基于最差子空间分解聚类的热工过程模糊辨识

以较少的模糊规则在全局范围内获得满意的辨识精度，是模糊辨识追求的重要目标。该文提出了一种基于最差子空间分解聚类的非线性系统模糊辨识方法。根据各子空间的“可线性化”程度，对聚类的有效性进行评判，进一步对有效性最差的子集进行重新分解聚类，并辨识新增子空间的模型参数，以此逐步完成整个样本空间的模糊划分和模型辨识过程，直至模型满足既定要求。文中给出了所提出的模糊辨识方法与其他相关模糊辨识方法的对比结果，并利用该方法对2个典型热工对象进行了模糊辨识。

Fuzzy Identification of Thermal Process Based on Decomposing Cluster of the Worst Subspace

The ultimate purpose for fuzzy identification is to obtain satisfactory accuracy over the whole range by using less fuzzy rules. This paper proposed a fuzzy identification method for nonlinear systems which were based on decomposing clustering of the worst subspace. The first step was to make a judgment of the clustering validity according to the linearizing level of each subspace, and then decomposed the subspace of the worst efficiency again and identified these model's parameters of new subspaces. In this way, the fuzzy partition of the entire sample space and the process of model identification were gradually achieved until the model had met requirement. The paper displays the comparative results of the proposed fuzzy identification method with other relative methods and makes fuzzy identification to two typical thermal objects by using this method.