Initial Investigation of Modal Parameter Estimation for Flexible Manipulator Using Hopfield Network

A type of recurrent artificial neural network (ANN) is studied for identification problems of modal parameter of linear structural systems. This recurrent ANN model is closely related to the analog Hopfield network operated in synchronous mode where the connection strengths of the ANN are determined from the system state measurements at each sampling time. The states of neurons represent the modal parameters of the linear structural system to be identified. The recurrent ANN model preserves both the parallelism and the distributed processing nature of the analog Hopfield model as well as asymptotically improving the least squares estimation error. Therefore, it is a good candidate for use in real-time control as a parametric estimator, in particular, for large dimensional distributed parameter systems. Simulation results for the estimation of modal parameters of a flexible beam are presented to determine the effectiveness of this ANN architecture.