Research On Intelligent Early Warning And Diagnosis Method Of Fan Bearing Fault Based On Deep Learning

With the rapid consumption of traditional fossil fuels and the deepening of environmental pollution,the development of renewable energy such as wind energy,solar energy and tidal energy has become a hot topic at home and abroad.Among them,wind energy is widely concerned and studied because of its rich source,wide distribution and pollution-free characteristics.Wind power technology is increasingly mature,and the assembly capacity of wind turbine is gradually increasing.But at the same time,the volume of the engine room of the fan is also gradually increasing,the hub is also increasing,and the challenges faced by the fan are also increasing.Among the various components of the fan,rolling bearing is the core component of the transmission system and generator system,which is widely used in rotating machinery and equipment.It is easy to cause the failure of each component of the bearing due to the continuous alternating load during operation.So,it is of great practical significance to carry out early warning and diagnosis of fan bearing fault.In this paper,the deep learning algorithm is used to study the early warning and diagnosis method of fan bearing fault.The early warning method of fan bearing fault based on long short-term memory(LSTM)and the improved stacked denoising autoencoders(SDAE)are proposed.Considering the need to view the location of the specific fault fan,based on the 3D visualization technology,the 3D full scene model of wind farm is established.Relying on the visual studio platform,the intelligent early warning and diagnosis system of wind turbine bearing fault is developed.The specific research contents are as follows:(1)In view of the actual engineering situation,the long-term and short-term memory network is used to establish the wind turbine bearing fault early warning method based on LSTM to realize the dynamic prediction of the root mean square value of the wind turbine bearing vibration data.Thro ugh the comparative analysis with the threshold value,the wind turbine bearing fault early warning is realized.Based on the rolling bearing life cycle data,the LSTM network,BP neural network and SVM are used to analyze the bearing fault early warning analysis verifies the superiority of the proposed method.(2)In view of the shortcomings of SDAE network superparameters not easy to be determined,considering the excellent performance of genetic algorithm in optimizing SDAE network superparameters and improving network accuracy,a fan bearing fault diagnosis method based on the improved SDAE network is proposed.Based on the actual rolling bearing fault classification data,the method proposed in this paper and various neural networks are respectively used for research and analysis,proving the effectiveness and superiority of the method proposed in this paper.(3)According to the actual system design,the system design is based on the system design and the actual requirement of ArcGIS,3ds Max and other software,the establishment of wind farm digital terrain and wind turbine three-dimensional model,the use of surface feature fusion method to complete the wind farm three-dimensional full scene model;design the system database framework,using Oracle database,complete the system database;relying on visual Studio platform,based on C#language and software secondary development technology,realizes online monitoring,time and frequency domain analysis,fault warning,fault diagnosis and other module functions,and completes the development of intelligent early warning and diagnosis system for fan bearing fault.