at least one known class utilizing said classifier;

said computer clustering a plurality of members associated with said at least one known class into two clusters including core class members and at least one outlier cluster in order to detect a presence of a new defect and a drift in the probabilistic distribution of the plurality of members between the core class and the at least one outlier cluster in a known defect in the device overtime, wherein the device is a rendering device and the known defect is an image quality defect; and

said computer determining if a probabilistic distribution of said plurality members of said outlier cluster associated with said class quantifies a change in distribution when a set of spurious features becomes sufficiently different from said corresponding distribution of said core class members to therefore indicate said new faults and said drift in said known defect and thereafter diagnose said new faults and said drift in said known defect in said device, wherein said new faults are unknown image quality defects; and

said computer retraining said classifier with updated characteristics of said new defect.

5. The method of claim 4 further comprising said computer performing a corrective action with respect to said new defect in said device.

6. The method of claim 5 wherein said corrective action comprises said computer re-training said classifier with updated expert identified classes based on a magnitude of a Kullback-Liebler divergence.

7. The method of claim 4 further comprising said computer assigning a label to said new defect based on a plurality of known faults utilizing said classifier.

8. The method of claim 4 further comprising said computer associating a support vector machine classifier with said classifier.

9. The method of claim 4 further comprising:

said computer extracting a raw image from a test pattern; and

said computer transmitting said raw image to said classifier in the event said new defect is identified with respect to an image quality associated with said device.

10. The method of claim 4 further comprising said computer configuring said classifier to generate a flag if said classifier detects statistical evidence for the existence of said new defect.

11. The method of claim 4 further comprising said computer training said classifier with respect to a plurality of images associated with at least one symptom.

12. A system for automatically diagnosing new faults in a device, said system comprising:

a processor;

a data bus coupled to said processor; and

a non-transitory computer-usable medium embodying computer program code, said non-transitory computer-usable medium being coupled to said data bus, said computer program code comprising instructions executable by said processor and configured for;

mapping known faults to user-defined classes to derive a plurality of known classes;

configuring a classifier based on training data, wherein said classifier assigns image quality based feature vectors to at least one known class among said plurality of known classes;

generating a measure of membership with respect to each feature vector belonging to said at least one known class utilizing said classifier;

clustering a plurality of members associated with said at least one known class into two clusters including core class members and at least one outlier cluster in order to detect a presence of a new defect and a drift in the probabilistic distribution of the plurality of members between the core class and the at least one outlier cluster in a known defect in the device overtime, wherein the device is a rendering device and the known defect is an image quality defect; and

determining if a probabilistic distribution of said plurality members of said outlier cluster associated with said class quantifies a change in distribution when a set of spurious features becomes sufficiently different from said corresponding distribution of said core class members to therefore indicate said new faults and said drift in said known defect and thereafter diagnose said new faults and said drift in said known defect in said device, wherein said new faults are unknown image quality defects; and

retraining said classifier with updated characteristics of said new defect.

13. The system of claim 12 wherein said instructions are further configured for performing a corrective action with respect to said new defect in said device.

14. The system of claim 13 wherein said corrective action comprises re-training said classifier with updated expert identified classes based on a magnitude of a Kullback-Liebler divergence.

15. The system of claim 12 wherein said instructions are further configured for assigning a label to said new defect based on a plurality of known faults utilizing said classifier.

16. The system of claim 12 wherein said instructions are further configured for associating a support vector machine classifier with said classifier.

17. The system of claim 12 wherein said instructions are further configured for:

extracting a raw image from a test pattern; and

transmitting said raw image to said classifier in the event said new defect is identified with respect to an image quality associated with said device.

18. The system of claim 12 wherein said instructions are further configured for configuring said classifier to generate a flag if said classifier detects statistical evidence for the existence of said new defect.

19. The system of claim 12 wherein said instructions are further configured for training said classifier with respect to a plurality of images associated with at least one symptom.

20. The system of claim 12 wherein said instructions are further configured for:

assigning a label to said new defect based on a plurality of known faults utilizing said classifier; and

associating a support vector machine classifier with said classifier.