Intelligent Fault Diagnosis Of Driving Drum Bearing Of Belt Conveyor

With the rise of big data in the field of industry,driving the development and upgrading of the coal industry,the dominant position for its productivity,transportation process,so participate in the mining transportation equipment is towards high speed,intelligent,and large-scale development,when the equipment work intensity gradually bigger,rolling bearing as the main component the probability of failure is gradually rising.As the main equipment of industrial transportation,the transmission roller state of the belt conveyor has a great impact on the efficiency of the equipment.The roller shaft bearing is the biggest safety hazard of the roller.If the rolling bearing of the roller fails and leads to failure,it will cause unimaginable losses.Therefore,the intelligent diagnosis of the belt conveyor drive roller shaft research,which improves the safety,reliability and production efficiency in the intelligent operation and maintenance process.For the rolling bearing of the early fault signal is weak and not easy to identify in complex working conditions,it is difficult to diagnose the fault type of bearings quickly accurate,first analyze the vibration mechanism of rolling bearing,starting from the failure force,establish the normal and state of dynamic model and the fault state of vibration equation,and study the characteristics of the fault vibration.The adaptive spectrum cliff processing vibration signal,obtain the fault features,the signal composition is more complex and modal aliasing bearing vibration signal analysis,through the short Fourier transform for adaptive improvement,to design adaptive spectrum cliff band pass filter for signal fault signal feature extraction,can verify the features from the simulation results.After spectral cliff analysis of vibration signals,The bearing faultdiagnosis method of deep learning is proposed,By comparing the identification efficiency of the original convolutional neural network,Select the more effective residual network,Based on this improvement,the network recognition speed and accuracy are also improved simultaneously,Multi-channel expansion of the residual blocks,To increase the resolution of the feature image,And add the upsampling module and the fusion feature module,The original feature image is enlarged without losing the detailed features,And integrate all the feature details and classify them,Finally,the model is quickly trained in combination with dense connections,Quick update of the parameters,Thefeatures were extracted repeatedly,Accurate diagnosis of experimental faults.The simulation test bench is built to verify that the identification efficiency of the improved network diagnosis model is better than the original network model.This paper has 39 figures,12tables and 108 references.