Reasearch On Wind Turbine Drive System Key Mechanical Components Condition Monitoring And Fault Diagnosis Technology

Under the background of the global ecological environmentaldeterioration and the shortage of fossil energy, wind power energy getsworldwide attention gradually. So the importance of maintenance services,such as condition monitoring and fault diagnosis, was heighted toguarantee the continuity and high efficiency of wind turbine. These factorslike tough working environment, uncertainty of wind speed and keepchanging load, make the wind turbine drive system susceptible to damagewhich lead to extremely high cost for a distributed power supply. So itmakes sense of doing some research about valid condition monitoring andfault diagnosis for some key components such as the gear and bearing inthe drive system.Firstly, jobs about researching the structure of the wind turbine drivesystem and common fault mechanism and characters of gear and bearing,analyzing the popular signal processing methods in time domain,frequency domain and time-frequency domain, were done in the first part.Secondly, a method was proposed based on the wavelet transformdomain correlation filter and principal component analysis to the topic ofcondition monitoring of the key components for the health monitoring ofmechanical parts by calculating principal components T2and SPEstatistical magnitudes. Another method was proposed based on theimproved extreme learning machine(IELM) to the shortcomings existed inthe traditional feed forward neutral network, including the deficiency ofprocessing information flow over time, and original fixed-size sequential learning machine couldn’t effectively reflect fault information with unifiedparametric proportion, whose training speed remains to be furtherimproved. IELM classification diagnosis model was set to realize faultclassification rapidly and precisely for wind turbine key mechanicalcomponents.Thirdly, gear and bearing faults would be simulated by designing andsetting up drive system fault test-bed under existing research condition,and condition monitoring and fault diagnosis methods would be validated.Degradation process from gear abrasion to broken teeth was simulated toverify principal component analysis condition monitoring method. Sevenkinds of fault data about gear and bearing totally was processed for faultmode identifications. Comparison in terms of training speed and testingaccuracy between BP, SVM, ELM, FSSELM, IELM were made wheregood results were obtained to verify the effectiveness of improved extremelearning machine network.Lastly, condition monitoring and fault diagnosis software prototypesystem was designed and developed for engineering application based oncommon signal analysis methods and related methods which wereproposed above, which includes acquisition, database storage, conditionmonitoring, fault diagnosis module. This system realizes the function ofcondition monitoring and fault diagnosis for key components in drivesystem, guarantees the safe and stable operation which makes sense forimproving the operation reliability.