Fault Analysis And Diagnosis Of The Wind Turbine Gearbox Based On Bayesian Network Model

Gearbox is the indispensable part of wind turbines, the role is to increase thespeed form wind wheel that it can provide enough energy to generate electricity.It is also a high incidence fault part of wind turbine. Economic losses caused byrepair gearbox and stop running are very high. And the gearbox failure may causeaccidents, major person injuries, equipment and other disasters. Gearbox installedspace is very narrow, and repair gearbox belongs to the aerial work，repair task isdifficulty after the failure. Therefore, how to improve the reliability of gearbox,predict before failure, accurate judgment and proper treatment timely after failureto minimize the losses becomes significant.Fault tree often used to analyze complex system’s failure.The tree structurecan clearly reflect the relationship between failure reason and failure type, thefailure mechanism of equipment are well reflected, but uncertainty also exists inthe analysis process, and the shortcoming such as the failure model independent.In this paper, the fault tree and Bayesian network model are integration; the graphdescription of model can make up the shortcomings of fault tree that eventcorrelation expression, only two states. The model’s nodes can representationinformation that fault related, avoid the lack of fault tree can only have one topevent. The model analyzes the fault combined with the mature software. Theefficiency of analysis, reasoning ability was higher than that of fault treeobviously. It is also superior to the vibration signal analysis, neural networkanalysis that the obvious deficiencies of dealing with uncertain problems, expressmultivariate information fusion.This paper introduces the wind turbine gearbox’s structure and workingprinciple in detail. Summarizes the failure reasons, failure patterns and failureeffects of wind turbine gearbox, established the fault tree based on the analysis offault mechanism, anglicizing the fault by fault tree. Introduction the theory of graph description, probabilistic reasoning of Bayesian network model and thesteps of how to constructed Bayesian networks model by the fault tree. Then useFullBNT-1.07to establish gearbox failure modeling, transformation fault tree intoBayesian networks model, and then through the known gear box factory data,literature data, wind farms feedback data we get the conditional probability, usematlab software and Bayesian network model to prediction beforefailure,analyze and diagnosis after failure. Based on Bayesian network gearboxis still in the exploratory stage, to be further improved.