OPTIMAL ADAPTIVE AND PREDICTIVE CONTROL SYSTEM OF BUILDINGS BY MULTI-LAYERED NEURAL NETWORK AND FUZZY THEORY

Abstract

This paper describes an optimal adaptive and predictive control system and its digital simulations for a single-degree-of-freedom system subjected to earthquake loading. In this system, an active mass driver system and an equivalent variable mass system are employed as an active control method. Prediction of earthquake input and structural identification are performed by using multi-layered neural networks based on the error back-propagation method. Off-line training are performed for these neural networks, and future earthquake inputs and structural responses of a structure are predicted Optimization is performed by means of maximizing decision. In maximizing decision, optimal target control variables are determined by using assumed membership functions of target responses. In this system, a certain interval is introduced as a control interval. Prediction of earthquake inputs and structural responses and the maximizing decision are performed with respect to the maximum values in every control interval. Results of digital simulations show the effectiveness of the proposed control system.