Research On Data-Driven Fault Diagnosis Method Of Centrifuge Spindle Bearing

As important industrial equipment,centrifuges are widely used in chemical,pharmaceutical,food and other fields.With the rapid development of China’s manufacturing industry in recent years,the demand for centrifuges in various industries has been increasing,while higher requirements have been put forward for the intelligence of the centrifuge industry.As the core component that supports the normal operation of the centrifuge,the health of the centrifuge spindle bearing directly determines whether the centrifuge can operate safely,stably and efficiently.Therefore,it is of great significance to realize the intelligent fault diagnosis of the centrifuge spindle bearing.This thesis takes the LLWZ horizontal spiral discharge filter centrifuge spindle bearing as the research object,and the fault diagnosis method of the data-driven centrifuge spindle bearing is studied,solving the problems of interference noise in the signal,incomplete signal feature extraction and low fault diagnosis efficiency in variable working conditions,which has important theoretical significance and application value for the intelligent diagnosis of centrifuge spindle bearings.The specific research contents are as follows:(1)Characteristic analysis of centrifuge spindle bearing and the construction of test platform.Taking the LLWZ type horizontal spiral discharge filter centrifuge spindle bearing as the research object,the structure of the centrifuge and spindle bearing are introduced,and the failure mode,maintenance strategy,vibration cause and failure frequency and other related characteristics of the bearing are analyzed.According to the characteristics of the transmission system of the centrifuge,a simulation test platform for the spindle bearing is built,and the bearing vibration data is collected to lay the foundation for the following research on the subject.(2)An improved combined signal pre-processing method is researched to address the problem of interference noise in the bearing data collected in practical engineering.Firstly,an EEMD is introduced for the problems of endpoint effects and modal mixing in EMD,and a method for distinguishing inner mode components through mutual information is proposed;Secondly,according to Singular Spectrum Analysis,the signal components are extracted from the high-frequency intrinsic mode functions,and then the low-frequency components are reconstructed with the processed high-frequency components as the denoised signals;finally,the denoising experiments are carried out on the real bearing signal and the simulated bearing signal respectively,and the results show that the proposed method can effectively remove the noise in the bearing vibration signal,and has a higher signal-to-noise ratio compared with other methods.(3)In view of the problems of incomplete feature extraction and low model generalization in the fault diagnosis of spindle bearings under constant and variable working conditions,a research on the fault diagnosis method of centrifuge spindle bearings was carried out.First of all,in the fault diagnosis of fixed working conditions,a deep feature extraction network combining a large-scale one-dimensional convolutional neural network and a long short-term memory network was built to realize the deep feature extraction of one-dimensional vibration signals and improve the fault diagnosis rate;secondly,to address the problem of low generalization of the model under variable working conditions,the DANN network is introduced to realize the knowledge migration between the source domain and the target domain data,and to improves the original network by MMD as a measure for the problem of inconsistent data distribution in the migration process,solving the problem of unsatisfactory cross-domain diagnosis capability under variable working conditions.Finally,the proposed method is verified on both public datasets and self-built datasets.The experimental results show that the method proposed in this thesis can effectively realize the deep feature extraction of bearing vibration signals and the knowledge migration between different domain data under variable working conditions,improving the cross-domain diagnostic capability.(4)Design and implementation of the centrifuge spindle bearing fault diagnosis system.According to the fault diagnosis requirements,the overall framework of the system is built,and the specific functions of each module are clarified.The Python language combined with the Lab VIEW virtual instrument software platform is used for joint programming and development to ensure the reliability and operability of the software system.Finally,an experiment was carried out on the built test bench to verify the feasibility of the system.