中文摘要：

针对在欠定的观测信号情况下，传统基于矩阵的盲源分离算法效果比较差的问题，提出一种基于极值域均值模式分解和盲源分离的单通道旋转机械信号故障特征提取方法，并应用于实际的故障诊断中。该方法先通过极值域均值模式分解法分解观测信号，把得到的固有模态函数和原观测信号一起组成新观测信号，从而实现了信号升维，使欠定问题转化为正定问题；然后，由奇异值分解和贝叶斯准则进行源数估计；最后，利用基于四阶累积量的特征矩阵联合对角化方法实现信号的盲分离。通过仿真，验证了该方法对旋转机械故障信号进行盲源分离的可行性。将提出的方法应用到齿轮和轴承系统的故障诊断中，进一步证明了该方法的有效性。

英文摘要：

Aiming at the problem that in underdetermined observed signal condition, the traditional blind source separation （BSS） method based on matrix calculation produces poor result, a new fault feature extraction method for rotating machinery is proposed, which is based on blind source separation and extremum field mean mode decomposition （EMMD） ;and then the new method is applied to practical fault diagnosis. Firstly, the new method uses extremum field mean mode decomposition （EMMD） to decompose the observed signal into a series of intrinsic mode functions （IMFs） ;then the obtained IMFs and the original observed signal are combined to compose new observed signal, the dimension of the new observed signal is increased;and the underdetermined BSS problem is transformed into a posi- tive definite problem. Secondly, singular value decomposition and Bayesian information criterion are used to estimate the number of source signals. Finally, the characteristic matrix joint diagonalization method based on fourth-order cumulant is adopted to solve the BSS problem. Simulation result shows the feasibility of the method in the BSS problem of the rotating machinery fault signal. The proposed method was applied to the fault diagnosis of a gear and bearing system, and the result further verifies its effectiveness.