采用改进多点最优最小熵反褶积的齿轮箱复合故障特征提取

Feature Extraction of Gearbox Composite Fault Based on the Improved Multipoint Optimal Minimum Entropy Deconvolution

针对强背景噪声下齿轮箱复合故障中的微弱故障特征难以提取的问题,提出了一种改进多点最优最小熵反褶积的齿轮箱复合故障特征提取方法。将小波降噪作为前置滤波器,克服了多点峭度谱在强背景噪声下难以有效识别信号中的微弱故障周期成分的缺点,使信号峭度增加了65.9%,突出了微弱故障周期成分;根据多点峭度谱识别出的故障周期成分设置合理的故障区间,利用多点最优最小熵反褶积突出了信号中的故障周期,避免了对信号直接包络解调而出现的漏诊现象;将差分能量算子解调应用于改进算法处理后的信号,与传统的Hilbert解调方法相比,该算法得到的解调谱中故障特征频率的峰值更加明显。通过对仿真信号与齿轮箱中齿轮点蚀磨损复合故障振动信号的研究结果表明,改进多点最优最小熵反褶积方法能够完整地提取信号中的故障特征频率成分,成功率达到了100%;与变分模态分解进行了对比分析,改进算法有效消除了模态混叠现象。仿真和试验结果表明,改进算法可以有效提取强背景噪声下齿轮箱复合故障中的微弱故障特征。

In order to solve the problem that it is difficult to extract the weak feature from gearbox complex fault signal under strong background noise,an improved multipoint optimal minimum entropy deconvolution method is proposed.The wavelet denoising is used as a pre-filter to effectively identify the weak fault period components in the multipoint kurtosis spectrum under strong background noise.The kurtosis of the signal is increased by 65.9%and the weak fault period components are highlighted.According to the fault period components identified by the multipoint kurtosis spectrum,a reasonable fault interval is set,and the fault period in the signal is enhanced by using the multipoint optimal minimum entropy deconvolution,so that the missed diagnosis caused by the direct envelope demodulation of the signal is avoided.The differential energy operator demodulation is applied to the signal enhanced by the improved algorithm.Compared with the Hilbert demodulation method,the improved algorithm obtains clearer peak value of fault characteristic frequency.Analysis on the simulation signal and vibration signal of gear pitting-wear composite fault shows that the proposed method can completely extract the fault characteristic frequency components in the signal and the success rate is up to 100%.Compared with the variational modal decomposition,the improved algorithm effectively eliminates the phenomenon of modal aliasing.Simulation and experimental results show that the improved algorithm can effectively extract the weak composite fault features of the gearbox under strong background noise.