Systems and methods for analyzing machine performance

Abstract

Disclosed are approaches for analyzing vibration data to determine and mitigate an occurrence of an anomaly with respect to an industrial machine. In one embodiment, time-series vibration data is received from a sensor measuring vibration of a machine. First outliers are detected in a short-term window of the time-series vibration data. Second outliers are detected in a long-term window of the time-series vibration data. The sensor is determined to be associated with an anomaly based at least in part on a quantity of the first outliers and a quantity of the second outliers.

1 - 30 . (canceled) 
     
     
         31 . A computer-implemented method, comprising:
 receiving time-series vibration data from a sensor measuring vibration of a machine;   detecting first outliers in a short-term window of the time-series vibration data;   detecting second outliers in a long-term window of the time-series vibration data; and   determining that the sensor is associated with an anomaly based at least in part on a quantity of the first outliers and a quantity of the second outliers.   
     
     
         32 . The computer-implemented method of  claim 31 , further comprising preprocessing the time-series vibration data by aggregating the time-series vibration data according to a time interval. 
     
     
         33 . The computer-implemented method of  claim 31 , wherein the long-term window encompasses the short-term window. 
     
     
         34 . The computer-implemented method of  claim 31 , wherein the first outliers and the second outliers are detected based at least in part on applying an isolation forest algorithm to the time-series vibration data. 
     
     
         35 . The computer-implemented method of  claim 31 , further comprising:
 determining a quantity of outliers in a detection window that are both in the first outliers of the short-term window and in the second outliers of the long-term window, the detection window being encompassed within the short-term window and the long-term window; and   wherein determining that the sensor is associated with the anomaly is further based at least in part on the quantity of outliers in the detection window.   
     
     
         36 . The computer-implemented method of  claim 31 , further comprising:
 normalizing the short-term window of the time-series vibration data; and   normalizing the long-term window of the time-series vibration data.   
     
     
         37 . The computer-implemented method of  claim 31 , further comprising generating a report indicating that the sensor is associated with the anomaly. 
     
     
         38 . The computer-implemented method of  claim 37 , wherein the report includes at least one of: a machine identification, a sensor type, a facility identification, or a sensor measurement. 
     
     
         39 . The computer-implemented method of  claim 31 , wherein detecting the first outliers further comprises identifying vibration data points in the short-term window having an amplitude meeting a first threshold, and detecting the second outliers further comprises identifying vibration data points in the long-term window having an amplitude meeting a second threshold. 
     
     
         40 . A system, comprising:
 at least one computing device configured to at least:
 receive time-series vibration data from a sensor measuring vibration of a machine; 
 detect first outliers in a short-term window of the time-series vibration data; 
 detect second outliers in a long-term window of the time-series vibration data; and 
 determine that the sensor is associated with an anomaly based at least in part on a quantity of the first outliers and a quantity of the second outliers. 
   
     
     
         41 . The system of  claim 40 , wherein the time-series vibration data is aggregated according to a time interval. 
     
     
         42 . The system of  claim 40 , wherein the first outliers and the second outliers are detected based at least in part on applying an isolation forest algorithm to the time-series vibration data. 
     
     
         43 . The system of  claim 40 , wherein the at least one computing device is further configured to at least determine a quantity of outliers in a detection window that are both in the first outliers of the short-term window and in the second outliers of the long-term window, the detection window being encompassed within the short-term window and the long-term window; and
 wherein the sensor is determined to be associated with the anomaly further based at least in part on the quantity of outliers in the detection window.   
     
     
         44 . The system of  claim 40 , wherein the at least one computing device is further configured to at least:
 normalize the short-term window of the time-series vibration data; and   normalize the long-term window of the time-series vibration data.   
     
     
         45 . The system of  claim 40 , wherein the at least one computing device is further configured to at least generate a report indicating that the sensor is associated with the anomaly, and the report includes at least one of: a machine identification, a sensor type, a facility identification, or a sensor measurement. 
     
     
         46 . The system of  claim 40 , wherein the at least one computing device is further configured to at least:
 identify vibration data points in the short-term window having an amplitude meeting a first threshold; and   identify vibration data points in the long-term window having an amplitude meeting a second threshold.   
     
     
         47 . A non-transitory computer-readable medium embodying instructions executable in at least one computing device, wherein when executed the instructions cause the at least one computing device to at least:
 receive time-series vibration data from a sensor measuring vibration of a machine;   detect first outliers in a short-term window of the time-series vibration data;   detect second outliers in a long-term window of the time-series vibration data; and   determine that the sensor is associated with an anomaly based at least in part on a quantity of the first outliers and a quantity of the second outliers.   
     
     
         48 . The non-transitory computer-readable medium of  claim 47 , wherein the first outliers and the second outliers are detected based at least in part on applying an isolation forest algorithm to the time-series vibration data. 
     
     
         49 . The non-transitory computer-readable medium of  claim 47 , wherein the instructions further cause the at least one computing device to at least generate a report indicating that the sensor is associated with the anomaly, and the report includes at least one of: a machine identification, a sensor type, a facility identification, or a sensor measurement. 
     
     
         50 . The non-transitory computer-readable medium of  claim 47 , wherein the instructions further cause the at least one computing device to at least:
 identify vibration data points in the short-term window having an amplitude meeting a first threshold; and   identify vibration data points in the long-term window having an amplitude meeting a second threshold.