Lateral-torsional seismic response control of two single-storey asymmetric plan buildings, interconnected using multiple magnetorheological (MR) dampers, is studied. LQR control is used to obtain desired control forces. The desired damper force, for the two-damper case studied herein, is obtained using least square minimization. Command voltages are predicted using either a Recurrent Neural Network (RNN) or a Clipped Voltage Law (CVL). Effective controllers are obtained based on performance criteria, by varying damper configurations. LQR-CVL and LQR-RNN prove considerably more effective than Passive-off control for response reduction of flexible building B1, but not so for rigid building B2. They yield a redistribution of base shear and torque between the buildings. When compared to Passive-on control their performance ranges from superior to comparable, except for torsional acceleration of B2 for which it is inferior. They yield considerable reduction in peak base shear/torque, and require much less power, compared to Passive-on control where saturation voltage is applied. LQR-RNN is somewhat more effective than LQR-CVL in response attenuation.

This content is only available via PDF.
You do not currently have access to this content.