On-line Multivariable System Identification and Adaptive Control of Industrial Processes

为了在快速变化的经济环境中保持竞争力，通过在线控制和优化来更有效地运营美国的生产设施是提高盈利能力的关键因素。人们普遍认识到，在许多工业对象中，存在许多关键控制回路，其中常规的单回路控制是不充分的，并且存在着对先进控制的显著的经济激励。在第一阶段的研究中，利用加州大学圣巴巴拉分校(UCSB)的计算机模拟和实验过程的数据，论证了使用典型变量分析(CVA)对工业过程进行在线多变量系统辨识的可行性。在第二阶段，整合在线辨识和控制设计将是一大重点。与在线控制设计相结合的CVA方法将在几种类型的过程上进行广泛的演示，并与现有方法进行比较。此外，还将提供将该方法扩展到时变系统的详细开发。系统辨识和控制设计方法的性能将在中试或全规模工业过程中进行实验评估。

To remain competitive in a rapidly changing economic environment, more efficient operation of production facilities in the United States via on-line control and optimization is a key factor in improving profitability. It is widely recognized that in many industrial plants there are a number of critical control loops where conventional single loop control is inadequate and there are significant economic incentives for advanced control. In the Phase I study, the feasibility of on-line multivariable system identification of industrial processes using canonical variate analysis (CVA) was demonstrated on data from computer simulations and experimental processes at the University of California, Santa Barbara (UCSB). The Phase II, the integration of on-line identification and control design will be a major focus. The CVA method integrated with on-line control design will be extensively demonstrated on several types of processes, and compared with existing methods. Furthermore, a detailed development of an extension of the approach to time varying systems will be provided. The performance of the system identification and control design methods will be experimentally evaluated on a pilot- or full- scale industrial process.