Study Of Vibration Monitoring And Fault Diagnosis Of The Vibrating Feeder Exciter Bearings

Sand aggregate is an important engineering material in the construction and development of China.Its production is moving towards large-scale,intelligent and green development.The main production equipment of sand aggregate is vibrating feeder,crusher,vibrating screen and so on.The vibration feeder was taken as the research object in this thesis.The fault form of the exciter bearing is analyzed.Combined with the actual production conditions of an aggregate mine in Yanshi,a monitoring system is constructed to collect signal data during the operation of the equipment,and a diagnostic model is constructed to realize intelligent bearing fault diagnosis.The specific contents are as follows.The structural composition and fault types of the double eccentric shaft vibratiing feeder were analyzed.Combined with the production conditions of the aggregate mine,the fault forms and causes of the exciter bearing were explored.An online monitoring system for the bearings based on vibration sensor,signal acquisition instrument and fault monitoring software was constructed.The monitoring device was selected and the measuring position were arranged,and the signal acquisition strategy was also developed.Then,the installation and debugging were carried out in the industrial field.According to the structure of the bearing,the typical fault characteristic frequency of the bearing was calculated by empirical formula.The three-dimensional model of normal and damaged bearings were drawn by Solid Works,and the vibration characteristics were simulated and analyzed by ADAMS software.The time domain and frequency domain characteristic signals of normal and different fault bearings under different speed conditions were obtained.The results were quantitatively analyzed to verify the reliability of the simulation and explore the influence of working condition changes on the feature distribution.By analyzing the influence of outliers in the numerical simulation signal on the kurtosis and standard deviation of the signal,a data cleaning algorithm based on binary Kmeans clustering was proposed to eliminate the outliers of the simulation signal,and the effectiveness of the proposed method was verified by the measured signal in industrial field.The wavelet threshold and empirical mode decomposition denoising method were introduced to decompose and reconstruct the vibration signals acquired in the industrial field.The noise reduction effect was compared and analyzed by using the evaluation indexes such as signal-to-noise ratio,mean square error and correlation coefficient,and the most suitable method for vibration signal denoising of exciter bearing was determined.Based on the ResNet50 neural network model as the basic framework,the attention mechanism was introduced to improve the network,and the domain adaptive metric criterion was added to construct a deep transfer domain adaptive learning model suitable for vibration exciter bearing fault identification.Different fault diagnosis experimental schemes were developed.The fault dataset was subjected to sliding window sampling and time-frequency conversion,and the training set and test set were divided.The accuracy of fault recognition was compared with other methods,and the recognition effect was visualized by confusion matrix and t SNE.There are 70 figures,17 tables and 82 references in this thesis.