System and method for detecting anomalies in sensory data of industrial machines located within a predetermined proximity

1. A computer-implemented method for detecting anomalies in industrial machine sensor data, comprising:

monitoring sensory input data of a plurality of industrial machines of a same type that are located within a predetermined proximity of each other, wherein monitoring the sensory input data further comprises tracking and aggregating a plurality of parameters associated with the sensory input data with respect to respective components of a plurality of components of the plurality of industrial machines, wherein the sensory input data is collected via a plurality of sensors each located in a predetermined proximity to one of the plurality of industrial machines, wherein the sensory input data is indicative of an operation of the plurality of components of the plurality of industrial machines;

analyzing the plurality of parameters associated with the sensory input data using a plurality of machine learning techniques, wherein different parameters of the plurality of parameters are analyzed using different machine learning techniques of the plurality of machine learning techniques, wherein the plurality of machine learning techniques includes applying at least one of: a neural network, a recurrent neural network, decision tree learning, a Bayesian network, and clustering;

computing, based on a suspected anomalous level value of each of a plurality of suspected anomalies of the sensory input data of the plurality of industrial machines, an average anomalous amount that is associated with a time interval for each of the plurality of suspected anomalies, wherein each suspected anomaly is a portion of the sensory input data;

generating a plurality of meta-models, each meta-model being generated for one of the plurality of components, wherein the meta-model generated for each component optimally indicates anomalies in the sensory input data for the component;

determining that at least one of the plurality of suspected anomalies is an anomaly, wherein a suspected anomaly is determined to be an anomaly when a result of a subtraction of the computed average anomalous amount from the suspected anomalous level value for the at least one of the plurality of suspected anomalies exceeds a predetermined threshold, wherein each of the plurality of meta-models is utilized to detect at least a portion of the determined at least one anomaly in behavior of a respective portion of one of the plurality of industrial machines; and

determining at least one predicted machine failure based on the at least one of the plurality of suspected anomalies determined to be an anomaly and the sensory input data of the plurality of industrial machines of the same type that are located within a predetermined proximity.

2. The computer-implemented method of claim 1 , further comprising:

monitoring a plurality of time stamped sensory input data related to the plurality of industrial machines; and

identifying, in the time interval corresponding to the plurality of time stamped sensory input data, the plurality of suspected anomalies of the sensory input data.

3. The computer-implemented method of claim 2 , wherein the identification of the plurality of suspected anomalies is achieved by unsupervised machine learning.

4. The computer-implemented method of claim 1 , further comprising:

generating, upon the determination that at least one of the plurality of suspected anomalies is an anomaly, a notification that is indicative of each determined anomaly.

5. The computer-implemented method of claim 1 , further comprising:

determining that at least one of the plurality of suspected anomalies is a reasonable behavior of at least one industrial machine of the plurality of industrial machines when the result of the subtraction of the average anomalous amount from each suspected anomalous level value of the plurality of sensory input data is equal to or less than the predetermined threshold.

6. The computer-implemented method of claim 1 , wherein the sensory input data is indicative of an operation of at least one component of the plurality of industrial machines.

7. The computer-implemented method of claim 1 , further comprising:

preprocessing raw sensory input data received from a plurality of sensors configured to collect sensory input data based on operation of the plurality of industrial machines; and

storing the preprocessed raw sensory input data in a data source.

8. The computer-implemented method of claim 7 , wherein the preprocessing includes at least one of: data cleansing, normalization, rescaling, re-trending, reformatting, and noise filtering.

9. A non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry to perform a process, the process comprising:

monitoring sensory input data of a plurality of industrial machines of a same type that are located within a predetermined proximity of each other, wherein monitoring the sensory input data further comprises tracking and aggregating a plurality of parameters associated with the sensory input data with respect to respective components of a plurality of components of the plurality of industrial machines, wherein the sensory input data is collected via a plurality of sensors each located in a predetermined proximity to one of the plurality of industrial machines, wherein the sensory input data is indicative of an operation of the plurality of components of the plurality of industrial machines;

analyzing the plurality of parameters associated with the sensory input data using a plurality of machine learning techniques, wherein different parameters of the plurality of parameters are analyzed using different machine learning techniques of the plurality of machine learning techniques, wherein the plurality of machine learning techniques includes applying at least one of: a neural network, a recurrent neural network, decision tree learning, a Bayesian network, and clustering;

computing, based on a suspected anomalous level value of each of a plurality of suspected anomalies of the sensory input data of the plurality of industrial machines, an average anomalous amount that is associated with a time interval for each of the plurality of suspected anomalies, wherein each suspected anomaly is a portion of the sensory input data;

generating a plurality of meta-models, each meta-model being generated for one of the plurality of components, wherein the meta-model generated for each component optimally indicates anomalies in the sensory input data for the component;

determining that at least one of the plurality of suspected anomalies is an anomaly, wherein a suspected anomaly is determined to be an anomaly when a result of a subtraction of the computed average anomalous amount from the suspected anomalous level value for the at least one of the plurality of suspected anomalies exceeds a predetermined threshold, wherein each of the plurality of meta-models is utilized to detect at least a portion of the determined at least one anomaly in behavior of a respective portion of one of the plurality of industrial machines; and

determining at least one predicted machine failure based on the at least one of the plurality of suspected anomalies determined to be an anomaly and the sensory input data of the plurality of industrial machines of the same type that are located within a predetermined proximity.

10. A system for detecting anomalies in industrial machine sensor data, comprising:

a processing circuitry; and

a memory, the memory containing instructions that, when executed by the processing circuitry, configure the system to:

monitor sensory input data of a plurality of industrial machines of a same type that are located within a predetermined proximity of each other, wherein monitoring the sensory input data further comprises tracking and aggregating a plurality of parameters associated with the sensory input data with respect to respective components of a plurality of components of the plurality of industrial machines, wherein the sensory input data is collected via a plurality of sensors each located in a predetermined proximity to one of the plurality of industrial machines, wherein the sensory input data is indicative of an operation of the plurality of components of the plurality of industrial machines;

analyze the plurality of parameters associated with the sensory input data using a plurality of machine learning techniques, wherein different parameters of the plurality of parameters are analyzed using different machine learning techniques of the plurality of machine learning techniques, wherein the plurality of machine learning techniques includes applying at least one of: a neural network, a recurrent neural network, decision tree learning, a Bayesian network, and clustering;

compute, based on a suspected anomalous level value of each of a plurality of suspected anomalies of the sensory input data of the plurality of industrial machines, an average anomalous amount that is associated with a time interval for each of the plurality of suspected anomalies, wherein each suspected anomaly is a portion of the sensory input data;

generate a plurality of meta-models, each meta-model being generated for one of the plurality of components, wherein the meta-model generated for each component optimally indicates anomalies in the sensory input data for the component;

determine that at least one of the plurality of suspected anomalies is an anomaly, wherein a suspected anomaly is determined to be an anomaly when a result of a subtraction of the computed average anomalous amount from the suspected anomalous level value for the at least one of the plurality of suspected anomalies exceeds a predetermined threshold, wherein each of the plurality of meta-models is utilized to detect at least a portion of the determined at least one anomaly in behavior of a respective portion of one of the plurality of industrial machines; and

determine at least one predicted machine failure based on the at least one of the plurality of suspected anomalies determined to be an anomaly and the sensory input data of the plurality of industrial machines of the same type that are located within a predetermined proximity.

11. The system of claim 10 , wherein the system is further configured to:

monitor a plurality of time stamped sensory input data related to the plurality of industrial machines; and

identify, in the time interval corresponding to the plurality of time stamped sensory input data, the plurality of suspected anomalies of the sensory input data.

12. The system of claim 11 , wherein the identification of the plurality of suspected anomalies is achieved by unsupervised machine learning.

13. The system of claim 10 , wherein the system is further configured to:

generate, upon the determination that at least one of the plurality of suspected anomalies is an anomaly, a notification that is indicative of each determined anomaly.

14. The system of claim 10 , wherein the system is further configured to:

determine that at least one of the plurality of suspected anomalies is a reasonable behavior of at least one industrial machine of the plurality of industrial machines when the result of the subtraction of the average anomalous amount from each suspected anomalous level value of the plurality of sensory input data is equal to or less than the predetermined threshold.

15. The system of claim 10 , wherein the sensory input data is indicative of an operation of at least one component of the plurality of industrial machines.

16. The system of claim 10 , wherein the system is further configured to:

preprocess raw sensory input data received from a plurality of sensors configured to collect sensory input data based on operation of the plurality of industrial machines; and

store the preprocessed raw sensory input data in a data source.

17. The system of claim 16 , wherein the preprocessing includes at least one of: data cleansing, normalization, rescaling, re-trending, reformatting, and noise filtering.

18. The method of claim 1 , further comprising:

selecting at least one corrective action with respect to at least one of the plurality of industrial machines based on the at least one predicted machine failure.