Abstract
The control of seismic response of buildings connected by a magnetorheological (MR) damper is studied. The desired control force is obtained using Linear Quadratic Gaussian (LQG) control based on the feedback of states estimated via measured outputs or Optimal Static Output Feedback (OSOF) control using the direct feedback of measured outputs. The damper input voltage is predicted using a Recurrent Neural Network (RNN), which proves more effective than the Clipped Voltage Law (CVL). Various sensor configurations and state weightings are considered to obtain effective control. LQG-RNN/OSOF-RNN yield significant reduction in response and base shear and require much less control effort compared to passive-on control with saturation voltage. Compared to passive-off control, they are very effective in attenuating maximum-peak/RMS responses and storeywise responses of the flexible building, except for max-peak accelerations that increase slightly. However, passive-off control is unable to transfer base shear to the stiffer building. LQG-RNN/OSOF-RNN also yield control at least as effective as LQR-RNN by deploying much fewer sensors but using a somewhat higher damper force. They are mostly comparable to each other, but OSOF-RNN requires an order-of-magnitude less CPU time for the control loop. Effective control is possible using few sensors.
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call