摘要：AbstractThe main goal of this paper is to present a new method for fault detection and isolation with a Bayesian network (BN). This method combines model-based and data-driven frameworks to detect and diagnose single, multiple and unknown faults. We propose an original BN structure with new decision rules. These rules are constructed to take advantage of the prior model knowledge and the available data. Our network presents new perspective to detect unknown fault and outperforms some recent work proposed in Bayesian networks literature. The performance of the method is illustrated on a heating water process simulating several scenarios of operating conditions.
