Method to use a receiver operator characteristics curve for model comparison in machine condition monitoring

• US 7,552,035 B2
• Filed: 10/28/2004
• Issued: 06/23/2009
• Est. Priority Date: 11/12/2003
• Status: Active Grant

• Current Assignee: Siemens Corp. (Siemens AG)
• Original Assignee: Siemens Corporate Research Incorporated (Siemens AG)
• Inventors: Cataltepe, Zehra, Neubauer, Claus, Yuan, Chao
• Primary Examiner(s): Rodriguez; Paul L
• Assistant Examiner(s): Thangavelu; Kandasamy
• Application Number:
  US10/977,220
• Publication Number:
  US 20050144537A1
• Time in Patent Office: 1,699 Days
• Field of Search: 703/2, 703/6, 703/13, 702/19, 702/183, 435/6, 435/7.23, 435/4, 382/132, 382/128, 706/46, 706/12, 536/24.3, 606/48, 606/50, 606/42, 606/98, 607/75, 602/74
• US Class Current: 703/2
• CPC Class Codes: G05B 23/0243 model based detection method, e.g. first-principles knowledge model

Abstract

A method to use a receiver operator characteristics curve for model comparison in machine condition monitoring. The method and systems of using this method may be used to evaluate different monitoring models. These models may be used to monitor a variety of different systems such as power plant systems or magnetic resonance imaging systems. The methods use training data and designate one or more points in the data as a false negative, thereby permitting a receiver operator characteristics analysis to be performed. Multiple receiver operator characteristics analyses may be performed either on different models or on different points within a single model, thereby permitting the receiver operator characteristics analyses to be used to select a beneficial model for monitoring a particular system.

Claims

1. A computer implemented method of evaluating a monitoring model of a power plant system or a magnetic resonance imaging system, comprising:

obtaining a first set of test data from the system through at least one monitoring sensor, the first set of test data including a first plurality of negative data points;

designating at least one of the first plurality of negative data points as a first positive data point; and

performing a first receiver operating characteristics (ROC) analysis on the first set of test data to obtain a first receiver operating characteristics analysis curve to be used as at least a portion of the monitoring model for the first set of test data;

selecting an evaluation criteria for the model wherein the evaluation criteria is based on the first ROC;

evaluating the monitoring model of the power plant system or magnetic resonance imaging system based upon the selected evaluation criteria from the first receiver operating characteristic analysis curve; and

monitoring operating condition of the power plant system or magnetic resonance imaging system using the evaluated monitoring model.

2. The method of claim 1, wherein the step of designating at least one of the first plurality of negative data points as the positive data point is performed using a first determinator selected from a trained technician, a statistical analysis of the first set of data, a review of prior training data, or a combination thereof.

3. The method of claim 1, further comprising the steps of:

obtaining a second set of test data from the system through at least one monitoring sensor, the second set of test data including a second plurality of negative data points;

designating at least one of the second plurality of negative data points as a second positive data point;

performing a second receiver operating characteristics analysis on the second set of test data to obtain a second receiver operating characteristics analysis curve for the second set of test data;

selecting a comparison criteria for comparing two or more receiver operating characteristics analysis curves based on the first ROC;

comparing the first receiver operating characteristics analysis curve with the second receiver operating characteristics analysis curve using the selected comparison criteria to select a monitoring model; and

monitoring operating condition of the power plant system or magnetic resonance imaging system using the selected monitoring model.

4. The method of claim 3, wherein the step of designating at least one of the first plurality of negative data points as the first positive data point is performed using a first determinator selected from a trained technician, a statistical analysis of the first set of test data, a review of prior training data, or a combination thereof;

further wherein the step of designating at least one of the second plurality of negative data points as the second positive data point is performed using a second determinator selected from a trained technician, a statistical analysis of the second set of test data, a review of prior training data, or a combination thereof;

wherein the first determinator is different than the second determinator.

5. The method of claim 1, further comprising the steps of:

designating another of at least one of the first plurality of negative data points as a second positive data point;

performing a second receiver operating characteristics analysis on the first set of test data to obtain a second receiver operating characteristics analysis curve for the first set of test data;

selecting a comparison criteria for comparing two or more receiver operating characteristics analysis curves based on the first ROC;

comparing the first receiver operating characteristics analysis curve with the second receiver operating characteristics analysis curve using the selected comparison criteria to select a monitoring model; and

monitoring operating condition of the power plant system or magnetic resonance imaging system using the selected monitoring model.

6. A machine-readable storage medium having stored thereon, computer-executable instructions which when executed by a computer perform a method to evaluate a monitoring model of a power plant system or a magnetic resonance imaging system, the method comprising:

obtaining a first set of test data from the system through at least one monitoring sensor, the first set of test data including a first plurality of negative data points;

designating at least one of the first plurality of negative data points as a first positive data point; and

performing a first receiver operating characteristics (ROC) analysis on the first set of test data to obtain a first receiver operating characteristics analysis curve to be used as at least a portion of the monitoring model for the first set of test data;

selecting an evaluation criteria for the model based on the first ROC;

evaluating the monitoring model of the power plant system or magnetic resonance imaging system based upon selected evaluation criteria from the first receiver operating characteristic analysis curve; and

monitoring operating condition of the power plant system or magnetic resonance imaging system using the evaluated monitoring model.

7. The machine-readable storage of claim 6, wherein the step of designating at least one of the first plurality of negative data points as the first positive data point is performed using a determinator selected from a trained technician, a statistical analysis of the first set of test data, a review of prior training data, or a combination thereof.

8. The machine-readable storage of claim 6, further comprising code sections executable by a machine for causing the machine to perform the steps of:

obtaining a second set of test data from the system through at least one monitoring sensor, the second set of test data including a second plurality of negative data points;

designating at least one of the second plurality of negative data points as a second positive data point;

performing a second receiver operating characteristics analysis on the second set of test data to obtain a second receiver operating characteristics analysis curve for the second set of test data;

selecting a comparison criteria for comparing two or more receiver operating characteristics analysis curves based on the first ROC;

comparing the first receiver operating characteristics analysis curve with the second receiver operating characteristics analysis curve using the selected comparison criteria to select a monitoring model; and

monitoring operating condition of the power plant system or magnetic resonance imaging system using the selected monitoring model.

9. The machine-readable storage of claim 8, wherein the step of designating at least one of the first plurality of negative data points as a the first positive data point is performed using a first determinator selected from a trained technician, a statistical analysis of the first set of test data, a review of prior training data, or a combination thereof;

further wherein the step of designating at least one of the second plurality of negative data points as the second positive data point is performed using a second determinator selected from a trained technician, a statistical analysis of the second set of test data, a review of prior training data, or a combination thereof;

wherein the first determinator is different than the second determinator.

10. The machine-readable storage of claim 6, further comprising code sections executable by a machine for causing the machine to perform the steps of:

designating another of at least one of the first plurality of negative data points as a second positive data point;

performing a second receiver operating characteristics analysis on the first set of test data to obtain a second receiver operating characteristics analysis curve for the first set of test data;

selecting a comparison criteria for comparing two or more receiver operating characteristics analysis curves based on the first ROC;

comparing the first receiver operating characteristics analysis curve with the second receiver operating characteristics analysis curve using the selected comparison criteria to select a monitoring model; and

monitoring operating condition of the power plant system or magnetic resonance imaging system using the selected monitoring model.

11. A computing system for evaluating a monitoring model of a power plant system or a magnetic resonance imaging system, comprising:

means for obtaining a first set of test data from the system through at least one monitoring sensor, the first set of test data including a first plurality of negative data points;

a first determinator for designating at least one of the first plurality of negative data points as a first positive data point; and

means for performing a first receiver operating characteristics (ROC) analysis on the set of test data to obtain a first receiver operating characteristics analysis curve to be used as at least a portion of the monitoring model for the first set of test data;

means for selecting an evaluation criteria for the model based on the first ROC;

means for evaluating the monitoring model of the power plant system or magnetic resonance imaging system based upon selected evaluation criteria from the first receiver operating analysis characteristic curve; and

means for monitoring operating condition of the power plant system or magnetic resonance imaging system using the evaluated monitoring model.

12. The computing system of claim 11, wherein the first determinator is selected from a trained technician, a statistical analysis of the first set of test data, a review of prior training data, or a combination thereof.

13. The computing system of claim 11, further comprising:

means for obtaining a second set of test data from the system through at least one monitoring sensor, the second set of test data including a second plurality of negative data points;

a second determinator for designating at least one of the second plurality of negative data points as a second positive data point;

means for performing a second receiver operating characteristics analysis on the second set of test data to obtain a second receiver operating characteristics analysis curve for the second set of test data;

means for selecting a comparison criteria for comparing two or more receiver operating characteristics analysis curves based on the first ROC;

means for comparing the first receiver operating characteristics analysis curve with the second receiver operating characteristics analysis curve using selected comparison criteria to select a monitoring model; and

means for monitoring operating condition of the power plant system or magnetic resonance imaging system using the selected monitoring model.

14. The computing system of claim 13, wherein the first determinator is selected from a trained technician, a statistical analysis of the first set of test data, a review of prior training data, or a combination thereof;

further wherein the second determinator is selected from a trained technician, a statistical analysis of the second set of test data, a review of prior training data, or a combination thereof;

wherein the first determinator is different than the second determinator.

15. The computing system of claim 11, further comprising:

a determinator for designating another of at least one of the first plurality of negative data points as a second positive data point;

means for performing a second receiver operating characteristics analysis on the first set of test data to obtain a second receiver operating characteristics analysis curve for the first set of test data;

means for selecting a comparison criteria for comparing two or more receiver operating characteristics analysis curves based on the first ROC;

means for comparing the first receiver operating characteristics analysis curve with the second receiver operating characteristics analysis curve using the selected comparison criteria to select a monitoring model; and

means for monitoring operating condition of the power plant system or magnetic resonance imaging system using the selected monitoring model.