论文部分内容阅读
针对基于配分函数的多重分形分析不利于局部标度特性突显的问题,把多重分形去趋势波动分析(MF-DFA)方法引入到振动诊断领域,提出对振动信号进行多重分形谱参数(|B|,α0,Δα和Δf)故障特征分析,并将α0用于故障诊断.首先分析了振动信号的多重分形特性;然后提取振动信号的4种多重分形谱参数特征,并进行了比较;最后用支持向量机算法实现振动故障诊断.研究表明:去除趋势后,振动信号的波动呈现显著多重分形特征,正常状态振动信号的α0明显大于故障状态,而振动信号的|B|,Δα和Δf特征变化规律则不明显;α0作为故障特征量,能有效地区分正常状态与故障状态,有效实现了振动故障诊断.
Aiming at the problem that the multi-fractal analysis based on the partition function is unfavorable to highlight the local scale features, the multi-fractal de-trend analysis (MF-DFA) method is introduced into the field of vibration diagnosis, and the multi-fractal spectral parameters (| B | , α0, Δα and Δf), and α0 is used in fault diagnosis.Firstly, the multifractal characteristics of vibration signals are analyzed. Then, the parameters of the four kinds of multifractal spectra of vibration signals are extracted and compared. Finally, The vector machine algorithm is used to realize the vibration fault diagnosis.The research shows that the vibration signal shows a significant multifractal characteristic after removing the trend, the α0 of the normal state vibration signal is obviously larger than the fault state, and the variation law of | B |, Δα and Δf of the vibration signal It is not obvious. As a fault feature quantity, α0 can effectively distinguish the normal state from the fault state and effectively realize the vibration fault diagnosis.