一类非方复杂工业系统的数据驱动控制方法研究

项目名称： 一类非方复杂工业系统的数据驱动控制方法研究

项目编号： No.61304035

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 王良勇

作者单位： 东北大学

项目金额： 25万元

中文摘要： 输入变量个数少于输出变量个数的复杂工业系统常具有机理不清、非方和多变量等综合复杂性，其建模和控制是一个具有挑战性的问题。目前很多复杂工业系统尚处于人工控制状态，或者从工程经验角度出发设计控制器结构，没有从理论上进行完整的分析与解释。另外，工业过程中每天都在产生着大量反映生产过程和设备运行的各种数据，还尚未被充分利用。针对上述情形，本项目首先针对单输入两输出的非方复杂工业系统，在无法深入了解被控对象过程机理的情况下，有效利用大量生产数据，研究建立面向控制目标的非方复杂工业系统数学模型；研究利用所建立非线性模型进行系统可操作性分析的方法，为控制结构的选择提供设计依据；进而研究基于虚拟未建模动态补偿和数据驱动的控制方法，并将所提方法在欠驱动机械臂系统和氧化铝制粉系统上进行验证。本项目的顺利实施将为复杂工业系统的控制方案设计提供借鉴，有助于提高复杂工业系统的自动化水平。

中文关键词： 非方；工业系统；数据驱动；可操作性；神经网络

英文摘要： Complex industrial systems with the number of the input variables less than that of output variables have comprehensive complexities such as unclear mechanism, non-square, and multi-variable.Therefore,the modeling and control for such systems is still a challenging problem. Many complex industrial systems are still in manual control stage, or the controller structures of such systems are designed from an engineering point. However, the choice of such control structures requires complete analysis and interpretation. Further, the industrial process generates a large number of data every day, which are related to the production process and have not been fully used. To deal with this problem, the project focuses on a single input two output non-square complex industrial systems.In the case of no insight into the mechanism of the plants, the control object oriented mathematical models are to be constructed for the non-square complex industrial systems through the effective use of the large number of offline and online data. Then the operability analysis method is required to provide design basis for the choice of the control structures and the virtual unmodeled dymamics compensation based data-driven control methods are introduced. Finally, the proposed methods are applied to underactuated robotic manipulators and pu

英文关键词： nonsquare；industrial systems；data driven；operability；neural networks