Fault Diagnosis Method Of Fan Blade With Crack Based On SOM Neural Networks

Fan are widely used in factories,warehouses,etc.,especially in industrial production sites such as mines,providing continuous fresh air to the working environment and reducing the concentration of harmful gases to ensure the safety of workers and smooth production.Blades as the main working parts,long-term withstand high centrifugal force and aerodynamic load,easy to crack,fracture failure and even lead to serious accidents,therefore,its condition monitoring and fault diagnosis,timely grasp the development trend of the blade crack It is the most important way to eliminate such accidents.Therefore,it is of great theoretical significance and important application value to realize the fault diagnosis of the blade crack with complex working environment.In view of this,this paper will establish a SOM neural network fan blade crack location diagnosis system,which will greatly improve the accuracy and efficiency of fault diagnosis.By analyzing the dynamics of the fan blade with crack,the differential equation of motion of the blade with crack is obtained,and the influence of the blade crack on the inherent characteristics of the blade is analyzed.The three-dimensional model of the fan blade is established and modeled.On the basis of state analysis,the mode analysis of the modal analysis is carried out by wavelet transform.The influence of different factors on the crack position of the fan blade is analyzed,and the corresponding variation law is obtained.The results show that the crack position of the fan blade is related to the wavelet coefficients cD and cV of the mode shape.The modal test method is used to measure the vibration signal of single cracked blades at different crack locations.The corresponding wavelet coefficients are obtained by two-dimensional wavelet transform,and the eigenvectors of the blade cracks are constructed.Based on the law between blade crack position and wavelet coefficient map,a crack location diagnosis system based on SOM(Self-Organizing Maps)neural network was constructed.By using part of the blade crack feature vector as the training data and the rest as the test data,the SOM neural network is trained and verified to realize the identification and interval classification of the fan blade crack position and complete the fault diagnosis.The actual blade is used to verify the diagnostic system built in this paper.It is proved that the proposed diagnostic system can effectively diagnose the fan blade crack position fault,and the efficiency is higher and more accurate.