ACTIVE CONTROL OF STRUCTURES USING NEURAL NETWORKS

A new neural-network-based control algorithm has been developed and te sted in the computer simulation of active control of a three-story fra me structure subjected to ground excitations. First, an emulator neura l network has been trained to forecast the future response of the stru cture from the immediate history of the system's response, which consi sts of the structure plus an actuator. The trained emulator has been u sed in predicting the future responses and in evaluating the sensitivi ties of the control signal with respect to those responses. At each ti me step of the simulation, the control signal has been adjusted to ind uce the required control force in the actuator based in a control crit erion. A controller neural network has been trained to learn the relat ion between the immediate history of response of the structure and act uator, and the adjusted control signals. The trained neurocontroller h as been used in controlling the structure for different dynamic loadin g conditions. Results of this initial study indicate that the neural-n etwork-based control algorithms have the promise of evolving into powe rful adaptive controllers after further research.