论文部分内容阅读
BP神经网络的非线性映射特性在故障诊断领域得到了广泛应用,但对于新增故障模式缺乏有效的识别,而ART2自适应共振网络能有效增添新的故障模式,却不能对输入量进行特征降维约简,而输入量太多时耗费时间太长,该文综合BP神经网络与ART2自适应共振网络的优点,研究了改进型BP-ART2神经网络故障诊断方法,在ART2结构的输入层增加隐层,通过非线性映射实现特征量降维约简,从而提高ART2神经网络的诊断效率.通过旋转机械故障诊断结果表明,该方法能够有效地实现特征约简和故障聚类.“,”BP neural networks has already been applied into diagnosis field,it can realize the nonlinearity mapping fault diagnosis,but can not identificate the new fault pattern effectivly. ART2 has an effect to increase the new fault pattern,but can not reduce the input characteristic dimension,when input amount is too large,it will consume too long time. Synthesized the merit of BP and ART2 networks,an improved BP-ART2 neural network fault diagnosis method has been reaserched,adding a hidden layer in input layer of ART2 structure,reduceing the input characteristic dimension by nonlinearity mapping,improving diagnosis efficiency of ART2 neural networks thereby. The experiment indicate in fault diagnosis of rotating machinery,the method can reduce the characteristic dimen-sion and realize fault clustering effectively.