基于密度聚类的模糊神经网络和CHP分解过程建模

Fuzzy Neural Network Design Based on Density Clustering and CHP Decomposing Process Model

针对基于样本数据的复杂系统建模问题,提出了基于密度聚类的模糊神经网络(DFNN)的建模方法,研究了利用密度聚类原理提取数据样本的内在规则的理论和方法,提取的规则能较好地反映样本数据输入输出的对应联系,根据提取的规则给出了模糊神经网络的模型结构.本文以化工生产过程过氧化氢异丙苯(CHP)分解反应过程为对象进行仿真建模,结果显示在模型精度和可靠性上均优于基于c均值聚类提取规则的模糊神经网络模型(CFNN).

According to modeling for complex systems only based on input - output data, a new building model of fuzzy neural network structure based on density clustering （DFNN） is presented. The author studied the theory, and method of density clustering, by which the inner rules about data , and relations between the data of inputs and outputs can be found. According to the rules, network structure was modified for better integration of them. Decomposing process of cumene hydroperoxide（CLIP） was sample model to build. Results showed that DFNN is better than fuzzy neural network based on c - means（CFNN） in precision and reliability.