Vibration Fault Diagnosis Of Wind Turbines Based On Variational Mode Decomposition And Firefly Optimized Probabilistic Neural Network

With the continuous development of the social economy,the contradictions of energy shortage,environmental pollution and greenhouse effect have become increasingly prominent.In order to solve these problems,all countries in the world are vigorously developing wind power,a renewable,clean and relatively mature new energy source..However,the wind turbines operate under normal conditions all year round,the.internal structure is complex and the coupling between the components is strong,which causes the vibration signal of the wind turbine to be non-stationary.The traditional signal processing and diagnosis methods can not meet the current fault diagnosis of wind turbines.Claim.Therefore,the vibration signal of wind turbine is non-stationary,and it is studied from the aspects of signal processing,feature extraction and fault diagnosis.This paper first introduces the background and significance of the topic,as well as the current research status of wind turbines in signal processing and diagnostic methods.Then,the basic composition of the wind turbine,the main fault type and the reasonable placement of the vibration sensor are explained,paving the way for subsequent research.Then,due to the non-stationary vibration signal of the,wind turbine,the time-frequency domain analysis method of the current recursive mode decomposition has a modal aliasing problem when processing the vibration signal.Therefore,the VMD is introduced and the simulation signal is used to verify the VMD.The EMD can suppress the modal aliasing well.However,the VMD has the problem of parameter selection when decomposing the vibration signal of wind turbine.A Pearson correlation coefficient method is proposed to determine the decomposition scale and penalty factor of VMD,and the simulated fault signal is used for simulation verification.Thirdly,in order to obtain the characteristics of the wind turbine operating state,a feature index system is constructed by extracting the information entropy characteristics of the VMD-decomposed vibration signals from the three angles of time domain,frequency domain and time-frequency domain.Then,the classification effect of PNN network is affected by the smoothing factor.The unique parameter is combined with the firefly algorithm to optimize the smoothing factor of the network.A vibration fault diagnosis model for wind turbine based on FA_PNN network is proposedFinally,the wind turbine diagnostic model based on VMD-FA_PNN is applied to the vibration data of the gearbox to verify the validity and feasibility of the model.The model is compared with the VMD-PNN model and the VMD-PSO-PNN model.The simulation results show that the VMD-FA_PNN diagnostic model has higher diagnostic accuracy and accuracy in wind turbine vibration fault diagnosis.