The Fault Prediction And Fault Diagnosis System Based On Gear Box Data Processing

The fault diagnosis of gear box has many characteristics of strong affinity, hugeamounts of SCADA data, various variables and so on. When analyzing the failure ofthe gear box, it not only considers the faulty parts for failure，but also needs to analyzethe correlations among parts of the gear box. Therefore, it is important and necessaryto process the enormous data and implement accurate and fast gear box fault diagnosis.The work done by this thesis is as follows:First, it introduces various monitoring methods and traditional fault diagnosistechnologies of different parts of wind turbines in recent years. Simultaneously, ittakes more time to learn for the neural network model and lacks evidence for theselection of learning samples. Moreover, fault diagnosis scheme based on the artificialneural network is very difficult to get the required training samples. Vibration signal isoften difficult to meet the needs of the high frequency vibration analysis; Faultdiagnosis scheme based on fault tree analysis can’t provide timely diagnosis.Second, it puts forward a new analysis method of gear box’s fault based on the oiltemperature collected easily and establishes the temperature model of gear box by(nonlinear state estimate technology) NSET method to predict the temperature of thegear box based on the problems in monitoring methods of gearbox. It is when gear boxruns abnormally that the statistical features of the temperature prediction residualchanged largely to detect the fault of gearbox early.Third, it proposes a scheme, which combines with FTA and BAM to implementfault diagnosis fast and accurately for gear box based on the oil temperature abnormalfailure diagnosed by the nonlinear state estimate technology, and to determine thespecific reason and measures for the failure. It not only can analyze and process eachfailure phenomenon and remove redundant fault data through fault tree, but also getindependent and orthogonal fault samples for BAM, which determines the cause of thefault and gives maintenance strategy.Forth, it designs a new gearbox diagnosis system on the basis of the conditionmonitoring and fault diagnosis, combined with expert system. It could completeintelligent fault alarm、fault query and BAM network diagnosis through the knowledgeacquisition, fault diagnosis and online diagnosis module.