Research On Wind Turbine Gearbox Fault Diagnosis Method Based On Information Fusio

Wind energy has become one of the most important energy sources in the world.It is abundant,renewable and low-carbon pollution-free,and is an important player in solving the world’s energy shortage and promoting China’s energy restructuring.Wind power has also become the most prospective renewable energy generation technology today,and China has been the world’s first in terms of cumulative installed capacity and new installed capacity of wind turbines for many years.As the installed amount of wind power continues to rise,unit fault diagnosis has been gradually become the focus of the people.As one of the most vital part of the wind turbine gearbox,its working condition directly affects the performance and lifetime of the wind turbine.Therefore,monitoring and diagnosing the operating status of the gearbox of wind turbines is of great significance for ensuring the safety and stability of the turbines,reducing operational costs,and improving power generation efficiency.In this thesis,the information fusion theory is applied to the fault diagnosis of wind turbine gearbox,and corresponding fault diagnosis methods are developed.The main research contents are as follows:(1)To address the issue of insufficient information in the single-domain characteristics of wind turbine gearbox,a multi-domain feature set for the gearbox was constructed using statistical analysis,Fourier transform,and Empirical Mode Decomposition methods to extract more comprehensive features.Firstly,the original waveform,frequency spectrum,and Intrinsic Mode Function components obtained by EMD decomposition were analyzed from the perspectives of time domain,frequency domain,and time-frequency domain of gearbox vibration signals.Then,14 timedomain features,5 frequency-domain features,and 5 time-frequency domain features were extracted,with the time-frequency features being the Singular Value Entropy(SVE)of the IMF components of the original gearbox signal.Finally,experimental results showed that the multi-domain features contained more comprehensive fault information compared to the single-domain features.(2)A fault diagnosis model was developed for single bearing faults in wind turbine gearboxes using feature selection with Random Forest and feature recognition optimization with Extreme Learning Machines.Firstly,redundant or ineffective features were eliminated from the multi-domain feature set of the bearing using the RF feature importance indicator to construct a low-dimensional and effective feature set.Then,the Whale Optimization Algorithm was used to optimize the ELM model and establish a bearing fault diagnosis model.Finally,the effectiveness of the proposed fault diagnosis model was verified through experiments.(3)A multi-sensor and multi-model fusion approach based on the DempsterShafer(D-S)evidence theory was employed to construct a fault diagnosis model for composite gear faults in wind turbine gearboxes.To ensure diversity and complementarity among the models,a triad of fault diagnosis models,namely Extreme Learning Machines,Support Vector Machines,and K-Nearest Neighbor algorithms,were utilized.The signals from each gear sensor were fed into each model for classification diagnosis.Subsequently,the classification probabilities from each model were integrated to yield preliminary diagnosis results for each sensor.Finally,the preliminary diagnosis results from all sensors were fused to yield a highly reliable decision result,enabling effective diagnosis of composite gear faults in wind turbine gearboxes.