基于ELMD奇异值和LDA的变速箱特征提取方法

Feature Extraction Method for Gearbox Based on ELMD Singular Value and LDA

针对变速箱故障时振动信号的复杂性和实际信号特征的冗余性,提出基于总体局部均值分解(ELMD)奇异值和线性判别分析(LDA)的故障特征提取方法,并利用该方法分析某装备实车试验中测取的变速箱正常、齿轮断齿和滚动轴承滚动体点蚀等3种状态下的振动信号。结果表明:该方法提取的特征能很好地将变速箱的各种状态区分开,可以实现变速箱的智能诊断。

Considering the complexity of gearbox vibration signal at failure and the redundancy of actual signal feature, the paper presents a fault feature extraction method based on ensemble local mean decomposition （ELMD） singular value and linear discriminant analysis （LDA） , and calculates the vibration signals at the state of normal gearbox, broken gear tooth and pitting rolling bearing respectively with this method in a real vehicle test. The result shows that the feature extracted with this method can distinguish conditions of gearbox and realize intelligent diagnosis.