旋转机械故障类型识别的神经网络方法研究

Research on Neural Network Method for Fault Type Identification of Rotating Machinery

广义回归神经网络(Generalized Regression Neural Network,GRNN)和概率神经网络(Probabilistic Neural Network,PNN)都是基于径向基函数的神经网络类型。广义回归神经网络通常用来实现函数逼近,而概率神经网络主要用于模式分类问题的研究。两者在机械设备故障诊断中均有广泛的应用。根据两种神经网络原理建立模型,对比分析广义回归神经网络和概率神经网络在旋转机械设备故障类型识别方面的优缺点。结果显示,两种神经网络在故障类型识别方面均取得了不错的效果,而概率神经网络相比广义回归神经网络而言,能应用更少的特征得到正确的结果。同时,将这两种神经网络得到的结果同BP神经网络和RBF神经网络得到的结果相比,发现GRNN神经网络和PNN神经网络具有更高的准确率和鲁棒性。

Generalized Regression Neural Network(GRNN)and Probabilistic Neural Network(PNN)are neural network types based on radial basis functions.Generalized regression neural networks are usually used to achieve function approximation,while probabilistic neural networks are mainly used for the study of pattern classification problems.Both neural networks have a wide range of applications in the diagnosis of mechanical equipment.This paper establishes a model based on two neural network principles,and compares the advantages and disadvantages of generalized regression neural network and probabilistic neural network in fault type identification of rotating machinery.The results show that both neural networks have achieved good results in fault type identification.While compared to generalized regression neural network,probabilistic neural network can get correct results with fewer features.At the same time,compared with BP neural network and RBF neural network,the results obtained by these two neural networks show that GRNN and PNN have higher accuracy and robustness.