SYSTEM AND METHODS FOR ADDRESSING DATA QUALITY ISSUES IN INDUSTRIAL DATA

• US 20160178414A1
• Filed: 12/17/2014
• Published: 06/23/2016
• Est. Priority Date: 12/17/2014
• Status: Abandoned Application

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Saha, Angshuman, Bhaskar, Nitika, Lindenberg, Janna, Vagliani, Richard, Dasaratha, Sridhar
• Application Number:
  US14/574,234
• Publication Number:
  US 20160178414A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G01D 21/00 Measuring or testing not otherwise provided for
• CPC Class Codes: G05B 23/0216 Human interface functionality, e.g. monitoring system providing help to the user in the selection of tests or in its configuration

Abstract

Embodiments allow data cleaning of industrial data gathered from at least one sensor. The data cleaning utilizes a workflow that defines at least one cleaning step to be performed. Each cleaning step comprises detecting defects based on at least one constraint such as various models and/or statistics. Potential defects are presented to a user for feedback. The data is cleaned based on the feedback. Multiple copies of the data are stored to track all the various cleaning choices. All choices can be rolled back at will so that cleaning decisions made can be eliminated and different choices applied. Intermediate data is captured to allow reporting and auditing of the cleaning process.

Claims

1. A device comprising at least one hardware implemented module configured to at least:

retrieve time series data measured from at least one sensor;

retrieve proximity data about the at least one sensor, the proximity data comprising;

sensor ID metadata for the at least one sensor;

or

sensor environment metadata for the at least one sensor;

or

both the sensor ID metadata for the at least one sensor and the sensor environment metadata for the at least one sensor;

detect a first set of defects in the retrieved time series data using constraints based upon at least one of;

at least one stored model of the at least one sensor, a location for the at least one sensor, or both;

or

a statistical model;

or

other constraints that relate to the time series data or proximity data;

present, via a user interface (UI), information relating to the first set of defects defects and receive, via the UI, feedback about the first set of defects;

clean the time series data based on the feedback;

capture information to allow reversal of changes in whole or in part made to the time series data based on the feedback.

2. The device of claim 1 wherein the information related to the first set of defects presented via the UI is selected by the device automatically based on the proximity data or other relevant domain information.

3. The device of claim 1, wherein the hardware implemented module is further configured to retrieve a work flow describing a series of cleaning operations to be applied to the time series data, the cleaning operations comprising:

at least one defect detection module to detect a set of defects in the time series data;

at least one visualization module to present information relating to detected defects and to receive feedback regarding the defects;

at least one data cleaning module to clean the time series data based on the received feedback; and

a versioning module to capture multiple versions of the time series data.

4. The device of claim 3, wherein the hardware implemented module is further configured to allow a user to modify the workflow by adding or deleting cleaning operations.

5. The device of claim 3, wherein the versioning module captures information allowing the multiple versions of the time series data to be created.

6. The device of claim 3, wherein the versioning module allows any changes to the time series data to be reversed in whole or in part.

7. The device of claim 3, wherein:

the at least one defect detection module detects the first set of defects;

the at least one visualization module presents, via a UI information relating to the detected defects and receive, via the UI, feedback about the first set of defects;

the at least one data cleaning module cleans the time series data based on the feedback; and

the versioning module captures information to allow reversal of changes in whole or in part made to the time series data based on the feedback.

8. A method performed by a device to clean time series data, the method comprising:

retrieving time series data measured from at least one sensor;

retrieving proximity data about the at least one sensor, the proximity data comprising sensor ID metadata or sensor environment metadata or both; and

performing at least one cleaning operation, each cleaning operation comprising;

detecting a set of defects in the retrieved time series data using at least one constraint based on any combination of;

the proximity data;

a statistical model;

a model relating to the proximity data;

or

a characteristic of the time series data or proximity data;

presenting, via a user interface (UI), information relating to the detected defects and receiving via the UI, feedback about the first set of defects;

cleaning the time series data based on the feedback; and

capturing information to allow reversal of changes in whole or in part made to the time series data.

9. The method of claim 8, further comprising performing multiple cleaning operations that result in multiple versions of the time series data.

10. The method of claim 8, wherein each cleaning operation further comprises capturing information for reporting.

11. The method of claim 10, wherein reporting comprises intermediate summaries that establish a clear audit trail of the data cleaning process.

12. The method of claim 10, wherein information captured for reporting and information captured to allow reversal of changes is the same set of information.

13. A computer storage medium comprising computer executable instructions that when executed configure a device to at least:

retrieve time series data measured from at least one sensor;

retrieve proximity data about the at least one sensor, the proximity data comprising sensor ID metadata or sensor environment metadata or both; and

perform at least one cleaning operation, each cleaning operation configuring the device to at least;

detect a set of defects in the retrieved time series data using at least one constraint based on any combination of;

the proximity data;

a statistical model;

a model relating to the proximity data;

or

a characteristic of the time series data or proximity data;

present, via a user interface (UI), information relating to the detected defects and receive, via the UI, feedback about the first set of defects;

clean the time series data based on the feedback; and

capture information to allow reversal of changes in whole or in part made to the time series data.

14. The computer storage medium of claim 13, further comprising instructions to configure the device to retrieve a workflow describing the at least one cleaning operation.

15. The computer storage medium of claim 13, further comprising instructions to configure the device to perform multiple cleaning operations that result in multiple versions of the time series data.

16. The computer storage medium of claim 13, wherein each cleaning operation further configures the device to capture information for reporting.

17. The computer storage medium of claim 16, wherein reporting comprises intermediate summaries that establish a clear audit trail of the data cleaning process.

18. The computer storage medium of claim 13, wherein information captured for reporting and information captured to allow reversal of changes is the same set of information.

19. The computer storage medium of claim 13, wherein the at least one constraint is based on a model of physical characteristics of the sensor and the environment of the sensor.

20. The computer storage medium of claim 13, wherein the at least one constraint is based on prior data.