"In this study, an adaptive critic-based neural network is developed for optimal control of structures and implemented on a Multi-Input Multi-output (MIMO) system in an attempt to minimize its vibrations. This MIMO system consists of two vertically suspended cantilever beams interconnected by three blocks of wood and is embedded with piezoelectric sensors and actuators and interfaced to a PC with a DAS 1602 data acquisition card. The system model is first acquired by experimental methods. The adaptive critic-based controller for the system model consists of two multi-layer perceptron networks called the action and the critic networks. The action network outputs the optimal control while the critic network outputs the co-state vector and critiques the control output from the action network. The networks are trained offline for a wide range of initial conditions using the equations obtained by Dynamic Programming methodology. After a required degree of convergence is achieved the resulting action network is used for optimal control in a feedback loop. Since all the states are not available for feedback, a Kalman estimator is designed for state estimation. Using the state estimates the neuro-controller is implemented on the structure and the experimental results are obtained. These results are compared with that of an optimal output feedback controller. A reduced order model of the system is then determined and a neurocontroller is designed for the same and implemented on the system. A robust controller is also designed for the second order system to provide “extra control” when there exists uncertainties in the system that makes the existing controller lose its optimality. The performance characteristic of the reduced order neuro-controller is compared with the full-order neuro-controller and the robustness of the neuro-controller is analyzed"--Abstract, page 3.
Balakrishnan, S. N.
Liou, Frank W.
Mechanical and Aerospace Engineering
M.S. in Mechanical Engineering
National Science Foundation (U.S.)
University of Missouri--Rolla
vii, 66 pages
© 2002 Vijayakumar Janardhan, All rights reserved.
Thesis - Restricted Access
Vibration -- Control
Neural networks (Computer science)
Print OCLC #
Link to Catalog Record
Electronic access to the full-text of this document is restricted to Missouri S&T users. Otherwise, request this publication directly from Missouri S&T Library or contact your local library.http://merlin.lib.umsystem.edu/record=b5072090~S5
Janardhan, Vijayakumar, "Implementation and validation of adaptive-critic based optimal neuro-controller on a multi-input multi-output system" (2002). Masters Theses. 2382.
Share My Thesis If you are the author of this work and would like to grant permission to make it openly accessible to all, please click the button above.