【摘要】 受环境噪声、传递路径、信号衰减以及源信号本身比较微弱的影响,滚动轴承早期微弱冲击性故障的信号特征难以提取。近年来,最小熵解卷积(Minimum Entropy Deconvolution,MED)已经成功应用在旋转机械故障检测中来提取振动冲击。MED方法的提取过程是一个迭代选择的过程,通过迭代选择一个有限脉冲响应使信号的熵最小,从而对信号进行滤波。但是该方法有一定的局限性:其对于单一冲击的信号解卷积效果良好,但是处理具有强噪声或者多个冲击源共同作用时的信号很困难。为了解决这个问题,提出新的解卷积方法:最大相关峭度解卷积(Maximum Correlated Kurtosis Deconvolution,MCKD),可有效利用滚动轴承故障周期性冲击的特点,其与MED相比,克服了单一冲击的限制,对两种冲击源甚至是多种共同卷积的解卷积具有更好的特征提取效果。仿真和实验对比验证了该方法具有良好的降噪和故障特征增强效果。更多还原

【Abstract】 Due to the influence of environmental noise, transmission path, signal attenuation and the weakness of the source signal itself, it is difficult to extract the signal characteristics of early weak impact faults of rolling bearings. In recent years, Minimum Entropy Deconvolution(MED) has been successfully applied to fault detection of rotating machinery to extract vibration shock. The MED method is an iterative selection process, which minimizes the entropy of the signal by selecting a finite impulse response iteratively to filter the signal. However, this method has some limitations: MED method has good deconvolution effect for single impulse signal, but it is difficult for signal processing with strong noise or multiple impulse sources. In order to solve this problem, this paper proposes a new deconvolution method, the maximum correlation kurtosis deconvolution(MCKD) method. This method effectively utilizes the characteristics of periodic impact of rolling bearing faults. Compared with MED, it overcomes the limitation of single impact. It even has better feature extraction effect for deconvolution of two or multi impact sources convolution. The comparison of simulation results with experiment results proves good noise reduction and fault feature enhancement effects of this method.更多还原