A multi-order FRFT self-adaptive filter based on segmental frequency fitting and early fault diagnosis in gears
A multi-order FRFT self-adaptive filter based on segmental frequency fitting and early fault diagnosis in gears
复制标题
DOI：
10.1016/j.measurement.2016.05.092
复制
发表时间：
2016-09
期刊：
影响因子：
5.6
通讯作者：
Jianmin Mei;Jia Jide;Ruili Zeng;Bin Zhou;Huimin Zhao
中科院分区：
文献类型：
--
作者：
Jianmin Mei;Jia Jide;Ruili Zeng;Bin Zhou;Huimin Zhao
为了有效地早期诊断齿轮故障，提出了一种基于分段频率拟合的多阶分数阶傅里叶变换（FRFT）自适应滤波器（MFSFF），从齿轮箱瞬态工况中分离出具有曲线频率的特征分量。首先，提出一种线性多尺度分割方法（LMSS），将频率曲线信号分割成频率近似线性的分段;然后，提出一种频率曲线拟合确定FRFT滤波器参数的方法（DFFPFF），计算每个分段信号的FRFT滤波器参数，并利用这些参数对每个分段信号进行滤波，确定MFSFF。利用MFSFF对齿轮箱加减速过程的振动进行了分析，并对滤波后的信号进行了解调。实验结果表明，LMSS能够将任意曲线频率的信号分割成频率接近线性的最小段，DFFPFF精确、快速，不受振源和分量个数的影响，能够准确地确定每段信号的FRFT滤波器参数; MFSFF准确地提取了齿轮箱瞬态工况的特征分量，同时去除了其它分量和噪声。对传统方法难以识别的齿轮早期故障进行解调，能够准确诊断齿轮早期故障。
To effectively diagnose gear failure at an early stage, a multi-order Fractional Fourier transform (FRFT) self-adaptive filter based on segmental frequency fitting (MFSFF) is proposed to separate the feature components with curved frequency from the gearbox’s transient conditions. First, a linear multi-scale segmentation method (LMSS) is developed to divide the signal with curved frequency into segments with nearly linear frequency; then, a method for determining the FRFT filter parameters by fitting the frequency curve (DFFPFF) is developed to calculate the FRFT filter parameters for each signal segment, and the signal in each segment is filtered by an FRFT filter using these parameters to determine the MFSFF. The vibration of the gearbox’s acceleration and deceleration process is analyzed using an MFSFF and the filtered signal is demodulated. The experimental results show that LMSS is able to divide any signal with curved frequency into minimal segments with nearly linear frequency; DFFPFF is exact, fast, not influenced by the vibration source or the number of components, and able to determine the FRFT filter parameters for each signal segment accurately; the feature component of the gearbox’s transient condition is accurately extracted by an MFSFF, and the other components and noise are removed simultaneously. Early gear failure is diagnosed exactly by demodulation of the extracted feature component, which is difficult to identify using the traditional method.