层次模块化模糊神经网络在铝电解故障诊断中的应用

目的针对铝电解故障发生机理及特点，提出采用层次模块化模糊神经网络实现对铝电解故障快速、有效检测的新方法．方法利用数据处理子模块，诊断子模块，决策子模块，从而对铝电解生产过程中的故障进行诊断．结果数据处理子网络实现了对数据的特征提取，优化了数据结构，将大结构的神经网络分解成多个小网络，并用决策神经网络对故障进行模式分类．结论该方法优化了大结构神经网络的学习性能，提高了故障检测的准确率，具有良好的实用价值．

The Application of Modular Fuzzy Neural Networks In the Aluminum Electrolysis Fault Diagnosis

This paper proposes a new method,which is a layer construction of Modular Fuzzy Neural Network according to the mechanism and characteristic of Aluminum Electrolysis Fault Diagnosis.This method through data processing sub module,diagnoses sub net module and decision fusion net,has realized fault diagnosis.With the neural network of big structure,task decomposing then changing the big structure into many neural networks of small structure,this method optimized the learning property of large structural neural network and improved the Performance of the Algorithm,when what is used in the Aluminum Electrolysis Fault Diagnosis.Tests show that the diagnosis results by this system are accurate,reliable and practicable.