Identification of a flexible manipulator using Neural Network

FAVIO MASSON; GUILLERMO SENTONI; EDUARDO NEBOT

Congreso; 1993 IFAC World Congress; 1993

The need for model identification and/or adaptive control of flexible structures arises because of the lack of knowledge of the values of important internal parameters and their variation under different load conditions. This paper presents a neural-network approach to olYadnthe model of a flexible manipulator that is able to capture its high-order dynamics. A technique using singular value decomposition is presented to optimise the network with respects to the number of inputs and nodes used. Finally an approach to accelerate the convergence of the standard backprnpagntion algorithm is also presented.