Robust Generalized Minimum Variance Controller Using Neural Network for Civil Engineering Problems

This paper presents a robustness of the proposed generalized minimum variance algorithm. The main idea is to use artificial neural network for generalization of the GMV. This will give a neural network-based control method wich can be applied to civil engineering structures. The neural network learns the control task from an already existing controller, which is the generalized minimum variance (GMV) controller. The objective is to take advantage of the generalization capabilities and the nonlinear behavior of neural networks in order to overcome the limitations of the existing controller and even to improve its performances. Simulation results demonstrate the robustness of this algorithm and its capability to compensate the structural parameter variations and seismic ground motion.