基于小波和神经网络的旋转机械故障诊断研究

发布时间：2019-07-09 08:07

【摘要】：旋转机械广泛应用于机械、冶金、电力、化工等行业。如果旋转机械发生故障而未及时控制和消除，则可能导致设备损坏，不仅造成巨大的经济损失，甚至会危及人身安全，，后果极为严重，因此对于旋转机械的故障诊断具有重要的意义。由于转子的轴心轨迹可以反映转子故障类型，本课题从轴心轨迹入手研究转子故障的自动诊断方法。具体内容包括： (1)轴心轨迹提纯：利用小波方法对轴心轨迹进行提纯。首先对比了选用不同的小波函数以及分解到不同的层数时提纯效果的差别，并选取出最佳的小波基和分解层数；然后根据转子振动信号的特点，分析了采样频率和转速对提纯效果的影响，并采用对信号重采样的方法降低这种影响，保证提纯效果； (2)轴心轨迹特征提取：在轴心轨迹提纯时重采样的基础上进行下采样，计算轴心轨迹在各个旋转周期内的极径（轴心轨迹上各点到基点O的距离）并对其进行平移、伸缩、旋转归一化，采用归一化的极径序列作为轴心轨迹的特征。通过仿真试验验证了归一化的极径序列对于不同轴心轨迹具有不同的变化特征，而且下采样后的极径序列在每个旋转周期有相同的样本长度，便于下一步识别； (3)轴心轨迹的识别：采用轴心轨迹的极径序列作为样本，讨论了建立相应的BP神经网络的过程，并用其对样本进行训练和识别，结果表明该网络能有效区分不同类型的轴心轨迹； (4)试验验证：在转子-轴承试验台上，测得几组实际故障的轴心轨迹，并对实测信号进行提纯、特征提取和识别，结果表明本文的方法是有效的。
[Abstract]:Rotating machinery is widely used in machinery, metallurgy, power, chemical and other industries. If the failure of rotating machinery is not controlled and eliminated in time, it may lead to equipment damage, which may not only cause huge economic losses, but even endanger personal safety, and the consequences are very serious, so it is of great significance for the fault diagnosis of rotating machinery. Because the axial trajectory of rotor can reflect the type of rotor fault, the automatic diagnosis method of rotor fault is studied in this paper. The main contents are as follows: (1) Axis trajectory purification: wavelet method is used to purify the axis trajectory. Firstly, the difference of purification effect when different wavelet functions are selected and decomposed into different layers is compared, and the best wavelet basis and decomposition layer number are selected. Then, according to the characteristics of rotor vibration signal, the influence of sampling frequency and rotating speed on purification effect is analyzed, and the method of signal resampling is used to reduce this effect and ensure the purification effect. (2) feature extraction of axis trajectory: on the basis of resampling when the axis trajectory is purified, the polar diameter of the axis trajectory in each rotation cycle (the distance from each point on the axis trajectory to the base point O) is calculated, and its translation, expansion and rotation normalization are carried out. The normalized polar diameter sequence is used as the characteristics of the axis trajectory. The simulation results show that the normalized polar diameter sequence has different variation characteristics for different axis trajectories, and the undersampled polar diameter sequence has the same sample length in each rotation cycle, which is convenient for the next step identification. (3) Identification of axis trajectory: using the polar diameter sequence of axis trajectory as sample, the process of establishing corresponding BP neural network is discussed, and the samples are trained and identified. The results show that the network can effectively distinguish different types of axis trajectory. (4) the experimental results show that the axial center trajectories of several groups of actual faults are measured on the rotor-bearing test-bed, and the measured signals are purified, feature extraction and recognition are carried out. the results show that the method in this paper is effective.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【参考文献】