Fault Diagnosis Of Wind Turbine Gearbox Based On Decision Tree Support Vector Machine

As one of the clean energy sources,wind power is becoming more and more popular in today's sustainable development.However,with the rapid development of wind power industry,because some technology is not mature,many problems arise in the late period operation and maintenance of the wind turbine.The problems cause the wind turbine itself shut down or even damage.Most of the problems happen in the gearbox of the wind turbine drive chain,so exploring the fault diagnosis of wind turbine gearbox in engineering implementation is of great significance for reducing the operation cost of wind power and improving the efficiency of the unit.Considering the actual operation and maintenance process,the diagnosis of the fault of the generator unit is mostly the screening and analysis of the vibration data collected by the engineers manually,and at the same time,because of the large data of the unit,it makes the engineer work hard and time consuming.In order to solve this problem,this thesis use the the actual vibration data which collected from the Longyuan power group as samples to analysis,and make fault diagnosis and machine learning algorithm combined,using support vector machine(SVM)to establish fault diagnosis model,instead of repetitive operation manual tc raise the efficiency and accuracy of fault diagnosis.Firstly,through the soft threshold filtering method based on wavelet packet of the vibration signals,the noise signal mixed in the acquisition process is filtered out.At the same time,the fault characteristic parameters are extracted in two aspects of the time domain and frequency domain,and the training group and the test group are established by using the extracted characteristic parameters.Secondly,based on the training and testing groups well,for training the fault classification model based on support vector machine,and the use of particle swarm algorithm(PSO)and decrease step fruit fly optimization algorithm(DS-FOA)for optimize the slack variable and the penalty varibale of support vector machine,so as to improve the rate of correct classification.Finally,based on the actual production situation,the vibration data collected from the actual wind farm are the characteristics of a single fault sample.A hybrid fault diagnosis model for wind turbine gear box based on decision tree support vector machine is proposed.The training samples and test samples were set up to verify the model through the actual collected vibration data.The results achieved a certain extent by training a single fault sample to complete the problem of hybrid fault classification.The experimental results verify the feasibility and effectiveness of the proposed algorithm.