模糊结构辨识的人工免疫聚类算法研究

Research on Artificial Immune Clustering Algorithm of Fuzzy Structure Identification

针对自适应神经模糊推理系统(ANFIS)在锌钡白干燥煅烧生产过程建模中出现的模糊结构辨识问题,采用了基于人工免疫系统(AIS)的聚类算法。该算法通过免疫网络对抗体及记忆数据集逐代克隆、变异及抑制操作,提取有用的模糊规则数目,避免ANFIS训练陷入局部极小点。本文详细探讨了AIS随机特性对聚类规模稳定性造成的影响以及AIS的聚类速度问题,对Castro算法做了必要修改。通过与减法聚类算法、模糊C-Means聚类算法(FCM)特性上的对比分析,得出AIS在复杂过程辨识中的实际应用价值。

In order to solve the problem of the fuzzy structure identification in Lithopone calcination process of the adaptive neuro-fuzzy inference system （ANFIS）, a new kind of artificial immune system （AIS） is proposed in this paper. The immune network takes clone, mutation and suppression actions to antibodies and memory data sets, then extracts useful fuzzy rules from them. So it can prevent the ANFIS from trapping in partial minimal ports. This paper discusses the clustering stability caused by randomness of AIS, analyses the clustering speed problem, and makes some necessary modifications to Castro algorithm. Through comparison with subtractive clustering and fuzzy C-Means clustering, practical value of AIS to complex process identification is obtained.