Abstract
This paper introduces a new robust neural network methodology for vibration mitigation of tall building under wind excitation. The building considered is a 76‐storey 306 m concrete office tower controlled by an active tuned mass damper (ATMD). The control strategy proposes two neural network (NN) models functioning together online to control the building. The first NN predicts the building's future responses and the second mimics the inverse dynamics of the building control force model, and operates on the future response with set of if–then rules to issue the required control force. Next, both NN models operate jointly to train a separate feed‐forward neural network model designated as the NN controller (NNC). The efficacy of the two neural networks as well as the NN controller are examined and compared with a linear quadratic Gaussian controller. The controllers' performance and robustness are verified and presented, with stiffness uncertainty, time‐delay, measurements' noise and some sensor failure. The results reveal the robustness and the effectiveness of the proposed method in alleviating wind‐induced vibration for tall buildings. Copyright © 2005 John Wiley & Sons, Ltd.

This publication has 15 references indexed in Scilit: