Research On Fault Monitoring And Intelligent Diagnosis System For Gearbox Of Blast Furnace Fan

At present,the maintenance mode of wind turbine gearboxes is still in the stages of "point inspection and regular repair" and "emergency repair",and its equipment status monitoring mainly adopts offline monitoring methods.However,the manual offline monitoring method cannot overcome the shortcomings of long monitoring period,small amount of monitoring data,and limited scope of monitoring equipment.It is difficult to comprehensively and timely grasp the operation status of equipment,and it cannot accurately and effectively monitor sudden equipment failures.It is difficult to break through the current situation.The bottleneck of equipment operation and maintenance.In order to improve the fault monitoring and intelligent diagnosis capabilities of the smelting blast furnace fan gearbox,this paper takes the smelting blast furnace fan gearbox as the research object and discusses the methods of gearbox fault monitoring and intelligent diagnosis.First,analyze the vibration characteristics of the gearbox fault signal.By analyzing the mesh stiffness characteristics of the faulty gear and the influence of different crack sizes on the stiffness,a dynamic model of the gear pair is established.The normal state of the gear and the early fault characteristics of the tooth root crack are analyzed by simulating the vibration signal.Secondly,an improved nonlinear wavelet threshold function is proposed to solve the problem of poor fault diagnosis effect in the complex environment of the gearbox structure.The nonlinear curve is used to replace the linear curve in the traditional threshold function,which solves the problem that the traditional threshold function is discontinuous at the threshold and has a constant deviation.By studying TWQT resonance sparse decomposition fault feature extraction,using particle swarm algorithm to optimize resonance sparse decomposition parameters,combined with MATLAB software to test the simulation signal using this method,it is concluded that high and low resonance components can extract features of different frequencies.Finally,build the VGG-16 convolutional neural network model,use the training set data for network training,and identify the accuracy of the convolutional neural network fault diagnosis model under different working conditions.The transfer learning model is introduced on the basis of the convolutional neural network model,and its network parameters are transferred to the gearbox fault diagnosis model.By studying the fault diagnosis model based on the deep transfer learning of the DDC theory,the underlying network is locked not to participate in the update,and the Adam algorithm is used Fine-tune high-level network parameters,achieve weight update and model optimization until the error is minimized,improve feature extraction capabilities,and obtain a deep migration learning fault diagnosis model.Set up laboratory simulation failure experiment platform and Machine Plus intelligent operation and maintenance cloud platform to collect blast furnace fan gearbox failure data for experimental verification.The model test is carried out through the signal acquisition of two test benches,which verifies the effectiveness of the proposed fault diagnosis system.This paper has 64 figures,19tables and 74 references.