一种基于分割的聚类算法用于振动信号解调

摘要

Resonance demodulation technique is widely employed in vibration signal analysis. In order to construct a proper bandpass filter, the prior knowledge, i.e., the resonance region of the mechanical system is required in the traditional resonance demodulation method. However, if the collected vibration signal is tainted by the background noise or interference, it is difficult to identify the center frequency and the bandwidth of the filter. This paper introduces a clustering-based segmentation method to determine these parameters automatically. Envelope analysis is then applied to demodulate the vibration data. According to the simulated cases, the presented scheme is robust to Gaussian noise and interferences.%在传统的共振解调技术中，为了构造一个合适的带通滤波器，机械系统的共振带是必须要知道的。但是，如果所采集的信号被背景噪声和干扰所污染，中心频率和频带就很难识别。本文介绍了一种基于分割的聚类算法。该算法可以自动识别以上参数。然后，就可以采用包络方法对振动信号进行解调。仿真实例证明该算法对高斯噪声和干扰鲁棒性较好。