Direct drive wind turbines,using the multipole connected directly driven by motor and impeller,is becoming the main installed wind model.It is more reliability and high efficiency for eliminating the gear box of the failure of parts.Condition monitoring and fault diagnosis is particularly important of the wind turbine,which can decline the unit to burn down,prevent the tower pole break and so on.Big data analysis method is necessary for the data is real-time,the volume is large,and data analysis process may include a unified analysis of multiple data sets.The main drive bearing fault and the magnetic steel falling off are most common phenomenon of wind turbine,which respects continuous characteristics and discrete characteristics.This paper analyzes and studies the relevant algorithms of health monitoring by using the distributed parallel feature of big data analysis method:According to continuous characteristics,fault diagnosis method based on EMD and data box is put forward.Firstly,the signal is decomposed by the EMD algorithm to the IMFs.Secondly,characteristics of amplitude domain parameters are extracted as feature vector.Lastly,the feature vector is entered into the fault classification in SVM verified by actual data.As a result,the algorithm has high accuracy and generalization ability,better solve practical problems.The operator of EMD and statistical description based on Spark is developed.According to discrete characteristics,this paper choose 'black powder' as a typical complex uncertainty problems.Firstly,use big data analysis method to research the basic association rules.Secondly,use GF-Growth association rules algorithm based on Spark to derive related rules,and establish library to forecast the phenomenon of black powder,the research on black powder problem is instructive. |