Curve fitting for signal estimation, prediction, and parametrization

• US 7,324,924 B2
• Filed: 01/27/2006
• Issued: 01/29/2008
• Est. Priority Date: 01/27/2006
• Status: Active Grant

• Current Assignee: GM Global Technology Operations LLC (General Motors Company)
• Original Assignee: GM Global Technology Operations Incorporated (General Motors Company)
• Inventors: Barajas, Leandro G., Xiao, Guoxian
• Primary Examiner(s): Cosimano; Edward R
• Application Number:
  US11/341,782
• Publication Number:
  US 20070179753A1
• Time in Patent Office: 732 Days
• Field of Search: 738/663, 324/500, 324/511, 340/500, 340/540, 340/679, 340/680, 340/691.1, 340/691.6, 702/127, 702/179, 702/180, 702/182, 702/183, 702/184, 702/185, 702/189, 715/500, 715/502, 715/512, 708/200, 708/208, 708/290
• US Class Current: 702/189
• CPC Class Codes: G05B 19/4183 characterised by data acquisition, e.g. workpiece identification
• CPC Class Codes: G05B 2219/32194 Quality prediction
• CPC Class Codes: G05B 2219/32371 Predict failure time by analysing history fault logs of same machines in databases
• CPC Class Codes: G05B 2219/34477 Fault prediction, analyzing signal trends
• CPC Class Codes: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

A method and system of automatically curve fit data series to generate data signal characterization and prediction is disclosed. The method includes receiving input data including a plurality of data series and set of input parameters. The method also includes preprocessing the data series to generate preprocessed data according to a plurality of input parameters. The method further includes performing sorting and prioritization of the data series and curve fitting by a plurality of models and search and optimization algorithms on the smoothed or raw data series according to a plurality of input parameters to generate text and graphic output, and storing text and graphic output. The system includes modules for carrying out steps of the method.

Claims

1. A method of a computer implemented process for automatically curve fitting data series to generate data signal characterization, the computer including a processor and a memory, the method comprising:

receiving from a data base to store in the memory input data including a plurality of data series and a set of input parameters to be utilized in subsequent preprocessing and curve fitting processes, the plurality of data series including performance indicators including at least one of machine fault codes, machine downtimes and machine maintenance data and the set of input parameters configurable by a user;

preprocessing by the processor the data series to generate preprocessed data according to a subset of the set of input parameters, the preprocessing configured to prepare the data series for a subsequent curve fitting process, the preprocessing comprising;

sorting by the processor the plurality of data series according to an input parameter to generate an ordered series; and

prioritizing by the processor the ordered data series according to an input parameter to generate an ordered and prioritized series;

performing curve fitting by the processor, the curve fitting by a plurality of curve fitting algorithms on the preprocessed data including the ordered and prioritized series according to a subset of the set of input parameters to generate text and graphic output to represent data signal characterizations including at least one of goodness of fit and prediction values, the text and graphic output configurable by the user; and

storing in memory the text and graphic output for subsequent display of representations of at least one of goodness of fit to at least one curve of the plurality of curve fitting algorithms and prediction values for future performance indicators including at least one of machine fault codes, machine downtimes and machine maintenance data;

wherein the data series comprise fault code data signals generated by machines in a production facility and wherein the data signal characterizations including at least one of goodness of fit and prediction values are utilized to generate work orders.

2. The method as recited in claim 1 wherein preprocessing comprises weighting the plurality of data series according to an input parameter to generate preprocessed data that is an ordered, prioritized and weighted series and wherein performing curve fitting further comprises:

performing curve fitting by the processor, the curve fitting by a plurality of curve fitting algorithms on the preprocessed data including the ordered, prioritized and weighted senes.

3. The method of claim 1, wherein performing curve fitting comprises performing curve fitting iteratively.

4. The method of claim 1, wherein performing curve fitting comprises performing curve fitting interactively.

5. The method of claim 1, wherein the data series are time series and wherein performing curve fitting on the data series further comprises:

generating by the processor values for evaluation of the goodness of fit of the data series to a plurality of curves of the curve fitting algorithms, the values for evaluation of the goodness of fit being for subsequent display.

6. The method of claim 1, wherein the data series are time series including values associated with past time values, and wherein performing curve fitting on the data series further comprises:

generating by the processor prediction values to extract trend information from the data series for subsequent display.

7. The method as recited in claim 1 wherein preprocessing comprises selecting a data series from the plurality of data series to generate preprocessed data that is a weighted, ordered, prioritized and selected data series and wherein performing curve fitting further comprises:

performing curve fitting by the processor, the curve fitting by a plurality of curve fitting algorithms on the preprocessed data including the weighted, ordered, prioritized and selected data series.

8. The method as recited in claim 7 wherein preprocessing comprises smoothing the selected data series according to an input parameter to generate preprocessed data that is an ordered, prioritized, selected and smoothed data series and wherein performing curve fitting further comprises:

performing curve fitting by the processor, the curve fitting by a plurality of curve fitting algorithms on the preprocessed data including the ordered, prioritized, selected and smoothed data series.

9. A method of a computer implemented process for automatically curve fitting data series to generate data signal characterization, the computer including a processor and a memory, the method comprising:

preprocessing by the processor, the preprocessing configured to prepare the data series including performance indicators including at least one of machine fault codes, machine downtimes and machine maintenance data for a subsequent curve fitting process, the preprocessing comprising;

receiving from a data base to store in the memory input data including a plurality of data series and a set of input parameters to be utilized in subsequent weighting, sorting, prioritizing, selecting, smoothing and curve fitting processes, the plurality of data series including performance indicators and the set of input parameters configurable by a user;

weighting by the processor the plurality of data series according to an input parameter to generate a weighted data series;

sorting by the processor the weighted data series according to an input parameter to generate a sorted data series;

prioritizing by the processor the sorted data series according to an input parameter to generate a prioritized data series;

selecting by the processor a data series from the prioritized data series to generate a selected data series;

smoothing by the processor the selected data series according to an input parameter to generate a smoothed data series; and

processing by the processor comprising;

performing curve fitting by a plurality of curve fitting algorithms on the smoothed data series according to a plurality of input parameters to generate text and graphic output to represent data signal characterizations including at least one of goodness of fit and predictions values, the text and graphic output configurable by the user; and

storing in memory text and graphic output for subsequent display of representations of at least one of goodness of fit to at least one curve of the plurality of curve fitting algorithms and prediction values for future performance indicators including at least one of machine fault codes, machine downtimes and machine maintenance data;

wherein the data series comprise fault code data signals generated by machines in a production facility and wherein the data signal characterizations including at least one of goodness of fit and prediction values are utilized to generate work orders.

10. The method of claim 9, wherein the data series are time series and wherein performing curve fitting on the data series generates values for evaluation of the goodness of fit of the data series to a plurality of curves generated by performing curve fitting on the data series by the processor, the values for evaluation of goodness of fit being for subsequent display.

11. The method of claim 9, wherein the data series are time series including values associated with past time values, wherein performing curve fitting generates prediction values to extract trend information from the data series for subsequent display.

12. The method of claim 9, wherein performing curve fitting comprises performing curve fitting iteratively.

13. The method of claim 9, wherein performing curve fitting comprises performing curve fitting interactively.

14. A computer implemented system for curve fitting data series to generate data signal characterization, the computer including a processor, a memory and modules including instructions and configured to cooperate with the processor, the system comprising:

a preprocessing module configured to prepare the data series including performance indicators for a subsequent curve fitting process, the preprocessing module comprising;

a data input module for receiving input data from a data base to store in the memory, the input data including a plurality of data series and a set of input parameters to be utilized by weighting, sorting, prioritizing, selecting smoothing and curve fitting modules, the plurality of data series including performance indicators including at least one of machine fault codes, machine downtimes and machine maintenance data and the set of input parameters configurable by a user;

a weighting module for weighting the plurality of data series according to an input parameter to generate weighted data series;

a sorting module for sorting the weighted data series according to an input parmeter to generate sorted data series;

a prioritization module for prioritizing the sorted data series according to an input parameter to generate prioritized data series;

a selection module for selecting a data series from the prioritized data series to generate a selected data series;

a smoothing module for smoothing the selected data series according to an input parameter to generate a smoothed data series; and

a processing module comprising;

a curve fitting module for performing curve fitting by a plurality of curve fitting models and curve fitting algorithms on the smoothed data series according to a plurality of input parameters to generate text and graphic output to represent data signal characterizations including at least one of goodness of fit to at least one curve of the plurality of curve fitting algorithms and prediction values for future performance indicators; and

a database module for storing text and graphic output in the memory text and graphic output for subsequent display of representations of at least one of goodness of fit and prediction values including at least one of machine fault codes, machine downtimes and machine maintenance data, the text and graphic output configurable by the user;

wherein the data series comprise fault code data signals generated by machines in a production facility and wherein the data signal characterizations including at least one of goodness of fit and prediction values are utilized to generate work orders.

15. The system of claim 14, wherein the curve fitting module is adapted to perform curve fitting iteratively and interactively.

16. The system of claim 14, wherein the data series are time series including values associated with past time values, wherein the curve fitting module is adapted to generate prediction values to extract trend information from the data series for subsequent display.

17. The system of claim 14, wherein the data series are time series and wherein the curve fitting module is adapted to generate values for goodness of fit of the data series to a plurality of curves generated by performing curve fitting on the data series by the processor, the values for evaluation of the goodness of fit being for subsequent display.