基于一型规则自主构建二型TSK神经模糊系统方法设计

Automatically constructing type–2 TSK neural fuzzy system based on type–1 fuzzy rules

中文关键词  二型模糊系统  神经模糊系统  类型转换  数据驱动  融合

英文关键词  type–2 fuzzy logic system  neural fuzzy system  type transformation  data driven  mergence

中文摘要

      本文提出了一种使用一型模糊规则生成区间二型TSK(Takagi-Sugeno-Kang)神经模糊系统的新方法. 该方 法以训练数据集与使用自组织方法由该训练集训练生成的一型模糊系统为驱动,通过新型模糊系统前件类型转换 算法与规则参数自适应学习算法的训练,在不高于原一型系统模糊集合总数前提下,自主构建区间二型TSK神经模 糊系统.此外, 针对两种典型的多输入单输出和多输入多输出系统, 在3种不同强度的系统扰动场景下进行了对比仿 真实验. 实验结果表明, 在含有不同扰动状态系统的建模与辨识中本方法较于对比方法具有更加优异的性能.

英文摘要

      This paper presents a novel approach to generating an interval type–2 TSK (Takagi-Sugeno-Kang) neural fuzzy system (IT2--TSK–NFS) by using type-1 TSK fuzzy (T1--TSK) rules. This method makes full use of training data sets and those T1 fuzzy rules generated from existing well-behaved self-organizing T1 methods to automatically generate a better performing IT2--TSK--NFS through novel antecedent type transformation and adaptive parameter training algorithms. Meanwhile, the rule number of the IT2--TSK--NFS stays the same as the original T1’ s whereas the total number of IT2--FSs in the antecedent is no more than that of the original ones. Two benchmark examples with three different disturbance scenarios are given in experiments. The comparison results show and validate the proposed IT2--TSK--NFS can perform better than original T1--TSK system, and in some cases better than other IT2 self-organizing methods in literature in dealing with system modelling and identi?cation issues under different disturbances.