System and method for selection of prediction tools

• US 20070288414A1
• Filed: 06/07/2006
• Published: 12/13/2007
• Est. Priority Date: 06/07/2006
• Status: Active Grant

• Current Assignee: GM Global Technology Operations LLC (General Motors Company)
• Original Assignee: Gem Global Technology Operations, Inc.
• Inventors: Barajas, Leandro G., Bandyopadhyay, Pulak, Xiao, Guoxian
• Granted Patent: US 7,558,771 B2
• Time in Patent Office: Days
• US Class Current: 706/46
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

A system and method for data analysis are disclosed. Data analysis may include a step of filtering the data to produce filtered data. The method may include processing a plurality of prediction algorithms to produce prediction values, the prediction values having associated historical and expected prediction confidence intervals. The method may also include evaluating performance of the prediction algorithms to generate performance indexes, the performance indexes having associated index confidence intervals. The method may also include generating relevance values of the prediction algorithms based on the performance indexes, and index confidence intervals. The method may further include applying the relevance values and prediction confidence intervals to determine how to combine prediction values, and applying multivariable data fusion to combine the prediction values. The form of output of the data analysis may be chosen from a list of output options, including predictions, reports, warnings and alarms, and other forms of reporting.

Claims

1. A method for data analysis, the method comprising:

processing data with a plurality of prediction algorithms to produce prediction values, the prediction values having associated prediction confidence intervals;

evaluating historical and expected performance of the prediction algorithms to generate performance indexes, the performance indexes having associated index confidence intervals;

generating relevance values of the prediction algorithms based on the performance indexes and index confidence intervals;

applying the relevance values and prediction confidence intervals to determine how to combine prediction values;

applying multivariable data fusion to combine the prediction values; and

producing output.

2. The method of claim 1, wherein processing data to produce prediction values comprises:

processing data with a plurality of prediction algorithms to produce prediction functions, the prediction functions having associated prediction confidence intervals; and

evaluating the prediction functions to produce associated values for predictions and indexes.

3. The method of claim 1, further comprising:

filtering data to produce filtered data;

wherein processing a plurality of prediction algorithms to produce prediction values comprises;

applying the plurality of prediction algorithms to the filtered data;

4. The method of claim 3, wherein filtering data comprises:

executing a filtering algorithm from a list of filtering options comprising removing irrelevant and redundant data, managing data integrity, and filtering data according to data type classification.

5. The method of claim 3, wherein filtering data comprises:

executing a filtering algorithm from a list of filtering options comprising filtering multiple episode events, filtering fault data according to real-time schedule detection, and filtering fault data according to fault type classification.

6. The method of claim 1, wherein processing a plurality of prediction algorithms comprises:

making a selection from a list of prediction options comprising a time-frequency analyzer, a wavelet analyzer, a curve fitting tool, an auto-regressive model, a neural networks model, a temporal data mining method, a frequent episode discovery method, and a Bayesian reasoning model.

7. The method of claim 1, wherein producing output comprises:

choosing from a list of output options comprising a short term prediction report, a long term prediction report, a prognostics report, a fault cross-correlation report, a maintenance prioritization report, fault trend estimation, maintenance warning and alarms, knowledge extraction, a fault forecasting report, a historical prediction-performance report, and a prediction confidence metric report.

8. The method of claim 1, wherein a first prediction algorithm produces a first prediction value of a first particular type and having a first magnitude of error and a second prediction algorithm produces a second prediction value of a second particular type and having a second magnitude of error, and applying multivariable sensor fusion comprises:

combining the first prediction value of the first particular type and the second prediction value of the second particular type to produce a combined prediction value having a combined magnitude of error less than the first magnitude of error and less than the second magnitude of error.

9. The method of claim 1, further comprising:

correcting for systematic errors in the prediction values.

10. The method of claim 9, wherein correcting for systematic errors comprises:

applying at least one of a proportional-integral-derivative and a non-linear controller.

11. A method for fault data analysis, the method comprising:

acquiring fault data using a real-time data acquisition system;

filtering the fault data to produce filtered fault data;

selecting a plurality of fault prediction methods;

applying the plurality of fault prediction methods to the filtered fault data to obtain a plurality of prediction values;

applying multivariable sensor fusion to combine the plurality of prediction values; and

producing output.

12. The method of claim 11, further comprising:

correcting the prediction values according to historical data and to expected model prediction accuracy.

13. The method of claim 11, wherein filtering fault data comprises:

executing a filtering algorithm from a list of filtering options comprising removing irrelevant time outs and idle time events from data, removing irrelevant and redundant data, managing data integrity, filtering multiple episode events, filtering data according to real-time schedule detection, and filtering data according to fault type classification.

14. The method of claim 11, wherein selecting a plurality of prediction methods comprises:

making selections from a list of prediction options comprising a time-frequency analyzer, a wavelet analyzer, a curve fitting tool, an auto-regressive model, a neural networks model, a temporal data mining method, a frequent episode discovery method, and a Bayesian reasoning model.

15. The method of claim 11, wherein the plurality of prediction values comprise a discrete-valued first prediction value having a first magnitude of error and a continuous-valued second prediction value having a second magnitude of error, and applying multivariable sensor fusion comprises:

combining the discrete-valued first prediction value and the continuous-valued second prediction value to produce a combined prediction value having a combined magnitude of error less than the first magnitude of error and less than the second magnitude of error.

16. The method of claim 11, wherein producing output comprises:

choosing from a list of output options comprising a short term prediction report, a long term prediction report, a prognostic report, a fault cross-correlation report, a maintenance prioritization report, fault trend estimation, maintenance warning and alarms, knowledge extraction, a fault forecasting report, a historical prediction performance report, a prediction confidence metric report, and a maintenance error code database correlation report.

17. A system for data analysis, comprising:

a filtering module for filtering data to produce filtered data;

a selection module for selecting a first prediction method and a second prediction method;

an application module for applying the first prediction method to the filtered data to obtain a first prediction value, and for applying the second prediction method to the filtered data to obtain a second prediction value;

a fusion module for applying multivariable sensor fusion to combine the first prediction value and the second prediction value; and

an output module for producing output.

18. The system of claim 17, wherein the filtering module is further for removing irrelevant and redundant data, managing data integrity, and filtering data according to data type classification.

19. The system of claim 17, wherein the selection module is further for selecting the first prediction method and the second prediction method from among a choice of a time-frequency analyzer, a wavelet analyzer, a curve fitting tool, an auto-regressive model, a neural networks model, a temporal data mining method, a frequent episode discovery method, and a Bayesian reasoning model.

20. The system of claim 17, wherein the first prediction value is a discrete-valued first prediction value having a first magnitude of error and the second prediction is a continuous-valued second prediction value having a second magnitude of error, and the fusion module is further for combining the discrete-valued first prediction value and the continuous-valued second prediction value to produce a combined prediction value having a combined magnitude of error less than the first magnitude of error and less than the second magnitude of error.