Skip to Main Content
Predicting failure in complex systems, such as satellite network systems, is a challenging problem. A satellite earth terminal contains many components, including high-powered amplifiers, signal converters, modems, routers, and generators, any of which may cause system failure. The ability to estimate accurately the probability of failure of any of these components, given the current state of the system, may help reduce the cost of operation. Probabilistic graphical models, in particular Bayesian networks, provide a consistent framework in which to address problems containing uncertainty and complexity. Building a Bayesian network for failure prediction in a complex system such as a satellite earth terminal requires a large quantity of data. Software monitoring systems have the potential to provide vast amounts of data related to the operating state of the satellite earth terminal. Measurable nodes of the Bayesian network correspond to states of measurable parameters in the system and unmeasurable nodes represent failure of various components. Nodes for environmental factors are also included. A description of Bayesian networks will be provided and a demonstration of inference on the Bayesian network, such as the calculation of the marginal probability of failure nodes given measurements and the maximum probability state of the system for failure diagnosis will be given. Using the data to learn local probabilities of the network will be covered. An interface between MaxView monitoring and control software and a Bayesian network API will also be described.