基于CCA和SOFM的轴承故障特征提取

英文篇名：Fault Feature Extraction of Bearings Based on CCA and SOFM
中文刊名：三峡大学学报(自然科学版)
英文刊名：Journal of China Three Gorges University(Natural Sciences)
作者：何涛 ; 万鹏 ; 谢卫容
英文作者：He Tao Wan Peng Xie Weirong(School of Mechanical Engineering ; Hubei Univ.of Technology ; Wuhan 430068 ; China)
中文关键词：数据挖掘 ; 曲元分析 ; 自组织映射神经网络 ; 聚类分析
英文关键词：data mining ; CCA ; self-organizing feature map ; clustering analysis
出版日期：2006-06-30
机构：湖北工业大学机械工程学院,湖北工业大学机械工程学院,湖北工业大学机械工程学院武汉430068,武汉430068,武汉430068
年：2006
期：03
出版单位：三峡大学学报(自然科学版)

摘要

提出曲元分析(CCA)和自组织特征映射(SOFM)相结合的方法用于轴承的故障诊断特征提取.首先通过传感器测得轴承在正常和非正常状态下的信号;然后对所得数据进行归一化;考虑到数据比较庞大,利用CCA进行降维;再利用SOFM进行训练,网络对不同状态下的输入具有明显不同的输出.利用Matlab神经网络工具箱来实现上述算法.实例仿真表明,这个算法可以快速正确地提取出轴承故障特征值,并通过聚类算法完成轴承的故障诊断.
The combination of curvilinear component analysis(CCA) and self-organizing feature map(SOFM) were applied to a diagnosis for fault feature extraction of bearing.Firstly regularizing the input signal of the bearing's normal and abnormal states obtained from sensors;and secondly dimension-reducing the data with CCA considering its hugeness;finally training it with SOFM,the networks have different output maps for different input states.The above method was implemented by the neural network toolbox in Matlab.The simulation results show that it can be used to extract fault features of bearing quickly and exactly and completed the inspecting of bearing by clustering analysis.

引文

[1]Demartines P,Herault J.Curvilinear Component Analy-sis:A Self-organizing Neural Network for NonlinearMapping of Data Sets[J].IEEE Transaction on NeuralNetworks,1997,8(1):148-154.
    [2]韩力群.人工神经网络理论、设计及应用[M].北京:化学工业出版社.2002:167-171.
    [3]Kohonen T.The Self-Organizing Map[J].Proc.ofIEEE,1990,78(8):312-317.
    [4]杨世明,孙龙德,杨新华.自组织神经网络地震岩相分析[J].新疆石油地质,2004,25(5):32-35.
    [5]杜华强,范文义.Matlab自组织神经网络在遥感图像分类中的应用[J].东北林业大学学报,2003,7:24-26.