階層型ニューラルネットワークによる構造-AMD連成システムの振動制御に関する研究

抄録

Multi-layer neural network has been regarded as a useful algorithm in identification and dynamic control of structures. As one of the existing researches, the active vibration control for structure-actuator coupled system consisted of a tendon system controlled by a signal hydraulic actuator using multi-layer neural networks has been studied. Active Mass Driver (AMD), as another kind of control device, has been studied on the concept of conventional control methods and control design. In this paper, a coupled system consisting of a four-story frame structure and an AMD driven by a electrohydraulic actuator was investigated. Firstly, considering the dynamics of electrohydraulic actuator, the dynamic equation of a structure-AMD coupled system was formulated. Secondly, an emulator neural network with suitably chosen input variables corresponding to the structure-AMD coupled system was established and trained for the purpose of non-parametric identification. Lastly, a neurocontroller was set up and trained with the aid ofthe trained emulator neural network to control the dynamic response of the structure-AMD coupled system subjected to earthquake loading. Because accelerometers can readily provide reliable and inexpensive measurement of absolute structural acceleration at strategic points on a structure, development of identification method based on acceleration feedback is presented. The effectiveness of the identification and control is evaluated through numerical simulations.