Fault Diagnosis Method Of Yaw Motor Bearing Based On Acoustic Signal

Against the backdrop of the development and utilization of green and clean energy,wind turbines have been widely used.Yaw motor is one of the most important components in wind turbines.In the event of a motor failure,this may result in the equipment being subjected to excessive wind loads.And excessive vibration of the equipment resulting in damage will eventually increase repair costs.As the most important mechanical component in the yaw motor,the rolling bearing bears various external forces such as impact,friction,and load during operation,which is prone to wear and damage,leading to malfunction and shutdown.Therefore,the fault diagnosis of the rolling bearings of the wind turbine yaw motor is of great significance.The specific research content of this article is as follows:Noise with high frequency components or high signal correlation may affect the filtering performance of the Least Mean Square(LMS)algorithm.Therefore,the noise reduction effect of this method will not be ideal.In response to the above issues.This article introduces kurtosis and correlation coefficients.The mixed signal components after empirical wavelet transform(EWT)decomposition are screened,and the noise signal is decomposed using the above empirical wavelet function.Then,the filtered corresponding frequency components are input into the adaptive filter for noise reduction.Next,the denoised signal is reconstructed to obtain the final denoised signal.Finally,through experimental comparison with the filtering method without filtering components,it was confirmed that this method has more advantages.When processing small sample data,the Support Vector Machine(SVM)performs well in the fault identification problem.There is no much requirement for hardware conditions and storage space,and the training can be completed in a short time.However,it is necessary to first select appropriate kernel functions and parameters,otherwise problems such as poor classification performance may occur.In response to the above situation,an improved slime mold algorithm optimized SVM based rolling bearing fault diagnosis model is proposed.Firstly,the signal is decomposed into a series of intrinsic mode functions using Adaptive Variational Modal Decomposition(AVMD),and the sample entropy is calculated to construct a feature vector as the input sequence.Finally,the improved slime mold optimization algorithm based on Levy’s flight strategy was used to optimize the penalty factor and kernel function parameters.The results indicate that this method can be effectively utilized in bearing fault diagnosis.At present,large-scale deep learning models such as convolutional neural networks and recurrent neural networks are prone to problems such as gradient vanishing,inability to achieve parallelization processing,and obtaining global perspectives.A fault diagnosis network model based on the visual Transformer framework is proposed to address this issue.Uniformly divide the wavelet transform time-frequency map containing time-frequency features into image blocks of the same size as input to the model.And use convolutional neural networks to process the output sequence of the visual Transformer encoder layer.Finally,a large number of easily available sound data of the test-bed are used as the source domain,and the actual yaw motor sound data is used as the target domain to diagnose the actual data through transfer learning.The experimental results show that this method can achieve higher accuracy in pattern classification compared to other models.