单通道盲源分离算法及其在工程机械振源分析中的应用

Single-channel Blind Source Separation and Its Application in Analyzing Vibration of Engineering Machinery

研究一种新的单通道盲源分离方法,解决了传统盲源分离方法因传感器数量不足而无法有效分离源信号的问题,同时源信号幅值也得到了准确的恢复。首先利用集合经验模态分解方法将一维测量信号分解为具有不同尺度特征的本征模态函数,而后与原测量信号组成多个二维矩阵,通过稀疏分量算法得到各源信号的真实估计。利用仿真信号,与已有方法进行对比,验证了提出方法的有效性。将提出的方法应用在挖掘机动力源附近的振动分析中,成功分离出了多个振源信号。利用时频分析对源信号分别进行特征识别,并将分离结果用于振源的贡献度与声信号传递特性分析。得到各振源对于测试位置的贡献度排序,以及对于挖掘机噪声信号的传递规律,为挖掘机的减振降噪措施提供了可靠的依据。

A novel blind source separation method is proposed, which can separate sources with single channel measurement signal. And the amplitude of sources can be recovered accurately. First of all, one-dimensional measurement signal is decomposed into several intrinsic mode functions（IMF） by ensemble empirical mode decomposition, and combined with corresponding IMF into multiple two-dimensional matrixes. Then the sources can be separated by sparse component analysis method effectively. The effectiveness of proposed method is verified by simulated data. In practical application, the proposed method is utilized for separating sources by one-dimensional measurement signal recorded from an excavator. Then based on the separated sources, the contribution characteristics and transfer behaviors are analyzed. The analysis results can be utilized to guide the reduction of excavator vibration and noise.