基于数据驱动的复杂工业过程输入空间边界求解及应用

项目名称： 基于数据驱动的复杂工业过程输入空间边界求解及应用

项目编号： No.61203114

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 自动化学科

项目作者： 梁炎明

作者单位： 西安理工大学

项目金额： 23万元

中文摘要： 输入空间边界是指各输入变量临界值的集合。针对复杂工业过程输入空间边界难以采用机理模型求解的问题，本项目在数据驱动的框架下，研究复杂工业过程连续变化和非连续变化情况下的输入空间边界求解方法。对于连续变化过程，将输入空间边界求解问题转化为过程的T-S模糊模型辨识问题，主要研究模糊输入空间划分与辨识精度、模型结构与动态性能和泛化能力之间的作用关系，为设计性能良好的动态T-S模糊模型提供理论依据。对于非连续变化过程，将输入空间边界求解问题转化为过程突变模型构建问题，提出利用测量数据和辨识模型构建突变模型的方法，突破了传统基于机理推导构建突变模型的局限。除此之外，利用输入空间边界求解的相关理论和方法对目前较难解决的单晶变晶问题和空气预热器着火问题进行研究，为正确设计它们的工艺参数奠定基础。本项目的实现将为过程的参数优化、鲁棒控制设计及故障诊断等提供正确的边界计算依据。

中文关键词： 工业过程；数据驱动；输入空间边界；智能辨识；突变理论

英文摘要： An input space boundary is the critical value set of each input variables. Aiming at an input space boundary that can not be solved by mechanism model in complex industrial processes, this project will study the solution method of the input space boundary of complex industrial processes with continuous variation and discontinuous variation under data-driven frame. For the continuous variation case, the problem of the input space boundary solution can be convert to the problem of T-S fuzzy model identification of industrial processes. In order to lay a theoretical basis for the design of dynamic T-S fuzzy model with good property, we will mainly study the relationship between fuzzy input space partition and identification accuracy and the relationship between model struct and property of dynamic and generalization. For the discontinuous variation case, the problem of the input space boundary solution can be convert to the problem of catastrophic model construction of industrial processes. A new method of constructing catastrophic model that will be achieved through measurement data and identification model is proposed. This method can break through the limitation of traditionarily constructing catastrophic model based on the deduce mechanism. In addition, the both problem of single crystal metacryst and air prehe

英文关键词： industrial processes；data driven；input spatial boundary；intelligent identification；catastrophe theory