Multi-sensor data summarization

• US 10,332,030 B2
• Filed: 03/02/2016
• Issued: 06/25/2019
• Est. Priority Date: 10/17/2015
• Status: Active Grant

• Current Assignee: TATA Consultancy Services Limited (Tata Sons Pvt Ltd.)
• Original Assignee: TATA Consultancy Services Limited (Tata Sons Pvt Ltd.)
• Inventors: Agarwal, Puneet, Shroff, Gautam, Saikia, Sarmimala, Srinivasan, Ashwin
• Primary Examiner(s): Waldron, Scott A.
• Assistant Examiner(s): Seck, Ababacar
• Application Number:
  US15/058,837
• Publication Number:
  US 20170109653A1
• Time in Patent Office: 1,210 Days
• CPC Class Codes: G06F 16/2228 Indexing structures
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/906 Clustering; Classification
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/025 Extracting rules from data
• CPC Class Codes: H04Q 9/00 Arrangements in telecontrol or telemetry systems for selectively calling a substation from a main station, in which substation desired apparatus is selected for applying a control signal thereto or for obtaining measured values therefrom

Abstract

This disclosure relates generally to multi-sensor data, and more particularly to summarizing multi-sensor data. In one embodiment, the method includes computing plurality of histograms from sensor data associated with a plurality of sensors. The respective histograms of each sensor are clustered into a first plurality of sensor-clusters, and a first set of rules is extracted therefrom. First set of rules defines patterns of histograms of a set of sensors occurring frequently over a time-period. Two or more sensor-clusters from amongst the first plurality of sensor-clusters are merged selectively to obtain a second plurality of sensor-clusters. Second set of rules are extracted from the second plurality of sensor-clusters, and a set of correlated sensors are identified therefrom based on the second set of rules. Third set of rules are extracted from the set of correlated sensors, the third set of rules summarizes the multi-sensor data to represent prominent co-occurring sensor behaviors.

Claims

1. A processor-implemented method for summarizing multi-sensor data associated with a machine or a device to summarize usage or behavior patterns of the machine or device, the method comprising:

computing, via one or more hardware processors, a plurality of histograms from sensor data associated with a plurality of sensors, wherein the sensor data is multi-sensor data received from machine or device and the sensor data is multi-dimensional data, and wherein the histograms are representative of each of the sensors'"'"' behavior for a time period of operation of the machine or the device;

clustering, via the one or more hardware processors and from the plurality of histograms, respective histograms of each of the plurality of sensors to obtain a first plurality of sensor-clusters based on shape of the respective histograms, each sensor-cluster of the first plurality of sensor-clusters comprising a centroid histogram representative of distinct sensor behavior for a distinct sensor of the plurality of sensors;

performing, via the one or more hardware processors, frequent pattern mining on the first plurality of sensor-clusters to extract a first set of rules, wherein a rule of the first set of rules being associated with a set of sensors of the plurality of sensors and comprising a set of sensor-clusters occurring frequently in the first plurality of sensor-clusters over the time period;

merging, via the one or more hardware processors, selectively two or more sensor-clusters from amongst the first plurality of sensor-clusters to obtain a second plurality of sensor-clusters, the two or more sensor-clusters selected corresponding to a sensor of the set of sensors, the two or more sensor-clusters being merged based on two or more rules from amongst the first set of rules associated with the two or more sensor-clusters and a distance measure between the two or more sensor-clusters of the sensor, wherein the two or more sensor-clusters of the sensors are merged based on co-occurrence of one or more other sensors of the plurality of sensors in the two or more sensor-clusters for a same time period;

extracting, via the one or more hardware processors, a second set of rules from the second plurality of sensor-clusters, the second set of rules indicative of distinct sensor behaviors associated with the second plurality of sensor-clusters;

identifying, via the one or more hardware processors, a plurality of sets of correlated sensors from the second plurality of sensor-clusters based on the second set of rules; and

extracting, via the one or more hardware processors, a third set of rules from the one or more sets of correlated sensors, the third set of rules summarizing the multi-sensor data to represent prominent co-occurring sensor behaviors, wherein step by step extraction of the first, second and third set of rules enables summarization of the multi-sensor data to summarize the usage or behavior patterns of the machine or the device.

2. The method of claim 1, wherein the respective histograms associated with a sensor-cluster of the first plurality of sensor-clusters are bounded by a first threshold distance measure.

3. The method of claim 2, wherein the first threshold distance measure includes one of an Euclidean distance, earth-mover distance, Kullback-Leibler divergence, Bhattacharyya distance, Manhattan distance, and Wasserstein metric.

4. The method of claim 1, wherein the clustering is performed based on a Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH) clustering model.

5. The method of claim 1, wherein the second set of rules are extracted by using a frequent pattern mining technique.

6. The method of claim 1, wherein the determination of the co-occurrence of the one or more other sensors, by using a graph based technique, comprises:

representing the first plurality of sensor-clusters and the first set of rules in form of a first graph, the first graph comprising a plurality of nodes and a plurality of edges connecting the plurality of nodes, wherein the plurality of nodes represents the first plurality of sensor-clusters, and the plurality of edges comprises a set of intra-cluster edge and a set of inter-cluster edge, wherein an inter-cluster edge comprises an edge between sensor-clusters associated with a rule of the first set of rules, and an intra-cluster edge comprises an edge between sensor-clusters of a sensor having the distance measure between the sensor-clusters less than a second threshold distance measure; and

identifying the second plurality of sensor-clusters from the first graph, wherein a sensor-cluster of the second plurality of the sensor-clusters is identified from a sub-graph of the first graph, wherein the sub-graph comprises the two or more sensor-clusters associated with the set of sensors, and wherein the two or more sensor-clusters associated with a sensor of the set of sensors in the sub-graph are merged on determination of presence of the intra-cluster edge between the two or more sensor-clusters.

7. The method of claim 1, wherein identifying the plurality of sets of correlated sensors associated with the second plurality of sensor-clusters comprises:

representing the second set of rules and the second plurality of sensor-clusters in form of a second graph, the second graph comprising a plurality of nodes and a plurality of edges connecting the plurality of nodes, wherein the plurality of nodes represents the second plurality of sensor-clusters, and the plurality of edges comprises inter-cluster edges between the second plurality of sensor-clusters; and

identifying the plurality of sets of correlated sensors associated with the second plurality of sensor-clusters from one or more sub-graphs of the second graph, wherein a plurality of unique sensors associated with each of the one or more sub-graphs represents the set of correlated sensors.

8. The method of claim 1, wherein extracting the third set of rules comprises:

applying frequent pattern mining to each of the plurality of sets of correlated sensors to obtain a set of frequent patterns; and

clustering the set of frequent patterns based on mutual overlap between the plurality of frequent patterns to obtain the third set of rules.

9. A computer implemented system for summarizing multi-sensor data associated with a machine or a device to summarize usage or behavior patterns of the machine or the device, the system comprising:

a memory storing instructions; and

one or more hardware processors coupled to said memory, wherein said one or more hardware processors configured by said instructions to;

compute a plurality of histograms from sensor data associated with a plurality of sensors, wherein the sensor data is multi-sensor data received from a machine or a device and the sensor data is multi-dimensional data, and wherein the histograms are representative of each of the sensors'"'"' behavior for a time period of operation of the machine or the device;

cluster, from the plurality of histograms, respective histograms of each of the plurality of sensors to obtain a first plurality of sensor-clusters based on shape of the respective histograms, each sensor-cluster of the first plurality of sensor-clusters comprising a centroid histogram representative of distinct sensor behavior for a distinct sensor of the plurality of sensors,

perform frequent pattern mining on the first plurality of sensor-clusters to extract a first set of rules, wherein a rule of the first set of rules being associated with a set of sensors of the plurality of sensors and comprising a set of sensor-clusters occurring frequently in the first plurality of sensor-clusters over the time period,

merge selectively two or more sensor-clusters from amongst the first plurality of sensor-clusters to obtain a second plurality of sensor-clusters, the two or more sensor-clusters selected corresponding to a sensor of the set of sensors, the two or more sensor-clusters being merged based on two or more rules from amongst the first set of rules associated with the two or more sensor-clusters and a distance measure between the two or more sensor-clusters of the sensor, wherein the two or more sensor-clusters of the sensors are merged based on co-occurrence of one or more other sensors of the plurality of sensors in the two or more sensor-clusters for a same time period,

extract a second set of rules from the second plurality of sensor-clusters, the second set of rules indicative of distinct sensor behaviors associated with the second plurality of sensor-clusters,

identify a plurality of sets of correlated sensors from the second plurality of sensor-clusters based on the second set of rules, and

extract a third set of rules from the one or more sets of correlated sensors, the third set of rules summarizing the multi-sensor data to represent prominent co-occurring sensor behaviors, wherein step by step extraction of the first, second and third set of rules enables summarization of the multi-sensor data to summarize the usage or behavior patterns of the machine or the device.

10. The system of claim 9, wherein the respective histograms associated with a sensor-cluster of the first plurality of sensor-clusters are bounded by a first threshold distance measure.

11. The system of claim 10, wherein the first threshold distance measure includes one of an Euclidean distance, earth-mover distance, Kullback-Leibler divergence, Bhattacharyya distance, Manhattan distance, and Wasserstein metric.

12. The system of claim 9, wherein the one or more hardware processors are further configured by the instructions to perform clustering based on a Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH) clustering model.

13. The system of claim 9, wherein one or more hardware processors are further configured by the instructions to extract the second set of rules by using a frequent pattern mining technique.

14. The system of claim 9, wherein to determine the co-occurrence of the one or more other sensors, by using a graph based technique, the one or more hardware processors are further configured by the instructions to:

represent the first plurality of sensor-clusters and the first set of rules in form of a first graph, the first graph comprising a plurality of nodes and a plurality of edges connecting the plurality of nodes, wherein the plurality of nodes represents the first plurality of sensor-clusters, and the plurality of edges comprises a set of intra-cluster edge and a set of inter-cluster edge, wherein an inter-cluster edge comprises an edge between sensor-clusters associated with a rule of the first set of rules, and an intra-cluster edge comprises an edge between sensor-clusters of a sensor having the distance measure between the sensor-clusters less than a second threshold distance measure; and

identify the second plurality of sensor-clusters from the first graph, wherein a sensor-cluster of the second plurality of the sensor-clusters is identified from a sub-graph of the first graph, wherein the sub-graph comprises the two or more sensor-clusters associated with the set of sensors, and wherein the two or more sensor-clusters associated with a sensor of the set of sensors in the sub-graph are merged on determination of presence of the intra-cluster edge between the two or more sensor-clusters.

15. The system of claim 9, wherein to identify the plurality of sets of correlated sensors associated with the second plurality of sensor-clusters, the one or more hardware processors are further configured by the instructions to:

represent the second set of rules and the second plurality of sensor-clusters in form of a second graph, the second graph comprising a plurality of nodes and a plurality of edges connecting the plurality of nodes, wherein the plurality of nodes represents the second plurality of sensor-clusters, and the plurality of edges comprises inter-cluster edges between the second plurality of sensor-clusters; and

identify the plurality of sets of correlated sensors associated with the second plurality of sensor-clusters from one or more sub-graphs of the second graph, wherein a plurality of unique sensors associated with each of the one or more sub-graphs represents the set of correlated sensors.

16. The system of claim 9, wherein to extract the third set of rules, the one or more hardware processors are further configured by the instructions to:

apply frequent pattern mining to each of the plurality of sets of correlated sensors to obtain a set of frequent patterns; and

cluster the set of frequent patterns based on mutual overlap between the plurality of frequent patterns to obtain the third set of rules.

17. A non-transitory computer-readable medium having embodied thereon a computer program for executing a method for summarizing multi-sensor data associated with a machine or a device to summarize usage or behavior patterns of the machine or the device, the method comprising:

computing a plurality of histograms from sensor data associated with a plurality of sensors, wherein the sensor data is multi-sensor data received from a machine or a device and the sensor data is multi-dimensional data, and wherein the histograms are representative of each of the sensors'"'"' behavior for a time period of operation of the machine or the device;

clustering from the plurality of histograms, respective histograms of each of the plurality of sensors to obtain a first plurality of sensor-clusters based on shape of the respective histograms, each sensor-cluster of the first plurality of sensor-clusters comprising a centroid histogram representative of distinct sensor behavior for a distinct sensor of the plurality of sensors;

performing frequent pattern mining on the first plurality of sensor-clusters to extract a first set of rules, wherein a rule of the first set of rules being associated with a set of sensors of the plurality of sensors and comprising a set of sensor-clusters occurring frequently in the first plurality of sensor-clusters over the time period;

merging selectively two or more sensor-clusters from amongst the first plurality of sensor-clusters to obtain a second plurality of sensor-clusters, the two or more sensor-clusters selected corresponding to a sensor of the set of sensors, the two or more sensor-clusters being merged based on two or more rules from amongst the first set of rules associated with the two or more sensor-clusters and a distance measure between the two or more sensor-clusters of the sensor, wherein the two or more sensor-clusters of the sensors are merged based on co-occurrence of one or more other sensors of the plurality of sensors in the two or more sensor-clusters for a same time period;

extracting a second set of rules from the second plurality of sensor-clusters, the second set of rules indicative of distinct sensor behaviors associated with the second plurality of sensor-clusters;

identifying a plurality of sets of correlated sensors from the second plurality of sensor-clusters based on the second set of rules; and

extracting a third set of rules from the one or more sets of correlated sensors, the third set of rules summarizing the multi-sensor data to represent prominent co-occurring sensor behaviors, wherein step by step extraction of the first, second and third set of rules enables summarization of the multi-sensor data to summarize the usage or behavior patterns of the machine or the device.