US 12,482,301 B2
Component maintenance prediction system with behavior modeling
Changzhou Wang, Bellevue, WA (US); Darren Puigh, Bellevue, WA (US); Darren Brian Macer, Seattle, WA (US); Jun Yuan, Sammamish, WA (US); Mark Mazarek, Bothell, WA (US); and Lesley Quach, Seattle, WA (US)
Assigned to The Boeing Company, Chicago, IL (US)
Filed by The Boeing Company, Chicago, IL (US)
Filed on Jun. 8, 2022, as Appl. No. 17/806,025.
Prior Publication US 2023/0401899 A1, Dec. 14, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G07C 5/00 (2006.01); G06F 18/214 (2023.01); G06N 20/00 (2019.01)
CPC G07C 5/006 (2013.01) [G06F 18/214 (2023.01); G07C 5/008 (2013.01); G06N 20/00 (2019.01)] 21 Claims
OG exemplary drawing
 
1. A method for managing a maintenance for a component in a vehicle, the method comprising:
training behavior machine learning models using first training data to output predicted values for target parameters for a normal behavior of the component operating in a tolerance, wherein the first training data comprises historical sensor data labeled with actual values for the target parameters to be predicted by the behavior machine learning models, wherein each behavior machine learning model in the behavior machine learning models predicts a target parameter in the target parameters for the component;
determining historical prediction metrics from the predicted values for the target parameters predicted by the behavior machine learning models in response to receiving the historical sensor data and actual values for the target parameters for the component;
training a maintenance machine learning model using second training data to predict whether the maintenance is needed for the component, wherein the second training data comprises the historical prediction metrics, wherein the maintenance machine learning model outputs a prediction as to whether the component will become out of the tolerance after a time period; and
determining whether the maintenance is needed for the component using new sensor data for the component, prediction metrics derived from the new sensor data for the component and the maintenance machine learning model.