基于变模式提取与去趋势波动分析的液压泵故障诊断研究

Hydraulic pump fault diagnosis method based on VME and DFA

由于液压泵故障识别难度大,并且检测到的液压泵故障振动信号中包含噪声等无关干扰信息,为此,提出了一种基于变模式提取和去趋势波动分析的方法,用于对液压泵故障进行识别和诊断。首先,通过变分模式提取方式,对采集到的液压泵振动信号进行了分解,得到了一系列固有模态函数分量;然后,通过去趋势波动分析方法,得到了不同模式分量的标度指数;利用标度指数的幅值阈值区分有用信号和噪声信号,并对含噪的模式分量进行了小波降噪,以最大程度地保留其中的有效信息;最后,将降噪处理后的模式分量和不含噪的模式分量进行了重构,并对重构信号进行了多统计学特征计算和局部保留投影降维。研究结果表明:变模式提取对信号进行分解后的结果,与原始不含噪信号的均方根误差仅为0.3251,该结果可为后续基于局部保留投影的液压泵不同故障类型的准确聚类提供数据支撑;同时,通过分析数值仿真信号和轴向柱塞泵液压实验系统所采集到的数据,进一步验证了所提出方法的有效性。

Aiming at the difficulty of identifying hydraulic pump faults,and the detected vibration signals containing noise and other irrelevant information,a method that combined variational mode extraction(VME)and detrended fluctuation analysis(DFA)to achieve identification and diagnosis of hydraulic pump faults was proposed.Firstly,the vibration signal collected by the hydraulic pump was decomposed by VME to obtain a series of intrinsic mode functions(IMFs),and the scale exponent of different mode components was obtained by DFA.The amplitude threshold of the scale exponent was used to distinguish useful signals and noise signals,and then the wavelet denoising was performed on the mode component with noise,so as to retain the effective information to the greatest extent.Finally,the denoised mode component and the non-noise mode component were reconstructed together.Through the Locality Preserving Projection(LPP),multiple statistical features of the reconstructed signal were calculated and the dimensionality was reduced.The results demonstrate that the root-mean-square-error(RMSE)between the decomposed signal and the original noise-free signal is only 0.3251,which provides data support for the subsequent accurate clustering of different fault types of hydraulic pumps based on LPP.At the same time,the effectiveness of the proposed method is verified by analyzing the numerical simulation signal and the data collected by the axial plunger pump hydraulic experiment system.