基于IITD样本熵和支持向量机的齿轮故障诊断方法

A Gear Fault Diagnosis Method Based on IITD Sample Entropy and SVM

基于ITD方法的线性变换和Akima插值,提出了一种改进的固有时间尺度分解方法(Improve Intrinsic Timescale Decomposition,简称IITD)方法。齿轮振动信号具有非平稳特征,其典型的故障样本难以获取,为此进一步提出了一种基于IITD样本熵和支持向量机的齿轮故障诊断方法。采用IITD法对非平稳的原始加速度振动信号进行分解,并提取包含主要故障特征信息的PR分量,将其样本熵值作为特征向量;然后将特征向量输入到支持向量机中识别齿轮的故障特征。实验分析结果表明:相比BP神经网络,能更有效地应用于齿轮的故障诊断。

An improve intrinsic time-scale decomposition （IITD） was proposed based on the linear transformation of ITD method and akima interpolation. Gear vibration signals has the characteristics of non-stationary, the typical fault samples are difficult to obtain, then a method of gear fault diagnose based on IITD sample entropy and support vector machine （SVM） was put forward. Firstly, the original acceleration vibration signal was decomposed by IITD; Then the RP containing the abundant fault characteristic information were chosen to calculate the sample entropy and form a feature vector; Finally SVM method was used as a classifier to identify different faults.Practical examples showed that the diagnosis approach proposed here can identify gear fault patterns effectively, compared to BP neural network.