Research On Load Identification Method Based On Vibration Information Of Rotor System

Rotating machinery is developing towards complication and large-scale,the rotor system plays a huge role in the machinery and also becomes one of the important factors affecting safe operations of the unit due to its vibration characteristics. And meanwhile, the shaft will bear complex and changeable loads in the process of its operation, accompanying with a new change to the bending-torsional coupling characteristics of rotor system. Therefore, studying the vibration characteristics of rotor system under loads can provide the theoretical basis for its load identification and fault diagnosis.To begin with, in order to realize the load identification by monitoring the vibration of rotor system, this paper ignores the influence of various factors such as oil-film forces of bearings with the consideration of the elastic rotational axis and the rigid disk. Linking loads and eccentricity to establish the mechanical bending-torsional coupling model of the system under the foundation of Lagrange arithmetic and creating the simulation model under the environment of MATLAB/Simulink, using ode45(Runge Kutta algorithm)to calculate and get the analytic solution. The regulation of time domainevolution for vibration signal under various loads can be got by simulation: the greater the amplitude of load is, the more deviation of its vibration displacement curve of balanced position is and changing characteristics are always in line with loading ones.Then, in order to achieve effective load identification precisely, five kinds of loads of rotor system have been analyzed qualitatively and quantitatively. At the same time, the combination of neural network, Singular Value Decomposition(SVD), Ensemble Empirical Mode Decomposition(EEMD)with qualitative identification method of rotor system and the quantitative identification method with the combination of inversion model and regressive forecast are put forward. The original signal is reconstructed and weakened by the SVD decomposition ingeniously, highlighting the loads' frequency components in vibration signal; The reconstructed signal is extracted and quantified by using EEMD; At last characteristics inputs to the network and thus conduct a classification and recognition as well so that the qualitative identification will be completed. During the procession, two methods are proposed. One is the inversion of math model, the load can be worked out by vibration signal; The other is the maximum characteristics value of signal can be extracted on the basis of qualitative identification and the load identification will be achieved by the return of the network capacity.Lastly, designing the test bench of rotor system and formulating a reasonable plan according to the load characteristics and experimental conditions can make the loading test conduct successively. And moreover, the comparison between the test data and simulation data suggests the correctness and validity of this model and load identification. Above all, the research of this subject can lay a theoretic support and test validation for the load identification and fault diagnosis of rotor system.