PROBABILISTIC IMPLEMENTATION OF SYSTEM HEALTH PROGNOSIS

Described is a system for health prognosis of a component based on the use of a graphical probabilistic model. The model comprises a layer of at least one component node representing the health of a component, a layer of at least one health observation node representing a health observation, and a layer of at least one usage node representing a usage observation. Additionally, the system is configured to collect component failure data, which is used in generating normalized failure histograms. The histograms provide failure densities for component and health observation failures. The system also computes at least one parameter for the model using component and health observation failure densities. A component prognosis is computed based on present observation of system usage and health with a history of past observation using a standard reasoning engine.