System and method using separators for developing training records for use in creating an empirical model of a process

• US 5,752,007 A
• Filed: 03/11/1996
• Issued: 05/12/1998
• Est. Priority Date: 03/11/1996
• Status: Expired due to Term

• Current Assignee: Fisher-Rosemount Systems Incorporated (Emerson Electric Company)
• Original Assignee: Fisher-Rosemount Systems Incorporated (Emerson Electric Company)
• Inventors: Morrison, Steven Michael
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): Fiul, Dan
• Application Number:
  US08/613,913
• Time in Patent Office: 792 Days
• Field of Search: 395/500, 395/2.11, 395/2.41, 395/2.68, 395/21, 364/578, 364/424.088, 364/972.4
• US Class Current: 703/2
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

A device and method develop a training record for use in creating an empirical model of a process from data associated with the process. Data from a continuous process in the form of a continuous file or data from a set of batch files associated with different runs of a batch process is used to develop a file having multiple values for each of a plurality of variables associated with the process. A flag or other indicator is placed within the file to indicate the location of a significant event within the process or the location at which the data from different batch files was concatenated together. A record is then developed from the file by selecting a multiplicity of values within the file. Thereafter, it is determined if any two of the selected values within the record are located within the file at positions which are on different sides of the position indicated by the flag or other indicator. If so, that record is prevented from being used as a training record for training an empirical model of the process.

Claims

1. A device for developing a training record, capable of being used to create a model of a process, from data pertaining to a plurality of variables associated with the process, comprising:

means for creating a file having multiple values for each of the plurality of variables from the data pertaining to the plurality of variables;

means for indicating a position within the file;

means for selecting first and second of the values within the file to define a record;

means for determining if the first value is associated with a position within the file which is on one side of the indicated position and if the second value is associated with a position within the file which is on another side of the indicated position to thereby determine if the first and second selected values are on different sides of the indicated position; and

means responsive to the determining means for preventing the use of the defined record as a training record when the first and second values are on different sides of the indicated position.

2. The device of claim 1, wherein the selecting means selects a value associated with each of the plurality of variables to define the record, the determining means determines if any of the selected values are associated with a position within the file which is on a different side of the indicated position as the position of any other of the selected values and further including means for classifying the defined record as a training record if all of the selected values within the defined record are associated with positions within the file which are on the same side of the indicated position.

3. The device of claim 1, wherein the indicating means includes means for placing a flag within the file at the indicated position.

4. The device of claim 1, wherein the first and second values each have a time-stamp associated therewith, the indicating means includes means for storing a time-stamp associated with the indicated position and the determining means includes means for comparing the time-stamp of the indicated position with the time-stamps associated with the first and the second values.

5. The device of claim 1, wherein the indicating means includes means for placing a set of fictitious data values within the file at the indicated position and the determining means includes means for checking whether one of the first and the second values comprises one of the fictitious data values.

6. The device of claim 1, wherein the process is a batch-type process, the creating means includes means for concatenating data values from two batch files associated with different batch runs of the batch process and the indicating means indicates the position between the data from the two batch files.

7. The device of claim 1, wherein the indicating means indicates a position within the file associated with a significant event within the process.

8. A device for creating a model of a process from a set of data values for each of a plurality of variables associated with the process, comprising:

means for creating a file having a multiplicity of values for each of a sub-set of the plurality of variables from the set of data values for each of the plurality of variables;

means for indicating a position within the file;

means for selecting a first value within the file and a second value within the file;

means for determining if the first value is at a position within the file which is on one side of the indicated position and if the second value is at a position within the file which is on the other side of the indicated position;

means for developing a training record from the first and second values if the position of the first value is on the same side of the indicated position as the position of the second value;

means for training a model of the process using the training record.

9. The device of claim 8, wherein the indicating means includes means for placing a flag within the file at the indicated position.

10. The device of claim 8, wherein the indicating means includes means for storing a time-stamp associated with the indicated position, wherein the first and second values each have a time-stamp associated therewith and wherein the determining means includes means for comparing the time-stamp associated with the indicated position with the time-stamps associated with the first and second values.

11. The device of claim 8, wherein the indicating means includes means for placing a set of fictitious data values within the file at the indicated position and the determining means includes means for ascertaining whether one of the first and second values comprises one of the fictitious data values.

12. The device of claim 8, wherein the process is a batch process, the set of data values for each of a plurality of variables associated with the process comprises a multiplicity of batch files associated with different runs of the batch process, the creating means includes means for concatenating data values from the multiplicity of batch files and the indicating means indicates the position within the file between the values of two of the multiplicity of concatenated batch files.

13. The device of claim 8, wherein the indicating means indicates a position within the file associated with the location of a significant event within the process.

14. A method of developing a training record for use in creating a model for a process which has a plurality of variables associated therewith, including the steps of:

creating a file having multiple values for each of the plurality of variables;

indicating a position within the file;

selecting multiplicity of values from the file to define a record;

determining if a first one of the selected values within the record is associated with a position within the file which is one side of the indicated position and if a second one of the selected values within the record is associated with a position within the file which is on another side the indicated position to determine if the first and second selected values are on the same side of the indicated position; and

preventing the use of the defined record as a training record if the first and second selected values are not on the same side of the indicated position.

15. The method of claim 14, further including the step of classifying the record as a training record if all of the selected values within the record are associated with positions within the file which are on one side of the indicated position.

16. The method of claim 14, wherein the step of indicating includes the step of placing a flag within the file at the indicated position.

17. The method of claim 14, wherein each of the selected values includes a time-stamp associated therewith, the step of indicating includes the step of creating a time-stamp for the indicated position and the step of determining includes the step of comparing the time-stamp of the indicated position with the timestamps associated with the first and second selected values.

18. The method of claim 14, wherein the step of indicating includes the step of placing a set of fictitious data values within the file at the indicated position and the step of determining includes the step of ascertaining whether any one of the selected values comprises one of the fictitious data values.

19. The method of claim 14, wherein the process is a batch process, the step of creating a file includes the step of concatenating data values associated with two batch runs of the batch process and the step of indicating a position within the file includes the step of indicating the position between the data values associated with the two batch runs of the batch process.

20. The method of claim 14, wherein the step of indicating a position within the file includes the step of indicating the location of a significant event within the process.

21. A method of creating a model of a process having a plurality of variables associated therewith, including the steps of;

developing a set of data values for each of the variables;

creating a file having a multiplicity of values for each of the variables from the set of data values;

indicating a position within the file;

selecting a first value within the file and a second value within the file;

determining if the first value is at a position within the file which is on one side of the indicated position and if the second value is at a position within the file which is on the other side of the indicated position;

using the first and second values to develop a training record if the position of the first value is on the same side of the indicated position as the position of the second value;

creating a model of the process using the training record.

22. The method of claim 21, wherein the step of indicating includes the step of placing a flag within the file at the indicated position.

23. The method of claim 21, wherein the step of indicating includes the step of creating a time-stamp for the identified position, wherein the first and second values have time-stamps associated therewith and wherein the step of determining includes the step of comparing the time-stamp of the identified position with the time-stamps associated with the first and second values.

24. The method of claim 21, wherein the step of indicating includes the step of placing a set of fictitious data values within the file at the indicated position and the step of determining includes the step of ascertaining whether one of the first and second values comprises one of the fictitious data values.

25. The method of claim 21, wherein the process is a batch process, the creating means includes means for concatenating data values from a multiplicity of batch files each of which is associated with a run of the batch process and the indicating means indicates the position within the file between the data values of two of the multiplicity of concatenated batch files.

26. The method of claim 21, wherein the step of indicating a position within the file includes the step of indicating a location of a significant event within the process.