Improved Bayesian network in steam turbine fault diagnosis

The fault diagnosis model of steam turbine based on Bayesian network is direct impacts on the complexity of the diagnostic process, so the construction of Bayesian network model is the primary problem. According actual fault diagnosis system of steam turbine containing redundancy and uncertain information, proposed attribute reduction method to fault feature, obtained the minimal diagnosis rules. Based on two-node union reverse inference, proposed an improved program in the construction of Bayesian network. The Bayesian network model based on the minimum fault decision table can effectively reduce the complexity of the network structure, while the using of improved Bayesian network can further reduce complexity of structure and improve the diagnosis speed. Finally, the effectiveness and fastness of this method are validated by the result of practical fault diagnosis example in Bently-RK4 rotor vibration bench.