基于盲信号提取的机械振动信号消噪方法研究

实际环境中，利用传感器测取的机械振动信号中，不可避免的混有来自相邻机械设备及周围环境的噪声干扰，这些噪声对于利用振动信号进行状态检测和故障诊断是很不利的。为了消除机械振动信号中的噪声，本文利用盲信号提取的特征提取特性，提出了一种新颖的噪声消除方法。该方法利用高阶累积量近似信号峭度建立代价函数，通过计算代价函数的梯度不断修正分离矩阵的估计，从而达到分离混合信号的目的。仿真试验在Matlab环境下进行，源信号采用故障轴承缺陷部位所产生的冲击振动信号。分离结果证明该方法能成功提取源信号，有效去除了噪声。

Research of the noise removal method for mechanical Vibration signal based on blind signal extraction

In actual, there exist inevitably a lot of interference from neighbor machine and noise from surrondings in mechanical vibration signal measured by sensor, which is disadvantageous for condition monitoring and fault diagnosis. In order to remove noise from mechanical vibration signal, this paper proposed a novel noise removed method using feature extraction property of blind signal extraction. This method constitutes cost function using high-order statistics to approximate kurtosis, modifies separating matrix by calculating grads of cost function, in order to separate mixed signal. Emulating experiments were carried out under the MATLAB software, original signals adopted vibration impulsion signal produced by vice position of faulty bear. Separation results confirm this method successfully extract original source, efficiently removes noise.