Estimation of abnormal sensors

• US 10,082,787 B2
• Filed: 08/28/2015
• Issued: 09/25/2018
• Est. Priority Date: 08/28/2015
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Hara, Satoshi
• Primary Examiner(s): Desta, Elias
• Application Number:
  US14/839,272
• Publication Number:
  US 20170060124A1
• Time in Patent Office: 1,124 Days
• CPC Class Codes: G05B 23/0221 Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods

Abstract

Provided is an estimation apparatus including a target data acquiring section operable to acquire target data serving as an examination target, the target data output by a plurality of sensors, a calculating section operable to calculate, for each of a plurality of sensor groups that each include two sensors among the plurality of sensors, a degree of outlierness of the target data relative to a reference data distribution of output from the sensor group, and an estimating section operable to estimate a sensor among the plurality of sensors to be a source of outlierness, based on a calculation result of the calculating section.

Claims

1. An apparatus comprising:

a target data acquiring section operable to acquire target data serving as an examination target, the target data output by a plurality of sensors;

a calculating section operable to calculate, for each of a plurality of sensor groups that each include at least two sensors among the plurality of sensors, a degree of outlierness of the target data relative to a reference data distribution of output from the sensor group; and

an estimating section operable to estimate at least one sensor among the plurality of sensors to be a source of outlierness, based on a comparison of the degrees of outlierness of the sensor groups that include the at least one sensor to the degrees of outlierness of the sensor groups that lack the at least one sensor.

2. The apparatus according to claim 1, wherein the calculating section is further operable to calculate, for each of the plurality of sensors, the degree of outlierness of each sensor group including the sensor, and the estimating section is further operable to calculate, for each sensor, a degree of association with which the sensor is associated with outlierness, from the at least two degrees of outlierness calculated for the groups including the sensor, and estimates whether the sensor is a source of outlierness based on the degree of association calculated for each sensor.

3. The apparatus according to claim 2, wherein the estimating section is further operable to estimate each sensor having a degree of association that is greater than or equal to a reference value relative to be a source of outlierness.

4. The apparatus according to claim 2, wherein the estimating section is further operable to sequentially select sensors to be sources of outlierness, in order of degree of association beginning with a sensor having a highest degree of association.

5. The apparatus according to claim 4, wherein the estimating section is further operable to estimate that sensors that have not yet been selected are not sensors that are sources of outlierness, if a difference between the degree of association of a most recently selected sensor and a next highest degree of association is greater than or equal to a reference difference.

6. The apparatus according to claim 4, wherein the estimating section is further operable to sequentially select a set number of sensors to be sources of outlierness.

7. The apparatus according to claim 4, wherein the estimating section is further operable to calculate, for each sensor that has not been selected, the degree of association from the degrees of outlierness calculated for sensor groups that do not include sensors that have already been selected.

8. The apparatus according to claim 4, wherein the estimating section is further operable to determine whether to further select a sensor, based on a total or an average of the degrees of outlierness calculated for sensor groups that do not include sensors that have already been selected.

9. The apparatus according to claim 2, wherein the estimating section is further operable to estimate that each sensor having a degree of association less than a reference value is not a source of outlierness.

10. The apparatus according to claim 2, wherein the estimating section is further operable to sequentially select sensors not to be sources of outlierness, in order of degree of association beginning with a sensor having the lowest degree of association.

11. The apparatus according claim 10, wherein the estimating section is further operable to estimate that sensors that have not yet been selected are sensors that are sources of outlierness, if a difference between the degree of association of the most recently selected sensor and a next lowest degree of association of is greater than or equal to a reference difference.

12. The apparatus according to claim 9, wherein the estimating section is further operable to sequentially select a set number of sensors not to be sources of outlierness, in order of degree of association beginning with a sensor having a lowest degree of association.

13. The apparatus according to claim 10, wherein the estimating section is further operable to calculate, for each sensor that has not been selected, the degree of association from the degrees of outlierness calculated for sensor groups that do not include sensors that have already been selected.

14. The apparatus according to claim 1, further comprising:

a reference data acquiring section operable to acquire reference data serving as a reference for output of the plurality of sensors; and

a distribution generating section operable to generate, for each of the plurality of sensor groups, the reference data distribution of output from the sensor group based on the reference data.

15. The apparatus according to claim 1, wherein the calculating section is further operable to calculate the degree of outlierness for each sensor group.

16. An apparatus comprising:

a reference data acquiring section operable to acquire reference data serving as a reference for output of a plurality of sensors;

a distribution generating section operable to generate, for each of a plurality of sensor groups that each include at least two sensors among the plurality of sensors, a reference data distribution of output from the sensor group based on the reference data; and

an output section that operable to output the reference data distribution generated for each of the sensor groups, to be used by an apparatus that calculates, for each of the plurality of sensor groups, a degree of outlierness of target data serving as an examination target output from the plurality of sensors relative to the reference data distribution, and estimates at least one sensor among the plurality of sensors to be a source of outlierness, based on a comparison of the degrees of outlierness of the sensor groups that include the at least one sensor to the degrees of outlierness of the sensor groups that lack the at least one sensor.

17. The apparatus according to claim 16, wherein the distribution generating section is further operable to generate, for each of the plurality of sensor groups that are each formed by selecting two sensors from among the plurality of sensors, a reference data distribution of output from the sensor group.

18. A method comprising:

acquiring target data serving as an examination target, the target data output by a plurality of sensors;

calculating, for each of a plurality of sensor groups that each include at least two sensors among the plurality of sensors, a degree of outlierness of the target data relative to a reference data distribution of output from the sensor group; and

estimating at least one sensor among the plurality of sensors to be a source of outlierness, based on a comparison of the degrees of outlierness of the sensor groups that include the at least one sensor to the degrees of outlierness of the sensor groups that lack the at least one sensor.

19. A computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform operations comprising:

acquiring target data serving as an examination target, the target data output by a plurality of sensors;

calculating, for each of a plurality of sensor groups that each include at least two sensors among the plurality of sensors, a degree of outlierness of the target data relative to a reference data distribution of output from the sensor group; and

estimating at least one sensor among the plurality of sensors to be a source of outlierness, based on a comparison of the degrees of outlierness of the sensor groups that include the at least one sensor to the degrees of outlierness of the sensor groups that lack the at least one sensor.

20. A computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform operations comprising:

acquiring reference data serving as a reference for output of a plurality of sensors;

generating, for each of a plurality of sensor groups that each include at least two sensors among the plurality of sensors, a reference data distribution of output from the sensor group based on the reference data; and

outputting the reference data distribution generated for each of the sensor groups, the output reference data distribution to be used to calculate, for each of the plurality of sensor groups, a degree of outlierness of target data serving as an examination target output from the plurality of sensors relative to the reference data distribution, and to estimate at least one sensor among the plurality of sensors to be a source of outlierness, based on a comparison of the degrees of outlierness of the sensor groups that include the at least one sensor to the degrees of outlierness of the sensor groups that lack the at least one sensor.