Accession Number : ADA320001

Title :   Study of Neuro-Controllers for Motion Control Systems with Distributed Mechanical Flexibility.

Descriptive Note : Technical rept. 15 Sep 92-14 Sep 93,

Corporate Author : ILLINOIS UNIV AT URBANA MECHANICAL ENGINEERING LAB

Personal Author(s) : Cetinkunt, Sabri

PDF Url : ADA320001

Report Date : 03 JAN 1994

Pagination or Media Count : 20

Abstract : Control of motion systems involving distributed mechanical flexibility is studied using artificial neural networks. Infinite dimensional nature of the problem due to distributed flexibility, nonlinear dynamics of mechanical structural systems, and fault tolerant operation requirements are taken into consideration. Three different neuro controller architectures are studied: 1) Hopfield nets for modal parameter estimation and real time solution of LQ optimal control problem, 2) Feedforward nets using EKF learning algorithm as a fast learning, trainable nonlinear adaptive controller, 3) CMAC neural network controller for high precision motion control. Thre results are summarized and details are presented in refereed publications.

Descriptors :   *NEURAL NETS, *ADAPTIVE CONTROL SYSTEMS, *MOTION, ALGORITHMS, REQUIREMENTS, MECHANICAL PROPERTIES, OPTIMIZATION, TRAINING, REAL TIME, DISTRIBUTION, DYNAMICS, PARAMETERS, ESTIMATES, NONLINEAR SYSTEMS, SOLUTIONS(GENERAL), PRECISION, TOLERANCE, ARCHITECTURE, LEARNING, MECHANICAL COMPONENTS, FAULTS.

Subject Categories : Cybernetics

Distribution Statement : APPROVED FOR PUBLIC RELEASE