A Study on the Theory and Practice of Nonlinear Process Control Using Functional Expansion (FEx) Models

摘要- Palazoglu - 9800073了解非线性系统的动力学对于许多化学过程的设计，操作和控制非常重要。 由于缺乏关于非线性过程行为的信息，导致化学工程师限制安全操作的范围，过度设计以试图软化非线性，或设计保守的控制系统以维持可接受的操作。 这样的工程解决方案可能产生次优的工厂操作，并且更经济有利的制度可能不必要地被认为是不安全或不可靠的。 本研究的目标是建立一个严格的设置，其中非线性过程控制问题可以分析，控制器可以设计和实施的实际过程。 这种设置将建立在一个新的范式，利用功能扩展（FEx）模型开发新的控制系统设计和分析工具。 这样的设置不仅有望拓宽人们对非线性过程控制问题的理解，而且还将促进这些新概念在真实的过程系统中的应用。 该方法利用类似的线性分析和控制概念，从而为工程师提供直观的设置。 需要开展的工作包括： (1)将使用函数展开来量化过程的“非线性”，并且将利用近似FEx模型结构来使用过程数据进行模型识别。 (2)控制设计工作将以FEx模型为基础，本着模型预测控制的精神。 这将产生新的观点的表示和操纵的非线性离散时间模型的控制。 (3)将制定一套标准，以帮助评估基于FEx模型的控制系统的稳定性和鲁棒性。 (4)这些方法将通过使用实验室规模的过程以及由商业分布式控制系统（DC S）运行的中试规模过程的实时应用进行测试。 最后一个领域将利用与悉尼大学Romagnoly教授的长期合作。

Abstract - Palazoglu - 9800073 Understanding the dynamics of nonlinear systems is important for the design, operation and control of many chemical processes. The lack of information regarding the behavior of nonlinear processes causes the chemical engineer to limit the region of safe operation, to overdesign in an attempt to soften nonlinearities, or to design conservative control systems such that an acceptable operation is maintained. Such engineering solutions may yield suboptimal plant operations, and more economically favorable regimes may be unnecessarily deemed unsafe or unreliable. The goal of this research is to establish a rigorous setting within which nonlinear process control problems can be analyzed, and controllers can be designed and implemented on actual processes. This setting will be built upon a new paradigm that exploits functional expansion (FEx) models to develop novel control system design and analysis tools. Such a setting is not only expected to broaden one's understanding of nonlinear process control problems, but will also promote the application of these new concepts to real process systems. The methods make use of analogous linear analysis and control concepts, thereby providing an intuitive setting for engineers. The work to be done includes: (1) Functional expansions will be used to quantify the "nonlinearity" of a process, and the approximate FEx model structure will be exploited for model identification using process data. (2) The control design effort will be based upon FEx models in the spirit of model predictive control. This will generate new viewpoints about the representation and manipulation of nonlinear discrete-time models for control. (3) A set of criteria will be developed to help assess the stability and robustness of FEx model based control systems. (4) The methods will be tested by real-time applications using a laboratory-scale process as well as pilot-scale processes that are run by commercial Distributed Control Systems(DC S). This last area will take advantage of a long-time collaboration with Prof. Romagnoly of the University of Sydney.