Nonlinear Modeling And Simulation Of AMB's Axial Trajectory

Active magnetic rotor-bearing system(AMB)has drawn great attention since it has several advantages such as frictionless and low energy consumption compared to existent traditional rotor-bearing system.Due to the strong nonlinearity of the AMB system,the chaotic phenomenon at the critical speed is a frontier problem.Therefore,how to provide a stable control algorithm and an accurate chaotic analysis method for the AMB have become a research focus of the system.For the control of the nonlinear bearing-rotor system,it can be realized by the traditional proportional differential control algorithm,sliding mode control algorithm,etc.However,unstable factors such as bifurcation caused by time delay may adversely affect the stability of system control.Moreover,the AMB system presents flexible features at the high-speed,and it is usually modeled and analyzed by the finite element method,whereas it has a large computational cost.Therefore,this dissertation proposes corresponding algorithms and deeply analyses the stability control of nonlinear AMB system and the chaotic phenomenon of flexible rotor,theoretical analysis platform for simulating calculation is also constructed.The axis trajectory data of the AMB system under normal and fault conditions are identified and classified.The main research work and innovations of this dissertation are as follows:1.The flexible rotor under high-speed rotation is analysed because of the complex nonlinear characteristics of AMB system.First,this dissertation does the analysis of the equilibrium base point of the system's characteristic equation,designs the Terminal sliding mode algorithm to stably control the system,proposes a nonlinear dynamic analysis method for active magnetic baering flexible rotor system based on finite element method.The maximum Lyapunov exponent and precise Runge-Kutta hybrid integration algorithm are used to analyze the chaos of rotor system under first-order and second-order supercritical speed.The numerical experiments show that the chaos under the second-order critical speed is more serious than the first-order critical speed,which is consistent with the reality.And compared with the explicit Runge-Kutta method,the precise Runge-Kutta hybrid integration method can improve the convergence step of calculation,avoid iterative solution and maintain high accuracy.2.In order to solve the identification problems in AMB system's faults,this dissertation uses classification methods in machine learning.Hu invariant moment of the rotor axis trajectory obtained from the experiment is as an eigenvector of classification,then the imperial competition algorithm is used to reduce the redundant data.Finally,the support vector machine is established to classify and to identify the rotor axis trajectory feature data.Compared with other traditional classification models such as C4.5 decision tree and back propagation neural network,the experimental results show that the support vector machine model based on Hu invariant moment and imperial competition has higher accuracy in small sample classification,which can indicate that it is suitable for the intelligent identification classification of AMB system faults.