最大重叠离散小波包变换边际谱特征在齿轮故障诊断中的应用

Application of Maximum Overlap Discrete Wavelet Packet Transform Marginal Spectrum Features in Gear Fault Diagnosis

针对齿轮故障振动信号多分量频带重叠引发的故障模式混淆问题,提出一种基于最大重叠离散小波包变换(MODWPT)边际谱特征和粒子群优化-支持向量机(PSO-SVM)的故障诊断方法。为了减少谐波及噪声对故障模式分量分离的干扰,首先利用MODWPT将采集到的实验信号进行5层分解,得到32个分量,通过频带能量占优方法,筛选出前16个分量,用来构造信号的希尔伯特边际谱;然后,将提取的边际谱特征代入PSO参数优化后的SVM,对故障类型进行识别。仿真信号分析结果表明,MODWPT边际谱在抗模式混叠、抗边界效应和频率提取准确性方面都要优于EMD方法。通过对6种不同类型的齿轮故障信号进行分析,MODWPT边际谱归一化特征具有明显的故障类型分层现象,对齿轮故障的识别准确率达到98%,说明该方法具有较强的故障诊断能力。

A fault diagnosis method based on the maximum overlap discrete wavelet packet transform(MODWPT)marginal spectrum features and the particle swarm optimization-support vector machine(PSO-SVM)is proposed to deal with the problem of fault mode confusion caused by multi-component band overlap of gear fault vibration signals.In order to reduce the influence of harmonics and noise on the fault mode component separation,MODWPT is used to decompose the collected experimental signals into 5 layers and to obtain 32 components.The first 16 components are selected based on the principle of the band energy dominant distribution to construct a Hilbert marginal spectrum.Then,the marginal spectrum features are extracted and considered as an input to the SVM with PSO parameter optimization for fault type identification.Analysis results of simulation signals show that the MODWPT marginal spectrum is superior to the EMD method in terms of anti-mode mixing,anti-end effect,and frequency extraction accuracy.The proposed method is applied to fault diagnosis of six different gear fault types.The normalized feature of MODWPT marginal spectrum has obvious hierarchical phenomenon of fault types,and the accuracy of identifying gear faults reaches 98%.These results show that the method has strong fault diagnosis capability.