摘要
提出一种基于模块化模糊神经网络的非线性系统故障诊断新方法 .该方法先使用模糊c 均值聚类法对测量空间进行模块分割 ,再利用模糊IF THEN规则对分割后的子空间分别采用局部BP模型进行逼近 .最后 ,通过离线学习获得不同子空间故障输出与测量输入的非线性动力特性 .试验表明该网络具有良好的泛化性能 ,可显著提高非线性系统故障检测的快速性、鲁棒性及准确率 .
A new approach to fault diagnosis based on modular fuzzy neural networks for nonlinear systems is proposed. Firstly,the measurement space has been divided into several subspaces by using fuzzy c means clustering. Secondly, according to the requirements of fuzzy rules, the subspaces have been fitted by local BP network respectively. Lastly, the characteristics between fault outputs and measuring inputs in different subspaces have been obtained by processing off line learning. Testing shows the network has good generalization performance and can distinctly improve the speediness,robustness and validity of fault diagnosis in nonlinear systems.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2001年第3期395-400,共6页
Control Theory & Applications
基金
supportedbySpecialScientificResearchFoundationforDoctoralSubjectofCollegesandUniversitiesinChina ( 5 10 80 60 )andNationalHigh
