VIBRATION CONTROL OF HIGH-RISE BUILDINGS USING NEURAL NETWORK TOKYO DENKI UNIVERSITY, DEPARTMENT OF MECHANICAL ENGINEERING

点击翻译按钮获取中文摘要

For mitigation of vibration induced in high-rise buildings subjected to wind or earthquake, an active mass damper was manufactured for a building model, which was equipped with a linear moter as an actuator and was digitally controlled by means of optimal feedback theory. In addition, an alternative control method using neural network model was considered in this program. In the first year of this project, the dynamic tests for the three story steel framed building model uisng a shaking table were carried out to investigate the performance of the optimally controlled active mass damper. The results showed that the responses of the building model were sufficiently reduced to about a half of the uncontrolled responses due to earthquake inputs. In the following year, an analytical model with the building model for the active mass damper was established to examine the dynamic responses of that subjected to several types of wind and earthquake. As a result, the fundamental design data on the active mass damper were obtained. Consequently, a neural network control system was built system was built on a 32 bit personal computer. This system comprises System Neural Network for system identification and Control Neural Network for obtaining optimal control forces by means of minimizing a criterion function in quadratic form. In computer simulation, even though the performance of vibration control was still worse than the result obtained by applying the optimal control theory, the usefulness of the present system was verified.