Method and system for implementing histogram-based alarms in a production system
View Patent ↗A method and system for implementing histogram-based alarms in a production system. Specifically, the method and system disclosed herein entail generating histograms overlaid with frequency (i.e., number of data samples) based class policies to serve as data classifiers for measurements, metrics, or information produced by physical and/or logical sensors. The accurateness of histograms to represent distributions of data, however, may depend on certain constraints—one of which may be the number of intervals or buckets employed. Therefore, disclosed herein is also a methodology for identifying an optimal number of buckets, for particular sensor specific datasets, based on a maximum samples trend and an empty buckets trend associated with the datasets.
1. A method for implementing distribution-based data classification, comprising:
receiving a new data sample (NDS) and a sensor identifier (SID);
obtaining a sampled dataset (SD) associated with the SID;
determining an optimal number of buckets (ONB) for the SD;
generating a distribution diagram (DD) representative of the SD using the ONB;
generating a sensor bucket classifier (SBC) by applying at least one class policy (CP) to the DD;
determining a classification for the NDS using the SBC; and
issuing an alarm appropriate to the classification,
wherein determining the ONB for the SD, comprises:
selecting a bucket list (BL) based on a cardinality of the SD;
generating a maximum samples dataset (MSD) using the BL;
deriving a first set of regression function parameters (RFPs) representative of the MSD;
generating an empty buckets dataset (EBD) using the BL;
deriving a second set of RFPs representative of the EBD; and
determining the ONB based on the first set of RFPs and the second set of RFPs.
2. The method of claim 1 , wherein generating the MSD using the BL, comprises:
for each number of buckets specified in the BL:
generating a histogram representative of the SD, wherein the histogram comprises a set of buckets, wherein a cardinality of the set of buckets equals the number of buckets;
identifying a bucket of the set of buckets associated with a highest frequency; and
setting the highest frequency as an element of the MSD that corresponds to the number of buckets.
3. The method of claim 1 , wherein generating the EBD using the BL, comprises:
for each number of buckets specified in the BL:
generating a histogram representative of the SD, wherein the histogram comprises a set of buckets, wherein a cardinality of the set of buckets equals the number of buckets;
identifying a subset of the set of buckets associated with a frequency of zero; and
setting a cardinality of the subset of the set of buckets as an element of the EBD that corresponds to the number of buckets.
4. The method of claim 1 , wherein the first set of RFPs comprises arguments to one selected from a group consisting of a linear regression function and an exponential decay regression function.
5. The method of claim 1 , wherein the second set of RFPs comprises arguments to one selected from a group consisting of a linear regression function and an exponential growth regression function.
6. The method of claim 1 , wherein the ONB relates to a point of intersection between a first regression function best fitting the MSD and a second regression function best fitting the EBD.
7. The method of claim 1 , wherein the DD is a histogram, wherein the histogram comprises a set of buckets, wherein a cardinality of the set of buckets equals the ONB.
8. The method of claim 1 , wherein the classification is one selected from a group consisting of a normal classification, a warning classification, and a critical classification.
9. A system, comprising:
a computer processor;
a data classifier (DC) executing on the computer processor, and programmed to:
receive a new data sample (NDS) and a sensor identifier (SID);
obtain a sampled dataset (SD) associated with the SID;
obtain an optimal number of buckets (ONB) for the SD from a bucket optimizer (BO), wherein determining the ONB for the SD, comprises:
selecting a bucket list (BL) based on a cardinality of the SD;
generating a maximum samples dataset (MSD) using the BL;
deriving a first set of regression function parameters (RFPs) representative of the MSD;
generating an empty buckets dataset (EBD) using the BL;
deriving a second set of RFPs representative of the EBD; and
determining the ONB based on the first set of RFPs and the second set of RFPs;
generate a distribution diagram (DD) representative of the SD using the ONB;
generate a sensor bucket classifier (SBC) by applying at least one class policy (CP) to the DD; and
determine a classification for the NDS using the SBC,
a sensor operatively connected to the computer processor and programmed to monitor a portion of a production system (PS), wherein the sensor is one selected from a group consisting of a physical sensor mounted on the PS and a logical sensor executing on the PS, and
an alarm generator (AG) executing on the computer processor and operatively connected to the DC, wherein the AG is programmed to issue an alarm appropriate to the classification.
10. The system of claim 9 , wherein after the alarm is issued, the PS is programmed to:
receive the alarm, wherein the alarm comprises the SID and a class label associated with the classification;
identify the sensor as being associated with the SID;
determine that the sensor is responsible for monitoring the portion of the PS; and
adjust, based on the class label, a factor that controls the portion of the PS.
11. The system of claim 9 , further comprising:
an analysis system (AS) comprising the computer processor and a data evaluation agent (DEA) executing thereon,
wherein the DEA comprises the DC, the BO, and the AG.
12. The system of claim 10 , further comprising:
a sensor operatively connected to the AS and programmed to monitor a portion of a production system (PS),
wherein the sensor is one selected from a group consisting of a physical sensor mounted on the PS and a logical sensor executing on the PS.
13. A non-transitory computer readable medium (CRM) comprising computer readable program code, which when executed by a computer processor, enables the computer processor to:
receive a new data sample (NDS) and a sensor identifier (SID);
obtain a sampled dataset (SD) associated with the SID;
determine an optimal number of buckets (ONB) for the SD;
generate a distribution diagram (DD) representative of the SD using the ONB;
generate a sensor bucket classifier (SBC) by applying at least one class policy (CP) to the DD;
determine a classification for the NDS using the SBC; and
issue an alarm appropriate to the classifications:
wherein to determine the ONB, the computer readable program code, which when executed by the computer processor, enables the computer processor to:
select a bucket list (BL) based on a cardinality of the SD;
generate a maximum samples dataset (MSD) using the BL;
derive a first set of regression function parameters (RFPs) representative of the MSD;
generate an empty buckets dataset (EBD) using the BL;
derive a second set of RFPs representative of the EBD; and
determine the ONB based on the first set of RFPs and the second set of RFPs.
14. The non-transitory CRM of claim 13 , wherein to generate the MSD, further comprises computer readable program code, which when executed by the computer processor, enables the computer processor to:
for each number of buckets specified in the BL:
generate a histogram representative of the SD, wherein the histogram comprises a set of buckets, wherein a cardinality of the set of buckets equals the number of buckets;
identify a bucket of the set of buckets associated with a highest frequency; and
set the highest frequency as an element of the MSD that corresponds to the number of buckets.
15. The non-transitory CRM of claim 13 , wherein to generate the EBD, further comprises computer readable program code, which when executed by the computer processor, enables the computer processor to:
for each number of buckets specified in the BL:
generate a histogram representative of the SD, wherein the histogram comprises a set of buckets, wherein a cardinality of the set of buckets equals the number of buckets;
identify a subset of the set of buckets associated with a frequency of zero; and
set a cardinality of the subset of the set of buckets as an element of the EBD that corresponds to the number of buckets.
16. The non-transitory CRM of claim 13 , wherein the ONB relates to a point of intersection between a first regression function best fitting the MSD and a second regression function best fitting the EBD.
17. The non-transitory CRM of claim 13 , wherein the DD is a histogram, wherein the histogram comprises a set of buckets, wherein a cardinality of the set of buckets equals the ONB.