NEURAL NETWORKS FOR STRUCTURAL CONTROL OF A BENCHMARK PROBLEM, ACTIVETENDON SYSTEM

Methodology for active structural control using neural networks has be en proposed by Ghaboussi and his co-workers(1-8) in the past several y ears. The control algorithm in the mathematically formulated methods i s replaced by a neural network controller (neuro-controller). Neuro-co ntrollers have been developed and applied in linear and nonlinear stru ctural control. Neuro-controllers are trained with the aid of the emul ator neural networks. The emulator neural network is trained to learn the transfer function between the actuator signal and the sensor readi ng and it uses that past values of these quantities to predict the fut ure values of the sensor readings. In this paper, we apply the previou sly developed neuro-control method in the benchmark problem of the act ive tendon system. The emulator neural network is developed and traine d using the evaluation model given in the benchmark problem which is c onsidered to be the true representation of the active tendon system. H owever, a reduced-order model has been developed and used, along with the emulator neural network, to train the neuro-controller. The evalua tion model represents the three story steel frame structure, including the actuator dynamics. The absolute acceleration of the first floor a nd the actuator piston displacement are used as feedback. Three neuro- controllers, with different control criteria, have been developed and their performances have been evaluated with the prescribed performance s indexes. The robustness of the neuro-controllers in the presence of some severe uncertainties, has also been evaluated. (C) 1998 John Wile y & Sons, Ltd.