Soil-structure interaction effects on the semi-active control effectiveness of structures using MR damper: Shaking table investigation

Abstract

To investigate the soil-structure interaction (SSI) effects on the semi-active control effectiveness, shaking table tests were carried out on a two-story steel frame structure with magnetorheological (MR) damper located on a fixed-base and soft soil foundation. First, a novel online learning deep neural network fuzzy control method (OL-DFNN) was used to control the structure equipped with MR dampers. The controller was made up of three modules, namely the offline trained deep neural network fuzzy control (DFNN), the online training module and the control structure system with SSI effects. Second, other control methods (passive ON, ON-OFF, and DFNN) were also examined to investigate the control efficiency when considering the SSI. Three different earthquake excitations were selected as the seismic inputs, recorded from El Centro, Kobe, and Tianjin earthquakes. Finally, shaking table tests of the structures with and without considering the SSI effect were carried out. The displacement and acceleration test results were comparatively evaluated using two performance indices: peak and root mean square (RMS) responses. The test results indicated that OL-DFNN control achieved the best control efficiency among the four control strategies for both fixed-base structures and soft soil foundation structures. Compared to fixed-base structures, the semi-active control effectiveness of the structures where the SSI effect is taken into account was reduced.