Research On Fault Diagnosis Of Wind Turbine Bearing Based On Compressed Sensing And Time-frequency Analysis

In the context of global carbon neutrality,the large-scale,high-quality wind power development is a key strategic decision for implementing the "double carbon" goals and duties.As the global wind turbine unit capacity grows and installed capacity rises,so does the demand for wind power operation and maintenance.Bearing,being a key component of the wind turbine,has a significant impact on the stable operation of the entire wind turbine.Therefore,condition monitoring and fault diagnosis of wind turbine bearings have significant engineering practical importance.The massive quantity of data created by the monitoring process places a huge burden on remote diagnosis,and it is of major academic research significance as well as practical application value to improve the system’s capacity to handle the massive amount of data.This dissertation addresses the urgent requirement for data compression in wind turbine status monitoring and fault diagnostic systems,the effects of different observation matrices and compression rates on signal reconstruction are investigated based on the theoretical model of compressive sensing,and the time-frequency feature extraction method of fault signals is investigated to realize bearing fault diagnosis within the framework of compressive sensing,which provides the theoretical basis and technical support for intelligent operation and maintenance and health management of wind turbines.The main research work is as follows:1.The high amount and complexity of wind turbine bearing failure vibration signal data places tremendous pressure on data storage and signal processing.Analyze the mathematical framework model of compression sensing,investigate the reconstruction effect of the measurement matrix at different compression rates,incorporate the Part Hadamard matrix into feature dimensionality reduction based on the approximate projection isometric property,propose a fault diagnosis model based on low-dimensional features,extract time-domain features as fault feature vectors directly from the low-dimensional observations sampled by compression,and then realize the bearing fault diagnosis under the framework of compression sensing.2.The wind turbine bearing vibration signal is a typical multi-component signal with clear nonlinear and non-stationary characteristics,and the variable modal decomposition method is used to extract the characteristic components of the fault.Considering the fact that parameter selection has a significant impact on signal decomposition results,an improved variational modal decomposition method based on the gray wolf algorithm and singular value decomposition is proposed.Using the gray wolf algorithm,the scale number and the quadratic penalty factor are preliminarily determined.The scale number is optimized by applying singular value decomposition or central frequency quadratic to maximize the optimization impact,avoid signal over-decomposition,and overcome difficulties such as mutual interference of signal components and difficulty in identifying weak errors.Further,the entropy features of the eigenmodal components are extracted to construct feature vectors,and fuzzy C clustering is used to achieve fault classification.3.The vibration signal from the wind turbine bearing offers a wealth of status information.The non-extensive Tsallis entropy principle is studied,and a method based on improved variational modal decomposition and Tsallis entropy feature extraction is proposed to quantify the time series and characterize different fault states.The multi-scale entropy theory is studied in order to measure the complexity and randomness of vibration signals at different scales,and the multi-scale Tsallis entropy and multi-scale continuous wavelet Tsallis singular entropy are constructed to comprehensively explore the dynamics of wind turbine bearing signals.The proposed approach is applied to Part Hadamard matrix low-dimensional compressive sampling to improve data processing efficiency for bearing fault diagnosis,and the practicality and superiority of compressive-aware sampling in the field of wind turbine bearing fault diagnosis is subsequently validated.