基于VMD-DE的坦克行星变速箱故障诊断方法研究

为了提高坦克行星变速箱齿轮故障模式识别准确率,将变分模态分解(VMD)与散布熵(DE)结合提出故障特征提取新方法。利用波形法确定VMD分解层数,VMD分解振动信号得到一组固有模态分量(IMF);根据归一化互信息准则筛选若干IMF重构信号,计算重构信号的散布熵;将重构信号散布熵作为特征值输入到粒子群优化(PSO)的多分类支持向量机(SVM)中实现故障模式识别。通过对坦克行星变速箱的正常、行星轮故障和太阳轮故障三种状态进行模式识别,分类准确率达到100%,且计算时间较短。与基于原始振动信号DE、VMD-SE(样本熵)、VMD-PE(排列熵)及EMD-DE(经验模态分解与DE结合)等方法比较,综合考虑准确率和计算时间两个因素,基于VMD-DE的方法故障诊断性能最佳。

Research on fault diagnosis method of tank planetary gearbox based on VMD-DE

In order to improve fault pattern recognition accuracy of tank planetary gearbox, this work proposed a new fault feature extraction method based on variational mode decomposition (VMD) and dispersion entropy (DE).First, intrinsic mode function (IMF) number was determined by waveform method.Next, VMD was used to decompose vibration signal to obtain a set of IMFs.Then, normalized mutual information criterion was used to screen several IMF for signal reconstruction.And then, dispersion entropy of reconstructed signal was calculated.Last, dispersion entropy was input into support vector machine (SVM), which was optimized by particle swarm optimization (PSO), to realize fault pattern recognition.The pattern recognition accuracy of the normal, planetary gear fault and sun gear fault of planetary gearbox was 100% with a relatively short computing time.Compared with methods of original signal DE, VMD-SE, VMD-PE and EMD-DE, the results show that VMD-DE methods has the best fault pattern recognition performance.