中文摘要：

在分析时域平均的基础上，通过将齿轮箱输入轴瞬时速度的三次曲线拟合和信号重采样相结合，提出了非同步特征信号的对域平均提取方法。时域平均对从噪声中提取特定周期信号是比较有效的方法，但当信号的故障特征频率未知或存在多种故障时，该方法不能获得满意的诊断效果，而频域平均能较好的实现该类故障的特征提取。基于早期故障信号的特点，提出了将多分辨分析的经验模态分解方法与频域平均相结合的故障特征提取方法。试验结果表明，所提出的时频域平均方法能有效地实现齿轮箱非同步特征信号的提取。

英文摘要：

The vibration signal of rotating machineries possesses the characteristics of ergodic cycle, time-frequency domain average can deal with the cycle domain of this kind of stationary signal better and extract the characteristic signal. The fault characteristics of vibration signal would be more obvious and easy to detect when rotating machineries are at nonsynchronous status. Via the analysis of time synchronous averaging, a new method for the rime domain averaging fault diagnosis of nonsynchronous feature signal is proposed, which is based on cubic curves fitting of the instant speed of the gearbox input shaft and the data resampting. Time domain averaging will not achieve satisfactory consequence if the fault characteristic frequency is unknown or there are various faults, while frequency average can extract this kind of fault effectively. Based on the characteristic of earlier faulty signal, this paper proposes a new fault diagnosis method, which combines the method of multi-resolution empirical mode decomposition （MEMD） and frequency domain average. The application results show that the proposed time-frequency domain average method can extract the nonsynchronous feature signal of gearbox effectively.