【作者】 罗跃纲；

【导师】 闻邦椿；

【作者基本信息】 东北大学 ， 机械设计及理论， 2002， 博士

【摘要】 旋转机械是指大型汽轮发电机组、水轮发电机组、核电机组、航空航天发动机、高速压缩机、离心机、离心泵和高精度机床等以转子系统为工作主体的机械设备,它们广泛地应用于电力、石化、冶金、机械、航空等各工业部门。随着科学技术与现代化工业的发展,旋转机械正朝着大型化、连续化、高速化、轻型化、集中化、自动化和大功率、大载荷方向发展。这一方面提高了生产率,降低了生产成本;但另一方面,这些设备一旦发生故障,所造成的经济损失将会成倍的增加。最近几十年来由于机械设备故障导致的灾难性事件时有发生,造成的经济损失、人员伤亡和社会影响也是难以估量的。因此,一方面对于旋转机械在速度、容量、效率和安全可靠性等方面提出了更高的要求;另一方面使得发展并应用先进的状态监测与诊断技术对设备故障进行检测和诊断显得尤为重要。本文以旋转机械的转子系统和工程结构为主要研究对象,首先系统阐述了转子系统中转轴的非线性刚度问题、转静子碰摩和基础松动等非线性故障转子动力学问题及研究方法、基于神经网络的智能诊断技术的研究目的、意义与研究概况,存在的问题与不足。在此基础上,系统、深入地研究了非线性转子系统由于转轴非线性刚度、碰摩和松动耦合故障引起的分岔与混沌行为,以及若干相关问题;结构损伤智能诊断识别中特征(敏感)参数的选取问题、智能诊断方法的改进措施,以及在设备与结构故障诊断中的应用问题。本文的主要工作有以下几个方面: 1. 建立了具有非线性刚度轴支撑的Jeffcott 转子系统动力学方程,利用多尺度法对弱非线性刚度系统的非共振、主共振、超谐共振和亚谐共振响应进行了分析;并应用数值分析方法研究了具有强非线性刚度系统响应的复杂动力学行为,系统参数变化对系统动力学响应的影响以及混沌运动的激变特性。2. 真实的转子系统的刚度通常是非线性的,本文建立了具有非线性刚度的转子系统局部碰摩的动力学微分方程,并应用数值分析方法研究了此类系统响应的复杂动力学行为,利用转子响应的分岔图、最大Lyapunov 指数曲线图、Poincaré截面映射图、时域波形图、相轨线图、轴心轨迹图、幅值谱图和功率谱图等图形分析了系统响应的周期运动、拟周期运动、更多还原

【Abstract】 Rotating machinery includes large scale of turbine generator, hydraulic motor, nuclear motor, aerospace motor, high-speed compressor, centrifugal engine, centrifugal pump and high accuracy machine that the rotor system is main working body. It is applied in the field of electric power industry, petrochemical industry, metallurgy, machinery, aerospace industry broadly. With the development of science and technology, rotating machinery is having the tendency of large scale, serialization, high-speed, light-duty, centralization, automation, big power and big load. This not only enhances the productivity but also reduces the cost. But once the machinery fault happens, the loss is large. There were disastrous accidents and casualties and the influence are large. So on one hand, the higher need is put forward in speed, content, efficiency, safe side of rotating machinery, on the other hand, it is important to develop advanced state inspect and diagnosis technology. In this paper, the rotor system and engineering structure are studied. The research overview, research goal and research significance of nonlinear rigid, impact and rubbing, foundation looseness, coupling faults, the bifurcation and chaos motion, intelligence fault diagnosis based on the neural networks are discussed and the shortcomings of them are discussed, too. On the basis of them, the bifurcation and chaos motions caused by the nonlinear rigid, impact and rubbing, foundation looseness, coupling faults in rotor system and other corresponding problems are discussed. The problems of sensitive parameter selection in the identification of structure damage fault diagnosis, the improvement measures of diagnosis method and the applications are discussed. The main works in this paper are as follows: 1. The Jeffcott rotor system dynamics equation was constructed which using the linear and cubic to express the strength of center shaft and material physics nonlinear factors. The multiple-scale method was used to research the complex dynamics motions of this kind of rotor system about non-resonance, main-resonance, super-harmonic and sub-harmonic responses. The numerical value analysis method was used to research the complex dynamics motions of the rotor system, and the influence caused by the change of system parameter and chaos motion excitation character was also analyzed. 2. The rigidity of real rotor system usually is nonlinear. In this paper, the rotor system local impact and rubbing dynamics differential equations having the nonlinear rigidity was constructed. The bifurcation and chaos behavior especially the influence on bifurcation and chaos behavior of impact and rubbing fault rotor system caused by the parameters of nonlinear rigidity, rotor rotating speed, eccentric mass was analyzed, using the numerical value analysis method. The bifurcation diagrams, maximum Lyapunov exponent diagrams, Poincar émaps, phase plane portraits, trajectories of journal center, time-history curve, amplitude spectra and power spectrum diagrams of the rotor motion were used. The convert and evolution course of periodical response, quasi-periodical response, double-periodical bifurcation, chaos of the system response was analyzed. 3. The influence on nonlinear dynamics of impact and rubbing fault of rotor system caused by foundation looseness was researched. The local impact and rubbing fault dynamics model and differential equation having the foundation looseness was put forward and constructed. The numerical value simulation of the rotor system was used to analysis the nonlinear dynamic characteristics. The bifurcation and chaos behavior under the different frequency rate and eccentric mass were researched. The influence of system dynamic responses caused by loose mass, impact and rubbing rigidity was analyzed, and the changing characteristics with the nonlinear rigidity of the shaft were researched, too. 4. The rotor system local impact and rubbing dynamics model that considering the stator mass and supporting rigidity and having the foundation looseness fault was put forward. The nonlinear dynamics equation was constructed. Bifurcation and chaos behavior of the rotor system was analyzed. The influence caused by different rotate speed and eccentric mass on the rotor system bifurcation and chaos behavior was researched. 5. The main problems existed in the neural networks, the improvement measures, the advantage and disadvantage and the features were inquired in this paper. The momentum coefficient, learn rate and the number of hidden units were provided utilizing the improvement BP neural networks. A new character parameter in structure damage fault diagnosis neural networks was put forward as an input parameter which overcoming the shortage of using one parameter. In the mean time, a parameter related with damage extent was added. The structure damage problems were analyzed using these parameters. A local minimum differential equation using genetic neural network algorithm was put forward in order to improve the convergence speed and avoid the local minimum. Through the practical fault diagnosis in air-compressor, the validity of it was proved. 6. From the practice, application of integrated neural networks in fault diagnosis were researched and integrated neural networks based on information fusion were established in this paper. The implement measure and principle of sub-neural networks were inquired, and its validity and feasibility were proved through the diagnosis examples.更多还原