Abstract
With the real-time remote seismic monitoring system being established in several countries including Japan, a new preview-type active control system of structural systems subject to seismic disturbances has been proposed by the authors(1). The preview-type control uses the estimated future seismic waveform obtained by seismic wave data in remote observation sites and an AI-based future wave estimation system. In this study, we consider that multiplexing the preview-type control law with the remote waveform data in multiple monitoring sites improves not only the robustness of the control system in the case of the system down in some sites, but also the control performance by switching or scheduling multiple preview-type control laws. We propose a hierarchical method for scheduling multiple preview-type control laws so that the above system down does not impact crucially in carrying out the switching or scheduling control. We refer to the control method with the proposed scheduling law as the ensemble control. In the ensemble control based on the scheduling method, some schedulers to decide the switching or scheduling are needed. In this study, the schedulers are designed as artificial neural networks whose input data are the state-vector, the controlled output signal of the controlled structural system, and the available seismic disturbance, etc. We show the advantage of the present control methodology with a simulation example using a recorded seismic event.
