一种基于ILCD融合与多重分形去趋势波动分析的退化特征提取方法

Method for degradation feature extraction based on the ILCD fusion and multi-fractal detrended fluctuation analysis

液压泵性能退化过程中,振动信号非线性强,导致退化特征提取困难、表征能力有限,为此,提出一种基于改良型局部特征尺度分解(ILCD)融合与多重分形去趋势波动分析(MF-DFA)的退化特征提取方法。在对信号进行ILCD分解的基础上,通过构建敏感因子从各内禀尺度分量(ISCs)筛选出包含关键故障信息的敏感分量,并依据融合规则实现多通道振动信号的融合处理,以改善重构信号中的特征信息;在此基础上,利用带有加窗分割的MF-DFA方法对融合信号作进一步处理,选取多重分形谱敏感参数作为液压泵性能退化特征向量;利用液压泵实测振动信号,验证了该方法的有效性。

In the process of hydraulic pump degradation,the vibration signal is nonlinear,from which it is hard to extract effective degradation features. Therefore,a novel method based on the Improved Local Characteristic-scale Decomposition (ILCD) fusion and Multi-fractal Detrended Fluctuation Analysis (MF-DFA) was proposed. Vibration signals were initially decomposed by ILCD. Intrinsic Scale Components were selected based on the presented sensitive factor to obtain important feature informations effectively. Multi-dimensional vibration signals were fused according to the fusion rules for improving the feature information in the reconstructed signal. Furthermore,the fused signal was processed by MF-DFA which contains the window division,and the multi-fractal spectral parameters were extracted as degradation features. Finally,the proposed method was verified by the real sampled vibration signals of a hydraulic pump.