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 method of fault diagnosis for nonlinear systems based on modularized fuzzy neural network is proposed.The method uses the fuzzy c-means clustering method to divide the measurement space into modules, and then uses the fuzzy IF THEN rule to separate the partitioned subspaces by using local BP model.Finally, the nonlinear dynamic characteristics of the output and measurement inputs of different subspaces are obtained by offline learning.Experiments show that this network has good generalization performance, which can significantly improve the fastness and robustness of fault detection in nonlinear systems Sex and accuracy.