基于双谱分析的齿轮故障诊断方法

Bi-spectrum analysis method for gear fault diagnosis

齿轮啮合过程中的振动信号往往呈现出非线性、非高斯性,加上强噪声的干扰,给故障特征的提取带来了较大的困难。为实现齿轮单一故障的分类和诊断,采用时序参数化的双谱分析方法,对齿轮故障模拟试验台上采集的正常状态和3种故障状态的振动信号进行了分析,根据双谱谱峰的分布及数目的差异性,实现了齿轮正常、裂纹、磨损、剥落4种状态的识别和分类。结果表明,双谱分析可以抑制背景噪声,并有效提取信号中的非高斯成分,是一种有效的故障诊断方法。

The non-linear and non-gaussian properties of vibration signals are usually exhibited in gear meshing. It made the extraction of fault feature difficult plus the strong noises influence. In order to classify and diagnose gear single fault, firstly normal and three different abnormal vibration signals were collected on simulant experiment table. Then the bi-spectrum estimated with time series parameters method was adopted for diagnosing. Finally four modes of normal, crack, wear and spalling were effective- ly classified and diagnosed, according to the number and distribution of spectrum peaks. The results show that noises can be inhibited and non-gaussian can be extracted by bi-spectrum, and bi-speetrum analysis is efficient method for gear diagnosing.