	US 12,451,241 B2
	Predicting failure of a medical device part using operational and non-operational data
Mohammed Guller, Danville, CA (US); and Deepak Nailwal, Bengaluru (IN)
Assigned to Glassbeam, Inc., Santa Clara, CA (US)
Filed by Glassbeam, Inc., Santa Clara, CA (US)
Filed on Feb. 16, 2024, as Appl. No. 18/444,507.
Claims priority of provisional application 63/446,704, filed on Feb. 17, 2023.
Prior Publication US 2024/0290475 A1, Aug. 29, 2024
Int. Cl. G16H 40/20 (2018.01); G16H 40/40 (2018.01)

CPC G16H 40/20 (2018.01) [G16H 40/40 (2018.01)]

22 Claims

1. A system, comprising:

a communication interface configured to receive data associated with a plurality of devices, the plurality of devices comprising a population of devices of a same device type each having a same target part subject to failure and the data for each device comprising machine data comprising a hundred or more logged events per device per day and non-machine data, for ninety or more days and fifty or more devices; and

a processor coupled to the communication interface and configured to:

determine based on the received data, for each of at least a subset of the plurality of devices, a part failure date on which the target part failed in that device;

create a dataset S comprising for each device in which the target part failed a set of one or both of errors and warnings logged in one or more windows W preceding the failure date on which the part failed combined with non-machine data associated with use of the device and part in the one or more windows W preceding the failure date;

use the dataset S to programmatically engineer a set of features usable to predict failure of the target part, the set of features including one or more features that are not based on logged warning or error events;

for each device in which the target part failed, programmatically labeling data associated with a first window of time prior to part failure as “positive” and labeling data associated with a second window of time prior to the first window of time as “negative”;

for each device in which the target part has not failed, programmatically labeling data associated with a first window of time prior to a buffer period as “negative”;

aggregating at least a subset of the labeled data according to an aggregation rule; and

use the labeled and aggregated data to train a machine learning model configured to be used to predict failure of the target part in a device based on data from that device, including by computing from the data features corresponding to the programmatically engineered set of features.