广义稀疏解卷算法研究及其轴承故障诊断应用

Generalized sparse deconvolution algorithm and its application of bearing fault diagnosis

机械系统中轴承局部故障会导致振动信号中出现瞬态冲击成分,而瞬态成分的有效提取是实现轴承故障诊断的关键。最小熵解卷积是一种基于峭度准则的微弱特征提取方法,然而其在强背景噪声下的稳定性较差。因此,提出一种基于广义P算子稀疏准则的解卷积方法。首先理论推导出广义P算子稀疏准则下的优化解卷积表达式,然后建立以归一化频率能量比为指标的广义稀疏准则下的轴承故障特征识别方法,最后利用仿真信号与实验数据对提出方法进行验证。仿真信号分析结果表明了提出方法能够识别出强背景噪声下轴承微弱故障特征;同时实验信号分析结果也证明了提出方法能可靠地检测出轴承微弱故障,并优于现有最小熵解卷积等故障诊断方法。

The weak repetitive transients are usually induced by the localized defect of the component of rotating machinery. Thus,the effective feature extraction method plays an important role in the fault diagnosis of rotating machinery. Minimum entropy deconvolution( MED) is considered as a popular way to extract weak faulty feature. However,it is sensitive to the strong background noise because it is based on the kurtosis criterion. In this paper,the deconvolution algorithm based on the generalized sparse criteria is proposed to overcome the shortcoming of the MED and diagnose the faulty bearing. Firstly,the optimal deconvolution expression is derived based on the generalized P operator. Then,the bearing faulty feature extraction method is constructed by combining the generalized sparse criterion and the normalized frequency energy ratio. Furthermore,the simulated signal and experimental data are employed to verify the proposed method,respectively. The analysis results of the simulated signals reveal that the proposed method can well extract the bearing fault feature under the strong background noise. Moreover,the experimental result demonstrates that the proposed method can accurately identify the weak faulty and outperform some existed method for bearing fault diagnosis such as Hilbert transformation and MED.