Bayesian Reliability Prognosis for Systems with Heterogeneous Information

摘要: A Bayesian methodology for prognosis of system reliability with heterogeneous information is presented. Available may be in the form physics-based or experiment-based mathematical models, historical data, expert opinion. Such typically describes failure rates individual components and does not provide on dependencies between them. The presented this paper addresses concern by learning conditional probabilities Bayes network as observations about are made. First, component faults defined event tree established. priors both events them established using various types experimental opinion, simulation data. Both updated new data collected, leading to an reliability. demonstrated automobile startup system. 1