组合式神经网络在转子故障诊断中的应用  

Application of Modular ANN in Rotor Diagnosis

作　　者：陈思维[1] 陈作炳[1] 李如强[1] 

机构地区：[1]武汉理工大学机电工程学院,武汉430070

出　　处：《拖拉机与农用运输车》2007年第3期84-86,共3页Tractor & Farm Transporter

摘　　要：在组合式神经网络的故障诊断中,模糊积分是一种对各神经子网的诊断结果进行信息融合的方法,其中需要计算模糊密度。本文首先介绍组合式神经网络故障诊断的思想,给出了基于模糊C均值聚类的模糊划分方法;其次分析一种不可取的模糊密度计算方法,基于模糊粗糙函数,给出计算组合式神经网络重要性的方法;然后,阐述了组合式神经网络的网络训练和诊断方法;最后,用故障模拟数据验证了诊断方法的有效性。In the diagnosis of modular ANN,fuzzy integral is an approach to fuse information of the diagnostic result of every single ANN, where fuzzy density must be calculated. Firstly, the idea of modular ANN is introduced and a fuzzy partition method based on FCM is proposed. Secondly, after the analysis of shortage of a computation method of fuzzy density, a computation approach of the significance of modular ANN is proposed. Thirdly, the training and diagnostic approaches of modular ANN is .expatiated. At last, the effectiveness of diagnosis based on modular ANN has been validated by the simulative data from rotor test rig.

关 键 词：转子 故障诊断 组合式神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP277[自动化与计算机技术—控制科学与工程]