• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1995.10b
• /
• pp.243-248
• /
• 1995

본 논문에서는 신경회로망의 학습능력을 이용하여 AC 모터의 속도제어에 이용된 기 존의 PI제어기의 문제점을 보완하고자 한다. 기존의 아날로그 PI제어기에서는 각 비례, 적분 파라메타를 개발자가 조정하여 고정하면 부하가 변동될 경우 적응성이 떨어지는 문제점을 안고 있었다. 본 논문에서 제시된 디지털 신경망제어기는 학습을 통해 새로운 환경에 적응 가능하다는 점에 가정하여 설계하고 성능을 비교 평가하였다. 본 논문에서 사용된 신경회로 망의 구조는 신경망중에서 가장 범용적으로 사용되는 다층 퍼셉트론 모델구조를 선택하였 다. 신경망 제어기장치로는 인텔 8097 마이크로 콘트롤러를 이용하였다.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1997.04a
• /
• pp.101-106
• /
• 1997

This paper proposes a stable adaptive neural network control(NNC) for fixable joint manipulators. For designing the stable adaptive NNC, the flexible system dynamics is separated into fast and slow subdynamics according to singular perturbation concept. For the slow subdynamics, an adaptive NNC is designed to warrant the system stability and NN learning by lyapunov stability criterion. And to stabilize the fast dynamics, derivative control loop is installed. Through numerical simulation, the performance of the proposed NNC was compared to that of an adaptive controller designed based on the knowledge of the system dynamics. The proposed NNC shows much improvement over the conventional adaptive controller.

• Proceedings of the KSME Conference
• /
• 2003.11a
• /
• pp.1602-1607
• /
• 2003

Vibration causes noise and sometimes makes structure unstable. Especially, due to the efforts of lightening, deformation of flexible structure is increased in its shape. Just a little disturbance can cause vibration and low damping ratio makes residual vibration last long time. This research is concerned with the model reference neuro-controller design for the vibration suppression of smart structures. By using a model reference neurocontroller, which is one of the algorithms of adaptive control, we performed an adaptive control of flexible cantilever plate and opened box structure with piezoelectric materials. The proposed adaptive vibration control algorithm, a model reference neuro-controller, was proved in its effectiveness by applying to an opened box structure. The model reference neuro-controller is implemented with DSP, and the real-time adaptive vibration control experiment results confirm that the model reference neuro-controller is reliable.