涡轮钻具滚动轴承故障诊断系统的研究
发布时间：2018-11-05 14:33
【摘要】：在旋转机械中,轴承的失效可能会引起系统故障。到目前为止,有很多基于振动的方法来监测轴承状态,在这些方法中都很少考虑到轴承振动的自身特性。在学位论文中,对正常轴承系统进行了研究分析,对轴承振动的不同类型有了全新的认识。在研究过程中,将轴承的滚动体与滚道之间的接触设定为非线性弹簧,将系统转换成一个2-自由度模型。通过研究分析确定了内圈的振动特性主要取决于轴承的内间隙。轴承故障产生的周期行为会影响到它的混沌行为,并在庞加莱映射中得以体现。同时,混沌的监测指标如李雅普诺夫指数、关联维数以及归一化信息熵都会发生变化。为了比较故障诊断技术中监测指标的灵敏度和鲁棒性,进行了全面的实验分析。实验结果也充分说明了关联维数、归一化信息熵和小波近似最大系数是轴承故障监测的可靠指标。在论文中,提出了模糊神经诊断系统。为了提高轴承故障诊断的可靠性,在新的诊断系统中,对上述的监测指标进行整合。同时,基于自适应模糊神经推理系统,提出了轴承预测方案,结合事先约定好的逻辑准则,通过理论与实验研究,证实了这种预测方案可以用来评估轴承的下一步工作状态。并通过实验验证了该预测方案在变转速和变载荷工况下的有效性和可靠性。通过本学位论文的研究,主要成果总结如下:1)证实了轴承运动平衡点的个数取决于轴承内间隙。小间隙轴承存在着周期运动,并具有唯一一个平衡点;而对于大间隙的轴承来说,在每一时帧都会有3个平衡点,而且会将相空间划分为一个不稳定的区域和两个稳定的区域。对于高转速的工况,轴承的内圈会发生跳动,从一个稳定的区域跳动到另一个区域,并存在着混沌行为。2)实验与数值模拟仿真的结果证实了正常间隙的滚动轴承和推力球轴承在变转速的工况下存在着宽频混沌振动。此外,对于正常的轴承系统来说,随着脉冲的倍增和混沌振动的干扰,轴承会出现故障。实验结果与数值模拟仿真结果的一致性也证实了轴承故障会严重的影响到混沌监测指标:李雅普诺夫指数、关联维数和归一化信息熵。3)基于神经网络提出轴承故障的诊断系统。将上述监测指标作为诊断层而输入,而输出的结果就对应于轴承的相关工作状态或故障模式。对比分析指出,自适应模糊神经推理系统能更有效的映射出轴承的工作状态。4)针对涡轮钻具滚动轴承系统,研究了神经网络两种可行的方法(回归神经网络和自适应模糊神经推理系统),并评估了它们对轴承系统工作状态预判性能的高低。研究表明,一旦被轴承振动数据训练后的自适应模糊神经推理系统,神经网络可以准确的获取轴承故障扩散的信息。这种被训练过的方法可以有效的用于对轴承未来的工作状态进行预测,而且适用于变转速变载荷的工况中。同时,结合所提出的轴承预判模型,对测试轴承的305个案例进行未来工作状态的评估,准确率达到98%以上。
[Abstract]:Failure of a bearing may cause a system failure in a rotating machine. So far, there are a number of vibration-based methods to monitor the bearing state, in which little consideration is given to the self-characteristics of the bearing vibration. In this paper, the normal bearing system is studied and analyzed, and the different types of bearing vibration have a new understanding. During the study, the contact between the rolling body of the bearing and the raceway is set as a non-linear spring, and the system is converted into a 2-degree-of-freedom model. Through the research and analysis, it is determined that the vibration characteristics of the inner ring mainly depend on the internal clearance of the bearing. The periodic behavior of bearing fault can affect its dynamic behavior, and it can be embodied in the Poincare map. At the same time, the monitoring indexes such as Lyapunov exponent, correlation dimension number and normalized information entropy change. In order to compare the sensitivity and robustness of the monitoring indexes in the fault diagnosis technology, a comprehensive experimental analysis was carried out. The experimental results also show that the correlation dimension number, normalized information entropy and wavelet approximate maximum coefficient are reliable indexes of bearing fault monitoring. In this paper, a fuzzy neural diagnostic system is proposed. In order to improve the reliability of bearing fault diagnosis, in the new diagnosis system, the above-mentioned monitoring indexes are integrated. At the same time, based on the adaptive fuzzy neural reasoning system, a bearing prediction scheme is put forward, combining with the pre-agreed logical criterion, and through theoretical and experimental research, it is confirmed that this kind of prediction scheme can be used to evaluate the next working state of the bearing. The validity and reliability of the prediction scheme under varying rotating speed and variable load condition are verified by experiments. Through the study of this degree thesis, the main results are summarized as follows: 1) The number of the bearing motion balance points is confirmed to be dependent on the bearing inner clearance. The small-gap bearing has a periodic motion and has a unique balance point; and for large-gap bearings, there are three equilibrium points at each time frame, and the phase space is divided into an unstable region and two stable regions. for high rotational speeds, the inner ring of the bearing will bounce and jump from a stable region to the other, The results of experiment and numerical simulation show that the rolling bearing and thrust ball bearing of normal gap have broadband vibration under the condition of variable speed. In addition, for normal bearing systems, the bearings will fail with the multiplication of the pulses and the interference of the backlash. The consistency between the experimental results and the numerical simulation results also confirms that the bearing fault can seriously affect the fault monitoring index: Lyapunov exponent, correlation dimension number and normalized information entropy. 3) Fault diagnosis system based on neural network is proposed. The above-mentioned monitoring index is input as a diagnostic layer, and the result of the output corresponds to the relevant operating state or failure mode of the bearing. According to the comparative analysis, the adaptive fuzzy neural reasoning system can map out the working state of the bearing more effectively. 4) According to the rolling bearing system of the drilling tool, two feasible methods of neural network (regression neural network and adaptive fuzzy neural inference system) are studied. The pre-judging performance of bearing system working state is evaluated and evaluated. The research shows that, once the adaptive fuzzy neural reasoning system after the bearing vibration data is trained, the neural network can accurately acquire the information about the bearing fault diffusion. The trained method can be used for predicting the working state of the bearing in the future, and is suitable for working conditions of variable speed variable load. At the same time, in combination with the proposed bearing prejudging model, 305 cases of the test bearing are evaluated in the future, and the accuracy is above 98%.
【学位授予单位】：西南石油大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TE921.2;TH133.33
本文编号：2312375
[Abstract]:Failure of a bearing may cause a system failure in a rotating machine. So far, there are a number of vibration-based methods to monitor the bearing state, in which little consideration is given to the self-characteristics of the bearing vibration. In this paper, the normal bearing system is studied and analyzed, and the different types of bearing vibration have a new understanding. During the study, the contact between the rolling body of the bearing and the raceway is set as a non-linear spring, and the system is converted into a 2-degree-of-freedom model. Through the research and analysis, it is determined that the vibration characteristics of the inner ring mainly depend on the internal clearance of the bearing. The periodic behavior of bearing fault can affect its dynamic behavior, and it can be embodied in the Poincare map. At the same time, the monitoring indexes such as Lyapunov exponent, correlation dimension number and normalized information entropy change. In order to compare the sensitivity and robustness of the monitoring indexes in the fault diagnosis technology, a comprehensive experimental analysis was carried out. The experimental results also show that the correlation dimension number, normalized information entropy and wavelet approximate maximum coefficient are reliable indexes of bearing fault monitoring. In this paper, a fuzzy neural diagnostic system is proposed. In order to improve the reliability of bearing fault diagnosis, in the new diagnosis system, the above-mentioned monitoring indexes are integrated. At the same time, based on the adaptive fuzzy neural reasoning system, a bearing prediction scheme is put forward, combining with the pre-agreed logical criterion, and through theoretical and experimental research, it is confirmed that this kind of prediction scheme can be used to evaluate the next working state of the bearing. The validity and reliability of the prediction scheme under varying rotating speed and variable load condition are verified by experiments. Through the study of this degree thesis, the main results are summarized as follows: 1) The number of the bearing motion balance points is confirmed to be dependent on the bearing inner clearance. The small-gap bearing has a periodic motion and has a unique balance point; and for large-gap bearings, there are three equilibrium points at each time frame, and the phase space is divided into an unstable region and two stable regions. for high rotational speeds, the inner ring of the bearing will bounce and jump from a stable region to the other, The results of experiment and numerical simulation show that the rolling bearing and thrust ball bearing of normal gap have broadband vibration under the condition of variable speed. In addition, for normal bearing systems, the bearings will fail with the multiplication of the pulses and the interference of the backlash. The consistency between the experimental results and the numerical simulation results also confirms that the bearing fault can seriously affect the fault monitoring index: Lyapunov exponent, correlation dimension number and normalized information entropy. 3) Fault diagnosis system based on neural network is proposed. The above-mentioned monitoring index is input as a diagnostic layer, and the result of the output corresponds to the relevant operating state or failure mode of the bearing. According to the comparative analysis, the adaptive fuzzy neural reasoning system can map out the working state of the bearing more effectively. 4) According to the rolling bearing system of the drilling tool, two feasible methods of neural network (regression neural network and adaptive fuzzy neural inference system) are studied. The pre-judging performance of bearing system working state is evaluated and evaluated. The research shows that, once the adaptive fuzzy neural reasoning system after the bearing vibration data is trained, the neural network can accurately acquire the information about the bearing fault diffusion. The trained method can be used for predicting the working state of the bearing in the future, and is suitable for working conditions of variable speed variable load. At the same time, in combination with the proposed bearing prejudging model, 305 cases of the test bearing are evaluated in the future, and the accuracy is above 98%.
【学位授予单位】：西南石油大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TE921.2;TH133.33
【参考文献】
相关期刊论文 前10条
1 邱占芝;吴婷婷;;基于小波变换和支持向量机缆力预测[J];大连交通大学学报;2016年05期
2 陈江;张飞虎;;基于磨削速度的K9光学玻璃平面磨削亚表面裂纹深度研究[J];金刚石与磨料磨具工程;2016年04期
3 卢树泉;曹炳元;;含二次隶属函数的模糊二次规划模型求解[J];湖南文理学院学报(自然科学版);2016年03期
4 黄建明;瞿合祚;李晓明;;基于短时傅里叶变换及其谱峭度的电能质量混合扰动分类[J];电网技术;2016年10期
5 莫佳亮;周储伟;雷晓恒;;一种微裂纹摩擦阻尼分析模型[J];机械制造与自动化;2016年04期
6 马朝永;张学飞;胥永刚;;形态分量分析和谱峭度在轴承故障诊断中的应用[J];噪声与振动控制;2016年04期
7 张颖;屈剑锋;任浩;;传感器网络同步态的节点故障诊断算法[J];重庆大学学报;2016年04期
8 李刚;赵刚;;基于更新径向基函数网络模型的广义Pareto分布函数拟合[J];计算力学学报;2016年04期
9 吕中亮;汤宝平;周忆;孟杰;;基于网格搜索法优化最大相关峭度反卷积的滚动轴承早期故障诊断方法[J];振动与冲击;2016年15期
10 刘畅安;胡芳仁;刘昕;;基于小波变换的载波相位恢复算法的研究[J];微型机与应用;2016年14期
，本文编号：2312375
本文链接：https://www.wllwen.com/jixiegongchenglunwen/2312375.html
