【摘要】 阐述径向基函数(radial base function,RBF)神经网络的基本原理和算法,将其应用于齿轮箱故障诊断与识别,建立齿轮箱的BRF故障诊断模型,并与BP(back propagation)神经网络、学习率自适应BP神经网络进行对比分析研究。结果表明,RBF神经网络性能优于BP神经网络,具有较快的训练速度、较强的非线性映射能力和精度较高的故障识别能力,非常适用于齿轮箱的状态监测和故障诊断。但在具体应用中应当注意,RBF网络的训练样本必须含有一定的噪声,以提高网络的容噪性能;各类故障的训练样本数不能太少,否则RBF网络的故障分类能力很差。更多还原

【Abstract】 The basic theory and arithmetic of RBF(radial basis function) neural network were expatiated,which was applied successfully to gearbox fault diagnosis,the RBF fault diagnosis model of gearbox was constructed,and was analyzed contrastively with the BP neural network and learning rate self-adaptive BP(back propagation) neural network. The study result shows that RBF neural network’s well performance,with the quick training pace,strong nonlinear mapped capability and highly accurate capability of fault identification,is superior to the BP neural network,and it is very suitable to the condition monitoring and fault diagnosis of gearbox. But during the practical application,which must notice that,the trained samples must include some noise in order to improve network’s capability of noise-tolerant; trained samples of each type fault can’t be few,otherwise,the fault classification capability of RBF network is worse.更多还原