风力机齿轮箱轴承故障信号具有典型非线性及非平稳特性,采用自适应变分模态分解对四种状态下振动信号进行分解,提出基于分形盒维数-峭度阈值法对AVMD处理所得分量进行筛选,选取富含故障信息的分量进行信号重构,采用多重分形去趋势波分析方法,分析重构信号的分形特征并识别其工作状态,结果表明：基于多重分形去趋势波分析法对非稳定轴承可进行有效地故障识别。轴承振动信号具有典型分形特征,在不同时间尺度下,标度指数、广义Hurst指数与多重分形谱均可反应轴承工作状态。三种多重分形谱参数对故障类型敏感度不同,谱函数最大值对应的奇异指数对内圈故障较为敏感,峰值占比对外圈故障较为敏感,分形谱宽对滚珠故障较为敏感。

The bearing fault signal of wind turbine gearbox has typical nonlinear and non-stationary characteristics. Firstly, the vibration signal of four states is decomposed by adaptive variational mode decomposition, and the component obtained by decomposition based on the fractal-dimensional kurtosis threshold method is proposed. The screening is performed, and the component with rich fault information is selected for signal reconstruction. The multi-fractal and trend wave analysis method is used to analyze the fractal characteristics of the reconstructed signal and identify its working state. The results show that the multi-fractal de-warping wave analysis method is not Stable bearings provide effective fault identification. The bearing vibration signal has typical fractal characteristics. The scale index, generalized Hurst index and multifractal spectrum can reflect the working state of the bearing at different time scales. The three multifractal spectral parameters have different sensitivity to the fault type. The singular exponent corresponding to the spectral function"s maximum value is sensitive to the inner loop fault, and the peak ratio is more sensitive to the outer loop fault. The fractal spectrum width is more sensitive to the ball fault.

国家自然科学基金资助项目