TSUJINO Taro Faculty of Engineering, Fukuoka Institute of Technology, Associate Professor, 工学部, 助教授 (00227406)
下村 卓 大阪大学, 大学院・基礎工学研究科, 助手 (40243191)
小原 敦美 大阪大学, 大学院・基礎工学研究科, 助教授 (90221168)
|Budget Amount *help
¥3,500,000 (Direct Cost : ¥3,500,000)
Fiscal Year 1999 : ¥500,000 (Direct Cost : ¥500,000)
Fiscal Year 1998 : ¥500,000 (Direct Cost : ¥500,000)
Fiscal Year 1997 : ¥2,500,000 (Direct Cost : ¥2,500,000)
For the last three years we have tackled several problems in order to obtain a robust control design method for vibration control of a magnetic levitation system from the viewpoint of practicality and inverse problem.
First, as its basis, we have established a consistent method to estimate unmodeled dynamics using closed loop identification, and then design a vibration control system using HィイD2∞ィエD2 control theory. As a result, it turned out that it is vital for suppression of spillover to shape the frequency gain characteristics of the closed-loop transfer function from sensor noise to output, called as the complementary sensitivity function, so as to have that of a notch filter.
Based on this observation, we have reconsidered the basic results obtained above form the viewpoint of practicality, and tried to extend the existing useful ILQ design method so as to treat the above frequency shaping problem. The key to it lies in the extension of a usual state observer used in ILQ design to linear function observer by introducing an additional design freedom. In both cases of full-order and minimal-order state observers, we have clarified analytically the class of complementary sensitivity functions that can be shaped by the ILQ controller with such a design freedom of observer. We then proposed a novel frequency shaping method using this type of controllers, and confirmed its usefulness experimentally. As a result, we have succeeded in designing a low-order controller of dimension as reduced by one third, compared with the HィイD2∞ィエD2 controller. In the case of minimal order observer, moreover, we have proposed a two-step design method achieving nominal performance first and robust performance afterward by utilizing an additional design freedom. In this design, the resulting controller turned out to have lower dimension, compared with the one based on a full-order observer.