摘要
轴心轨迹是转子系统故障诊断的重要依据,将整周期重采样、归一化的极半径序列引入轴心轨迹自动识别系统。首先对振动信号进行整周期重采样以降低转速和采样频率对小波去噪效果的影响,然后利用小波变换对其去噪并合成提纯的轴心轨迹,最后计算具有平移、伸缩和旋转不变性的极半径序列作为轴心轨迹特征,采用BP神经网络进行识别。实验结果表明该方法具有良好的识别效果。
Rotor orbit is an important basis for rotor system fault diagnosis. Full period re sampling and normalized polar radius sequence are introduced into the automatic recognition system. Firstly, vibration signals are processed with full period resampling to reduce the influence on the wavelet de-noising effect by rotation speed and sampling frequency. Then, they are de-noised with wavelet transform to synthetic purified orbit. Lastly, normalized polar radius sequence which is invariant to translation, scaling and rotation of the rotor orbit is calculated and used as the orbit feature. Rotor orbits are identified using the BP neural network. Experimental results show that the method has good recognition effect.
出处
《中国测试》
CAS
北大核心
2014年第1期110-114,共5页
China Measurement & Test
基金
国家自然科学基金项目(11072190)
关键词
轴心轨迹
整周期重采样
特征提取
极半径
自动识别
rotor orbit
full period re-sampling
feature extraction
polar radius
automatic identification
