基于改进变分模态分解与支持向量机的风力机轴承故障诊断

Nonlinear Characteristic Analysis of Wind Turbine Bearings by SVM based on Optimized Variational Mode Decomposition

为研究风力机齿轮箱轴承振动信号非线性及故障诊断问题,采用改进变分模态分解方法对四种状态轴承振动信号进行处理,提出无量纲参数多重分形谱值因子,联合峭度值对分解所得模态分量进行选取,剔除无效信息分量并进行信号重组,采用分形维数研究重组信号的分形特征,并通过支持向量机进行模式识别。结果表明:基于样本熵优化的改进变分模态分解方法可获得高质量的模态信息;通过多重分形谱值因子及峭度选取并重组的信号具有良好的振动特性,其分形维数可通过信号非线性程度定量区分轴承工作状态;采用支持向量机对不同轴承工作状态的重组信号进行分类,结果具有较高的准确度。

In order to study the nonlinear and non-stable characteristics of the vibration signal of the wind turbine gearbox bearing,four state bearing vibration signals are processed by the optimized variational mode decomposition method(OVMD),and the dimensionless parameter multifractal spectral value factor(Mc)is proposed.The kurtosis is used to select the BIMF component obtained by decomposition,eliminate the invalid information component and perform signal recombination.And then MF-DFA method is employed to analyze the fractal characteristics of the recombined signal and perform pattern recognition.The results show that the proposed Mc results in effective BIMF which is culminated by kurtosis and improves the accuracy of component reconstruction after OVMD processing.The relationship between scale index and generalized Hurst index and time scale is reflected by the wind turbine gearbox with MF-DFA analysis.The bearing signal has fractal characteristics.Under different weighting factors,the scale index,generalized Hurst and multifractal spectrum can effectively distinguish the bearing state,and the multi-fractal feature value can quantitatively classify the bearing working state.