具有呼吸裂纹的转子动力学特征提取及预诊方法研究

发布时间：2018-02-28 04:22

本文关键词： 呼吸裂纹转子动力学智能预诊经验模式分解动态预测　出处：《哈尔滨工业大学》2014年博士论文　论文类型：学位论文

【摘要】：旋转机械是现代化工业社会必不可少的装备，对国民经济的重要行业和航空航天等国防重点领域有极其重要的影响。旋转设备一旦发生突然故障，设备的维修和非计划停机所带来的经济损失和社会危害是巨大的。转子裂纹故障是旋转机械中最常见的故障形式之一，因此对裂纹转子的故障性质和发展过程进行深入研究，并对裂纹转子剩余寿命进行预测，从而采取相应的措施实现预防性维护，降低事故发生率和计划外停机次数，对于保证系统实现高效、安全、平稳、低成本运行具有重要的经济价值和社会意义。本文针对具有横向呼吸裂纹的Jeffcott转子系统的故障特征提取、性能评价和剩余寿命预测等预诊关键技术进行研究，提出了一种融合动力学分析、断裂力学理论和基于数据驱动预测方法的裂纹转子预诊新途径。在对重力占主要作用的转子系统裂纹呼吸变化过程分析的基础上，，利用独立完整的数学函数对裂纹呼吸过程进行描述，建立了裂纹转子系统的动力学模型，并对呼吸裂纹转子系统特有的动力学行为进行了分析。研究了转子在经过1/3和1/2亚临界转速区时的振动位移、轴心轨迹随转速的变化规律，并利用Floquet理论对转子的转速、裂纹深度和阻尼等参数对系统周期运动稳定性的影响进行了分析。针对裂纹转子在经过1/3和1/2亚临界转速区时的振动响应，利用经验模式分解（Empirical Mode Decomposition, EMD）法进行分解，提取特定高倍频振动成分作为裂纹故障的典型特征，有助于深入理解转子裂纹的变化规律，使特征提取具有明确的物理意义。在标准转子实验台上对裂纹转子经过亚临界转速区时的轴心运动轨迹演化规律和特定高倍频成分变化规律进行了实验验证。实验结果与仿真分析结果一致，验证了呼吸裂纹函数建模的正确性和基于EMD方法的裂纹故障特征提取方法的有效性。在完成裂纹转子动力学研究和特征提取之后，综合考虑裂纹转子动力学和断裂力学理论信息，建立了基于BP神经网络（Back-propagation Neural Network, BPNN）模型的性能评价模型。利用断裂力学理论分析了裂纹产生和发展的基本原理及过程，研究了裂纹尖端的应力强度因子理论及计算方法；建立了呼吸裂纹转子系统的有限元模型，并将动力学响应与断裂力学理论相结合求解出了在转子旋转过程中考虑裂纹面接触问题的动态应力强度因子，研究了整周期内应力强度因子的变化规律，为进一步全面理解呼吸裂纹的特性提供了条件；在求得动态应力强度因子的变化幅度后，利用断裂力学中经典的Paris裂纹扩展模型对特定裂纹深度的疲劳扩展寿命进行评估，作为性能评价的基础；将由EMD方法所提取的1/3和1/2亚临界转速下的特定高倍频幅值及其均值作为神经网络的联合特征输入，并将由疲劳劣化程度和疲劳扩展寿命得到性能值作为神经网络的输出，从而建立了综合考虑动力学振动特征和断裂力学信息的基于BP神经网络的性能评价模型，实现了振动特征与具有断裂力学意义的性能值之间的非线性映射。以所评价的性能为基础，根据动态预测思想，分别建立了多尺度马尔可夫剩余寿命动态预测模型和BP神经网络剩余寿命动态预测模型，针对两种模型各自的预测特性，最终建立了基于多尺度马尔可夫模型与BP神经网络模型相结合的动态预测模型。通过对裂纹转子的剩余寿命仿真验证表明，本文所提出的混合预测模型具有良好的动态预测性能，是一种行之有效的预测方法。本文以裂纹转子动力学分析为基础，结合断裂力学理论从机理上认识故障产生、发展过程和表现特征，并采用智能方法实现了性能的评价和剩余寿命预测，最终提出了一种新的裂纹转子故障预诊途径，将传统的基于模型的劣化过程分析理论引入预诊领域，为基于数据挖掘的预诊技术在工程中的应用提供理论支撑。
[Abstract]:Rotary machinery is an essential equipment for modern industrial society . It has an important influence on the important industries of national economy and national defense priorities such as aerospace and so on . In the event of a sudden failure of rotating equipment , the economic loss and social harm caused by equipment maintenance and unplanned downtime are enormous . The failure of rotor crack is one of the most common fault forms in the rotating machinery . Therefore , the failure property and the development process of the crack rotor are predicted . Therefore , corresponding measures are taken to realize preventive maintenance , reduce the accident rate and the number of unplanned downtime , and have important economic value and social significance for ensuring the system to realize high efficiency , safe , stable and low cost operation . In this paper , the key technologies of fault feature extraction , performance evaluation and residual life prediction of Jeffcott rotor system with transverse breathing crack are studied . A new method of fusion dynamics analysis , fracture mechanics theory and crack rotor pre - diagnosis based on data driving prediction method is presented . Based on the analysis of the crack breathing process of the rotor system with the main function of gravity , the dynamic model of the crack rotor system is established by using the independent and complete mathematical function , and the dynamic behavior of the rotor system is analyzed . The experimental results are consistent with the results of the simulation analysis . The validity of the modeling of the breathing crack function and the method of feature extraction based on EMD method are verified . Based on the fracture mechanics theory , the stress intensity factor theory and the calculation method of the crack tip are analyzed . The dynamic response and fracture mechanics theory are used to evaluate the fatigue life of the crack tip . The dynamic response and the fatigue propagation life are used as the output of the neural network . Based on the evaluated performance , the dynamic prediction model of residual life of multi - scale Markov model and the residual life dynamic prediction model of BP neural network are established according to the dynamic prediction idea . Based on the prediction characteristics of the two models , a dynamic prediction model based on multi - scale Markov model and BP neural network model is established . The simulation of the residual life of the crack rotor shows that the hybrid prediction model proposed in this paper has good dynamic prediction performance and is an effective prediction method . In this paper , based on the dynamic analysis of the crack rotor , the fault generation , the development process and the performance characteristic are realized by combining the theory of fracture mechanics , and the performance evaluation and the residual life prediction are realized by using the intelligent method . The traditional model - based deterioration process analysis theory is introduced into the pre - diagnosis field , which provides a theoretical support for the application of the data mining based pre - diagnosis technology in the engineering .

【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TH113;TH165.3

【参考文献】