基于满意聚类的非线性系统多模型建模方法

针对一类生产过程中存在严重非线性的系统，基于系统运行中积累的可靠的输入／输出数据，提出了一种新的多模型建模方法．根据对各种指标的满意要求，对数据进行二次聚类，不仅得到了更有效的系统多模型，而且得到了每个模型的适用域．与模糊聚类和建立T—S模型方法相比，本方法不依赖系统的先验知识和预先定义模糊隶属度，具有良好的泛化性．以pH中和过程为例进行了仿真研究，验证了该方法简单易用，有很高的建模精度，并对存在的数据不确定性具有一定的鲁棒性．

New Multi-Model Approach to Nonlinear Systems Based on Satisfying Cluster

For some manufacturing processes in which the systems are seriously non linear, based on the accumulated I/O data which is reliable from the run time of the systems, this paper proposed a new multi modeling method. According to the satisfactory demands of many factors, it clusters the data twice. Then the better models of the system as well as their valid area are got. In contrast to the methods of fuzzy clustering and T S models' constructing, this method does not depend on the prior knowledge or define the fuzzy subjection, so it has a favorable universality. The simulating research based on the pH nuetralization testifies that this method is very simple and practical. It has a satisfied modeling accuracy and is robust for the uncertainty of the existing data.