Reliability Analysis Of Wind Turbine Based On Evidence Network

Wind turbine has complex structure and many components.It appears blade damage,gear wear and other faults in the operation process,which seriously affect its normal operation.It is of great significance to analyze the reliability of wind turbine.There are often many states in the process of wind turbine degradation.The fault tree analysis can’t effectively deal with the polymorphism and complexity of the system.Evidence network is the combination of evidence theory and graph theory.It has rich probability expression ability and uncertainty reasoning ability,which can effectively deal with the uncertainty and polymorphism in the system.Therefore,this paper takes the complexity,uncertainty and polymorphism of wind turbine into account,and uses evidence network to analyze its reliability.(1)Considering the difficulties of complex components and many fault modes of wind turbine,the construction of evidence network model based on fault tree analysis is studied.Firstly,the components,functions and failure modes of wind turbine are analyzed.And wind turbine is divided into six subsystems according to different functions.Secondly,the fault tree is established according to the fault causes of each subsystem.Finally,the evidence network of wind turbine is constructed through the mapping relationship between fault tree and evidence network.(2)Considering mixed uncertainty in wind turbine components,the reliability analysis of structural mixed uncertainty is studied.It provides the confidence interval of root node for the evidence network.Firstly,the multi-fidelity modeling is constructed by using multi fidelity metamodeling method,which solves the complex process of structural reliability analysis under mixed uncertainty.Low-precision samples are used to represent the change trend of the proxy model,and a small number of high-precision samples are used to improve the proxy model.Secondly,in order to reduce the number of agent model calls and improve the operation efficiency,the dimension reduction decomposition and first order second moment are adopted.The dimension reduction decomposition is used to deal with the evidence variables of the outer layer.The first order second moment is used to deal with the probability variables of the inner layer.The failure probability at the joint configuration point is obtained by the failure probability at the edge configuration point.Finally,by setting the decomposition threshold of the state,the confidence interval of the root node in each state is obtained.(3)The evidence network of wind turbine has the problems of complex network and many root node states.The state importance method is used to analyze the reliability of wind turbine.Firstly,in the conditional reliability quality table,the probability value is used to measure the uncertain logical relationship between components and systems in wind turbine.Secondly,the importance of each state is sorted.The state of the root node takes the maximum value in the desirable range according to the constraint condition of the sum of state probabilities.Different evidence networks often have different structures and conditional reliability quality tables.It is difficult to form a unified theory to measure the importance of each state of the root node,thus,all the importance ranking are taken into account.The evidence network is spread according same importance ranking.Thus,the confidence interval of wind turbine in each state is obtained.