CPC G07C 5/0816 (2013.01) [G05B 23/0283 (2013.01); G06Q 10/20 (2013.01); G07C 5/006 (2013.01)] | 28 Claims |
1. A method for managing a platform, the method comprising:
receiving, into a platform manager in a computer system, historical sensor information indicating a historical platform health and historical context information corresponding to the historical sensor information in which the historical context information is for a set of operating conditions and comprises timescales;
forming, by the platform manager, context profiles in a training dataset from the historical context information;
using historical sensor information and a desired level of performance for a component of the platform in a machine learning algorithm for transforming the context profiles into coefficient matrices while training a time-varying multi-task machine learning model for survival analysis to: handle time-varying feature vectors, obtain a sparse coefficient matrix, and derive a remaining useful life for a component of the platform;
receiving, by the platform manager, from a sensor system for the platform, sensor information comprising context information and platform health information for a platform health of the platform and forming a context profile;
sending, by the computer system, the sensor information for the platform health of the platform and the context profile into a trained time-varying multi-task machine learning model for survival analysis;
transforming, in the trained time-varying multi-task machine learning model for survival analysis, the context profile into coefficient matrices and therefrom deriving a remaining useful life vector for the component of the platform; and
procuring, using the remaining useful life vector, resources comprising components for service and repair of the component in the platform.
|