针对强噪声环境下滚动轴承微弱信号易被淹没，其识别缺乏数学理论基础的问题，基于分形理论提出一种改进变分模态分解(Improved Variational Mode Decomposition,IVMD)与最大相关峭度解卷积(Maximum correlated kurtosis deconvolution,MCKD)相结合的轴承早期故障识别方法。采用灰狼算法(Grey Wolf Optimizer,GWO)优化VMD参数，分形筛选最优分量，MCKD算法突显信号中的冲击成分，对其进行包络谱分析实现故障诊断。与其它方法相比，IVMDMCKD方法可较好突显故障特征频率及其倍频，实现滚动轴承早期微弱故障诊断。

In order to solve the problems that weak signals of rolling bearings are easily drowned in strong noise environment and its identification lacks the mathematical theoritical foundation, a novel early fault identification method of rolling bearings combining with improved variational mode decomposition (IVMD) and maximum correlated kurtosis deconvolution (MCKD) was proposed based on fractal theory. Grey wolf optimizer (GWO) was used to optimize the VMD parameters and filter the optimal fractal components. MCKD algorithm was used to highlight the impact components in the signal, and the envelope spectrum was used to analyze the VMD parameters to realize fault diagnosis. Compared with other methods, the IVMDMCKD method can better emphasize the fault characteristic frequency and its frequency multiplier, realizing the early weak fault diagnosis of rolling bearings.

TH133

国家自然科学基金（51976131，52006148）；上海市“科技创新行动计划”地方院校能力建设项目（19060502200）