基于盲解卷积和聚类的机械弱冲击声信号提取

Weak transient impulse signal extraction based on blind deconvolution and cluster in acoustical machine diagnosis

针对对比函数和紧缩方法的时域盲解卷积算法在分离机械弱冲击信号时,其结果易受解卷积滤波器长度影响的缺点,提出结合分层聚类的改进算法。该算法通过设置一个变长度滤波器组,对获得的多个盲解卷积结果进行聚类分析,解决了单次盲解卷积结果不确定问题,获得了可靠性高的估计信号。计算机仿真和实际环境下故障轴承声信号提取实验验证了算法的有效性。

The time-domain blind deconvolution algorithm based on contrast function and deflation has recently become the focus of intensive research work due to its potential in many applications. However, it has a disadvantage that the separation results are easily influenced by the length of deconvolution filters when the signals come from machine sound. In this paper, an improved blind deconvolution algorithm based on hierarchical cluster is proposed. Hierarchical cluster is applied to analysis the results which are obtained by using a group of deconvolution filters of various lengths. Therefore, the improved algorithm can be employed to receive more reliable and better estimated signals. Computer simulation and acoustical transient impulse signal extraction of faulty bearing in a real-world situation are used to verify the validity of the proposed algorithm.