高阶统计量与RBF网络结合用于齿轮故障分类

Classification of Gear Faults Using RBF Network Combined with Higher-Order Statistics

提出一种基于高阶统计量特征提取的径向基函数网络齿轮故障分类方法。以齿轮箱振动信号的高阶统计量估计值作为齿轮故障特征，以径向基函数神经网络作为分类器，成功地对齿轮故障进行了分类。研究表明。

An approach of gear fault classification,combing higher-order statistics with radial basis function (RBF) networks,is proposed. Higher-order statistics can eliminate additive Gaussian measurement noise, boost signal-to-noise ratio. The radial basis function (RBF) network, as an alternative to the BP network, preserves several advantages, such as fast convergence and best approximation to non-linear functions, etc. In this paper, the higher-order statistics calculated from vibration signals of a gearbox are used as input features, and an RBF network as the classifier, gear faults are successfully recognized. The experimental results show that the method of fault classification combining higher-order statistics with RBF network is very effective.