过程控制中的智能集成建模方法

针对复杂工业过程中广泛存在的对象难于辨识的问题,本文讨论了过程控制中的一类集成建模方法.通过研究铅锌冶炼过程的两个对象,分别建立了预测烧结块成分的集成模型和鼓风炉透气性模型.然后阐述了以神经网络为基础,结合其它方法的智能集成建模策略,这种方法由于充分利用了对象的各类信息,能够有效解决复杂对象的建模问题.

An Intellectual Integrated Modeling Method in Process Control

This paper presents a kind of integrated modeling method in process control according to the universally existing difficulties of discriminating a plant in complex industrial processes. After discussing two examples in Pb-Zn smelting process, an integrated model to predict compositions of sinters and a permeability model of blast furnace are set up respectively. The predictive model for compositions of sinters combines a neural network model with a mathematic model by means of fuzzy classification. The permeability model introduces experience to decide the weights of different sample data before the data are used for the training of neural network. Simulation results prove the superiority of integrated models to common neural network models. The paper concludes that the intellectual integrated modeling strategy based on neural network and combined with other methods, which involve known experience and mechanism, is able to effectively solve modeling problem of complex plant because all sorts of information of the plant are fully studied and utilized.