基于GWO-TVF-EMD方法的行星齿轮箱齿面剥落故障诊断

针对时变滤波经验模态分解（TVF-EMD）方法的不足之处，将样本熵作为适应度函数，采用灰狼优化（GWO）算法对带宽阈值和B样条阶数核心参数进行寻优，得到最优组合解，对不同的故障冲击试验振动信号进行分解。对本征模态函数（IMF）分量选取过程进行优化，采用多个加权指标对所有IMF分量进行计算，最终选取最优IMF分量，再通过包络谱分析提取出行星轮齿面剥落故障特征。在行星齿轮箱故障试验中，利用方均根法对剥落故障进行初步识别，根据GWO-TVF-EMD法分解得到各剥落故障信号最优IMF分量，使用包络谱分析明显判断出行星齿轮的故障频率。该方法能够提取3种不同程度齿面剥落故障的细节特征，理论值与实际值的相对误差为1.68%。 

Fault diagnosis of tooth surface spalling of planetary gearbox based on GWO-TVF-EMD method

In order to overcome the shortcomings of the time varying filter for empirical mode decomposition （TVF-EMD） method，the sample entropy was used as the fitness function.The gray wolf optimization （GWO） algorithm was used to optimize the bandwidth threshold and B-spline order core parameters，and the optimal combination solution was obtained to decompose the vibration signals of different fault impact tests.The selection process of intrinsic mode function （IMF） components was optimized，and multiple weighted indexes were used to calculate all the IMF components，and finally the optimal IMF component was selected.Then，the spalling fault features of planetary gear were extracted by envelope spectrum analysis.In the planetary gear box fault test，the root mean square method was used to identify the spalling fault initially，and the optimal IMF component of each spalling fault signal was obtained by GWO-TVF-EMD，and the envelope spectrum analysis was used to judge the fault frequency of the planetary gear clearly.The method can extract the detailed features of three different degrees of tooth surface spalling faults，and the relative error between the theoretical value and the actual value was 1.68%.