中文摘要：

炼胶机齿轮箱发生早期故障时,其振动信号一般很微弱,且隐含的冲击成分常被淹没在强烈的噪声中,导致齿轮故障诊断异常艰难。论文介绍了M步时延相关峭度的概念,并引出了最大相关峭度解卷积方法。该方法通过计算故障信号的最大相关峭度值来估算出感兴趣的解卷积故障周期T,然后选择合适的时延步数M对故障信号做最大相关峭度解卷积,最后对最大相关峭度解卷积滤波后的信号进行包络解调以提取出其故障特征,并诊断出了该齿轮箱轴V上的齿轮8（Z_8=28）的微弱裂纹故障。最后还将最大相关峭度解卷积方法与谱峭度方法进行了对比分析。应用实例结果与对比分析验证了最大相关峭度解卷积方法应用于齿轮箱早期故障诊断的有效性。

英文摘要：

When refiner gearbox has incipient failure, its vibration signal, with hidden impulse component veiled by strong background noise, is usually very weak, which results in difficultly of gear fault diagnosis. In this paper, the concept of M-shift correlated kurtosis （Kc,M ） was introduced, and the method of maximum correlated kurtosis deconvolution （MCKD） was derived. Firstly, estimate interest period T of deeonvolution through calculating the maximum Kc,M value for the faulty signal. Then, select the appropriate M-shift for the maximum correlated kurtosis deconvolution. Finally, through envelope demodulation analysis to the MCKD filtered signal, the fault feature of root slight crack from gear 8 （ Z8 = 28 ） on the shaft V was successfully extracted. In addition, the MCKD method was compared to the method of fast spectral kurtosis. The application result and comparison analysis verify the effectiveness of the method of maximum correlated kurtosis deconvolution （MCKD） applied to incipient fault diagnosis of gearbox.