Research On Vibration Fault Diagnosis Method Of Wind Turbine Based On Information Fusion By Convolutional Neural Networks

Wind turbines are usually operated under complex terrain and weather conditions.Poor operating conditions can cause their rotating mechanical parts to be prone to failure.This will bring huge safety hazards to the equipment.A huge economic loss.Therefore,real-time operation status monitoring and periodic fault diagnosis of the rotating mechanical parts of the fan have practical significance.Aiming at the operating characteristics and fault characteristics of rotating mechanical parts of wind turbines,this paper proposes a design scheme of remote fault monitoring and diagnosis(CMS)system for wind turbine drive train with two-layer structure.At the same time,it is based on the problem of CMS system relying on expert diagnosis experience.A vibration state recognition method based on Fusion Feature Convolution Learning Network(FF-CNN)is proposed.Through experimental research,it is found that the recognition accuracy of the recognition model based on this method reaches over 97%.The main research work of this paper is as follows:1.Based on the structure and working principle of large-scale wind turbine,the main fault types and causes of rotating fan equipment are analyzed,and the common fault diagnosis methods and real-time remote condition monitoring system are studied one by one.2.According to the requirements of the condition monitoring and fault diagnosis analysis of the fan drive chain,the corresponding logical framework and development process of the CMS system are determined,and the location of the vibration sensor,signal acquisition and pre-processing are determined accordingly.At the same time,the corresponding diagnosis and analysis algorithm is determined according to the fault characteristics of each rotating part of the fan,and the validity of the diagnosis algorithm is analyzed in detail from three aspects of vibration signal time domain,frequency domain and time frequency domain.3.Aiming at the shortcomings of CMS system and the fault characteristics of bearings and gearboxes on the fan drive chain,a vibration state recognition method based on fusion feature convolution learning network is proposed.The method adopts Resonance-based sparse signal decomposition(RSSD)algorithm to separate the periodic and impact components of different fault characteristics of the vibration signal.The Teager energy operator is used to amplify the weak impact characteristics,and then the signal period,the impact components are input into the CNN to perform deep learning in a targeted manner to complete the fault classification task.4.Based on the recognition model proposed in this paper,the influence of model structure and parameters on recognition accuracy is discussed.The model is optimized experimentally and the optimized model is trained.The final model classification accuracy rate is 97.6%.At the same time,it is confirmed by comparative experimental research that the recognition model proposed in this paper has higher diagnostic accuracy than other recognition models.