US 12,175,434 B2
Prognostics and health management service
Srikanth Venkata Tenneti, Sunnyvale, CA (US); Arvindh Krishnaswamy, Palo Alto, CA (US); Karim Helwani, Mountain View, CA (US); Mehmet Umut Isik, Menlo Park, CA (US); Ritwik Giri, Sunnyvale, CA (US); Fangzhou Cheng, Mountain View, CA (US); and Aparna Pandey, Cupertino, CA (US)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Sep. 30, 2020, as Appl. No. 17/039,649.
Prior Publication US 2022/0101270 A1, Mar. 31, 2022
Int. Cl. G06Q 10/20 (2023.01); G06F 16/21 (2019.01); G06F 16/906 (2019.01)
CPC G06Q 10/20 (2013.01) [G06F 16/219 (2019.01); G06F 16/906 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving a request to perform machine health anomaly detection for a machine using clusters;
performing the machine health anomaly detection, the performing comprising:
initializing a first one or more clusters on a set of samples that have been labeled as normal, wherein the initialized one or more clusters are associated with normal operating modes of the machine;
receiving a data point from a sensor; and
determining when the received data point is a part of one of the first one or more clusters utilizing a distance to centers of the one or more clusters, wherein the determining further comprises:
when the received data point is determined to belong to one of the first one or more clusters associated with normal operating modes of the machine, assigning the received data point to the determined cluster, updating the cluster, and updating a history for the cluster,
when the received data point is determined to belong to one of a second one or more clusters associated with anomalous operating modes of the machine, raising an anomaly, updating the cluster, and updating a history for the cluster, and
when the received data point is determined to not belong to any existing cluster, creating a new cluster associated with the received data point, raising an anomaly, and, based at least in part on user feedback related to the raised anomaly, identifying the received data point as being associated with one of a new anomalous operating mode of the machine or a new normal operating mode of the machine, or, if the user feedback indicates that the raised anomaly is not associated with either a new anomalous operating mode or a new normal operating mode, incorporating the received data point into an existing cluster; and
based at least in part on determining that the received data point belongs to one of the second one or more clusters, alerting a user to one of a predicted need for maintenance associated with the machine or a predicted failure of the machine.