基于GAPSO-SVM的多级齿轮箱故障诊断新方法

多级齿轮箱是机械传动的重要部件,针对运行过程中的状态识别问题,研究并提出一种基于振动信号的小波包分解能量谱特征提取和支持向量机(support vector machine,SVM)的智能评估新方法。用小波包分解算法对振动信号进行分解,提取时频信号的能量谱构建多级齿轮箱状态特征集,训练SVM模型。针对SVM的惩罚因子C和高斯核参数g选择困难的问题,结合遗传算法(genetic algorithm,GA)和粒子群算法(particle swarm optimization,PSO)的基因粒子群算法(genetic algorithm-particle swarm optimization,GAPSO)优化SVM参数。GAPSO同时具有GA全局搜索的性能和PSO快速收敛特点。将优化后的SVM算法应用于多级齿轮箱故障诊断,结果表明,GAPSO-SVM模型故障识别精度为98.55%,高于基本的SVM、PSO-SVM和BP神经网络,而且泛化能力强,该方法更适合多级齿轮箱故障诊断。

Application of wavelet packet decomposition and its energy spectrum on the coal-rock interface identification

The theory and method of wavelet packet decomposition and its energy spectrum dealing with the coal rock Interface Identification are presented in the paper. The characteristic frequency band of the coal rock signal could be identified by wavelet packet decomposition and its energy spectrum conveniently, at the same time, quantification analysis were performed. The result demonstrates that this method is more advantageous and of practical value than traditional Fourier analysis method.