Research On Vibration Characteristics Of Self-Generated Hydrostatic Oil Film Bearing-Rotor System

Hydrodynamic plain bearings are important components for supporting rotor shafts of large rotating machinery such as steam turbines,fans and compressors.They often work under harsh conditions such as heavy load,high speed and variable working conditions.Their lubricating properties directly determine whether rotating machinery can operate reliably and continuously.Compared with conventional fixed bearing structure,such as round tile bearing.The tilting pad oil film bearing has advantages of low friction power consumption,strong vibration absorbing,and automatically self-aligning.However,traditional tilting tile bearings have shortcomings of complex protection,easy wear manufacturing,and higher cost and so on.It's hot-topic that research about pivot location theory and innovation design of tilting tile bearing.In this paper,the static and dynamic characteristics of a kind of auto-hydrostatic film bearing with double layer oil film structure and tilting pad characteristic combined with theoretical modeling and numerical calculation are studied.On this basis,the dynamic characteristics of the actual auto-hydrostatic film bearing-rotor system are analyzed.Furthermore,a bearing-rotor system test platform is built to test the self-generated hydrostatic oil film bearing,providing a new research method and means for the vibration characteristics of self-generated hydrostatic oil film bearing.The main contents of this paper include:Based on the theory of fluid dynamic pressure and the geometric relation of bearing bush,the expression of oil film thickness is derived,and the dimensionless Reynolds equation derived from the inner and outer oil film of the hydrostatic oil film bearing is derived by combining with the basic form of the conventional radial bearing's Reynolds equation.An improved finite difference method based on known bearing design parameters is proposed to solve the Reynolds equation iteratively to obtain the distribution characteristics of oil film pressure,and then to solve the static characteristics of oil film such as dimensionless bearing capacity,flow power consumption and temperature rise.Based on the assumption of small pressure perturbation,the stiffness and damping coefficients of the oil film at different rotating speeds are solved,and their variation rules are analyzed.On the basis of solving the dynamic characteristic coefficient of the auto-hydrostatic film bearing,the dynamic characteristic of the actual auto-hydrostatic film bearing-rotor system is analyzed based on the finite element analysis theory.Firstly,the bearing rotor system should be discretized into finite element nodes,the motion equation based on Lagrangian equation should be established,and then assembled into the motion differential equation of the bearing rotor system considering gyroscopic effect,and the Newmark numerical integration method should be used to solve the equation.The dynamic characteristics of the rotor under the support of autohydrostatic oil film bearing are calculated,such as the critical speed,modal shape,and unbalanced response.Then,the action law of shear effect,gyro effect and other factors on the dynamic characteristics of the system are analyzed.The dynamic differential equations of the elastic shaft under the action of swinging are derived.On this basis,the dynamic differential equations of the rotor system supported by oil film bearing under verticaly and horizontaly swinging are assembled respectively according to the sequence of the finite element nodes.Newmark numerical integration method is used to solve the dynamic equations of rotor system under verticaly swinging.The effects of swinging frequency and amplitude on the vibration characteristics of rotor subsystem are studied.Newmark numerical integration method is used to solve the dynamic equations of rotor system under horizontaly swinging.The effects of swinging frequency and amplitude on the vibration characteristics of rotor subsystem are studied.The variation law of displacement response,gravity load,additional load,and axis locus are analyzed.A test platform is established,which can be employed to test vibration performance of bearings with different structural parameters.Rotor signals supported with two different bearings are analyzed.In order to extract or separate the main vibration characteristics from the measured bearing-rotor vibration displacement signals containing multiple excitation sources and strong background noise,three algorithms of sparse characterization,singular value decomposition and principal component analysis are studied respectively.The effect of the equal interval of phase angle and frequency parameters in cosine dictionary on the signal processing effect is studied,and the signal-to-noise ratio,root-mean-square error and running time are used to measure the effect.The results show that the interval of frequency and phase parameters is appropriate within [0.05,0.5],while too large value will make the filtering effect worse,and too small value will greatly increase the calculation amount.The quantity rule of effective singular value and the order rule of amplitude value are derived theoretically,namely the amplitude filtering characteristic of singular value decomposition.Whose validity of the algorithm is verified by the analysis of simulation signal and actual rotor vibration signals.The properties of the covariance matrix eigenvalue difference spectrum are studied,an improved principal component analysis algorithm based on it is put forward.The number of effective principal components can be selected automatically,by means of maximum peak position of covariance matrix eigenvalues of difference spectrum.What's more,different frequency components can be extracted by the combination of component signals between different spectral peaks.Aiming at the problem that the traditional intelligent recognition method of rotating machinery requires artificial feature extraction and low diagnostic accuracy,a deep convolutional neural network fault diagnosis model is proposed for axis trajectory recognition.This method integrates feature extraction with classification recognition,and the overall recognition effect is better than Hu+ BP and Hu+SVM.In addition,the effectiveness of the proposed model in the feature extraction of the axis trajectory is verified by principal component analysis results of the features of the fully connected layer.DCAE-FDM model is prosed to automatic recognition of axis trajectories.Combined AE and CNN,a fault diagnosis model DCAE-FDM of deep convolution automatic encoder(DCAE)is constructed,which is used to extract the automatic coding features of axis trajectory.On the other hand,aiming at the problem of low recognition rate of single feature,it is proposed to fuse Hu invariant moment features with deep automatic encoding features.The fusion features are adopted to the training and prediction of DCAE-FDM model.The results show that the proposed method is significantly better than the deep learning method(DAE+BP,DAE+SVM)and the traditional recognition method(Hu+SVM,Hu+BP)in accuracy,recall rate,and F1 score value.