基于LMS算法的广义形态滤波器应用于转子分形故障状态识别的效果研究

Result about Rotor Fractal Fault Identification of Applying Generalized Morphological Filter Based on LMS Algorithm

将描述混沌运动的关联维数用于转子系统振动信号分析,并针对现场实测振动信号中存在噪声污染,提出了一种基于自适应形态滤波对信号进行降噪处理。该滤波器综合了不同的结构元素且在开—闭、闭—开滤波器权系数的确定上采用了基于最小均方误差算法的自适应技术,具有更好的滤波性能和细节信息保留能力;分析比较降噪前后转子4种状态(正常、油膜涡动、局部碰摩、全周碰摩)下信号的关联维数,结果表明,用含有噪声信号的关联维数描述系统的特征行为不可靠,有必要对实测信号进行降噪处理。自适应形态滤波器具有良好的降噪效果,将降噪后的关联维数作为特征量,在转子故障诊断中进行识别故障、判断运行状态具有可行性。

The correlation dimension describing chaotic motion is used in analysis of rotor vibration signal.Aiming at the noise pollution existing in the on-site measured vibration signal,the article presents a new method of noise reduction using self-adaptive morphological filter.The filter combines with the different structural elements and uses adaptive technology based on minimum mean square error algorithm on the determination of weight coefficient of the open-close,close-open filter;it has better filtering performance and the ability to retain detailed information.The paper analyzes and compares the correlation dimension of information under the four conditions such as normal,oil whirl,local rub-impact and all-round rub-impact,and the result shows that it is not reliable to describe the characteristic behavior of system by correlation dimension containing noise information;it is necessary to conduct noise-reduction on the measured signal.The effect about noise reduction of adaptive morphological filter is good.It is feasible to take the correlation dimension after the noise reduction as the characteristic quantity for identifying fault in the rotor fault diagnosis,determining the operating state.